{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.7943758192000635,
  "global_step": 40000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.984126984126984e-06,
      "loss": 11.0771,
      "theoretical_loss": 20.841282884190854,
      "tokens_seen": 65536
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.968253968253968e-06,
      "loss": 11.0714,
      "theoretical_loss": 17.594666186972347,
      "tokens_seen": 131072
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.9523809523809525e-06,
      "loss": 10.9477,
      "theoretical_loss": 15.967942175480273,
      "tokens_seen": 196608
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.936507936507936e-06,
      "loss": 10.6658,
      "theoretical_loss": 14.920781838632275,
      "tokens_seen": 262144
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.92063492063492e-06,
      "loss": 10.3389,
      "theoretical_loss": 14.16468202040668,
      "tokens_seen": 327680
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.1904761904761905e-05,
      "loss": 10.0572,
      "theoretical_loss": 13.581026555193928,
      "tokens_seen": 393216
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.3888888888888888e-05,
      "loss": 9.8012,
      "theoretical_loss": 13.110274084044894,
      "tokens_seen": 458752
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.5873015873015872e-05,
      "loss": 9.6269,
      "theoretical_loss": 12.718594708127029,
      "tokens_seen": 524288
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.7857142857142855e-05,
      "loss": 9.5127,
      "theoretical_loss": 12.385057546517329,
      "tokens_seen": 589824
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.984126984126984e-05,
      "loss": 9.3822,
      "theoretical_loss": 12.095877689678783,
      "tokens_seen": 655360
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.1825396825396824e-05,
      "loss": 9.3134,
      "theoretical_loss": 11.841531313745381,
      "tokens_seen": 720896
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.380952380952381e-05,
      "loss": 9.19,
      "theoretical_loss": 11.615184291350435,
      "tokens_seen": 786432
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.5793650793650793e-05,
      "loss": 9.1376,
      "theoretical_loss": 11.411779222830493,
      "tokens_seen": 851968
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.7777777777777776e-05,
      "loss": 9.0561,
      "theoretical_loss": 11.227476784755577,
      "tokens_seen": 917504
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.9761904761904762e-05,
      "loss": 8.9674,
      "theoretical_loss": 11.059299000621618,
      "tokens_seen": 983040
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.1746031746031745e-05,
      "loss": 8.9181,
      "theoretical_loss": 10.904893169100655,
      "tokens_seen": 1048576
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.373015873015873e-05,
      "loss": 8.9953,
      "theoretical_loss": 10.76237082270643,
      "tokens_seen": 1114112
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.571428571428571e-05,
      "loss": 8.7793,
      "theoretical_loss": 10.630194958873984,
      "tokens_seen": 1179648
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.76984126984127e-05,
      "loss": 8.7566,
      "theoretical_loss": 10.507099255000346,
      "tokens_seen": 1245184
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.968253968253968e-05,
      "loss": 8.7115,
      "theoretical_loss": 10.392029026407034,
      "tokens_seen": 1310720
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.1666666666666665e-05,
      "loss": 8.487,
      "theoretical_loss": 10.28409730354045,
      "tokens_seen": 1376256
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.365079365079365e-05,
      "loss": 8.5783,
      "theoretical_loss": 10.182551635913724,
      "tokens_seen": 1441792
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.563492063492063e-05,
      "loss": 8.5185,
      "theoretical_loss": 10.086748644376303,
      "tokens_seen": 1507328
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.761904761904762e-05,
      "loss": 8.4859,
      "theoretical_loss": 9.996134261483984,
      "tokens_seen": 1572864
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.96031746031746e-05,
      "loss": 8.3094,
      "theoretical_loss": 9.910228209036813,
      "tokens_seen": 1638400
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.1587301587301586e-05,
      "loss": 8.4173,
      "theoretical_loss": 9.828611674184266,
      "tokens_seen": 1703936
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.357142857142857e-05,
      "loss": 8.3191,
      "theoretical_loss": 9.750917429551151,
      "tokens_seen": 1769472
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.555555555555555e-05,
      "loss": 8.3262,
      "theoretical_loss": 9.67682184172525,
      "tokens_seen": 1835008
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.753968253968254e-05,
      "loss": 8.1858,
      "theoretical_loss": 9.606038353789344,
      "tokens_seen": 1900544
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.9523809523809524e-05,
      "loss": 8.1585,
      "theoretical_loss": 9.538312129408556,
      "tokens_seen": 1966080
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.15079365079365e-05,
      "loss": 8.1146,
      "theoretical_loss": 9.47341562027564,
      "tokens_seen": 2031616
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.349206349206349e-05,
      "loss": 8.0534,
      "theoretical_loss": 9.41114487355416,
      "tokens_seen": 2097152
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.547619047619048e-05,
      "loss": 7.962,
      "theoretical_loss": 9.351316436877685,
      "tokens_seen": 2162688
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.746031746031745e-05,
      "loss": 7.8543,
      "theoretical_loss": 9.29376474930398,
      "tokens_seen": 2228224
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.944444444444444e-05,
      "loss": 8.0138,
      "theoretical_loss": 9.238339930088461,
      "tokens_seen": 2293760
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.142857142857142e-05,
      "loss": 7.7162,
      "theoretical_loss": 9.184905895151996,
      "tokens_seen": 2359296
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.341269841269842e-05,
      "loss": 7.7563,
      "theoretical_loss": 9.133338745057056,
      "tokens_seen": 2424832
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.53968253968254e-05,
      "loss": 7.5792,
      "theoretical_loss": 9.083525379177438,
      "tokens_seen": 2490368
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.738095238095239e-05,
      "loss": 7.591,
      "theoretical_loss": 9.035362299287534,
      "tokens_seen": 2555904
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.936507936507937e-05,
      "loss": 7.6518,
      "theoretical_loss": 8.988754572553061,
      "tokens_seen": 2621440
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.134920634920635e-05,
      "loss": 7.4322,
      "theoretical_loss": 8.943614929284315,
      "tokens_seen": 2686976
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.333333333333333e-05,
      "loss": 7.4551,
      "theoretical_loss": 8.899862975121929,
      "tokens_seen": 2752512
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.531746031746032e-05,
      "loss": 7.24,
      "theoretical_loss": 8.857424500797205,
      "tokens_seen": 2818048
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.73015873015873e-05,
      "loss": 7.2072,
      "theoretical_loss": 8.816230875422118,
      "tokens_seen": 2883584
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.928571428571429e-05,
      "loss": 7.3823,
      "theoretical_loss": 8.776218511555161,
      "tokens_seen": 2949120
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.126984126984126e-05,
      "loss": 7.1742,
      "theoretical_loss": 8.737328392164539,
      "tokens_seen": 3014656
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.325396825396825e-05,
      "loss": 7.1599,
      "theoretical_loss": 8.699505651152553,
      "tokens_seen": 3080192
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.523809523809524e-05,
      "loss": 7.1427,
      "theoretical_loss": 8.66269920037918,
      "tokens_seen": 3145728
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.722222222222223e-05,
      "loss": 7.0218,
      "theoretical_loss": 8.62686139718003,
      "tokens_seen": 3211264
    },
    {
      "epoch": 0.0,
      "objective/train/advantage_avg": 0.49531280994415283,
      "objective/train/docs_used": 8371,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 6.728449821472168,
      "objective/train/original_loss": 6.728449821472168,
      "objective/train/theoretical_loss": 8.591947747254773,
      "objective/train/tokens_used": 23736800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24548619985580444,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507793426513672,
      "objective/train/weighted_lm_loss": 7.069992542266846,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0216598510742188,
      "theoretical_loss": 8.591947747254773,
      "tokens_seen": 3276800
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.92063492063492e-05,
      "loss": 6.9774,
      "theoretical_loss": 8.591947747254773,
      "tokens_seen": 3276800
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001011904761904762,
      "loss": 6.8521,
      "theoretical_loss": 8.55791663853882,
      "tokens_seen": 3342336
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00010317460317460317,
      "loss": 6.7841,
      "theoretical_loss": 8.524729102289708,
      "tokens_seen": 3407872
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00010515873015873016,
      "loss": 6.8152,
      "theoretical_loss": 8.492348598140893,
      "tokens_seen": 3473408
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00010714285714285714,
      "loss": 6.7547,
      "theoretical_loss": 8.460740820316484,
      "tokens_seen": 3538944
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00010912698412698413,
      "loss": 6.7581,
      "theoretical_loss": 8.429873522574422,
      "tokens_seen": 3604480
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001111111111111111,
      "loss": 6.7221,
      "theoretical_loss": 8.399716359763914,
      "tokens_seen": 3670016
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00011309523809523809,
      "loss": 6.6347,
      "theoretical_loss": 8.370240744154781,
      "tokens_seen": 3735552
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00011507936507936508,
      "loss": 6.6158,
      "theoretical_loss": 8.341419714929035,
      "tokens_seen": 3801088
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00011706349206349207,
      "loss": 6.4159,
      "theoretical_loss": 8.313227819424887,
      "tokens_seen": 3866624
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00011904761904761905,
      "loss": 6.4367,
      "theoretical_loss": 8.285641004895568,
      "tokens_seen": 3932160
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00012103174603174604,
      "loss": 6.5093,
      "theoretical_loss": 8.25863651969382,
      "tokens_seen": 3997696
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.000123015873015873,
      "loss": 6.3352,
      "theoretical_loss": 8.232192822921675,
      "tokens_seen": 4063232
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.000125,
      "loss": 6.3951,
      "theoretical_loss": 8.206289501696752,
      "tokens_seen": 4128768
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00012698412698412698,
      "loss": 6.4355,
      "theoretical_loss": 8.180907195283321,
      "tokens_seen": 4194304
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00012896825396825398,
      "loss": 6.2604,
      "theoretical_loss": 8.156027525421205,
      "tokens_seen": 4259840
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00013095238095238096,
      "loss": 6.2779,
      "theoretical_loss": 8.131633032259415,
      "tokens_seen": 4325376
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00013293650793650793,
      "loss": 6.3027,
      "theoretical_loss": 8.107707115366345,
      "tokens_seen": 4390912
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001349206349206349,
      "loss": 6.2102,
      "theoretical_loss": 8.084233979345122,
      "tokens_seen": 4456448
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001369047619047619,
      "loss": 6.1439,
      "theoretical_loss": 8.061198583632713,
      "tokens_seen": 4521984
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001388888888888889,
      "loss": 6.1627,
      "theoretical_loss": 8.03858659610554,
      "tokens_seen": 4587520
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00014087301587301586,
      "loss": 5.9802,
      "theoretical_loss": 8.016384350153084,
      "tokens_seen": 4653056
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00014285714285714284,
      "loss": 6.1413,
      "theoretical_loss": 7.9945788049155055,
      "tokens_seen": 4718592
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00014484126984126984,
      "loss": 6.2189,
      "theoretical_loss": 7.973157508411693,
      "tokens_seen": 4784128
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00014682539682539685,
      "loss": 6.0561,
      "theoretical_loss": 7.952108563311223,
      "tokens_seen": 4849664
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00014880952380952382,
      "loss": 6.1682,
      "theoretical_loss": 7.931420595127772,
      "tokens_seen": 4915200
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001507936507936508,
      "loss": 6.0328,
      "theoretical_loss": 7.911082722632908,
      "tokens_seen": 4980736
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001527777777777778,
      "loss": 5.9525,
      "theoretical_loss": 7.891084530308333,
      "tokens_seen": 5046272
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00015476190476190478,
      "loss": 6.0668,
      "theoretical_loss": 7.871416042671642,
      "tokens_seen": 5111808
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00015674603174603175,
      "loss": 6.054,
      "theoretical_loss": 7.8520677003260175,
      "tokens_seen": 5177344
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00015873015873015873,
      "loss": 5.8692,
      "theoretical_loss": 7.83303033759787,
      "tokens_seen": 5242880
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00016071428571428573,
      "loss": 5.9204,
      "theoretical_loss": 7.8142951616387775,
      "tokens_seen": 5308416
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001626984126984127,
      "loss": 5.8646,
      "theoretical_loss": 7.795853732879069,
      "tokens_seen": 5373952
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00016468253968253969,
      "loss": 6.0267,
      "theoretical_loss": 7.777697946730324,
      "tokens_seen": 5439488
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00016666666666666666,
      "loss": 6.0271,
      "theoretical_loss": 7.759820016443023,
      "tokens_seen": 5505024
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00016865079365079366,
      "loss": 5.9168,
      "theoretical_loss": 7.742212457033625,
      "tokens_seen": 5570560
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00017063492063492064,
      "loss": 5.8456,
      "theoretical_loss": 7.724868070202639,
      "tokens_seen": 5636096
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00017261904761904762,
      "loss": 5.7787,
      "theoretical_loss": 7.707779930171886,
      "tokens_seen": 5701632
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001746031746031746,
      "loss": 5.7974,
      "theoretical_loss": 7.690941370375033,
      "tokens_seen": 5767168
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001765873015873016,
      "loss": 5.8043,
      "theoretical_loss": 7.674345970940984,
      "tokens_seen": 5832704
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00017857142857142857,
      "loss": 5.7159,
      "theoretical_loss": 7.6579875469145655,
      "tokens_seen": 5898240
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00018055555555555555,
      "loss": 5.8569,
      "theoretical_loss": 7.641860137163401,
      "tokens_seen": 5963776
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00018253968253968252,
      "loss": 5.8218,
      "theoretical_loss": 7.6259579939239845,
      "tokens_seen": 6029312
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00018452380952380953,
      "loss": 5.8325,
      "theoretical_loss": 7.610275572943593,
      "tokens_seen": 6094848
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001865079365079365,
      "loss": 5.7266,
      "theoretical_loss": 7.5948075241780755,
      "tokens_seen": 6160384
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001884920634920635,
      "loss": 5.7308,
      "theoretical_loss": 7.579548683008654,
      "tokens_seen": 6225920
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00019047619047619048,
      "loss": 5.8053,
      "theoretical_loss": 7.564494061943624,
      "tokens_seen": 6291456
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00019246031746031748,
      "loss": 5.7645,
      "theoretical_loss": 7.549638842773467,
      "tokens_seen": 6356992
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00019444444444444446,
      "loss": 5.6472,
      "theoretical_loss": 7.534978369150202,
      "tokens_seen": 6422528
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00019642857142857144,
      "loss": 5.6673,
      "theoretical_loss": 7.520508139563954,
      "tokens_seen": 6488064
    },
    {
      "epoch": 0.0,
      "objective/train/advantage_avg": 0.4908171594142914,
      "objective/train/docs_used": 9704,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 5.399485111236572,
      "objective/train/original_loss": 5.3994855880737305,
      "objective/train/theoretical_loss": 7.5062238006917354,
      "objective/train/tokens_used": 27013600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2412356734275818,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503079891204834,
      "objective/train/weighted_lm_loss": 5.670967102050781,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 7.5062238006917354,
      "tokens_seen": 6553600
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0001984126984126984,
      "loss": 5.7328,
      "theoretical_loss": 7.5062238006917354,
      "tokens_seen": 6553600
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00020039682539682542,
      "loss": 5.7643,
      "theoretical_loss": 7.492121141095167,
      "tokens_seen": 6619136
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002023809523809524,
      "loss": 5.5097,
      "theoretical_loss": 7.478196085245621,
      "tokens_seen": 6684672
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00020436507936507937,
      "loss": 5.6864,
      "theoretical_loss": 7.464444687856724,
      "tokens_seen": 6750208
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00020634920634920634,
      "loss": 5.7318,
      "theoretical_loss": 7.45086312850561,
      "tokens_seen": 6815744
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00020833333333333335,
      "loss": 5.4749,
      "theoretical_loss": 7.43744770652558,
      "tokens_seen": 6881280
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00021031746031746032,
      "loss": 5.5339,
      "theoretical_loss": 7.42419483615405,
      "tokens_seen": 6946816
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002123015873015873,
      "loss": 5.6535,
      "theoretical_loss": 7.4111010419207375,
      "tokens_seen": 7012352
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00021428571428571427,
      "loss": 5.7321,
      "theoretical_loss": 7.398162954262078,
      "tokens_seen": 7077888
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00021626984126984128,
      "loss": 5.5421,
      "theoretical_loss": 7.385377305348793,
      "tokens_seen": 7143424
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00021825396825396825,
      "loss": 5.5598,
      "theoretical_loss": 7.3727409251143925,
      "tokens_seen": 7208960
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00022023809523809523,
      "loss": 5.5974,
      "theoretical_loss": 7.360250737473224,
      "tokens_seen": 7274496
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002222222222222222,
      "loss": 5.5628,
      "theoretical_loss": 7.347903756717382,
      "tokens_seen": 7340032
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002242063492063492,
      "loss": 5.5724,
      "theoretical_loss": 7.335697084082529,
      "tokens_seen": 7405568
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00022619047619047618,
      "loss": 5.6969,
      "theoretical_loss": 7.323627904473269,
      "tokens_seen": 7471104
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00022817460317460316,
      "loss": 5.3875,
      "theoretical_loss": 7.311693483339357,
      "tokens_seen": 7536640
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00023015873015873016,
      "loss": 5.4919,
      "theoretical_loss": 7.299891163694537,
      "tokens_seen": 7602176
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00023214285714285717,
      "loss": 5.5971,
      "theoretical_loss": 7.288218363270328,
      "tokens_seen": 7667712
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00023412698412698414,
      "loss": 5.5551,
      "theoretical_loss": 7.276672571797575,
      "tokens_seen": 7733248
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00023611111111111112,
      "loss": 5.5232,
      "theoretical_loss": 7.265251348408967,
      "tokens_seen": 7798784
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002380952380952381,
      "loss": 5.6741,
      "theoretical_loss": 7.253952319156202,
      "tokens_seen": 7864320
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002400793650793651,
      "loss": 5.5124,
      "theoretical_loss": 7.242773174635825,
      "tokens_seen": 7929856
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00024206349206349207,
      "loss": 5.6728,
      "theoretical_loss": 7.2317116677181055,
      "tokens_seen": 7995392
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00024404761904761905,
      "loss": 5.3354,
      "theoretical_loss": 7.22076561137372,
      "tokens_seen": 8060928
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.000246031746031746,
      "loss": 5.5085,
      "theoretical_loss": 7.2099328765932205,
      "tokens_seen": 8126464
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.000248015873015873,
      "loss": 5.5368,
      "theoretical_loss": 7.1992113903946695,
      "tokens_seen": 8192000
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00025,
      "loss": 5.4071,
      "theoretical_loss": 7.188599133914977,
      "tokens_seen": 8257536
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.000251984126984127,
      "loss": 5.4527,
      "theoretical_loss": 7.178094140580843,
      "tokens_seen": 8323072
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00025396825396825396,
      "loss": 5.36,
      "theoretical_loss": 7.167694494355343,
      "tokens_seen": 8388608
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00025595238095238096,
      "loss": 5.3949,
      "theoretical_loss": 7.157398328056499,
      "tokens_seen": 8454144
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00025793650793650796,
      "loss": 5.4218,
      "theoretical_loss": 7.147203821744321,
      "tokens_seen": 8519680
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00025992063492063497,
      "loss": 5.4023,
      "theoretical_loss": 7.1371092011730495,
      "tokens_seen": 8585216
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002619047619047619,
      "loss": 5.3711,
      "theoretical_loss": 7.127112736305475,
      "tokens_seen": 8650752
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002638888888888889,
      "loss": 5.2684,
      "theoretical_loss": 7.117212739886398,
      "tokens_seen": 8716288
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00026587301587301587,
      "loss": 5.2483,
      "theoretical_loss": 7.107407566072454,
      "tokens_seen": 8781824
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00026785714285714287,
      "loss": 5.5102,
      "theoretical_loss": 7.097695609115671,
      "tokens_seen": 8847360
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002698412698412698,
      "loss": 5.3072,
      "theoretical_loss": 7.0880753020982725,
      "tokens_seen": 8912896
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002718253968253968,
      "loss": 5.1466,
      "theoretical_loss": 7.0785451157163735,
      "tokens_seen": 8978432
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002738095238095238,
      "loss": 5.439,
      "theoretical_loss": 7.06910355711033,
      "tokens_seen": 9043968
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00027579365079365083,
      "loss": 5.2897,
      "theoretical_loss": 7.059749168739638,
      "tokens_seen": 9109504
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002777777777777778,
      "loss": 5.3691,
      "theoretical_loss": 7.050480527300383,
      "tokens_seen": 9175040
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002797619047619048,
      "loss": 5.3555,
      "theoretical_loss": 7.041296242683314,
      "tokens_seen": 9240576
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00028174603174603173,
      "loss": 5.3703,
      "theoretical_loss": 7.032194956970776,
      "tokens_seen": 9306112
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00028373015873015873,
      "loss": 5.3154,
      "theoretical_loss": 7.023175343470762,
      "tokens_seen": 9371648
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002857142857142857,
      "loss": 5.3311,
      "theoretical_loss": 7.014236105786485,
      "tokens_seen": 9437184
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002876984126984127,
      "loss": 5.2603,
      "theoretical_loss": 7.005375976919894,
      "tokens_seen": 9502720
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002896825396825397,
      "loss": 5.3003,
      "theoretical_loss": 6.99659371840772,
      "tokens_seen": 9568256
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002916666666666667,
      "loss": 5.3527,
      "theoretical_loss": 6.987888119488599,
      "tokens_seen": 9633792
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002936507936507937,
      "loss": 5.3101,
      "theoretical_loss": 6.979257996300014,
      "tokens_seen": 9699328
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00029563492063492064,
      "loss": 5.27,
      "theoretical_loss": 6.970702191103734,
      "tokens_seen": 9764864
    },
    {
      "epoch": 0.0,
      "objective/train/advantage_avg": 0.49107539653778076,
      "objective/train/docs_used": 11015,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 5.094058036804199,
      "objective/train/original_loss": 5.094058036804199,
      "objective/train/theoretical_loss": 6.962219571538605,
      "objective/train/tokens_used": 30290400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24188564717769623,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503370761871338,
      "objective/train/weighted_lm_loss": 5.350419521331787,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 6.962219571538605,
      "tokens_seen": 9830400
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00029761904761904765,
      "loss": 5.3296,
      "theoretical_loss": 6.962219571538605,
      "tokens_seen": 9830400
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002996031746031746,
      "loss": 5.3081,
      "theoretical_loss": 6.953809029899522,
      "tokens_seen": 9895936
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003015873015873016,
      "loss": 5.2714,
      "theoretical_loss": 6.945469482441503,
      "tokens_seen": 9961472
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00030357142857142855,
      "loss": 5.2838,
      "theoretical_loss": 6.937199868707852,
      "tokens_seen": 10027008
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003055555555555556,
      "loss": 5.3265,
      "theoretical_loss": 6.928999150881393,
      "tokens_seen": 10092544
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00030753968253968255,
      "loss": 5.2493,
      "theoretical_loss": 6.920866313157865,
      "tokens_seen": 10158080
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00030952380952380956,
      "loss": 5.3069,
      "theoretical_loss": 6.912800361140576,
      "tokens_seen": 10223616
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003115079365079365,
      "loss": 5.4451,
      "theoretical_loss": 6.904800321255455,
      "tokens_seen": 10289152
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003134920634920635,
      "loss": 5.1631,
      "theoretical_loss": 6.896865240185702,
      "tokens_seen": 10354688
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00031547619047619046,
      "loss": 5.2823,
      "theoretical_loss": 6.888994184325256,
      "tokens_seen": 10420224
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00031746031746031746,
      "loss": 5.2868,
      "theoretical_loss": 6.881186239250335,
      "tokens_seen": 10485760
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003194444444444444,
      "loss": 5.2119,
      "theoretical_loss": 6.873440509208347,
      "tokens_seen": 10551296
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00032142857142857147,
      "loss": 5.414,
      "theoretical_loss": 6.8657561166234995,
      "tokens_seen": 10616832
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003234126984126984,
      "loss": 5.0445,
      "theoretical_loss": 6.858132201618454,
      "tokens_seen": 10682368
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003253968253968254,
      "loss": 5.1709,
      "theoretical_loss": 6.8505679215514235,
      "tokens_seen": 10747904
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00032738095238095237,
      "loss": 5.1904,
      "theoretical_loss": 6.8430624505681,
      "tokens_seen": 10813440
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00032936507936507937,
      "loss": 5.2185,
      "theoretical_loss": 6.835614979167886,
      "tokens_seen": 10878976
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003313492063492063,
      "loss": 5.0221,
      "theoretical_loss": 6.82822471378385,
      "tokens_seen": 10944512
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003333333333333333,
      "loss": 5.1341,
      "theoretical_loss": 6.8208908763759295,
      "tokens_seen": 11010048
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003353174603174603,
      "loss": 4.9061,
      "theoretical_loss": 6.813612704036853,
      "tokens_seen": 11075584
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00033730158730158733,
      "loss": 5.1869,
      "theoretical_loss": 6.806389448610343,
      "tokens_seen": 11141120
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00033928571428571433,
      "loss": 5.1124,
      "theoretical_loss": 6.799220376321128,
      "tokens_seen": 11206656
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003412698412698413,
      "loss": 5.2221,
      "theoretical_loss": 6.79210476741633,
      "tokens_seen": 11272192
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003432539682539683,
      "loss": 5.035,
      "theoretical_loss": 6.785041915817828,
      "tokens_seen": 11337728
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00034523809523809523,
      "loss": 5.1263,
      "theoretical_loss": 6.778031128785191,
      "tokens_seen": 11403264
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00034722222222222224,
      "loss": 5.2328,
      "theoretical_loss": 6.771071726588795,
      "tokens_seen": 11468800
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003492063492063492,
      "loss": 5.1011,
      "theoretical_loss": 6.76416304219278,
      "tokens_seen": 11534336
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003511904761904762,
      "loss": 5.0111,
      "theoretical_loss": 6.757304420947466,
      "tokens_seen": 11599872
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003531746031746032,
      "loss": 5.1246,
      "theoretical_loss": 6.750495220290931,
      "tokens_seen": 11665408
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003551587301587302,
      "loss": 5.0826,
      "theoretical_loss": 6.743734809459394,
      "tokens_seen": 11730944
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00035714285714285714,
      "loss": 4.9246,
      "theoretical_loss": 6.737022569206117,
      "tokens_seen": 11796480
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00035912698412698415,
      "loss": 4.9429,
      "theoretical_loss": 6.730357891528529,
      "tokens_seen": 11862016
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003611111111111111,
      "loss": 5.1395,
      "theoretical_loss": 6.723740179403264,
      "tokens_seen": 11927552
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003630952380952381,
      "loss": 5.1806,
      "theoretical_loss": 6.717168846528885,
      "tokens_seen": 11993088
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00036507936507936505,
      "loss": 5.0837,
      "theoretical_loss": 6.710643317075979,
      "tokens_seen": 12058624
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00036706349206349205,
      "loss": 5.0545,
      "theoretical_loss": 6.70416302544442,
      "tokens_seen": 12124160
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00036904761904761905,
      "loss": 5.051,
      "theoretical_loss": 6.6977274160275195,
      "tokens_seen": 12189696
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00037103174603174606,
      "loss": 4.936,
      "theoretical_loss": 6.691335942982863,
      "tokens_seen": 12255232
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.000373015873015873,
      "loss": 5.2205,
      "theoretical_loss": 6.684988070009584,
      "tokens_seen": 12320768
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.000375,
      "loss": 5.1985,
      "theoretical_loss": 6.6786832701318755,
      "tokens_seen": 12386304
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.000376984126984127,
      "loss": 5.0665,
      "theoretical_loss": 6.672421025488536,
      "tokens_seen": 12451840
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00037896825396825396,
      "loss": 5.2492,
      "theoretical_loss": 6.666200827128331,
      "tokens_seen": 12517376
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00038095238095238096,
      "loss": 5.0901,
      "theoretical_loss": 6.660022174811009,
      "tokens_seen": 12582912
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00038293650793650797,
      "loss": 5.0826,
      "theoretical_loss": 6.653884576813754,
      "tokens_seen": 12648448
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00038492063492063497,
      "loss": 4.9834,
      "theoretical_loss": 6.647787549742928,
      "tokens_seen": 12713984
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003869047619047619,
      "loss": 5.0116,
      "theoretical_loss": 6.641730618350916,
      "tokens_seen": 12779520
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003888888888888889,
      "loss": 5.0323,
      "theoretical_loss": 6.6357133153579175,
      "tokens_seen": 12845056
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00039087301587301587,
      "loss": 5.0115,
      "theoretical_loss": 6.629735181278515,
      "tokens_seen": 12910592
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003928571428571429,
      "loss": 4.9832,
      "theoretical_loss": 6.623795764252885,
      "tokens_seen": 12976128
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003948412698412698,
      "loss": 5.0157,
      "theoretical_loss": 6.61789461988249,
      "tokens_seen": 13041664
    },
    {
      "epoch": 0.0,
      "objective/train/advantage_avg": 0.4912474751472473,
      "objective/train/docs_used": 12170,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 4.883951187133789,
      "objective/train/original_loss": 4.883951187133789,
      "objective/train/theoretical_loss": 6.612031311070119,
      "objective/train/tokens_used": 33567200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2418125718832016,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05035400390625,
      "objective/train/weighted_lm_loss": 5.129752159118652,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 6.612031311070119,
      "tokens_seen": 13107200
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0003968253968253968,
      "loss": 5.0655,
      "theoretical_loss": 6.612031311070119,
      "tokens_seen": 13107200
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00039880952380952383,
      "loss": 4.8914,
      "theoretical_loss": 6.6062054078641355,
      "tokens_seen": 13172736
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00040079365079365083,
      "loss": 5.1053,
      "theoretical_loss": 6.600416487306803,
      "tokens_seen": 13238272
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004027777777777778,
      "loss": 4.9361,
      "theoretical_loss": 6.594664133286568,
      "tokens_seen": 13303808
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004047619047619048,
      "loss": 5.1403,
      "theoretical_loss": 6.588947936394168,
      "tokens_seen": 13369344
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00040674603174603173,
      "loss": 4.9681,
      "theoretical_loss": 6.58326749378245,
      "tokens_seen": 13434880
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00040873015873015874,
      "loss": 5.061,
      "theoretical_loss": 6.577622409029796,
      "tokens_seen": 13500416
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004107142857142857,
      "loss": 4.8943,
      "theoretical_loss": 6.5720122920070345,
      "tokens_seen": 13565952
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004126984126984127,
      "loss": 4.9159,
      "theoretical_loss": 6.566436758747731,
      "tokens_seen": 13631488
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004146825396825397,
      "loss": 4.8941,
      "theoretical_loss": 6.560895431321773,
      "tokens_seen": 13697024
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004166666666666667,
      "loss": 4.8609,
      "theoretical_loss": 6.5553879377121245,
      "tokens_seen": 13762560
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004186507936507937,
      "loss": 4.9139,
      "theoretical_loss": 6.5499139116946825,
      "tokens_seen": 13828096
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00042063492063492065,
      "loss": 4.9303,
      "theoretical_loss": 6.544472992721121,
      "tokens_seen": 13893632
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00042261904761904765,
      "loss": 4.9147,
      "theoretical_loss": 6.539064825804655,
      "tokens_seen": 13959168
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004246031746031746,
      "loss": 5.0321,
      "theoretical_loss": 6.533689061408615,
      "tokens_seen": 14024704
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004265873015873016,
      "loss": 4.9443,
      "theoretical_loss": 6.528345355337773,
      "tokens_seen": 14090240
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00042857142857142855,
      "loss": 4.9354,
      "theoretical_loss": 6.523033368632323,
      "tokens_seen": 14155776
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004305555555555556,
      "loss": 4.9561,
      "theoretical_loss": 6.51775276746445,
      "tokens_seen": 14221312
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00043253968253968256,
      "loss": 4.9391,
      "theoretical_loss": 6.512503223037404,
      "tokens_seen": 14286848
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00043452380952380956,
      "loss": 4.85,
      "theoretical_loss": 6.507284411487015,
      "tokens_seen": 14352384
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004365079365079365,
      "loss": 4.9346,
      "theoretical_loss": 6.502096013785574,
      "tokens_seen": 14417920
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004384920634920635,
      "loss": 4.8673,
      "theoretical_loss": 6.496937715648013,
      "tokens_seen": 14483456
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00044047619047619046,
      "loss": 4.7758,
      "theoretical_loss": 6.491809207440321,
      "tokens_seen": 14548992
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00044246031746031746,
      "loss": 4.8789,
      "theoretical_loss": 6.486710184090135,
      "tokens_seen": 14614528
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004444444444444444,
      "loss": 4.8879,
      "theoretical_loss": 6.481640344999435,
      "tokens_seen": 14680064
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00044642857142857147,
      "loss": 4.9789,
      "theoretical_loss": 6.476599393959298,
      "tokens_seen": 14745600
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004484126984126984,
      "loss": 4.8676,
      "theoretical_loss": 6.471587039066644,
      "tokens_seen": 14811136
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004503968253968254,
      "loss": 4.8724,
      "theoretical_loss": 6.466602992642919,
      "tokens_seen": 14876672
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00045238095238095237,
      "loss": 4.9345,
      "theoretical_loss": 6.461646971154669,
      "tokens_seen": 14942208
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00045436507936507937,
      "loss": 5.0052,
      "theoretical_loss": 6.456718695135944,
      "tokens_seen": 15007744
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004563492063492063,
      "loss": 4.899,
      "theoretical_loss": 6.45181788911249,
      "tokens_seen": 15073280
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004583333333333333,
      "loss": 4.7496,
      "theoretical_loss": 6.446944281527674,
      "tokens_seen": 15138816
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00046031746031746033,
      "loss": 4.7412,
      "theoretical_loss": 6.442097604670096,
      "tokens_seen": 15204352
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00046230158730158733,
      "loss": 4.6114,
      "theoretical_loss": 6.437277594602845,
      "tokens_seen": 15269888
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00046428571428571433,
      "loss": 4.7578,
      "theoretical_loss": 6.43248399109436,
      "tokens_seen": 15335424
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004662698412698413,
      "loss": 4.763,
      "theoretical_loss": 6.427716537550838,
      "tokens_seen": 15400960
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004682539682539683,
      "loss": 4.7541,
      "theoretical_loss": 6.422974980950157,
      "tokens_seen": 15466496
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00047023809523809523,
      "loss": 4.8955,
      "theoretical_loss": 6.41825907177728,
      "tokens_seen": 15532032
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00047222222222222224,
      "loss": 4.9811,
      "theoretical_loss": 6.413568563961092,
      "tokens_seen": 15597568
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004742063492063492,
      "loss": 4.7225,
      "theoretical_loss": 6.408903214812629,
      "tokens_seen": 15663104
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004761904761904762,
      "loss": 4.8861,
      "theoretical_loss": 6.404262784964672,
      "tokens_seen": 15728640
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004781746031746032,
      "loss": 4.7362,
      "theoretical_loss": 6.399647038312671,
      "tokens_seen": 15794176
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004801587301587302,
      "loss": 4.7356,
      "theoretical_loss": 6.395055741956952,
      "tokens_seen": 15859712
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00048214285714285715,
      "loss": 4.887,
      "theoretical_loss": 6.390488666146185,
      "tokens_seen": 15925248
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.00048412698412698415,
      "loss": 4.8904,
      "theoretical_loss": 6.3859455842220765,
      "tokens_seen": 15990784
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004861111111111111,
      "loss": 4.952,
      "theoretical_loss": 6.381426272565258,
      "tokens_seen": 16056320
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004880952380952381,
      "loss": 4.917,
      "theoretical_loss": 6.376930510542334,
      "tokens_seen": 16121856
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004900793650793651,
      "loss": 4.8962,
      "theoretical_loss": 6.372458080454071,
      "tokens_seen": 16187392
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.000492063492063492,
      "loss": 4.8315,
      "theoretical_loss": 6.368008767484675,
      "tokens_seen": 16252928
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0004940476190476191,
      "loss": 4.8416,
      "theoretical_loss": 6.363582359652165,
      "tokens_seen": 16318464
    },
    {
      "epoch": 0.0,
      "objective/train/advantage_avg": 0.49277380108833313,
      "objective/train/docs_used": 13382,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 4.31853723526001,
      "objective/train/original_loss": 4.318537712097168,
      "objective/train/theoretical_loss": 6.359178647759789,
      "objective/train/tokens_used": 36844000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2433290332555771,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505143404006958,
      "objective/train/weighted_lm_loss": 4.537021636962891,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 6.359178647759789,
      "tokens_seen": 16384000
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.000496031746031746,
      "loss": 4.7369,
      "theoretical_loss": 6.359178647759789,
      "tokens_seen": 16384000
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.000498015873015873,
      "loss": 4.7574,
      "theoretical_loss": 6.354797425348455,
      "tokens_seen": 16449536
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005,
      "loss": 4.61,
      "theoretical_loss": 6.350438488650175,
      "tokens_seen": 16515072
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000501984126984127,
      "loss": 4.6561,
      "theoretical_loss": 6.3461016365424685,
      "tokens_seen": 16580608
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000503968253968254,
      "loss": 4.7203,
      "theoretical_loss": 6.341786670503726,
      "tokens_seen": 16646144
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005059523809523809,
      "loss": 4.7175,
      "theoretical_loss": 6.33749339456949,
      "tokens_seen": 16711680
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005079365079365079,
      "loss": 4.7454,
      "theoretical_loss": 6.333221615289645,
      "tokens_seen": 16777216
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005099206349206349,
      "loss": 4.9234,
      "theoretical_loss": 6.328971141686479,
      "tokens_seen": 16842752
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005119047619047619,
      "loss": 4.7527,
      "theoretical_loss": 6.324741785213623,
      "tokens_seen": 16908288
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005138888888888888,
      "loss": 4.7599,
      "theoretical_loss": 6.320533359715803,
      "tokens_seen": 16973824
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005158730158730159,
      "loss": 4.7576,
      "theoretical_loss": 6.316345681389436,
      "tokens_seen": 17039360
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005178571428571429,
      "loss": 4.7286,
      "theoretical_loss": 6.312178568744001,
      "tokens_seen": 17104896
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005198412698412699,
      "loss": 4.7687,
      "theoretical_loss": 6.308031842564212,
      "tokens_seen": 17170432
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005218253968253968,
      "loss": 4.4944,
      "theoretical_loss": 6.303905325872933,
      "tokens_seen": 17235968
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005238095238095238,
      "loss": 4.9019,
      "theoretical_loss": 6.2997988438948465,
      "tokens_seen": 17301504
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005257936507936508,
      "loss": 4.6807,
      "theoretical_loss": 6.295712224020842,
      "tokens_seen": 17367040
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005277777777777778,
      "loss": 4.7674,
      "theoretical_loss": 6.2916452957731135,
      "tokens_seen": 17432576
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005297619047619047,
      "loss": 4.6331,
      "theoretical_loss": 6.287597890770951,
      "tokens_seen": 17498112
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005317460317460317,
      "loss": 4.5051,
      "theoretical_loss": 6.283569842697203,
      "tokens_seen": 17563648
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005337301587301587,
      "loss": 4.4246,
      "theoretical_loss": 6.279560987265395,
      "tokens_seen": 17629184
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005357142857142857,
      "loss": 4.6671,
      "theoretical_loss": 6.275571162187505,
      "tokens_seen": 17694720
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005376984126984126,
      "loss": 4.7847,
      "theoretical_loss": 6.271600207142356,
      "tokens_seen": 17760256
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005396825396825396,
      "loss": 4.7324,
      "theoretical_loss": 6.26764796374462,
      "tokens_seen": 17825792
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005416666666666666,
      "loss": 4.6055,
      "theoretical_loss": 6.263714275514436,
      "tokens_seen": 17891328
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005436507936507936,
      "loss": 4.5236,
      "theoretical_loss": 6.259798987847603,
      "tokens_seen": 17956864
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005456349206349206,
      "loss": 4.599,
      "theoretical_loss": 6.255901947986348,
      "tokens_seen": 18022400
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005476190476190477,
      "loss": 4.664,
      "theoretical_loss": 6.25202300499066,
      "tokens_seen": 18087936
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005496031746031747,
      "loss": 4.5562,
      "theoretical_loss": 6.248162009710162,
      "tokens_seen": 18153472
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005515873015873017,
      "loss": 4.7609,
      "theoretical_loss": 6.244318814756527,
      "tokens_seen": 18219008
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005535714285714287,
      "loss": 4.6061,
      "theoretical_loss": 6.240493274476407,
      "tokens_seen": 18284544
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005555555555555556,
      "loss": 4.5964,
      "theoretical_loss": 6.236685244924882,
      "tokens_seen": 18350080
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005575396825396826,
      "loss": 4.5874,
      "theoretical_loss": 6.232894583839405,
      "tokens_seen": 18415616
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005595238095238096,
      "loss": 4.4101,
      "theoretical_loss": 6.229121150614235,
      "tokens_seen": 18481152
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005615079365079366,
      "loss": 4.7411,
      "theoretical_loss": 6.225364806275344,
      "tokens_seen": 18546688
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005634920634920635,
      "loss": 4.5962,
      "theoretical_loss": 6.2216254134558024,
      "tokens_seen": 18612224
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005654761904761905,
      "loss": 4.5972,
      "theoretical_loss": 6.217902836371607,
      "tokens_seen": 18677760
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005674603174603175,
      "loss": 4.6108,
      "theoretical_loss": 6.214196940797967,
      "tokens_seen": 18743296
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005694444444444445,
      "loss": 4.582,
      "theoretical_loss": 6.210507594046018,
      "tokens_seen": 18808832
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005714285714285714,
      "loss": 4.7575,
      "theoretical_loss": 6.206834664939976,
      "tokens_seen": 18874368
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005734126984126984,
      "loss": 4.7611,
      "theoretical_loss": 6.203178023794688,
      "tokens_seen": 18939904
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005753968253968254,
      "loss": 4.6871,
      "theoretical_loss": 6.199537542393618,
      "tokens_seen": 19005440
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005773809523809524,
      "loss": 4.6562,
      "theoretical_loss": 6.195913093967214,
      "tokens_seen": 19070976
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005793650793650794,
      "loss": 4.5739,
      "theoretical_loss": 6.192304553171669,
      "tokens_seen": 19136512
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005813492063492064,
      "loss": 4.5292,
      "theoretical_loss": 6.188711796068077,
      "tokens_seen": 19202048
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005833333333333334,
      "loss": 4.3859,
      "theoretical_loss": 6.185134700101948,
      "tokens_seen": 19267584
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005853174603174604,
      "loss": 4.6491,
      "theoretical_loss": 6.181573144083102,
      "tokens_seen": 19333120
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005873015873015874,
      "loss": 4.633,
      "theoretical_loss": 6.178027008165916,
      "tokens_seen": 19398656
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005892857142857143,
      "loss": 4.6405,
      "theoretical_loss": 6.174496173829926,
      "tokens_seen": 19464192
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005912698412698413,
      "loss": 4.602,
      "theoretical_loss": 6.170980523860769,
      "tokens_seen": 19529728
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005932539682539683,
      "loss": 4.4713,
      "theoretical_loss": 6.167479942331466,
      "tokens_seen": 19595264
    },
    {
      "epoch": 0.01,
      "objective/train/advantage_avg": 0.49178165197372437,
      "objective/train/docs_used": 14656,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 4.28451681137085,
      "objective/train/original_loss": 4.284516334533691,
      "objective/train/theoretical_loss": 6.163994314584031,
      "objective/train/tokens_used": 40120800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24251613020896912,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504111051559448,
      "objective/train/weighted_lm_loss": 4.500372409820557,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 6.163994314584031,
      "tokens_seen": 19660800
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005952380952380953,
      "loss": 4.5733,
      "theoretical_loss": 6.163994314584031,
      "tokens_seen": 19660800
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005972222222222222,
      "loss": 4.5806,
      "theoretical_loss": 6.16052352721141,
      "tokens_seen": 19726336
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0005992063492063492,
      "loss": 4.5064,
      "theoretical_loss": 6.157067468039726,
      "tokens_seen": 19791872
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006011904761904762,
      "loss": 4.4136,
      "theoretical_loss": 6.153626026110843,
      "tokens_seen": 19857408
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006031746031746032,
      "loss": 4.5153,
      "theoretical_loss": 6.150199091665225,
      "tokens_seen": 19922944
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006051587301587301,
      "loss": 4.5327,
      "theoretical_loss": 6.146786556125103,
      "tokens_seen": 19988480
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006071428571428571,
      "loss": 4.4621,
      "theoretical_loss": 6.143388312077918,
      "tokens_seen": 20054016
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006091269841269841,
      "loss": 4.5998,
      "theoretical_loss": 6.140004253260065,
      "tokens_seen": 20119552
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006111111111111112,
      "loss": 4.6879,
      "theoretical_loss": 6.136634274540901,
      "tokens_seen": 20185088
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006130952380952381,
      "loss": 4.5875,
      "theoretical_loss": 6.13327827190704,
      "tokens_seen": 20250624
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006150793650793651,
      "loss": 4.5669,
      "theoretical_loss": 6.129936142446908,
      "tokens_seen": 20316160
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006170634920634921,
      "loss": 4.475,
      "theoretical_loss": 6.12660778433556,
      "tokens_seen": 20381696
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006190476190476191,
      "loss": 4.4535,
      "theoretical_loss": 6.123293096819758,
      "tokens_seen": 20447232
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000621031746031746,
      "loss": 4.6288,
      "theoretical_loss": 6.119991980203299,
      "tokens_seen": 20512768
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000623015873015873,
      "loss": 4.3293,
      "theoretical_loss": 6.11670433583259,
      "tokens_seen": 20578304
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000625,
      "loss": 4.5311,
      "theoretical_loss": 6.113430066082468,
      "tokens_seen": 20643840
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000626984126984127,
      "loss": 4.4825,
      "theoretical_loss": 6.1101690743422505,
      "tokens_seen": 20709376
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000628968253968254,
      "loss": 4.4968,
      "theoretical_loss": 6.106921265002029,
      "tokens_seen": 20774912
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006309523809523809,
      "loss": 4.4781,
      "theoretical_loss": 6.103686543439177,
      "tokens_seen": 20840448
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006329365079365079,
      "loss": 4.6214,
      "theoretical_loss": 6.100464816005092,
      "tokens_seen": 20905984
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006349206349206349,
      "loss": 4.5962,
      "theoretical_loss": 6.097255990012153,
      "tokens_seen": 20971520
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006369047619047619,
      "loss": 4.5147,
      "theoretical_loss": 6.094059973720896,
      "tokens_seen": 21037056
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006388888888888888,
      "loss": 4.4541,
      "theoretical_loss": 6.090876676327392,
      "tokens_seen": 21102592
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006408730158730159,
      "loss": 4.4346,
      "theoretical_loss": 6.087706007950839,
      "tokens_seen": 21168128
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006428571428571429,
      "loss": 4.5464,
      "theoretical_loss": 6.084547879621354,
      "tokens_seen": 21233664
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006448412698412699,
      "loss": 4.5378,
      "theoretical_loss": 6.081402203267963,
      "tokens_seen": 21299200
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006468253968253968,
      "loss": 4.5303,
      "theoretical_loss": 6.078268891706788,
      "tokens_seen": 21364736
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006488095238095238,
      "loss": 4.4849,
      "theoretical_loss": 6.075147858629419,
      "tokens_seen": 21430272
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006507936507936508,
      "loss": 4.3543,
      "theoretical_loss": 6.072039018591484,
      "tokens_seen": 21495808
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006527777777777778,
      "loss": 4.5641,
      "theoretical_loss": 6.068942287001391,
      "tokens_seen": 21561344
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006547619047619047,
      "loss": 4.4829,
      "theoretical_loss": 6.065857580109258,
      "tokens_seen": 21626880
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006567460317460317,
      "loss": 4.3724,
      "theoretical_loss": 6.0627848149960135,
      "tokens_seen": 21692416
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006587301587301587,
      "loss": 4.5818,
      "theoretical_loss": 6.059723909562683,
      "tokens_seen": 21757952
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006607142857142857,
      "loss": 4.3904,
      "theoretical_loss": 6.0566747825198295,
      "tokens_seen": 21823488
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006626984126984126,
      "loss": 4.4051,
      "theoretical_loss": 6.0536373533771695,
      "tokens_seen": 21889024
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006646825396825396,
      "loss": 4.5032,
      "theoretical_loss": 6.050611542433352,
      "tokens_seen": 21954560
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006666666666666666,
      "loss": 4.3651,
      "theoretical_loss": 6.047597270765904,
      "tokens_seen": 22020096
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006686507936507936,
      "loss": 4.5418,
      "theoretical_loss": 6.044594460221319,
      "tokens_seen": 22085632
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006706349206349207,
      "loss": 4.4352,
      "theoretical_loss": 6.041603033405314,
      "tokens_seen": 22151168
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006726190476190477,
      "loss": 4.3779,
      "theoretical_loss": 6.038622913673237,
      "tokens_seen": 22216704
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006746031746031747,
      "loss": 4.2616,
      "theoretical_loss": 6.035654025120612,
      "tokens_seen": 22282240
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006765873015873017,
      "loss": 4.319,
      "theoretical_loss": 6.032696292573847,
      "tokens_seen": 22347776
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006785714285714287,
      "loss": 4.2787,
      "theoretical_loss": 6.029749641581073,
      "tokens_seen": 22413312
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006805555555555556,
      "loss": 4.343,
      "theoretical_loss": 6.02681399840313,
      "tokens_seen": 22478848
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006825396825396826,
      "loss": 4.4465,
      "theoretical_loss": 6.023889290004692,
      "tokens_seen": 22544384
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006845238095238096,
      "loss": 4.4453,
      "theoretical_loss": 6.020975444045523,
      "tokens_seen": 22609920
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006865079365079366,
      "loss": 4.28,
      "theoretical_loss": 6.018072388871873,
      "tokens_seen": 22675456
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006884920634920635,
      "loss": 4.4475,
      "theoretical_loss": 6.015180053507997,
      "tokens_seen": 22740992
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006904761904761905,
      "loss": 4.2961,
      "theoretical_loss": 6.012298367647816,
      "tokens_seen": 22806528
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006924603174603175,
      "loss": 4.3297,
      "theoretical_loss": 6.009427261646688,
      "tokens_seen": 22872064
    },
    {
      "epoch": 0.01,
      "objective/train/advantage_avg": 0.49345701932907104,
      "objective/train/docs_used": 15907,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 4.267792701721191,
      "objective/train/original_loss": 4.267792701721191,
      "objective/train/theoretical_loss": 6.006566666513313,
      "objective/train/tokens_used": 43397600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24379104375839233,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505850315093994,
      "objective/train/weighted_lm_loss": 4.483581066131592,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 6.006566666513313,
      "tokens_seen": 22937600
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006944444444444445,
      "loss": 4.4127,
      "theoretical_loss": 6.006566666513313,
      "tokens_seen": 22937600
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006964285714285714,
      "loss": 4.09,
      "theoretical_loss": 6.003716513901762,
      "tokens_seen": 23003136
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0006984126984126984,
      "loss": 4.2815,
      "theoretical_loss": 6.000876736103618,
      "tokens_seen": 23068672
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007003968253968254,
      "loss": 4.4355,
      "theoretical_loss": 5.998047266040238,
      "tokens_seen": 23134208
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007023809523809524,
      "loss": 4.4496,
      "theoretical_loss": 5.995228037255137,
      "tokens_seen": 23199744
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007043650793650794,
      "loss": 4.3207,
      "theoretical_loss": 5.992418983906469,
      "tokens_seen": 23265280
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007063492063492064,
      "loss": 4.1968,
      "theoretical_loss": 5.989620040759641,
      "tokens_seen": 23330816
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007083333333333334,
      "loss": 4.2346,
      "theoretical_loss": 5.986831143180019,
      "tokens_seen": 23396352
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007103174603174604,
      "loss": 4.2309,
      "theoretical_loss": 5.9840522271257495,
      "tokens_seen": 23461888
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007123015873015874,
      "loss": 4.2863,
      "theoretical_loss": 5.981283229140683,
      "tokens_seen": 23527424
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007142857142857143,
      "loss": 4.2112,
      "theoretical_loss": 5.978524086347409,
      "tokens_seen": 23592960
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007162698412698413,
      "loss": 4.2578,
      "theoretical_loss": 5.975774736440385,
      "tokens_seen": 23658496
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007182539682539683,
      "loss": 4.2903,
      "theoretical_loss": 5.973035117679164,
      "tokens_seen": 23724032
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007202380952380953,
      "loss": 4.3281,
      "theoretical_loss": 5.970305168881732,
      "tokens_seen": 23789568
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007222222222222222,
      "loss": 4.4398,
      "theoretical_loss": 5.967584829417934,
      "tokens_seen": 23855104
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007242063492063492,
      "loss": 4.2352,
      "theoretical_loss": 5.964874039202989,
      "tokens_seen": 23920640
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007261904761904762,
      "loss": 4.2132,
      "theoretical_loss": 5.96217273869112,
      "tokens_seen": 23986176
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007281746031746032,
      "loss": 4.1317,
      "theoretical_loss": 5.959480868869245,
      "tokens_seen": 24051712
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007301587301587301,
      "loss": 4.3792,
      "theoretical_loss": 5.956798371250791,
      "tokens_seen": 24117248
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007321428571428571,
      "loss": 4.1713,
      "theoretical_loss": 5.954125187869572,
      "tokens_seen": 24182784
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007341269841269841,
      "loss": 4.2422,
      "theoretical_loss": 5.951461261273764,
      "tokens_seen": 24248320
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007361111111111112,
      "loss": 4.137,
      "theoretical_loss": 5.948806534519972,
      "tokens_seen": 24313856
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007380952380952381,
      "loss": 4.1628,
      "theoretical_loss": 5.9461609511673625,
      "tokens_seen": 24379392
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007400793650793651,
      "loss": 4.2607,
      "theoretical_loss": 5.943524455271907,
      "tokens_seen": 24444928
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007420634920634921,
      "loss": 4.2082,
      "theoretical_loss": 5.940896991380676,
      "tokens_seen": 24510464
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007440476190476191,
      "loss": 4.3838,
      "theoretical_loss": 5.938278504526239,
      "tokens_seen": 24576000
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000746031746031746,
      "loss": 4.2242,
      "theoretical_loss": 5.935668940221127,
      "tokens_seen": 24641536
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000748015873015873,
      "loss": 4.1961,
      "theoretical_loss": 5.933068244452382,
      "tokens_seen": 24707072
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00075,
      "loss": 4.1996,
      "theoretical_loss": 5.930476363676177,
      "tokens_seen": 24772608
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000751984126984127,
      "loss": 4.2341,
      "theoretical_loss": 5.927893244812513,
      "tokens_seen": 24838144
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000753968253968254,
      "loss": 4.2983,
      "theoretical_loss": 5.92531883523999,
      "tokens_seen": 24903680
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007559523809523809,
      "loss": 4.144,
      "theoretical_loss": 5.922753082790654,
      "tokens_seen": 24969216
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007579365079365079,
      "loss": 4.1519,
      "theoretical_loss": 5.920195935744903,
      "tokens_seen": 25034752
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007599206349206349,
      "loss": 4.3062,
      "theoretical_loss": 5.917647342826485,
      "tokens_seen": 25100288
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007619047619047619,
      "loss": 4.164,
      "theoretical_loss": 5.915107253197538,
      "tokens_seen": 25165824
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007638888888888888,
      "loss": 4.2066,
      "theoretical_loss": 5.91257561645372,
      "tokens_seen": 25231360
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007658730158730159,
      "loss": 4.236,
      "theoretical_loss": 5.910052382619399,
      "tokens_seen": 25296896
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007678571428571429,
      "loss": 4.2914,
      "theoretical_loss": 5.907537502142899,
      "tokens_seen": 25362432
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007698412698412699,
      "loss": 4.2342,
      "theoretical_loss": 5.905030925891829,
      "tokens_seen": 25427968
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007718253968253968,
      "loss": 4.1399,
      "theoretical_loss": 5.9025326051484575,
      "tokens_seen": 25493504
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007738095238095238,
      "loss": 4.3571,
      "theoretical_loss": 5.900042491605162,
      "tokens_seen": 25559040
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007757936507936508,
      "loss": 4.313,
      "theoretical_loss": 5.897560537359936,
      "tokens_seen": 25624576
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007777777777777778,
      "loss": 4.1413,
      "theoretical_loss": 5.895086694911951,
      "tokens_seen": 25690112
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007797619047619047,
      "loss": 4.2811,
      "theoretical_loss": 5.892620917157192,
      "tokens_seen": 25755648
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007817460317460317,
      "loss": 4.1719,
      "theoretical_loss": 5.8901631573841335,
      "tokens_seen": 25821184
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007837301587301587,
      "loss": 4.2168,
      "theoretical_loss": 5.887713369269493,
      "tokens_seen": 25886720
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007857142857142857,
      "loss": 4.2532,
      "theoretical_loss": 5.88527150687402,
      "tokens_seen": 25952256
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007876984126984126,
      "loss": 4.298,
      "theoretical_loss": 5.88283752463836,
      "tokens_seen": 26017792
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007896825396825396,
      "loss": 4.0948,
      "theoretical_loss": 5.880411377378962,
      "tokens_seen": 26083328
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007916666666666666,
      "loss": 4.0214,
      "theoretical_loss": 5.87799302028405,
      "tokens_seen": 26148864
    },
    {
      "epoch": 0.01,
      "objective/train/advantage_avg": 0.4926624894142151,
      "objective/train/docs_used": 17033,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 4.048245906829834,
      "objective/train/original_loss": 4.048246383666992,
      "objective/train/theoretical_loss": 5.8755824089096285,
      "objective/train/tokens_used": 46674400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24356918036937714,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050504446029663,
      "objective/train/weighted_lm_loss": 4.252634048461914,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 5.8755824089096285,
      "tokens_seen": 26214400
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007936507936507937,
      "loss": 4.3117,
      "theoretical_loss": 5.8755824089096285,
      "tokens_seen": 26214400
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007956349206349207,
      "loss": 3.9992,
      "theoretical_loss": 5.873179499175565,
      "tokens_seen": 26279936
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007976190476190477,
      "loss": 4.0696,
      "theoretical_loss": 5.87078424736171,
      "tokens_seen": 26345472
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0007996031746031747,
      "loss": 4.1271,
      "theoretical_loss": 5.868396610104066,
      "tokens_seen": 26411008
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008015873015873017,
      "loss": 4.1637,
      "theoretical_loss": 5.866016544391016,
      "tokens_seen": 26476544
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008035714285714287,
      "loss": 4.0211,
      "theoretical_loss": 5.863644007559595,
      "tokens_seen": 26542080
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008055555555555556,
      "loss": 4.1509,
      "theoretical_loss": 5.861278957291807,
      "tokens_seen": 26607616
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008075396825396826,
      "loss": 4.0588,
      "theoretical_loss": 5.858921351611005,
      "tokens_seen": 26673152
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008095238095238096,
      "loss": 4.0002,
      "theoretical_loss": 5.856571148878293,
      "tokens_seen": 26738688
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008115079365079366,
      "loss": 4.1366,
      "theoretical_loss": 5.854228307788999,
      "tokens_seen": 26804224
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008134920634920635,
      "loss": 4.1982,
      "theoretical_loss": 5.851892787369178,
      "tokens_seen": 26869760
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008154761904761905,
      "loss": 4.044,
      "theoretical_loss": 5.849564546972166,
      "tokens_seen": 26935296
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008174603174603175,
      "loss": 3.8531,
      "theoretical_loss": 5.847243546275179,
      "tokens_seen": 27000832
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008194444444444445,
      "loss": 4.0057,
      "theoretical_loss": 5.844929745275948,
      "tokens_seen": 27066368
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008214285714285714,
      "loss": 4.1101,
      "theoretical_loss": 5.842623104289412,
      "tokens_seen": 27131904
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008234126984126984,
      "loss": 3.9774,
      "theoretical_loss": 5.840323583944434,
      "tokens_seen": 27197440
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008253968253968254,
      "loss": 3.8177,
      "theoretical_loss": 5.838031145180573,
      "tokens_seen": 27262976
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008273809523809524,
      "loss": 4.1632,
      "theoretical_loss": 5.835745749244896,
      "tokens_seen": 27328512
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008293650793650794,
      "loss": 4.1018,
      "theoretical_loss": 5.833467357688816,
      "tokens_seen": 27394048
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008313492063492064,
      "loss": 4.1213,
      "theoretical_loss": 5.8311959323649925,
      "tokens_seen": 27459584
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008333333333333334,
      "loss": 4.1399,
      "theoretical_loss": 5.82893143542425,
      "tokens_seen": 27525120
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008353174603174604,
      "loss": 4.0058,
      "theoretical_loss": 5.826673829312552,
      "tokens_seen": 27590656
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008373015873015874,
      "loss": 4.1258,
      "theoretical_loss": 5.824423076767996,
      "tokens_seen": 27656192
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008392857142857143,
      "loss": 3.9901,
      "theoretical_loss": 5.8221791408178625,
      "tokens_seen": 27721728
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008412698412698413,
      "loss": 4.0441,
      "theoretical_loss": 5.81994198477569,
      "tokens_seen": 27787264
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008432539682539683,
      "loss": 4.0542,
      "theoretical_loss": 5.817711572238393,
      "tokens_seen": 27852800
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008452380952380953,
      "loss": 4.0306,
      "theoretical_loss": 5.815487867083405,
      "tokens_seen": 27918336
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008472222222222222,
      "loss": 3.9466,
      "theoretical_loss": 5.813270833465875,
      "tokens_seen": 27983872
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008492063492063492,
      "loss": 3.9006,
      "theoretical_loss": 5.811060435815881,
      "tokens_seen": 28049408
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008511904761904762,
      "loss": 4.0148,
      "theoretical_loss": 5.808856638835682,
      "tokens_seen": 28114944
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008531746031746032,
      "loss": 3.9268,
      "theoretical_loss": 5.806659407497016,
      "tokens_seen": 28180480
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008551587301587301,
      "loss": 4.1383,
      "theoretical_loss": 5.804468707038416,
      "tokens_seen": 28246016
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008571428571428571,
      "loss": 4.0713,
      "theoretical_loss": 5.802284502962563,
      "tokens_seen": 28311552
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008591269841269841,
      "loss": 4.0327,
      "theoretical_loss": 5.8001067610336765,
      "tokens_seen": 28377088
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008611111111111112,
      "loss": 4.1369,
      "theoretical_loss": 5.797935447274931,
      "tokens_seen": 28442624
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008630952380952381,
      "loss": 3.9127,
      "theoretical_loss": 5.795770527965905,
      "tokens_seen": 28508160
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008650793650793651,
      "loss": 3.959,
      "theoretical_loss": 5.793611969640068,
      "tokens_seen": 28573696
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008670634920634921,
      "loss": 3.7726,
      "theoretical_loss": 5.791459739082283,
      "tokens_seen": 28639232
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008690476190476191,
      "loss": 4.043,
      "theoretical_loss": 5.789313803326356,
      "tokens_seen": 28704768
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000871031746031746,
      "loss": 3.9936,
      "theoretical_loss": 5.787174129652598,
      "tokens_seen": 28770304
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000873015873015873,
      "loss": 4.0503,
      "theoretical_loss": 5.785040685585437,
      "tokens_seen": 28835840
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000875,
      "loss": 3.9107,
      "theoretical_loss": 5.782913438891035,
      "tokens_seen": 28901376
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000876984126984127,
      "loss": 3.9906,
      "theoretical_loss": 5.780792357574953,
      "tokens_seen": 28966912
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000878968253968254,
      "loss": 4.0754,
      "theoretical_loss": 5.778677409879835,
      "tokens_seen": 29032448
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008809523809523809,
      "loss": 3.9564,
      "theoretical_loss": 5.7765685642831155,
      "tokens_seen": 29097984
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008829365079365079,
      "loss": 4.0196,
      "theoretical_loss": 5.7744657894947675,
      "tokens_seen": 29163520
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008849206349206349,
      "loss": 4.0888,
      "theoretical_loss": 5.772369054455065,
      "tokens_seen": 29229056
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008869047619047619,
      "loss": 3.9354,
      "theoretical_loss": 5.770278328332372,
      "tokens_seen": 29294592
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008888888888888888,
      "loss": 4.0387,
      "theoretical_loss": 5.768193580520972,
      "tokens_seen": 29360128
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008908730158730159,
      "loss": 3.8796,
      "theoretical_loss": 5.766114780638902,
      "tokens_seen": 29425664
    },
    {
      "epoch": 0.01,
      "objective/train/advantage_avg": 0.49078458547592163,
      "objective/train/docs_used": 18174,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 3.943504571914673,
      "objective/train/original_loss": 3.9435043334960938,
      "objective/train/theoretical_loss": 5.7640418985258295,
      "objective/train/tokens_used": 49951200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2413143813610077,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503051280975342,
      "objective/train/weighted_lm_loss": 4.141692161560059,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 5.7640418985258295,
      "tokens_seen": 29491200
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008928571428571429,
      "loss": 4.0427,
      "theoretical_loss": 5.7640418985258295,
      "tokens_seen": 29491200
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008948412698412699,
      "loss": 3.9764,
      "theoretical_loss": 5.761974904240946,
      "tokens_seen": 29556736
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008968253968253968,
      "loss": 3.8796,
      "theoretical_loss": 5.759913768060882,
      "tokens_seen": 29622272
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0008988095238095238,
      "loss": 3.9728,
      "theoretical_loss": 5.757858460477662,
      "tokens_seen": 29687808
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009007936507936508,
      "loss": 4.0234,
      "theoretical_loss": 5.755808952196658,
      "tokens_seen": 29753344
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009027777777777778,
      "loss": 3.8761,
      "theoretical_loss": 5.753765214134593,
      "tokens_seen": 29818880
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009047619047619047,
      "loss": 4.0411,
      "theoretical_loss": 5.7517272174175496,
      "tokens_seen": 29884416
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009067460317460317,
      "loss": 4.0655,
      "theoretical_loss": 5.7496949333790095,
      "tokens_seen": 29949952
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009087301587301587,
      "loss": 3.8995,
      "theoretical_loss": 5.747668333557913,
      "tokens_seen": 30015488
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009107142857142857,
      "loss": 3.9744,
      "theoretical_loss": 5.745647389696742,
      "tokens_seen": 30081024
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009126984126984126,
      "loss": 4.072,
      "theoretical_loss": 5.743632073739626,
      "tokens_seen": 30146560
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009146825396825396,
      "loss": 3.9495,
      "theoretical_loss": 5.741622357830466,
      "tokens_seen": 30212096
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009166666666666666,
      "loss": 3.9492,
      "theoretical_loss": 5.7396182143110845,
      "tokens_seen": 30277632
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009186507936507937,
      "loss": 4.0284,
      "theoretical_loss": 5.737619615719392,
      "tokens_seen": 30343168
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009206349206349207,
      "loss": 3.892,
      "theoretical_loss": 5.735626534787584,
      "tokens_seen": 30408704
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009226190476190477,
      "loss": 3.8598,
      "theoretical_loss": 5.7336389444403375,
      "tokens_seen": 30474240
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009246031746031747,
      "loss": 3.8676,
      "theoretical_loss": 5.731656817793057,
      "tokens_seen": 30539776
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009265873015873017,
      "loss": 3.8372,
      "theoretical_loss": 5.729680128150115,
      "tokens_seen": 30605312
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009285714285714287,
      "loss": 3.7756,
      "theoretical_loss": 5.727708849003127,
      "tokens_seen": 30670848
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009305555555555556,
      "loss": 3.9913,
      "theoretical_loss": 5.72574295402924,
      "tokens_seen": 30736384
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009325396825396826,
      "loss": 3.9164,
      "theoretical_loss": 5.723782417089441,
      "tokens_seen": 30801920
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009345238095238096,
      "loss": 3.8756,
      "theoretical_loss": 5.721827212226889,
      "tokens_seen": 30867456
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009365079365079366,
      "loss": 3.7424,
      "theoretical_loss": 5.719877313665254,
      "tokens_seen": 30932992
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009384920634920635,
      "loss": 3.8924,
      "theoretical_loss": 5.717932695807093,
      "tokens_seen": 30998528
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009404761904761905,
      "loss": 3.9166,
      "theoretical_loss": 5.715993333232223,
      "tokens_seen": 31064064
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009424603174603175,
      "loss": 3.8553,
      "theoretical_loss": 5.7140592006961315,
      "tokens_seen": 31129600
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009444444444444445,
      "loss": 3.832,
      "theoretical_loss": 5.712130273128388,
      "tokens_seen": 31195136
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009464285714285714,
      "loss": 3.7882,
      "theoretical_loss": 5.710206525631086,
      "tokens_seen": 31260672
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009484126984126984,
      "loss": 4.0133,
      "theoretical_loss": 5.708287933477296,
      "tokens_seen": 31326208
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009503968253968254,
      "loss": 4.0449,
      "theoretical_loss": 5.706374472109536,
      "tokens_seen": 31391744
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009523809523809524,
      "loss": 3.8148,
      "theoretical_loss": 5.704466117138258,
      "tokens_seen": 31457280
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009543650793650794,
      "loss": 3.91,
      "theoretical_loss": 5.702562844340358,
      "tokens_seen": 31522816
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009563492063492064,
      "loss": 3.7582,
      "theoretical_loss": 5.700664629657688,
      "tokens_seen": 31588352
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009583333333333334,
      "loss": 4.045,
      "theoretical_loss": 5.698771449195602,
      "tokens_seen": 31653888
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009603174603174604,
      "loss": 3.8957,
      "theoretical_loss": 5.696883279221504,
      "tokens_seen": 31719424
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009623015873015874,
      "loss": 3.8502,
      "theoretical_loss": 5.695000096163421,
      "tokens_seen": 31784960
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009642857142857143,
      "loss": 3.7878,
      "theoretical_loss": 5.693121876608577,
      "tokens_seen": 31850496
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009662698412698413,
      "loss": 3.9058,
      "theoretical_loss": 5.691248597302009,
      "tokens_seen": 31916032
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009682539682539683,
      "loss": 3.8435,
      "theoretical_loss": 5.689380235145171,
      "tokens_seen": 31981568
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009702380952380953,
      "loss": 3.8857,
      "theoretical_loss": 5.687516767194563,
      "tokens_seen": 32047104
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009722222222222222,
      "loss": 3.8421,
      "theoretical_loss": 5.685658170660382,
      "tokens_seen": 32112640
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009742063492063492,
      "loss": 3.9891,
      "theoretical_loss": 5.683804422905179,
      "tokens_seen": 32178176
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009761904761904762,
      "loss": 3.7935,
      "theoretical_loss": 5.6819555014425305,
      "tokens_seen": 32243712
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009781746031746032,
      "loss": 3.7785,
      "theoretical_loss": 5.680111383935727,
      "tokens_seen": 32309248
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009801587301587302,
      "loss": 3.9492,
      "theoretical_loss": 5.678272048196479,
      "tokens_seen": 32374784
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009821428571428572,
      "loss": 3.8507,
      "theoretical_loss": 5.6764374721836255,
      "tokens_seen": 32440320
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000984126984126984,
      "loss": 3.7337,
      "theoretical_loss": 5.674607634001871,
      "tokens_seen": 32505856
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009861111111111112,
      "loss": 3.9994,
      "theoretical_loss": 5.672782511900524,
      "tokens_seen": 32571392
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009880952380952382,
      "loss": 3.9901,
      "theoretical_loss": 5.670962084272254,
      "tokens_seen": 32636928
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009900793650793652,
      "loss": 3.9945,
      "theoretical_loss": 5.669146329651861,
      "tokens_seen": 32702464
    },
    {
      "epoch": 0.01,
      "objective/train/advantage_avg": 0.49417802691459656,
      "objective/train/docs_used": 19239,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 3.745575189590454,
      "objective/train/original_loss": 3.745574951171875,
      "objective/train/theoretical_loss": 5.667335226715059,
      "objective/train/tokens_used": 53228000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2445439100265503,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506609678268433,
      "objective/train/weighted_lm_loss": 3.935272216796875,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 5.667335226715059,
      "tokens_seen": 32768000
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000992063492063492,
      "loss": 3.9393,
      "theoretical_loss": 5.667335226715059,
      "tokens_seen": 32768000
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000994047619047619,
      "loss": 3.8418,
      "theoretical_loss": 5.66552875427727,
      "tokens_seen": 32833536
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000996031746031746,
      "loss": 3.8581,
      "theoretical_loss": 5.663726891292432,
      "tokens_seen": 32899072
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998015873015873,
      "loss": 3.6826,
      "theoretical_loss": 5.6619296168518245,
      "tokens_seen": 32964608
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.001,
      "loss": 3.9296,
      "theoretical_loss": 5.6601369101828904,
      "tokens_seen": 33030144
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009999799398194584,
      "loss": 3.8296,
      "theoretical_loss": 5.658348750648093,
      "tokens_seen": 33095680
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009999598796389167,
      "loss": 3.7304,
      "theoretical_loss": 5.656565117743766,
      "tokens_seen": 33161216
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000999939819458375,
      "loss": 3.7348,
      "theoretical_loss": 5.654785991098984,
      "tokens_seen": 33226752
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009999197592778335,
      "loss": 3.807,
      "theoretical_loss": 5.6530113504744435,
      "tokens_seen": 33292288
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000999899699097292,
      "loss": 3.7786,
      "theoretical_loss": 5.651241175761357,
      "tokens_seen": 33357824
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009998796389167502,
      "loss": 3.79,
      "theoretical_loss": 5.649475446980356,
      "tokens_seen": 33423360
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009998595787362088,
      "loss": 3.9043,
      "theoretical_loss": 5.647714144280403,
      "tokens_seen": 33488896
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000999839518555667,
      "loss": 3.8137,
      "theoretical_loss": 5.645957247937725,
      "tokens_seen": 33554432
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009998194583751255,
      "loss": 3.7915,
      "theoretical_loss": 5.644204738354742,
      "tokens_seen": 33619968
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009997993981945839,
      "loss": 3.837,
      "theoretical_loss": 5.642456596059025,
      "tokens_seen": 33685504
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009997793380140422,
      "loss": 3.7575,
      "theoretical_loss": 5.6407128017022465,
      "tokens_seen": 33751040
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009997592778335006,
      "loss": 3.8424,
      "theoretical_loss": 5.638973336059157,
      "tokens_seen": 33816576
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000999739217652959,
      "loss": 3.6526,
      "theoretical_loss": 5.637238180026563,
      "tokens_seen": 33882112
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009997191574724173,
      "loss": 3.6857,
      "theoretical_loss": 5.6355073146223145,
      "tokens_seen": 33947648
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009996990972918757,
      "loss": 3.6679,
      "theoretical_loss": 5.633780720984311,
      "tokens_seen": 34013184
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000999679037111334,
      "loss": 3.7352,
      "theoretical_loss": 5.632058380369512,
      "tokens_seen": 34078720
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009996589769307924,
      "loss": 3.6703,
      "theoretical_loss": 5.630340274152954,
      "tokens_seen": 34144256
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009996389167502508,
      "loss": 3.8389,
      "theoretical_loss": 5.62862638382679,
      "tokens_seen": 34209792
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009996188565697091,
      "loss": 3.7032,
      "theoretical_loss": 5.626916690999319,
      "tokens_seen": 34275328
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009995987963891675,
      "loss": 3.6495,
      "theoretical_loss": 5.625211177394046,
      "tokens_seen": 34340864
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009995787362086258,
      "loss": 3.6542,
      "theoretical_loss": 5.6235098248487425,
      "tokens_seen": 34406400
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009995586760280842,
      "loss": 3.5303,
      "theoretical_loss": 5.621812615314505,
      "tokens_seen": 34471936
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009995386158475428,
      "loss": 3.752,
      "theoretical_loss": 5.620119530854848,
      "tokens_seen": 34537472
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000999518555667001,
      "loss": 3.7169,
      "theoretical_loss": 5.618430553644782,
      "tokens_seen": 34603008
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009994984954864595,
      "loss": 3.6834,
      "theoretical_loss": 5.616745665969917,
      "tokens_seen": 34668544
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009994784353059177,
      "loss": 3.6833,
      "theoretical_loss": 5.615064850225565,
      "tokens_seen": 34734080
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009994583751253762,
      "loss": 3.7607,
      "theoretical_loss": 5.613388088915855,
      "tokens_seen": 34799616
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009994383149448346,
      "loss": 3.707,
      "theoretical_loss": 5.611715364652864,
      "tokens_seen": 34865152
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000999418254764293,
      "loss": 3.6116,
      "theoretical_loss": 5.610046660155737,
      "tokens_seen": 34930688
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009993981945837513,
      "loss": 3.6668,
      "theoretical_loss": 5.608381958249839,
      "tokens_seen": 34996224
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009993781344032097,
      "loss": 3.6732,
      "theoretical_loss": 5.6067212418659,
      "tokens_seen": 35061760
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000999358074222668,
      "loss": 3.6028,
      "theoretical_loss": 5.605064494039176,
      "tokens_seen": 35127296
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009993380140421264,
      "loss": 3.6866,
      "theoretical_loss": 5.603411697908612,
      "tokens_seen": 35192832
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009993179538615848,
      "loss": 3.6574,
      "theoretical_loss": 5.601762836716016,
      "tokens_seen": 35258368
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009992978936810431,
      "loss": 3.4094,
      "theoretical_loss": 5.600117893805246,
      "tokens_seen": 35323904
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009992778335005015,
      "loss": 3.5417,
      "theoretical_loss": 5.598476852621397,
      "tokens_seen": 35389440
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009992577733199599,
      "loss": 3.5647,
      "theoretical_loss": 5.596839696710004,
      "tokens_seen": 35454976
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009992377131394182,
      "loss": 3.7131,
      "theoretical_loss": 5.59520640971624,
      "tokens_seen": 35520512
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009992176529588766,
      "loss": 3.5971,
      "theoretical_loss": 5.5935769753841384,
      "tokens_seen": 35586048
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000999197592778335,
      "loss": 3.9092,
      "theoretical_loss": 5.591951377555809,
      "tokens_seen": 35651584
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009991775325977935,
      "loss": 3.7656,
      "theoretical_loss": 5.590329600170675,
      "tokens_seen": 35717120
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009991574724172517,
      "loss": 3.8054,
      "theoretical_loss": 5.588711627264699,
      "tokens_seen": 35782656
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009991374122367102,
      "loss": 3.6167,
      "theoretical_loss": 5.587097442969634,
      "tokens_seen": 35848192
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009991173520561684,
      "loss": 3.7877,
      "theoretical_loss": 5.585487031512276,
      "tokens_seen": 35913728
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000999097291875627,
      "loss": 3.7781,
      "theoretical_loss": 5.583880377213717,
      "tokens_seen": 35979264
    },
    {
      "epoch": 0.01,
      "objective/train/advantage_avg": 0.4903879463672638,
      "objective/train/docs_used": 20470,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 3.424363613128662,
      "objective/train/original_loss": 3.424363613128662,
      "objective/train/theoretical_loss": 5.582277464488612,
      "objective/train/tokens_used": 56504800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2408606857061386,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502631664276123,
      "objective/train/weighted_lm_loss": 3.596794605255127,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 5.582277464488612,
      "tokens_seen": 36044800
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009990772316950853,
      "loss": 3.5566,
      "theoretical_loss": 5.582277464488612,
      "tokens_seen": 36044800
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009990571715145437,
      "loss": 3.8534,
      "theoretical_loss": 5.58067827784446,
      "tokens_seen": 36110336
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000999037111334002,
      "loss": 3.6709,
      "theoretical_loss": 5.579082801880871,
      "tokens_seen": 36175872
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009990170511534604,
      "loss": 3.7757,
      "theoretical_loss": 5.577491021288861,
      "tokens_seen": 36241408
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009989969909729188,
      "loss": 3.5704,
      "theoretical_loss": 5.57590292085014,
      "tokens_seen": 36306944
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009989769307923771,
      "loss": 3.7544,
      "theoretical_loss": 5.574318485436422,
      "tokens_seen": 36372480
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009989568706118355,
      "loss": 3.512,
      "theoretical_loss": 5.572737700008718,
      "tokens_seen": 36438016
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009989368104312939,
      "loss": 3.5182,
      "theoretical_loss": 5.5711605496166605,
      "tokens_seen": 36503552
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009989167502507522,
      "loss": 3.785,
      "theoretical_loss": 5.569587019397816,
      "tokens_seen": 36569088
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009988966900702106,
      "loss": 3.6573,
      "theoretical_loss": 5.5680170945770175,
      "tokens_seen": 36634624
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998876629889669,
      "loss": 3.6389,
      "theoretical_loss": 5.56645076046569,
      "tokens_seen": 36700160
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009988565697091273,
      "loss": 3.4898,
      "theoretical_loss": 5.5648880024612,
      "tokens_seen": 36765696
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009988365095285857,
      "loss": 3.6444,
      "theoretical_loss": 5.5633288060461865,
      "tokens_seen": 36831232
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009988164493480443,
      "loss": 3.6373,
      "theoretical_loss": 5.561773156787925,
      "tokens_seen": 36896768
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009987963891675024,
      "loss": 3.667,
      "theoretical_loss": 5.5602210403376775,
      "tokens_seen": 36962304
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998776328986961,
      "loss": 3.5937,
      "theoretical_loss": 5.558672442430058,
      "tokens_seen": 37027840
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009987562688064191,
      "loss": 3.7541,
      "theoretical_loss": 5.557127348882403,
      "tokens_seen": 37093376
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009987362086258777,
      "loss": 3.6276,
      "theoretical_loss": 5.555585745594145,
      "tokens_seen": 37158912
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998716148445336,
      "loss": 3.6938,
      "theoretical_loss": 5.554047618546193,
      "tokens_seen": 37224448
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009986960882647944,
      "loss": 3.6552,
      "theoretical_loss": 5.552512953800322,
      "tokens_seen": 37289984
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009986760280842528,
      "loss": 3.5607,
      "theoretical_loss": 5.550981737498564,
      "tokens_seen": 37355520
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009986559679037112,
      "loss": 3.7794,
      "theoretical_loss": 5.549453955862602,
      "tokens_seen": 37421056
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009986359077231695,
      "loss": 3.7612,
      "theoretical_loss": 5.547929595193182,
      "tokens_seen": 37486592
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009986158475426279,
      "loss": 3.5894,
      "theoretical_loss": 5.5464086418695135,
      "tokens_seen": 37552128
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009985957873620862,
      "loss": 3.5894,
      "theoretical_loss": 5.544891082348692,
      "tokens_seen": 37617664
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009985757271815446,
      "loss": 3.6129,
      "theoretical_loss": 5.543376903165111,
      "tokens_seen": 37683200
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998555667001003,
      "loss": 3.5844,
      "theoretical_loss": 5.5418660909298945,
      "tokens_seen": 37748736
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009985356068204613,
      "loss": 3.6763,
      "theoretical_loss": 5.540358632330321,
      "tokens_seen": 37814272
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009985155466399197,
      "loss": 3.4916,
      "theoretical_loss": 5.538854514129265,
      "tokens_seen": 37879808
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998495486459378,
      "loss": 3.4597,
      "theoretical_loss": 5.5373537231646335,
      "tokens_seen": 37945344
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009984754262788366,
      "loss": 3.5356,
      "theoretical_loss": 5.535856246348814,
      "tokens_seen": 38010880
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998455366098295,
      "loss": 3.6021,
      "theoretical_loss": 5.534362070668127,
      "tokens_seen": 38076416
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009984353059177534,
      "loss": 3.5672,
      "theoretical_loss": 5.532871183182278,
      "tokens_seen": 38141952
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009984152457372117,
      "loss": 3.5533,
      "theoretical_loss": 5.53138357102382,
      "tokens_seen": 38207488
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00099839518555667,
      "loss": 3.6106,
      "theoretical_loss": 5.529899221397624,
      "tokens_seen": 38273024
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009983751253761284,
      "loss": 3.4664,
      "theoretical_loss": 5.528418121580339,
      "tokens_seen": 38338560
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009983550651955868,
      "loss": 3.3992,
      "theoretical_loss": 5.526940258919881,
      "tokens_seen": 38404096
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009983350050150452,
      "loss": 3.6064,
      "theoretical_loss": 5.5254656208349004,
      "tokens_seen": 38469632
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009983149448345035,
      "loss": 3.5422,
      "theoretical_loss": 5.523994194814273,
      "tokens_seen": 38535168
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998294884653962,
      "loss": 3.3976,
      "theoretical_loss": 5.522525968416592,
      "tokens_seen": 38600704
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009982748244734203,
      "loss": 3.6676,
      "theoretical_loss": 5.521060929269655,
      "tokens_seen": 38666240
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009982547642928786,
      "loss": 3.5906,
      "theoretical_loss": 5.519599065069971,
      "tokens_seen": 38731776
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998234704112337,
      "loss": 3.5077,
      "theoretical_loss": 5.518140363582252,
      "tokens_seen": 38797312
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009982146439317953,
      "loss": 3.4583,
      "theoretical_loss": 5.516684812638932,
      "tokens_seen": 38862848
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998194583751254,
      "loss": 3.6336,
      "theoretical_loss": 5.515232400139671,
      "tokens_seen": 38928384
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998174523570712,
      "loss": 3.667,
      "theoretical_loss": 5.5137831140508755,
      "tokens_seen": 38993920
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009981544633901706,
      "loss": 3.4816,
      "theoretical_loss": 5.512336942405216,
      "tokens_seen": 39059456
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009981344032096288,
      "loss": 3.5222,
      "theoretical_loss": 5.510893873301154,
      "tokens_seen": 39124992
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009981143430290874,
      "loss": 3.5284,
      "theoretical_loss": 5.509453894902469,
      "tokens_seen": 39190528
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009980942828485457,
      "loss": 3.483,
      "theoretical_loss": 5.508016995437791,
      "tokens_seen": 39256064
    },
    {
      "epoch": 0.01,
      "objective/train/advantage_avg": 0.4876997768878937,
      "objective/train/docs_used": 21760,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 3.0592448711395264,
      "objective/train/original_loss": 3.0592451095581055,
      "objective/train/theoretical_loss": 5.506583163200142,
      "objective/train/tokens_used": 59781600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23880939185619354,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0499838590621948,
      "objective/train/weighted_lm_loss": 3.212702512741089,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 5.506583163200142,
      "tokens_seen": 39321600
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000998074222668004,
      "loss": 3.3564,
      "theoretical_loss": 5.506583163200142,
      "tokens_seen": 39321600
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009980541624874625,
      "loss": 3.5219,
      "theoretical_loss": 5.505152386546474,
      "tokens_seen": 39387136
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009980341023069208,
      "loss": 3.6034,
      "theoretical_loss": 5.503724653897207,
      "tokens_seen": 39452672
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009980140421263792,
      "loss": 3.5868,
      "theoretical_loss": 5.502299953735795,
      "tokens_seen": 39518208
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009979939819458375,
      "loss": 3.43,
      "theoretical_loss": 5.5008782746082625,
      "tokens_seen": 39583744
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000997973921765296,
      "loss": 3.6451,
      "theoretical_loss": 5.499459605122775,
      "tokens_seen": 39649280
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009979538615847543,
      "loss": 3.3693,
      "theoretical_loss": 5.498043933949187,
      "tokens_seen": 39714816
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009979338014042126,
      "loss": 3.5735,
      "theoretical_loss": 5.496631249818616,
      "tokens_seen": 39780352
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000997913741223671,
      "loss": 3.4721,
      "theoretical_loss": 5.495221541523011,
      "tokens_seen": 39845888
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009978936810431294,
      "loss": 3.4922,
      "theoretical_loss": 5.493814797914718,
      "tokens_seen": 39911424
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009978736208625877,
      "loss": 3.6077,
      "theoretical_loss": 5.49241100790606,
      "tokens_seen": 39976960
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000997853560682046,
      "loss": 3.5721,
      "theoretical_loss": 5.491010160468919,
      "tokens_seen": 40042496
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009978335005015047,
      "loss": 3.399,
      "theoretical_loss": 5.489612244634316,
      "tokens_seen": 40108032
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009978134403209628,
      "loss": 3.4734,
      "theoretical_loss": 5.4882172494919965,
      "tokens_seen": 40173568
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009977933801404214,
      "loss": 3.542,
      "theoretical_loss": 5.486825164190026,
      "tokens_seen": 40239104
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009977733199598795,
      "loss": 3.4436,
      "theoretical_loss": 5.4854359779343795,
      "tokens_seen": 40304640
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009977532597793381,
      "loss": 3.5102,
      "theoretical_loss": 5.48404967998854,
      "tokens_seen": 40370176
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009977331995987965,
      "loss": 3.4231,
      "theoretical_loss": 5.482666259673101,
      "tokens_seen": 40435712
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009977131394182548,
      "loss": 3.6534,
      "theoretical_loss": 5.481285706365366,
      "tokens_seen": 40501248
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009976930792377132,
      "loss": 3.5207,
      "theoretical_loss": 5.479908009498965,
      "tokens_seen": 40566784
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009976730190571716,
      "loss": 3.4939,
      "theoretical_loss": 5.478533158563456,
      "tokens_seen": 40632320
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00099765295887663,
      "loss": 3.7361,
      "theoretical_loss": 5.477161143103945,
      "tokens_seen": 40697856
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009976328986960883,
      "loss": 3.4448,
      "theoretical_loss": 5.4757919527207,
      "tokens_seen": 40763392
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009976128385155466,
      "loss": 3.5947,
      "theoretical_loss": 5.474425577068777,
      "tokens_seen": 40828928
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000997592778335005,
      "loss": 3.4699,
      "theoretical_loss": 5.473062005857637,
      "tokens_seen": 40894464
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009975727181544634,
      "loss": 3.6217,
      "theoretical_loss": 5.471701228850778,
      "tokens_seen": 40960000
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009975526579739217,
      "loss": 3.6091,
      "theoretical_loss": 5.470343235865368,
      "tokens_seen": 41025536
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00099753259779338,
      "loss": 3.4589,
      "theoretical_loss": 5.468988016771869,
      "tokens_seen": 41091072
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009975125376128385,
      "loss": 3.4358,
      "theoretical_loss": 5.467635561493681,
      "tokens_seen": 41156608
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009974924774322968,
      "loss": 3.4588,
      "theoretical_loss": 5.466285860006783,
      "tokens_seen": 41222144
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009974724172517554,
      "loss": 3.6569,
      "theoretical_loss": 5.46493890233937,
      "tokens_seen": 41287680
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009974523570712135,
      "loss": 3.453,
      "theoretical_loss": 5.4635946785715,
      "tokens_seen": 41353216
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009974322968906721,
      "loss": 3.3928,
      "theoretical_loss": 5.462253178834744,
      "tokens_seen": 41418752
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009974122367101303,
      "loss": 3.3296,
      "theoretical_loss": 5.460914393311839,
      "tokens_seen": 41484288
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009973921765295888,
      "loss": 3.4467,
      "theoretical_loss": 5.4595783122363395,
      "tokens_seen": 41549824
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009973721163490472,
      "loss": 3.4153,
      "theoretical_loss": 5.458244925892275,
      "tokens_seen": 41615360
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009973520561685056,
      "loss": 3.4421,
      "theoretical_loss": 5.456914224613812,
      "tokens_seen": 41680896
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000997331995987964,
      "loss": 3.4096,
      "theoretical_loss": 5.455586198784916,
      "tokens_seen": 41746432
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009973119358074223,
      "loss": 3.4896,
      "theoretical_loss": 5.454260838839016,
      "tokens_seen": 41811968
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009972918756268807,
      "loss": 3.5514,
      "theoretical_loss": 5.452938135258673,
      "tokens_seen": 41877504
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000997271815446339,
      "loss": 3.3437,
      "theoretical_loss": 5.451618078575256,
      "tokens_seen": 41943040
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009972517552657974,
      "loss": 3.2631,
      "theoretical_loss": 5.450300659368612,
      "tokens_seen": 42008576
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009972316950852557,
      "loss": 3.3576,
      "theoretical_loss": 5.448985868266736,
      "tokens_seen": 42074112
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000997211634904714,
      "loss": 3.5492,
      "theoretical_loss": 5.447673695945468,
      "tokens_seen": 42139648
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009971915747241725,
      "loss": 3.406,
      "theoretical_loss": 5.446364133128155,
      "tokens_seen": 42205184
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009971715145436308,
      "loss": 3.4232,
      "theoretical_loss": 5.4450571705853505,
      "tokens_seen": 42270720
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009971514543630892,
      "loss": 3.1882,
      "theoretical_loss": 5.443752799134492,
      "tokens_seen": 42336256
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009971313941825476,
      "loss": 3.4595,
      "theoretical_loss": 5.442451009639596,
      "tokens_seen": 42401792
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009971113340020061,
      "loss": 3.3603,
      "theoretical_loss": 5.44115179301095,
      "tokens_seen": 42467328
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009970912738214643,
      "loss": 3.5864,
      "theoretical_loss": 5.439855140204802,
      "tokens_seen": 42532864
    },
    {
      "epoch": 0.01,
      "objective/train/advantage_avg": 0.49280866980552673,
      "objective/train/docs_used": 22938,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 3.297335386276245,
      "objective/train/original_loss": 3.297335624694824,
      "objective/train/theoretical_loss": 5.438561042223066,
      "objective/train/tokens_used": 63058400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24311578273773193,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505168437957764,
      "objective/train/weighted_lm_loss": 3.463775634765625,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 5.438561042223066,
      "tokens_seen": 42598400
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009970712136409229,
      "loss": 3.5201,
      "theoretical_loss": 5.438561042223066,
      "tokens_seen": 42598400
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000997051153460381,
      "loss": 3.3525,
      "theoretical_loss": 5.437269490113014,
      "tokens_seen": 42663936
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009970310932798396,
      "loss": 3.3086,
      "theoretical_loss": 5.435980474966981,
      "tokens_seen": 42729472
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000997011033099298,
      "loss": 3.3753,
      "theoretical_loss": 5.434693987922072,
      "tokens_seen": 42795008
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009969909729187563,
      "loss": 3.3824,
      "theoretical_loss": 5.433410020159865,
      "tokens_seen": 42860544
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009969709127382147,
      "loss": 3.2654,
      "theoretical_loss": 5.432128562906122,
      "tokens_seen": 42926080
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000996950852557673,
      "loss": 3.384,
      "theoretical_loss": 5.430849607430501,
      "tokens_seen": 42991616
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009969307923771314,
      "loss": 3.2841,
      "theoretical_loss": 5.429573145046266,
      "tokens_seen": 43057152
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009969107321965898,
      "loss": 3.422,
      "theoretical_loss": 5.428299167110012,
      "tokens_seen": 43122688
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009968906720160481,
      "loss": 3.3227,
      "theoretical_loss": 5.427027665021371,
      "tokens_seen": 43188224
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009968706118355065,
      "loss": 3.285,
      "theoretical_loss": 5.425758630222747,
      "tokens_seen": 43253760
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000996850551654965,
      "loss": 3.3776,
      "theoretical_loss": 5.424492054199025,
      "tokens_seen": 43319296
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009968304914744232,
      "loss": 3.267,
      "theoretical_loss": 5.423227928477306,
      "tokens_seen": 43384832
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009968104312938818,
      "loss": 3.3353,
      "theoretical_loss": 5.421966244626631,
      "tokens_seen": 43450368
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00099679037111334,
      "loss": 3.5885,
      "theoretical_loss": 5.42070699425771,
      "tokens_seen": 43515904
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009967703109327985,
      "loss": 3.4124,
      "theoretical_loss": 5.419450169022653,
      "tokens_seen": 43581440
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009967502507522569,
      "loss": 3.4499,
      "theoretical_loss": 5.418195760614707,
      "tokens_seen": 43646976
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009967301905717152,
      "loss": 3.395,
      "theoretical_loss": 5.416943760767989,
      "tokens_seen": 43712512
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009967101303911736,
      "loss": 3.4137,
      "theoretical_loss": 5.415694161257225,
      "tokens_seen": 43778048
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000996690070210632,
      "loss": 3.3275,
      "theoretical_loss": 5.414446953897489,
      "tokens_seen": 43843584
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009966700100300903,
      "loss": 3.1889,
      "theoretical_loss": 5.413202130543947,
      "tokens_seen": 43909120
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009966499498495487,
      "loss": 3.2611,
      "theoretical_loss": 5.411959683091602,
      "tokens_seen": 43974656
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000996629889669007,
      "loss": 3.339,
      "theoretical_loss": 5.410719603475034,
      "tokens_seen": 44040192
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009966098294884654,
      "loss": 3.5504,
      "theoretical_loss": 5.409481883668155,
      "tokens_seen": 44105728
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009965897693079238,
      "loss": 3.3063,
      "theoretical_loss": 5.408246515683956,
      "tokens_seen": 44171264
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009965697091273821,
      "loss": 3.145,
      "theoretical_loss": 5.407013491574263,
      "tokens_seen": 44236800
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009965496489468405,
      "loss": 3.3912,
      "theoretical_loss": 5.405782803429483,
      "tokens_seen": 44302336
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009965295887662989,
      "loss": 3.2364,
      "theoretical_loss": 5.404554443378366,
      "tokens_seen": 44367872
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009965095285857572,
      "loss": 3.3157,
      "theoretical_loss": 5.403328403587762,
      "tokens_seen": 44433408
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009964894684052158,
      "loss": 3.2918,
      "theoretical_loss": 5.402104676262381,
      "tokens_seen": 44498944
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000996469408224674,
      "loss": 3.2743,
      "theoretical_loss": 5.400883253644551,
      "tokens_seen": 44564480
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009964493480441325,
      "loss": 3.3715,
      "theoretical_loss": 5.399664128013988,
      "tokens_seen": 44630016
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009964292878635907,
      "loss": 3.2805,
      "theoretical_loss": 5.398447291687555,
      "tokens_seen": 44695552
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009964092276830492,
      "loss": 3.3267,
      "theoretical_loss": 5.397232737019031,
      "tokens_seen": 44761088
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009963891675025076,
      "loss": 3.3381,
      "theoretical_loss": 5.396020456398885,
      "tokens_seen": 44826624
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000996369107321966,
      "loss": 3.1136,
      "theoretical_loss": 5.394810442254032,
      "tokens_seen": 44892160
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009963490471414243,
      "loss": 3.2009,
      "theoretical_loss": 5.393602687047627,
      "tokens_seen": 44957696
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009963289869608827,
      "loss": 3.3298,
      "theoretical_loss": 5.392397183278822,
      "tokens_seen": 45023232
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000996308926780341,
      "loss": 3.1392,
      "theoretical_loss": 5.391193923482547,
      "tokens_seen": 45088768
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009962888665997994,
      "loss": 3.3421,
      "theoretical_loss": 5.38999290022929,
      "tokens_seen": 45154304
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009962688064192578,
      "loss": 3.2676,
      "theoretical_loss": 5.388794106124874,
      "tokens_seen": 45219840
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009962487462387161,
      "loss": 3.2214,
      "theoretical_loss": 5.387597533810238,
      "tokens_seen": 45285376
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009962286860581745,
      "loss": 3.245,
      "theoretical_loss": 5.386403175961223,
      "tokens_seen": 45350912
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009962086258776329,
      "loss": 3.177,
      "theoretical_loss": 5.385211025288349,
      "tokens_seen": 45416448
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009961885656970912,
      "loss": 3.1048,
      "theoretical_loss": 5.38402107453661,
      "tokens_seen": 45481984
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009961685055165496,
      "loss": 3.2362,
      "theoretical_loss": 5.382833316485255,
      "tokens_seen": 45547520
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000996148445336008,
      "loss": 3.3837,
      "theoretical_loss": 5.381647743947578,
      "tokens_seen": 45613056
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009961283851554665,
      "loss": 3.3365,
      "theoretical_loss": 5.3804643497707145,
      "tokens_seen": 45678592
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009961083249749247,
      "loss": 3.2244,
      "theoretical_loss": 5.379283126835426,
      "tokens_seen": 45744128
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009960882647943833,
      "loss": 3.3835,
      "theoretical_loss": 5.378104068055899,
      "tokens_seen": 45809664
    },
    {
      "epoch": 0.01,
      "objective/train/advantage_avg": 0.4905271530151367,
      "objective/train/docs_used": 24173,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.979590654373169,
      "objective/train/original_loss": 2.97959041595459,
      "objective/train/theoretical_loss": 5.37692716637954,
      "objective/train/tokens_used": 66335200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24144382774829865,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502800941467285,
      "objective/train/weighted_lm_loss": 3.129429817199707,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 5.37692716637954,
      "tokens_seen": 45875200
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009960682046138414,
      "loss": 3.1644,
      "theoretical_loss": 5.37692716637954,
      "tokens_seen": 45875200
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009960481444333,
      "loss": 3.2088,
      "theoretical_loss": 5.375752414786769,
      "tokens_seen": 45940736
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009960280842527583,
      "loss": 3.3534,
      "theoretical_loss": 5.374579806290823,
      "tokens_seen": 46006272
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009960080240722167,
      "loss": 3.1675,
      "theoretical_loss": 5.373409333937557,
      "tokens_seen": 46071808
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995987963891675,
      "loss": 3.1467,
      "theoretical_loss": 5.372240990805237,
      "tokens_seen": 46137344
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009959679037111334,
      "loss": 3.2232,
      "theoretical_loss": 5.37107477000435,
      "tokens_seen": 46202880
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009959478435305918,
      "loss": 3.2601,
      "theoretical_loss": 5.369910664677409,
      "tokens_seen": 46268416
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009959277833500502,
      "loss": 2.9964,
      "theoretical_loss": 5.368748667998759,
      "tokens_seen": 46333952
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009959077231695085,
      "loss": 3.1695,
      "theoretical_loss": 5.367588773174377,
      "tokens_seen": 46399488
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009958876629889669,
      "loss": 3.2412,
      "theoretical_loss": 5.366430973441693,
      "tokens_seen": 46465024
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009958676028084252,
      "loss": 3.4332,
      "theoretical_loss": 5.365275262069391,
      "tokens_seen": 46530560
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009958475426278836,
      "loss": 3.0728,
      "theoretical_loss": 5.364121632357221,
      "tokens_seen": 46596096
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995827482447342,
      "loss": 3.3624,
      "theoretical_loss": 5.36297007763582,
      "tokens_seen": 46661632
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009958074222668003,
      "loss": 3.2145,
      "theoretical_loss": 5.36182059126652,
      "tokens_seen": 46727168
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009957873620862587,
      "loss": 3.2506,
      "theoretical_loss": 5.360673166641163,
      "tokens_seen": 46792704
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009957673019057173,
      "loss": 3.229,
      "theoretical_loss": 5.359527797181926,
      "tokens_seen": 46858240
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009957472417251754,
      "loss": 3.2049,
      "theoretical_loss": 5.358384476341126,
      "tokens_seen": 46923776
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995727181544634,
      "loss": 3.129,
      "theoretical_loss": 5.357243197601056,
      "tokens_seen": 46989312
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009957071213640921,
      "loss": 3.2963,
      "theoretical_loss": 5.356103954473793,
      "tokens_seen": 47054848
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009956870611835507,
      "loss": 3.3261,
      "theoretical_loss": 5.35496674050103,
      "tokens_seen": 47120384
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995667001003009,
      "loss": 3.083,
      "theoretical_loss": 5.353831549253895,
      "tokens_seen": 47185920
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009956469408224674,
      "loss": 3.1944,
      "theoretical_loss": 5.352698374332774,
      "tokens_seen": 47251456
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009956268806419258,
      "loss": 3.2186,
      "theoretical_loss": 5.351567209367143,
      "tokens_seen": 47316992
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009956068204613842,
      "loss": 3.387,
      "theoretical_loss": 5.350438048015396,
      "tokens_seen": 47382528
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009955867602808425,
      "loss": 3.0156,
      "theoretical_loss": 5.349310883964664,
      "tokens_seen": 47448064
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995566700100301,
      "loss": 3.0316,
      "theoretical_loss": 5.348185710930661,
      "tokens_seen": 47513600
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009955466399197593,
      "loss": 3.2195,
      "theoretical_loss": 5.3470625226575,
      "tokens_seen": 47579136
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009955265797392176,
      "loss": 3.1434,
      "theoretical_loss": 5.345941312917536,
      "tokens_seen": 47644672
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995506519558676,
      "loss": 3.2259,
      "theoretical_loss": 5.344822075511196,
      "tokens_seen": 47710208
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009954864593781343,
      "loss": 3.0835,
      "theoretical_loss": 5.343704804266816,
      "tokens_seen": 47775744
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995466399197593,
      "loss": 3.1539,
      "theoretical_loss": 5.342589493040472,
      "tokens_seen": 47841280
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995446339017051,
      "loss": 3.1759,
      "theoretical_loss": 5.3414761357158245,
      "tokens_seen": 47906816
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009954262788365096,
      "loss": 3.0574,
      "theoretical_loss": 5.340364726203955,
      "tokens_seen": 47972352
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995406218655968,
      "loss": 3.0117,
      "theoretical_loss": 5.339255258443202,
      "tokens_seen": 48037888
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009953861584754264,
      "loss": 2.9506,
      "theoretical_loss": 5.338147726399006,
      "tokens_seen": 48103424
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009953660982948847,
      "loss": 2.8871,
      "theoretical_loss": 5.33704212406375,
      "tokens_seen": 48168960
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995346038114343,
      "loss": 3.272,
      "theoretical_loss": 5.3359384454566055,
      "tokens_seen": 48234496
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009953259779338015,
      "loss": 3.1474,
      "theoretical_loss": 5.334836684623371,
      "tokens_seen": 48300032
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009953059177532598,
      "loss": 3.013,
      "theoretical_loss": 5.333736835636325,
      "tokens_seen": 48365568
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009952858575727182,
      "loss": 2.9694,
      "theoretical_loss": 5.332638892594063,
      "tokens_seen": 48431104
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009952657973921765,
      "loss": 3.0434,
      "theoretical_loss": 5.331542849621357,
      "tokens_seen": 48496640
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995245737211635,
      "loss": 3.0976,
      "theoretical_loss": 5.330448700868995,
      "tokens_seen": 48562176
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009952256770310933,
      "loss": 3.132,
      "theoretical_loss": 5.329356440513633,
      "tokens_seen": 48627712
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009952056168505516,
      "loss": 3.1594,
      "theoretical_loss": 5.32826606275765,
      "tokens_seen": 48693248
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00099518555667001,
      "loss": 3.0358,
      "theoretical_loss": 5.327177561828993,
      "tokens_seen": 48758784
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009951654964894684,
      "loss": 3.04,
      "theoretical_loss": 5.326090931981036,
      "tokens_seen": 48824320
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995145436308927,
      "loss": 3.1244,
      "theoretical_loss": 5.3250061674924325,
      "tokens_seen": 48889856
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000995125376128385,
      "loss": 3.1184,
      "theoretical_loss": 5.323923262666968,
      "tokens_seen": 48955392
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009951053159478437,
      "loss": 3.13,
      "theoretical_loss": 5.32284221183342,
      "tokens_seen": 49020928
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009950852557673018,
      "loss": 3.1912,
      "theoretical_loss": 5.321763009345411,
      "tokens_seen": 49086464
    },
    {
      "epoch": 0.01,
      "objective/train/advantage_avg": 0.4820156991481781,
      "objective/train/docs_used": 25394,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.984027862548828,
      "objective/train/original_loss": 2.98402738571167,
      "objective/train/theoretical_loss": 5.3206856495812715,
      "objective/train/tokens_used": 69612000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2343626171350479,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0493927001953125,
      "objective/train/weighted_lm_loss": 3.1309571266174316,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 5.3206856495812715,
      "tokens_seen": 49152000
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009950651955867604,
      "loss": 3.0822,
      "theoretical_loss": 5.3206856495812715,
      "tokens_seen": 49152000
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009950451354062187,
      "loss": 3.0233,
      "theoretical_loss": 5.3196101269438945,
      "tokens_seen": 49217536
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009950250752256771,
      "loss": 2.9487,
      "theoretical_loss": 5.318536435860599,
      "tokens_seen": 49283072
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009950050150451355,
      "loss": 3.0769,
      "theoretical_loss": 5.317464570782989,
      "tokens_seen": 49348608
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009949849548645938,
      "loss": 3.0474,
      "theoretical_loss": 5.316394526186816,
      "tokens_seen": 49414144
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0009949648946840522,
      "loss": 2.7961,
      "theoretical_loss": 5.3153262965718415,
      "tokens_seen": 49479680
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009949448345035106,
      "loss": 2.9753,
      "theoretical_loss": 5.314259876461705,
      "tokens_seen": 49545216
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000994924774322969,
      "loss": 3.0748,
      "theoretical_loss": 5.3131952604037815,
      "tokens_seen": 49610752
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009949047141424273,
      "loss": 2.9606,
      "theoretical_loss": 5.312132442969055,
      "tokens_seen": 49676288
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009948846539618856,
      "loss": 2.9401,
      "theoretical_loss": 5.3110714187519825,
      "tokens_seen": 49741824
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000994864593781344,
      "loss": 2.9654,
      "theoretical_loss": 5.310012182370359,
      "tokens_seen": 49807360
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009948445336008024,
      "loss": 2.8686,
      "theoretical_loss": 5.308954728465188,
      "tokens_seen": 49872896
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009948244734202607,
      "loss": 3.0768,
      "theoretical_loss": 5.307899051700552,
      "tokens_seen": 49938432
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000994804413239719,
      "loss": 2.9913,
      "theoretical_loss": 5.306845146763486,
      "tokens_seen": 50003968
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009947843530591777,
      "loss": 2.9679,
      "theoretical_loss": 5.305793008363841,
      "tokens_seen": 50069504
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009947642928786358,
      "loss": 3.0225,
      "theoretical_loss": 5.304742631234162,
      "tokens_seen": 50135040
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009947442326980944,
      "loss": 3.04,
      "theoretical_loss": 5.3036940101295595,
      "tokens_seen": 50200576
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009947241725175525,
      "loss": 2.8716,
      "theoretical_loss": 5.302647139827585,
      "tokens_seen": 50266112
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009947041123370111,
      "loss": 2.9582,
      "theoretical_loss": 5.301602015128104,
      "tokens_seen": 50331648
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009946840521564695,
      "loss": 3.0407,
      "theoretical_loss": 5.300558630853171,
      "tokens_seen": 50397184
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009946639919759278,
      "loss": 2.9239,
      "theoretical_loss": 5.299516981846912,
      "tokens_seen": 50462720
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009946439317953862,
      "loss": 2.909,
      "theoretical_loss": 5.29847706297539,
      "tokens_seen": 50528256
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009946238716148446,
      "loss": 2.952,
      "theoretical_loss": 5.297438869126498,
      "tokens_seen": 50593792
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000994603811434303,
      "loss": 2.9815,
      "theoretical_loss": 5.296402395209828,
      "tokens_seen": 50659328
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009945837512537613,
      "loss": 2.9763,
      "theoretical_loss": 5.295367636156552,
      "tokens_seen": 50724864
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009945636910732197,
      "loss": 3.1347,
      "theoretical_loss": 5.294334586919307,
      "tokens_seen": 50790400
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000994543630892678,
      "loss": 3.0942,
      "theoretical_loss": 5.293303242472074,
      "tokens_seen": 50855936
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009945235707121364,
      "loss": 2.9039,
      "theoretical_loss": 5.29227359781006,
      "tokens_seen": 50921472
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009945035105315947,
      "loss": 3.0075,
      "theoretical_loss": 5.29124564794958,
      "tokens_seen": 50987008
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000994483450351053,
      "loss": 2.8655,
      "theoretical_loss": 5.290219387927946,
      "tokens_seen": 51052544
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009944633901705115,
      "loss": 2.8902,
      "theoretical_loss": 5.289194812803347,
      "tokens_seen": 51118080
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009944433299899698,
      "loss": 3.0833,
      "theoretical_loss": 5.288171917654733,
      "tokens_seen": 51183616
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009944232698094284,
      "loss": 2.801,
      "theoretical_loss": 5.287150697581708,
      "tokens_seen": 51249152
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009944032096288866,
      "loss": 2.7836,
      "theoretical_loss": 5.286131147704413,
      "tokens_seen": 51314688
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009943831494483451,
      "loss": 2.9264,
      "theoretical_loss": 5.285113263163414,
      "tokens_seen": 51380224
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009943630892678035,
      "loss": 2.9479,
      "theoretical_loss": 5.284097039119587,
      "tokens_seen": 51445760
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009943430290872619,
      "loss": 3.0413,
      "theoretical_loss": 5.283082470754016,
      "tokens_seen": 51511296
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009943229689067202,
      "loss": 2.8968,
      "theoretical_loss": 5.282069553267874,
      "tokens_seen": 51576832
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009943029087261786,
      "loss": 2.7193,
      "theoretical_loss": 5.2810582818823235,
      "tokens_seen": 51642368
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000994282848545637,
      "loss": 2.9009,
      "theoretical_loss": 5.2800486518383964,
      "tokens_seen": 51707904
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009942627883650953,
      "loss": 2.8084,
      "theoretical_loss": 5.279040658396896,
      "tokens_seen": 51773440
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009942427281845537,
      "loss": 2.9595,
      "theoretical_loss": 5.278034296838284,
      "tokens_seen": 51838976
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000994222668004012,
      "loss": 2.8929,
      "theoretical_loss": 5.27702956246258,
      "tokens_seen": 51904512
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009942026078234704,
      "loss": 2.8877,
      "theoretical_loss": 5.276026450589249,
      "tokens_seen": 51970048
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009941825476429288,
      "loss": 2.9187,
      "theoretical_loss": 5.275024956557102,
      "tokens_seen": 52035584
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009941624874623871,
      "loss": 2.7947,
      "theoretical_loss": 5.274025075724189,
      "tokens_seen": 52101120
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009941424272818455,
      "loss": 2.7595,
      "theoretical_loss": 5.273026803467695,
      "tokens_seen": 52166656
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009941223671013038,
      "loss": 2.9168,
      "theoretical_loss": 5.272030135183842,
      "tokens_seen": 52232192
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009941023069207622,
      "loss": 2.872,
      "theoretical_loss": 5.271035066287779,
      "tokens_seen": 52297728
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009940822467402206,
      "loss": 2.9237,
      "theoretical_loss": 5.27004159221349,
      "tokens_seen": 52363264
    },
    {
      "epoch": 0.02,
      "objective/train/advantage_avg": 0.49426913261413574,
      "objective/train/docs_used": 26541,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.603318452835083,
      "objective/train/original_loss": 2.603318691253662,
      "objective/train/theoretical_loss": 5.269049708413682,
      "objective/train/tokens_used": 72888800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24451129138469696,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506700277328491,
      "objective/train/weighted_lm_loss": 2.735368251800537,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 5.269049708413682,
      "tokens_seen": 52428800
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009940621865596791,
      "loss": 2.7216,
      "theoretical_loss": 5.269049708413682,
      "tokens_seen": 52428800
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009940421263791373,
      "loss": 2.857,
      "theoretical_loss": 5.268059410359697,
      "tokens_seen": 52494336
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009940220661985959,
      "loss": 2.9196,
      "theoretical_loss": 5.267070693541403,
      "tokens_seen": 52559872
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009940020060180542,
      "loss": 2.8979,
      "theoretical_loss": 5.2660835534671,
      "tokens_seen": 52625408
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009939819458375126,
      "loss": 2.862,
      "theoretical_loss": 5.265097985663418,
      "tokens_seen": 52690944
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000993961885656971,
      "loss": 2.9804,
      "theoretical_loss": 5.264113985675227,
      "tokens_seen": 52756480
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009939418254764293,
      "loss": 2.8927,
      "theoretical_loss": 5.263131549065529,
      "tokens_seen": 52822016
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009939217652958877,
      "loss": 2.786,
      "theoretical_loss": 5.262150671415373,
      "tokens_seen": 52887552
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000993901705115346,
      "loss": 2.7832,
      "theoretical_loss": 5.261171348323755,
      "tokens_seen": 52953088
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009938816449348044,
      "loss": 2.9547,
      "theoretical_loss": 5.2601935754075155,
      "tokens_seen": 53018624
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009938615847542628,
      "loss": 2.8122,
      "theoretical_loss": 5.259217348301259,
      "tokens_seen": 53084160
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009938415245737211,
      "loss": 2.7562,
      "theoretical_loss": 5.25824266265725,
      "tokens_seen": 53149696
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009938214643931795,
      "loss": 2.8343,
      "theoretical_loss": 5.257269514145324,
      "tokens_seen": 53215232
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000993801404212638,
      "loss": 2.7475,
      "theoretical_loss": 5.256297898452793,
      "tokens_seen": 53280768
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009937813440320962,
      "loss": 2.7761,
      "theoretical_loss": 5.2553278112843556,
      "tokens_seen": 53346304
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009937612838515548,
      "loss": 2.7001,
      "theoretical_loss": 5.254359248362006,
      "tokens_seen": 53411840
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000993741223671013,
      "loss": 2.8198,
      "theoretical_loss": 5.2533922054249365,
      "tokens_seen": 53477376
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009937211634904715,
      "loss": 2.7642,
      "theoretical_loss": 5.25242667822946,
      "tokens_seen": 53542912
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009937011033099299,
      "loss": 2.8012,
      "theoretical_loss": 5.251462662548908,
      "tokens_seen": 53608448
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009936810431293882,
      "loss": 2.7885,
      "theoretical_loss": 5.25050015417355,
      "tokens_seen": 53673984
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009936609829488466,
      "loss": 2.8942,
      "theoretical_loss": 5.2495391489104986,
      "tokens_seen": 53739520
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000993640922768305,
      "loss": 2.9782,
      "theoretical_loss": 5.248579642583628,
      "tokens_seen": 53805056
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009936208625877633,
      "loss": 2.8523,
      "theoretical_loss": 5.24762163103348,
      "tokens_seen": 53870592
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009936008024072217,
      "loss": 2.7695,
      "theoretical_loss": 5.246665110117186,
      "tokens_seen": 53936128
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00099358074222668,
      "loss": 2.8059,
      "theoretical_loss": 5.24571007570837,
      "tokens_seen": 54001664
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009935606820461384,
      "loss": 2.8622,
      "theoretical_loss": 5.244756523697072,
      "tokens_seen": 54067200
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009935406218655968,
      "loss": 2.8403,
      "theoretical_loss": 5.243804449989659,
      "tokens_seen": 54132736
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009935205616850551,
      "loss": 2.5313,
      "theoretical_loss": 5.242853850508741,
      "tokens_seen": 54198272
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009935005015045135,
      "loss": 2.8811,
      "theoretical_loss": 5.2419047211930865,
      "tokens_seen": 54263808
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009934804413239719,
      "loss": 2.7909,
      "theoretical_loss": 5.24095705799754,
      "tokens_seen": 54329344
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009934603811434302,
      "loss": 2.6954,
      "theoretical_loss": 5.240010856892939,
      "tokens_seen": 54394880
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009934403209628888,
      "loss": 2.7536,
      "theoretical_loss": 5.239066113866029,
      "tokens_seen": 54460416
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000993420260782347,
      "loss": 2.7655,
      "theoretical_loss": 5.238122824919387,
      "tokens_seen": 54525952
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009934002006018055,
      "loss": 2.8289,
      "theoretical_loss": 5.237180986071335,
      "tokens_seen": 54591488
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000993380140421264,
      "loss": 2.7578,
      "theoretical_loss": 5.236240593355863,
      "tokens_seen": 54657024
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009933600802407223,
      "loss": 2.6962,
      "theoretical_loss": 5.235301642822543,
      "tokens_seen": 54722560
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009933400200601806,
      "loss": 2.7935,
      "theoretical_loss": 5.234364130536457,
      "tokens_seen": 54788096
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000993319959879639,
      "loss": 2.7993,
      "theoretical_loss": 5.2334280525781125,
      "tokens_seen": 54853632
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009932998996990973,
      "loss": 2.8282,
      "theoretical_loss": 5.232493405043362,
      "tokens_seen": 54919168
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009932798395185557,
      "loss": 2.8451,
      "theoretical_loss": 5.231560184043333,
      "tokens_seen": 54984704
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000993259779338014,
      "loss": 2.7866,
      "theoretical_loss": 5.230628385704337,
      "tokens_seen": 55050240
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009932397191574724,
      "loss": 2.9565,
      "theoretical_loss": 5.229698006167809,
      "tokens_seen": 55115776
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009932196589769308,
      "loss": 2.5017,
      "theoretical_loss": 5.228769041590215,
      "tokens_seen": 55181312
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009931995987963892,
      "loss": 2.7794,
      "theoretical_loss": 5.227841488142982,
      "tokens_seen": 55246848
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009931795386158475,
      "loss": 2.6695,
      "theoretical_loss": 5.2269153420124255,
      "tokens_seen": 55312384
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009931594784353059,
      "loss": 2.7819,
      "theoretical_loss": 5.22599059939967,
      "tokens_seen": 55377920
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009931394182547642,
      "loss": 2.75,
      "theoretical_loss": 5.225067256520572,
      "tokens_seen": 55443456
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009931193580742226,
      "loss": 2.5643,
      "theoretical_loss": 5.224145309605651,
      "tokens_seen": 55508992
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000993099297893681,
      "loss": 2.7934,
      "theoretical_loss": 5.223224754900014,
      "tokens_seen": 55574528
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009930792377131395,
      "loss": 2.7763,
      "theoretical_loss": 5.222305588663275,
      "tokens_seen": 55640064
    },
    {
      "epoch": 0.02,
      "objective/train/advantage_avg": 0.4954495131969452,
      "objective/train/docs_used": 27758,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.493486166000366,
      "objective/train/original_loss": 2.493486166000366,
      "objective/train/theoretical_loss": 5.221387807169494,
      "objective/train/tokens_used": 76165600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24568335711956024,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507941246032715,
      "objective/train/weighted_lm_loss": 2.6202352046966553,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 5.221387807169494,
      "tokens_seen": 55705600
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009930591775325977,
      "loss": 2.6668,
      "theoretical_loss": 5.221387807169494,
      "tokens_seen": 55705600
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009930391173520563,
      "loss": 2.9847,
      "theoretical_loss": 5.2204714067070945,
      "tokens_seen": 55771136
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009930190571715146,
      "loss": 2.6578,
      "theoretical_loss": 5.219556383578795,
      "tokens_seen": 55836672
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992998996990973,
      "loss": 2.7279,
      "theoretical_loss": 5.218642734101538,
      "tokens_seen": 55902208
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009929789368104314,
      "loss": 2.6868,
      "theoretical_loss": 5.217730454606417,
      "tokens_seen": 55967744
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009929588766298897,
      "loss": 2.79,
      "theoretical_loss": 5.216819541438607,
      "tokens_seen": 56033280
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992938816449348,
      "loss": 2.7179,
      "theoretical_loss": 5.215909990957291,
      "tokens_seen": 56098816
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009929187562688064,
      "loss": 2.7788,
      "theoretical_loss": 5.215001799535598,
      "tokens_seen": 56164352
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009928986960882648,
      "loss": 2.739,
      "theoretical_loss": 5.214094963560519,
      "tokens_seen": 56229888
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009928786359077232,
      "loss": 2.5606,
      "theoretical_loss": 5.213189479432854,
      "tokens_seen": 56295424
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009928585757271815,
      "loss": 2.734,
      "theoretical_loss": 5.212285343567135,
      "tokens_seen": 56360960
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00099283851554664,
      "loss": 2.5668,
      "theoretical_loss": 5.211382552391555,
      "tokens_seen": 56426496
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009928184553660983,
      "loss": 2.7951,
      "theoretical_loss": 5.210481102347908,
      "tokens_seen": 56492032
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009927983951855566,
      "loss": 2.6594,
      "theoretical_loss": 5.2095809898915135,
      "tokens_seen": 56557568
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992778335005015,
      "loss": 2.7474,
      "theoretical_loss": 5.208682211491157,
      "tokens_seen": 56623104
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009927582748244733,
      "loss": 2.687,
      "theoretical_loss": 5.207784763629014,
      "tokens_seen": 56688640
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009927382146439317,
      "loss": 2.5976,
      "theoretical_loss": 5.206888642800598,
      "tokens_seen": 56754176
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009927181544633903,
      "loss": 2.6364,
      "theoretical_loss": 5.205993845514678,
      "tokens_seen": 56819712
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009926980942828484,
      "loss": 2.6975,
      "theoretical_loss": 5.205100368293225,
      "tokens_seen": 56885248
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992678034102307,
      "loss": 2.6616,
      "theoretical_loss": 5.204208207671343,
      "tokens_seen": 56950784
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009926579739217654,
      "loss": 2.6684,
      "theoretical_loss": 5.203317360197204,
      "tokens_seen": 57016320
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009926379137412237,
      "loss": 2.629,
      "theoretical_loss": 5.202427822431982,
      "tokens_seen": 57081856
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992617853560682,
      "loss": 2.6768,
      "theoretical_loss": 5.201539590949796,
      "tokens_seen": 57147392
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009925977933801405,
      "loss": 2.5754,
      "theoretical_loss": 5.200652662337637,
      "tokens_seen": 57212928
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009925777331995988,
      "loss": 2.5187,
      "theoretical_loss": 5.199767033195316,
      "tokens_seen": 57278464
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009925576730190572,
      "loss": 2.7063,
      "theoretical_loss": 5.198882700135387,
      "tokens_seen": 57344000
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009925376128385155,
      "loss": 2.7405,
      "theoretical_loss": 5.1979996597831,
      "tokens_seen": 57409536
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992517552657974,
      "loss": 2.6938,
      "theoretical_loss": 5.1971179087763275,
      "tokens_seen": 57475072
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009924974924774323,
      "loss": 2.6516,
      "theoretical_loss": 5.19623744376551,
      "tokens_seen": 57540608
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009924774322968906,
      "loss": 2.6739,
      "theoretical_loss": 5.195358261413589,
      "tokens_seen": 57606144
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009924573721163492,
      "loss": 2.5905,
      "theoretical_loss": 5.1944803583959525,
      "tokens_seen": 57671680
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009924373119358074,
      "loss": 2.6303,
      "theoretical_loss": 5.193603731400369,
      "tokens_seen": 57737216
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992417251755266,
      "loss": 2.6194,
      "theoretical_loss": 5.192728377126931,
      "tokens_seen": 57802752
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009923971915747243,
      "loss": 2.7367,
      "theoretical_loss": 5.191854292287992,
      "tokens_seen": 57868288
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009923771313941827,
      "loss": 2.7908,
      "theoretical_loss": 5.190981473608112,
      "tokens_seen": 57933824
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992357071213641,
      "loss": 2.5301,
      "theoretical_loss": 5.1901099178239924,
      "tokens_seen": 57999360
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009923370110330994,
      "loss": 2.5255,
      "theoretical_loss": 5.189239621684421,
      "tokens_seen": 58064896
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009923169508525577,
      "loss": 2.7432,
      "theoretical_loss": 5.188370581950215,
      "tokens_seen": 58130432
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009922968906720161,
      "loss": 2.7216,
      "theoretical_loss": 5.18750279539416,
      "tokens_seen": 58195968
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009922768304914745,
      "loss": 2.493,
      "theoretical_loss": 5.1866362588009505,
      "tokens_seen": 58261504
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009922567703109328,
      "loss": 2.6899,
      "theoretical_loss": 5.185770968967139,
      "tokens_seen": 58327040
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009922367101303912,
      "loss": 2.6763,
      "theoretical_loss": 5.184906922701076,
      "tokens_seen": 58392576
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009922166499498496,
      "loss": 2.5682,
      "theoretical_loss": 5.184044116822849,
      "tokens_seen": 58458112
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992196589769308,
      "loss": 2.5105,
      "theoretical_loss": 5.183182548164234,
      "tokens_seen": 58523648
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009921765295887663,
      "loss": 2.5142,
      "theoretical_loss": 5.182322213568634,
      "tokens_seen": 58589184
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009921564694082246,
      "loss": 2.7216,
      "theoretical_loss": 5.181463109891023,
      "tokens_seen": 58654720
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992136409227683,
      "loss": 2.8247,
      "theoretical_loss": 5.1806052339978965,
      "tokens_seen": 58720256
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009921163490471414,
      "loss": 2.5839,
      "theoretical_loss": 5.179748582767209,
      "tokens_seen": 58785792
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009920962888666,
      "loss": 2.6148,
      "theoretical_loss": 5.178893153088326,
      "tokens_seen": 58851328
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992076228686058,
      "loss": 2.5025,
      "theoretical_loss": 5.178038941861966,
      "tokens_seen": 58916864
    },
    {
      "epoch": 0.02,
      "objective/train/advantage_avg": 0.4918819069862366,
      "objective/train/docs_used": 29071,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.2515406608581543,
      "objective/train/original_loss": 2.2515406608581543,
      "objective/train/theoretical_loss": 5.1771859460001455,
      "objective/train/tokens_used": 79442400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24229572713375092,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504199266433716,
      "objective/train/weighted_lm_loss": 2.365377187728882,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 5.1771859460001455,
      "tokens_seen": 58982400
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009920561685055167,
      "loss": 2.4743,
      "theoretical_loss": 5.1771859460001455,
      "tokens_seen": 58982400
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000992036108324975,
      "loss": 2.5575,
      "theoretical_loss": 5.176334162426128,
      "tokens_seen": 59047936
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009920160481444334,
      "loss": 2.6685,
      "theoretical_loss": 5.175483588074371,
      "tokens_seen": 59113472
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009919959879638918,
      "loss": 2.5362,
      "theoretical_loss": 5.17463421989047,
      "tokens_seen": 59179008
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009919759277833501,
      "loss": 2.6476,
      "theoretical_loss": 5.1737860548311065,
      "tokens_seen": 59244544
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009919558676028085,
      "loss": 2.5998,
      "theoretical_loss": 5.172939089863998,
      "tokens_seen": 59310080
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009919358074222668,
      "loss": 2.6016,
      "theoretical_loss": 5.172093321967843,
      "tokens_seen": 59375616
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009919157472417252,
      "loss": 2.432,
      "theoretical_loss": 5.171248748132273,
      "tokens_seen": 59441152
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009918956870611836,
      "loss": 2.7428,
      "theoretical_loss": 5.170405365357794,
      "tokens_seen": 59506688
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000991875626880642,
      "loss": 2.5627,
      "theoretical_loss": 5.169563170655742,
      "tokens_seen": 59572224
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009918555667001003,
      "loss": 2.5943,
      "theoretical_loss": 5.16872216104823,
      "tokens_seen": 59637760
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009918355065195587,
      "loss": 2.4533,
      "theoretical_loss": 5.167882333568095,
      "tokens_seen": 59703296
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000991815446339017,
      "loss": 2.6007,
      "theoretical_loss": 5.167043685258852,
      "tokens_seen": 59768832
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009917953861584754,
      "loss": 2.5313,
      "theoretical_loss": 5.16620621317464,
      "tokens_seen": 59834368
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009917753259779337,
      "loss": 2.5483,
      "theoretical_loss": 5.1653699143801735,
      "tokens_seen": 59899904
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000991755265797392,
      "loss": 2.6718,
      "theoretical_loss": 5.164534785950695,
      "tokens_seen": 59965440
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009917352056168507,
      "loss": 2.4902,
      "theoretical_loss": 5.163700824971922,
      "tokens_seen": 60030976
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009917151454363088,
      "loss": 2.7512,
      "theoretical_loss": 5.162868028540002,
      "tokens_seen": 60096512
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009916950852557674,
      "loss": 2.6651,
      "theoretical_loss": 5.1620363937614595,
      "tokens_seen": 60162048
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009916750250752258,
      "loss": 2.6301,
      "theoretical_loss": 5.161205917753154,
      "tokens_seen": 60227584
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009916549648946841,
      "loss": 2.5769,
      "theoretical_loss": 5.160376597642223,
      "tokens_seen": 60293120
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009916349047141425,
      "loss": 2.5911,
      "theoretical_loss": 5.159548430566044,
      "tokens_seen": 60358656
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009916148445336009,
      "loss": 2.6131,
      "theoretical_loss": 5.158721413672179,
      "tokens_seen": 60424192
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009915947843530592,
      "loss": 2.5863,
      "theoretical_loss": 5.1578955441183325,
      "tokens_seen": 60489728
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009915747241725176,
      "loss": 2.5351,
      "theoretical_loss": 5.157070819072301,
      "tokens_seen": 60555264
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000991554663991976,
      "loss": 2.7029,
      "theoretical_loss": 5.156247235711929,
      "tokens_seen": 60620800
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009915346038114343,
      "loss": 2.6004,
      "theoretical_loss": 5.155424791225061,
      "tokens_seen": 60686336
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009915145436308927,
      "loss": 2.6118,
      "theoretical_loss": 5.1546034828094935,
      "tokens_seen": 60751872
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000991494483450351,
      "loss": 2.5272,
      "theoretical_loss": 5.153783307672935,
      "tokens_seen": 60817408
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009914744232698094,
      "loss": 2.6638,
      "theoretical_loss": 5.152964263032954,
      "tokens_seen": 60882944
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009914543630892678,
      "loss": 2.5694,
      "theoretical_loss": 5.152146346116936,
      "tokens_seen": 60948480
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009914343029087261,
      "loss": 2.5069,
      "theoretical_loss": 5.151329554162039,
      "tokens_seen": 61014016
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009914142427281847,
      "loss": 2.6197,
      "theoretical_loss": 5.150513884415149,
      "tokens_seen": 61079552
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009913941825476428,
      "loss": 2.6271,
      "theoretical_loss": 5.1496993341328325,
      "tokens_seen": 61145088
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009913741223671014,
      "loss": 2.5083,
      "theoretical_loss": 5.148885900581297,
      "tokens_seen": 61210624
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009913540621865596,
      "loss": 2.6221,
      "theoretical_loss": 5.148073581036338,
      "tokens_seen": 61276160
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009913340020060181,
      "loss": 2.5627,
      "theoretical_loss": 5.14726237278331,
      "tokens_seen": 61341696
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009913139418254765,
      "loss": 2.5517,
      "theoretical_loss": 5.146452273117065,
      "tokens_seen": 61407232
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009912938816449349,
      "loss": 2.7035,
      "theoretical_loss": 5.1456432793419244,
      "tokens_seen": 61472768
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009912738214643932,
      "loss": 2.5791,
      "theoretical_loss": 5.144835388771626,
      "tokens_seen": 61538304
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009912537612838516,
      "loss": 2.6198,
      "theoretical_loss": 5.144028598729285,
      "tokens_seen": 61603840
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00099123370110331,
      "loss": 2.6126,
      "theoretical_loss": 5.1432229065473525,
      "tokens_seen": 61669376
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009912136409227683,
      "loss": 2.48,
      "theoretical_loss": 5.142418309567569,
      "tokens_seen": 61734912
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009911935807422267,
      "loss": 2.5514,
      "theoretical_loss": 5.141614805140927,
      "tokens_seen": 61800448
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000991173520561685,
      "loss": 2.6091,
      "theoretical_loss": 5.140812390627624,
      "tokens_seen": 61865984
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009911534603811434,
      "loss": 2.5793,
      "theoretical_loss": 5.1400110633970275,
      "tokens_seen": 61931520
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009911334002006018,
      "loss": 2.7038,
      "theoretical_loss": 5.139210820827625,
      "tokens_seen": 61997056
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009911133400200601,
      "loss": 2.5758,
      "theoretical_loss": 5.138411660306989,
      "tokens_seen": 62062592
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009910932798395185,
      "loss": 2.4817,
      "theoretical_loss": 5.137613579231737,
      "tokens_seen": 62128128
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009910732196589769,
      "loss": 2.6186,
      "theoretical_loss": 5.136816575007483,
      "tokens_seen": 62193664
    },
    {
      "epoch": 0.02,
      "objective/train/advantage_avg": 0.4942206144332886,
      "objective/train/docs_used": 30276,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.677966356277466,
      "objective/train/original_loss": 2.677966594696045,
      "objective/train/theoretical_loss": 5.136020645048806,
      "objective/train/tokens_used": 82719200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24524354934692383,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506689548492432,
      "objective/train/weighted_lm_loss": 2.8136417865753174,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 5.136020645048806,
      "tokens_seen": 62259200
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009910531594784354,
      "loss": 2.6662,
      "theoretical_loss": 5.136020645048806,
      "tokens_seen": 62259200
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009910330992978936,
      "loss": 2.4777,
      "theoretical_loss": 5.135225786779204,
      "tokens_seen": 62324736
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009910130391173522,
      "loss": 2.6514,
      "theoretical_loss": 5.134431997631053,
      "tokens_seen": 62390272
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009909929789368103,
      "loss": 2.4079,
      "theoretical_loss": 5.1336392750455735,
      "tokens_seen": 62455808
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009909729187562689,
      "loss": 2.5303,
      "theoretical_loss": 5.132847616472784,
      "tokens_seen": 62521344
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009909528585757272,
      "loss": 2.5487,
      "theoretical_loss": 5.132057019371466,
      "tokens_seen": 62586880
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009909327983951856,
      "loss": 2.4935,
      "theoretical_loss": 5.1312674812091235,
      "tokens_seen": 62652416
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990912738214644,
      "loss": 2.4875,
      "theoretical_loss": 5.130478999461939,
      "tokens_seen": 62717952
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009908926780341023,
      "loss": 2.5383,
      "theoretical_loss": 5.129691571614747,
      "tokens_seen": 62783488
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009908726178535607,
      "loss": 2.4761,
      "theoretical_loss": 5.128905195160981,
      "tokens_seen": 62849024
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990852557673019,
      "loss": 2.589,
      "theoretical_loss": 5.128119867602646,
      "tokens_seen": 62914560
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009908324974924774,
      "loss": 2.601,
      "theoretical_loss": 5.127335586450272,
      "tokens_seen": 62980096
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009908124373119358,
      "loss": 2.634,
      "theoretical_loss": 5.126552349222884,
      "tokens_seen": 63045632
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009907923771313941,
      "loss": 2.5431,
      "theoretical_loss": 5.12577015344796,
      "tokens_seen": 63111168
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009907723169508525,
      "loss": 2.5072,
      "theoretical_loss": 5.124988996661393,
      "tokens_seen": 63176704
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990752256770311,
      "loss": 2.5105,
      "theoretical_loss": 5.124208876407453,
      "tokens_seen": 63242240
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009907321965897692,
      "loss": 2.4315,
      "theoretical_loss": 5.1234297902387524,
      "tokens_seen": 63307776
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009907121364092278,
      "loss": 2.4456,
      "theoretical_loss": 5.12265173571621,
      "tokens_seen": 63373312
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009906920762286862,
      "loss": 2.5354,
      "theoretical_loss": 5.121874710409012,
      "tokens_seen": 63438848
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009906720160481445,
      "loss": 2.526,
      "theoretical_loss": 5.121098711894572,
      "tokens_seen": 63504384
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990651955867603,
      "loss": 2.4092,
      "theoretical_loss": 5.120323737758502,
      "tokens_seen": 63569920
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009906318956870613,
      "loss": 2.443,
      "theoretical_loss": 5.119549785594572,
      "tokens_seen": 63635456
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009906118355065196,
      "loss": 2.5223,
      "theoretical_loss": 5.118776853004677,
      "tokens_seen": 63700992
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990591775325978,
      "loss": 2.5233,
      "theoretical_loss": 5.118004937598793,
      "tokens_seen": 63766528
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009905717151454363,
      "loss": 2.4831,
      "theoretical_loss": 5.1172340369949545,
      "tokens_seen": 63832064
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009905516549648947,
      "loss": 2.5658,
      "theoretical_loss": 5.116464148819206,
      "tokens_seen": 63897600
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990531594784353,
      "loss": 2.5614,
      "theoretical_loss": 5.115695270705579,
      "tokens_seen": 63963136
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009905115346038114,
      "loss": 2.3751,
      "theoretical_loss": 5.114927400296046,
      "tokens_seen": 64028672
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009904914744232698,
      "loss": 2.4838,
      "theoretical_loss": 5.114160535240492,
      "tokens_seen": 64094208
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009904714142427282,
      "loss": 2.5281,
      "theoretical_loss": 5.113394673196682,
      "tokens_seen": 64159744
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009904513540621865,
      "loss": 2.4431,
      "theoretical_loss": 5.112629811830217,
      "tokens_seen": 64225280
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990431293881645,
      "loss": 2.4407,
      "theoretical_loss": 5.111865948814511,
      "tokens_seen": 64290816
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009904112337011032,
      "loss": 2.4222,
      "theoretical_loss": 5.111103081830752,
      "tokens_seen": 64356352
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009903911735205618,
      "loss": 2.545,
      "theoretical_loss": 5.1103412085678634,
      "tokens_seen": 64421888
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00099037111334002,
      "loss": 2.5683,
      "theoretical_loss": 5.10958032672248,
      "tokens_seen": 64487424
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009903510531594785,
      "loss": 2.572,
      "theoretical_loss": 5.108820433998908,
      "tokens_seen": 64552960
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990330992978937,
      "loss": 2.381,
      "theoretical_loss": 5.108061528109092,
      "tokens_seen": 64618496
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009903109327983953,
      "loss": 2.4737,
      "theoretical_loss": 5.107303606772582,
      "tokens_seen": 64684032
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009902908726178536,
      "loss": 2.4322,
      "theoretical_loss": 5.106546667716508,
      "tokens_seen": 64749568
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990270812437312,
      "loss": 2.592,
      "theoretical_loss": 5.1057907086755305,
      "tokens_seen": 64815104
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009902507522567704,
      "loss": 2.5258,
      "theoretical_loss": 5.105035727391828,
      "tokens_seen": 64880640
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009902306920762287,
      "loss": 2.6152,
      "theoretical_loss": 5.104281721615046,
      "tokens_seen": 64946176
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990210631895687,
      "loss": 2.5266,
      "theoretical_loss": 5.103528689102281,
      "tokens_seen": 65011712
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009901905717151454,
      "loss": 2.5043,
      "theoretical_loss": 5.102776627618036,
      "tokens_seen": 65077248
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009901705115346038,
      "loss": 2.5268,
      "theoretical_loss": 5.102025534934196,
      "tokens_seen": 65142784
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009901504513540622,
      "loss": 2.5599,
      "theoretical_loss": 5.10127540882999,
      "tokens_seen": 65208320
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009901303911735205,
      "loss": 2.403,
      "theoretical_loss": 5.100526247091967,
      "tokens_seen": 65273856
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990110330992979,
      "loss": 2.4966,
      "theoretical_loss": 5.099778047513957,
      "tokens_seen": 65339392
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009900902708124373,
      "loss": 2.5187,
      "theoretical_loss": 5.099030807897048,
      "tokens_seen": 65404928
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009900702106318958,
      "loss": 2.5642,
      "theoretical_loss": 5.098284526049547,
      "tokens_seen": 65470464
    },
    {
      "epoch": 0.02,
      "objective/train/advantage_avg": 0.49179795384407043,
      "objective/train/docs_used": 31526,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.5606729984283447,
      "objective/train/original_loss": 2.5606727600097656,
      "objective/train/theoretical_loss": 5.097539199786951,
      "objective/train/tokens_used": 85996000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24245329201221466,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504124164581299,
      "objective/train/weighted_lm_loss": 2.6897695064544678,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9753099083900452,
      "theoretical_loss": 5.097539199786951,
      "tokens_seen": 65536000
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000990050150451354,
      "loss": 2.5972,
      "theoretical_loss": 5.097539199786951,
      "tokens_seen": 65536000
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009900300902708126,
      "loss": 2.5759,
      "theoretical_loss": 5.096794826931921,
      "tokens_seen": 65601536
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009900100300902707,
      "loss": 2.5983,
      "theoretical_loss": 5.096051405314247,
      "tokens_seen": 65667072
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009899899699097293,
      "loss": 2.6292,
      "theoretical_loss": 5.095308932770816,
      "tokens_seen": 65732608
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009899699097291876,
      "loss": 2.4965,
      "theoretical_loss": 5.094567407145588,
      "tokens_seen": 65798144
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000989949849548646,
      "loss": 2.4289,
      "theoretical_loss": 5.093826826289559,
      "tokens_seen": 65863680
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009899297893681044,
      "loss": 2.5533,
      "theoretical_loss": 5.093087188060739,
      "tokens_seen": 65929216
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009899097291875627,
      "loss": 2.4959,
      "theoretical_loss": 5.09234849032411,
      "tokens_seen": 65994752
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000989889669007021,
      "loss": 2.4689,
      "theoretical_loss": 5.09161073095161,
      "tokens_seen": 66060288
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009898696088264795,
      "loss": 2.4506,
      "theoretical_loss": 5.090873907822093,
      "tokens_seen": 66125824
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009898495486459378,
      "loss": 2.408,
      "theoretical_loss": 5.090138018821309,
      "tokens_seen": 66191360
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009898294884653962,
      "loss": 2.4846,
      "theoretical_loss": 5.089403061841864,
      "tokens_seen": 66256896
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009898094282848545,
      "loss": 2.502,
      "theoretical_loss": 5.0886690347832015,
      "tokens_seen": 66322432
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000989789368104313,
      "loss": 2.4468,
      "theoretical_loss": 5.087935935551566,
      "tokens_seen": 66387968
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009897693079237713,
      "loss": 2.4499,
      "theoretical_loss": 5.0872037620599775,
      "tokens_seen": 66453504
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009897492477432296,
      "loss": 2.4336,
      "theoretical_loss": 5.0864725122282035,
      "tokens_seen": 66519040
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000989729187562688,
      "loss": 2.5683,
      "theoretical_loss": 5.0857421839827275,
      "tokens_seen": 66584576
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009897091273821466,
      "loss": 2.4402,
      "theoretical_loss": 5.085012775256727,
      "tokens_seen": 66650112
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009896890672016047,
      "loss": 2.441,
      "theoretical_loss": 5.084284283990035,
      "tokens_seen": 66715648
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009896690070210633,
      "loss": 2.4647,
      "theoretical_loss": 5.083556708129126,
      "tokens_seen": 66781184
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009896489468405214,
      "loss": 2.3144,
      "theoretical_loss": 5.082830045627072,
      "tokens_seen": 66846720
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00098962888665998,
      "loss": 2.4836,
      "theoretical_loss": 5.08210429444353,
      "tokens_seen": 66912256
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009896088264794384,
      "loss": 2.4112,
      "theoretical_loss": 5.081379452544702,
      "tokens_seen": 66977792
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009895887662988967,
      "loss": 2.5422,
      "theoretical_loss": 5.080655517903319,
      "tokens_seen": 67043328
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000989568706118355,
      "loss": 2.4475,
      "theoretical_loss": 5.079932488498602,
      "tokens_seen": 67108864
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009895486459378135,
      "loss": 2.6382,
      "theoretical_loss": 5.079210362316245,
      "tokens_seen": 67174400
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009895285857572718,
      "loss": 2.4146,
      "theoretical_loss": 5.078489137348383,
      "tokens_seen": 67239936
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009895085255767302,
      "loss": 2.4907,
      "theoretical_loss": 5.077768811593563,
      "tokens_seen": 67305472
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009894884653961886,
      "loss": 2.3637,
      "theoretical_loss": 5.077049383056725,
      "tokens_seen": 67371008
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000989468405215647,
      "loss": 2.3986,
      "theoretical_loss": 5.076330849749166,
      "tokens_seen": 67436544
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009894483450351055,
      "loss": 2.5128,
      "theoretical_loss": 5.0756132096885205,
      "tokens_seen": 67502080
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009894282848545636,
      "loss": 2.6255,
      "theoretical_loss": 5.0748964608987315,
      "tokens_seen": 67567616
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009894082246740222,
      "loss": 2.5026,
      "theoretical_loss": 5.074180601410026,
      "tokens_seen": 67633152
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009893881644934804,
      "loss": 2.3306,
      "theoretical_loss": 5.073465629258886,
      "tokens_seen": 67698688
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000989368104312939,
      "loss": 2.4388,
      "theoretical_loss": 5.072751542488024,
      "tokens_seen": 67764224
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009893480441323973,
      "loss": 2.4484,
      "theoretical_loss": 5.072038339146357,
      "tokens_seen": 67829760
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009893279839518557,
      "loss": 2.3594,
      "theoretical_loss": 5.0713260172889845,
      "tokens_seen": 67895296
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000989307923771314,
      "loss": 2.4074,
      "theoretical_loss": 5.070614574977157,
      "tokens_seen": 67960832
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009892878635907724,
      "loss": 2.4488,
      "theoretical_loss": 5.069904010278254,
      "tokens_seen": 68026368
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009892678034102308,
      "loss": 2.3926,
      "theoretical_loss": 5.069194321265758,
      "tokens_seen": 68091904
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009892477432296891,
      "loss": 2.2627,
      "theoretical_loss": 5.068485506019231,
      "tokens_seen": 68157440
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009892276830491475,
      "loss": 2.3413,
      "theoretical_loss": 5.0677775626242845,
      "tokens_seen": 68222976
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009892076228686058,
      "loss": 2.4958,
      "theoretical_loss": 5.067070489172562,
      "tokens_seen": 68288512
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009891875626880642,
      "loss": 2.3923,
      "theoretical_loss": 5.066364283761709,
      "tokens_seen": 68354048
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009891675025075226,
      "loss": 2.4773,
      "theoretical_loss": 5.06565894449535,
      "tokens_seen": 68419584
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000989147442326981,
      "loss": 2.3944,
      "theoretical_loss": 5.064954469483064,
      "tokens_seen": 68485120
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009891273821464393,
      "loss": 2.4673,
      "theoretical_loss": 5.064250856840359,
      "tokens_seen": 68550656
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009891073219658977,
      "loss": 2.5153,
      "theoretical_loss": 5.06354810468865,
      "tokens_seen": 68616192
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009890872617853562,
      "loss": 2.367,
      "theoretical_loss": 5.06284621115523,
      "tokens_seen": 68681728
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009890672016048144,
      "loss": 2.4625,
      "theoretical_loss": 5.062145174373255,
      "tokens_seen": 68747264
    },
    {
      "epoch": 0.02,
      "objective/train/advantage_avg": 0.49227413535118103,
      "objective/train/docs_used": 32629,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.2441399097442627,
      "objective/train/original_loss": 2.2441399097442627,
      "objective/train/theoretical_loss": 5.061444992481711,
      "objective/train/tokens_used": 89272800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24276849627494812,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504614114761353,
      "objective/train/weighted_lm_loss": 2.3572661876678467,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 5.061444992481711,
      "tokens_seen": 68812800
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000989047141424273,
      "loss": 2.4288,
      "theoretical_loss": 5.061444992481711,
      "tokens_seen": 68812800
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000989027081243731,
      "loss": 2.5382,
      "theoretical_loss": 5.060745663625397,
      "tokens_seen": 68878336
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009890070210631897,
      "loss": 2.5068,
      "theoretical_loss": 5.060047185954893,
      "tokens_seen": 68943872
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000988986960882648,
      "loss": 2.3632,
      "theoretical_loss": 5.059349557626547,
      "tokens_seen": 69009408
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009889669007021064,
      "loss": 2.346,
      "theoretical_loss": 5.058652776802443,
      "tokens_seen": 69074944
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009889468405215648,
      "loss": 2.4677,
      "theoretical_loss": 5.057956841650385,
      "tokens_seen": 69140480
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009889267803410231,
      "loss": 2.471,
      "theoretical_loss": 5.057261750343864,
      "tokens_seen": 69206016
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009889067201604815,
      "loss": 2.3835,
      "theoretical_loss": 5.0565675010620454,
      "tokens_seen": 69271552
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009888866599799399,
      "loss": 2.319,
      "theoretical_loss": 5.05587409198974,
      "tokens_seen": 69337088
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009888665997993982,
      "loss": 2.5456,
      "theoretical_loss": 5.055181521317383,
      "tokens_seen": 69402624
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009888465396188566,
      "loss": 2.4407,
      "theoretical_loss": 5.0544897872410095,
      "tokens_seen": 69468160
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000988826479438315,
      "loss": 2.3936,
      "theoretical_loss": 5.053798887962236,
      "tokens_seen": 69533696
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009888064192577733,
      "loss": 2.3531,
      "theoretical_loss": 5.053108821688232,
      "tokens_seen": 69599232
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009887863590772317,
      "loss": 2.4097,
      "theoretical_loss": 5.052419586631703,
      "tokens_seen": 69664768
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00098876629889669,
      "loss": 2.4689,
      "theoretical_loss": 5.051731181010866,
      "tokens_seen": 69730304
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009887462387161484,
      "loss": 2.3541,
      "theoretical_loss": 5.051043603049426,
      "tokens_seen": 69795840
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000988726178535607,
      "loss": 2.3945,
      "theoretical_loss": 5.0503568509765575,
      "tokens_seen": 69861376
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009887061183550651,
      "loss": 2.3833,
      "theoretical_loss": 5.049670923026879,
      "tokens_seen": 69926912
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009886860581745237,
      "loss": 2.3977,
      "theoretical_loss": 5.048985817440432,
      "tokens_seen": 69992448
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009886659979939818,
      "loss": 2.4533,
      "theoretical_loss": 5.048301532462662,
      "tokens_seen": 70057984
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009886459378134404,
      "loss": 2.5417,
      "theoretical_loss": 5.047618066344392,
      "tokens_seen": 70123520
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009886258776328988,
      "loss": 2.3964,
      "theoretical_loss": 5.046935417341806,
      "tokens_seen": 70189056
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009886058174523571,
      "loss": 2.4526,
      "theoretical_loss": 5.046253583716425,
      "tokens_seen": 70254592
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009885857572718155,
      "loss": 2.425,
      "theoretical_loss": 5.045572563735084,
      "tokens_seen": 70320128
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009885656970912739,
      "loss": 2.3847,
      "theoretical_loss": 5.0448923556699174,
      "tokens_seen": 70385664
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009885456369107322,
      "loss": 2.458,
      "theoretical_loss": 5.044212957798328,
      "tokens_seen": 70451200
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009885255767301906,
      "loss": 2.3902,
      "theoretical_loss": 5.043534368402973,
      "tokens_seen": 70516736
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000988505516549649,
      "loss": 2.2822,
      "theoretical_loss": 5.042856585771746,
      "tokens_seen": 70582272
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009884854563691073,
      "loss": 2.3476,
      "theoretical_loss": 5.0421796081977455,
      "tokens_seen": 70647808
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009884653961885657,
      "loss": 2.4328,
      "theoretical_loss": 5.041503433979264,
      "tokens_seen": 70713344
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000988445336008024,
      "loss": 2.5019,
      "theoretical_loss": 5.040828061419762,
      "tokens_seen": 70778880
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009884252758274824,
      "loss": 2.4293,
      "theoretical_loss": 5.040153488827851,
      "tokens_seen": 70844416
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009884052156469408,
      "loss": 2.5365,
      "theoretical_loss": 5.039479714517272,
      "tokens_seen": 70909952
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009883851554663991,
      "loss": 2.4693,
      "theoretical_loss": 5.038806736806871,
      "tokens_seen": 70975488
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009883650952858577,
      "loss": 2.3275,
      "theoretical_loss": 5.038134554020587,
      "tokens_seen": 71041024
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009883450351053159,
      "loss": 2.3314,
      "theoretical_loss": 5.037463164487425,
      "tokens_seen": 71106560
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009883249749247744,
      "loss": 2.3667,
      "theoretical_loss": 5.036792566541441,
      "tokens_seen": 71172096
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009883049147442326,
      "loss": 2.3339,
      "theoretical_loss": 5.036122758521715,
      "tokens_seen": 71237632
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009882848545636912,
      "loss": 2.5154,
      "theoretical_loss": 5.03545373877234,
      "tokens_seen": 71303168
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009882647943831495,
      "loss": 2.4845,
      "theoretical_loss": 5.034785505642397,
      "tokens_seen": 71368704
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009882447342026079,
      "loss": 2.4743,
      "theoretical_loss": 5.034118057485935,
      "tokens_seen": 71434240
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009882246740220662,
      "loss": 2.5096,
      "theoretical_loss": 5.033451392661957,
      "tokens_seen": 71499776
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009882046138415246,
      "loss": 2.3763,
      "theoretical_loss": 5.032785509534391,
      "tokens_seen": 71565312
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000988184553660983,
      "loss": 2.4245,
      "theoretical_loss": 5.032120406472082,
      "tokens_seen": 71630848
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009881644934804413,
      "loss": 2.4523,
      "theoretical_loss": 5.031456081848761,
      "tokens_seen": 71696384
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009881444332998997,
      "loss": 2.2822,
      "theoretical_loss": 5.030792534043039,
      "tokens_seen": 71761920
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000988124373119358,
      "loss": 2.4078,
      "theoretical_loss": 5.030129761438376,
      "tokens_seen": 71827456
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009881043129388164,
      "loss": 2.4901,
      "theoretical_loss": 5.029467762423069,
      "tokens_seen": 71892992
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009880842527582748,
      "loss": 2.4351,
      "theoretical_loss": 5.028806535390228,
      "tokens_seen": 71958528
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009880641925777331,
      "loss": 2.4051,
      "theoretical_loss": 5.028146078737766,
      "tokens_seen": 72024064
    },
    {
      "epoch": 0.02,
      "objective/train/advantage_avg": 0.49358925223350525,
      "objective/train/docs_used": 33987,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.276097536087036,
      "objective/train/original_loss": 2.2760977745056152,
      "objective/train/theoretical_loss": 5.0274863908683685,
      "objective/train/tokens_used": 92549600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24399147927761078,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050599455833435,
      "objective/train/weighted_lm_loss": 2.3911280632019043,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 5.0274863908683685,
      "tokens_seen": 72089600
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009880441323971915,
      "loss": 2.258,
      "theoretical_loss": 5.0274863908683685,
      "tokens_seen": 72089600
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009880240722166499,
      "loss": 2.4077,
      "theoretical_loss": 5.026827470189487,
      "tokens_seen": 72155136
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009880040120361084,
      "loss": 2.4191,
      "theoretical_loss": 5.026169315113311,
      "tokens_seen": 72220672
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009879839518555668,
      "loss": 2.3911,
      "theoretical_loss": 5.025511924056753,
      "tokens_seen": 72286208
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009879638916750252,
      "loss": 2.5566,
      "theoretical_loss": 5.024855295441432,
      "tokens_seen": 72351744
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009879438314944835,
      "loss": 2.3835,
      "theoretical_loss": 5.024199427693654,
      "tokens_seen": 72417280
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000987923771313942,
      "loss": 2.4704,
      "theoretical_loss": 5.023544319244391,
      "tokens_seen": 72482816
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009879037111334003,
      "loss": 2.3699,
      "theoretical_loss": 5.0228899685292685,
      "tokens_seen": 72548352
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009878836509528586,
      "loss": 2.1989,
      "theoretical_loss": 5.022236373988544,
      "tokens_seen": 72613888
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000987863590772317,
      "loss": 2.4072,
      "theoretical_loss": 5.021583534067085,
      "tokens_seen": 72679424
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009878435305917753,
      "loss": 2.3943,
      "theoretical_loss": 5.020931447214364,
      "tokens_seen": 72744960
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009878234704112337,
      "loss": 2.4555,
      "theoretical_loss": 5.020280111884426,
      "tokens_seen": 72810496
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000987803410230692,
      "loss": 2.3937,
      "theoretical_loss": 5.01962952653588,
      "tokens_seen": 72876032
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009877833500501504,
      "loss": 2.2974,
      "theoretical_loss": 5.018979689631879,
      "tokens_seen": 72941568
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009877632898696088,
      "loss": 2.2553,
      "theoretical_loss": 5.018330599640101,
      "tokens_seen": 73007104
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009877432296890674,
      "loss": 2.3621,
      "theoretical_loss": 5.017682255032735,
      "tokens_seen": 73072640
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009877231695085255,
      "loss": 2.5088,
      "theoretical_loss": 5.017034654286462,
      "tokens_seen": 73138176
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000987703109327984,
      "loss": 2.4084,
      "theoretical_loss": 5.0163877958824346,
      "tokens_seen": 73203712
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009876830491474422,
      "loss": 2.4194,
      "theoretical_loss": 5.0157416783062665,
      "tokens_seen": 73269248
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009876629889669008,
      "loss": 2.4157,
      "theoretical_loss": 5.015096300048009,
      "tokens_seen": 73334784
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009876429287863592,
      "loss": 2.4174,
      "theoretical_loss": 5.0144516596021385,
      "tokens_seen": 73400320
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009876228686058175,
      "loss": 2.205,
      "theoretical_loss": 5.013807755467537,
      "tokens_seen": 73465856
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000987602808425276,
      "loss": 2.2951,
      "theoretical_loss": 5.013164586147477,
      "tokens_seen": 73531392
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009875827482447343,
      "loss": 2.3731,
      "theoretical_loss": 5.012522150149602,
      "tokens_seen": 73596928
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009875626880641926,
      "loss": 2.4183,
      "theoretical_loss": 5.011880445985916,
      "tokens_seen": 73662464
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000987542627883651,
      "loss": 2.3,
      "theoretical_loss": 5.011239472172758,
      "tokens_seen": 73728000
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009875225677031094,
      "loss": 2.5542,
      "theoretical_loss": 5.010599227230792,
      "tokens_seen": 73793536
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009875025075225677,
      "loss": 2.5132,
      "theoretical_loss": 5.009959709684992,
      "tokens_seen": 73859072
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000987482447342026,
      "loss": 2.409,
      "theoretical_loss": 5.009320918064615,
      "tokens_seen": 73924608
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009874623871614844,
      "loss": 2.4014,
      "theoretical_loss": 5.008682850903201,
      "tokens_seen": 73990144
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009874423269809428,
      "loss": 2.4533,
      "theoretical_loss": 5.00804550673854,
      "tokens_seen": 74055680
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009874222668004012,
      "loss": 2.309,
      "theoretical_loss": 5.007408884112671,
      "tokens_seen": 74121216
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009874022066198595,
      "loss": 2.5961,
      "theoretical_loss": 5.006772981571855,
      "tokens_seen": 74186752
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009873821464393181,
      "loss": 2.4414,
      "theoretical_loss": 5.006137797666559,
      "tokens_seen": 74252288
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009873620862587763,
      "loss": 2.2648,
      "theoretical_loss": 5.005503330951454,
      "tokens_seen": 74317824
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009873420260782348,
      "loss": 2.4416,
      "theoretical_loss": 5.004869579985381,
      "tokens_seen": 74383360
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000987321965897693,
      "loss": 2.3453,
      "theoretical_loss": 5.004236543331345,
      "tokens_seen": 74448896
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009873019057171516,
      "loss": 2.3843,
      "theoretical_loss": 5.0036042195565,
      "tokens_seen": 74514432
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00098728184553661,
      "loss": 2.3351,
      "theoretical_loss": 5.002972607232129,
      "tokens_seen": 74579968
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009872617853560683,
      "loss": 2.3041,
      "theoretical_loss": 5.002341704933631,
      "tokens_seen": 74645504
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009872417251755266,
      "loss": 2.4829,
      "theoretical_loss": 5.001711511240506,
      "tokens_seen": 74711040
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000987221664994985,
      "loss": 2.3724,
      "theoretical_loss": 5.0010820247363394,
      "tokens_seen": 74776576
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009872016048144434,
      "loss": 2.3896,
      "theoretical_loss": 5.000453244008782,
      "tokens_seen": 74842112
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009871815446339017,
      "loss": 2.4098,
      "theoretical_loss": 4.9998251676495435,
      "tokens_seen": 74907648
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00098716148445336,
      "loss": 2.425,
      "theoretical_loss": 4.999197794254371,
      "tokens_seen": 74973184
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009871414242728185,
      "loss": 2.5423,
      "theoretical_loss": 4.998571122423035,
      "tokens_seen": 75038720
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009871213640922768,
      "loss": 2.2918,
      "theoretical_loss": 4.997945150759316,
      "tokens_seen": 75104256
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009871013039117352,
      "loss": 2.3774,
      "theoretical_loss": 4.997319877870986,
      "tokens_seen": 75169792
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009870812437311935,
      "loss": 2.3024,
      "theoretical_loss": 4.9966953023697975,
      "tokens_seen": 75235328
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000987061183550652,
      "loss": 2.4583,
      "theoretical_loss": 4.99607142287147,
      "tokens_seen": 75300864
    },
    {
      "epoch": 0.02,
      "objective/train/advantage_avg": 0.49300244450569153,
      "objective/train/docs_used": 35047,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.2772889137268066,
      "objective/train/original_loss": 2.2772884368896484,
      "objective/train/theoretical_loss": 4.995448237995667,
      "objective/train/tokens_used": 95826400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24339433014392853,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505375862121582,
      "objective/train/weighted_lm_loss": 2.392406463623047,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.995448237995667,
      "tokens_seen": 75366400
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009870411233701103,
      "loss": 2.4079,
      "theoretical_loss": 4.995448237995667,
      "tokens_seen": 75366400
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009870210631895688,
      "loss": 2.3811,
      "theoretical_loss": 4.994825746365992,
      "tokens_seen": 75431936
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000987001003009027,
      "loss": 2.4155,
      "theoretical_loss": 4.994203946609964,
      "tokens_seen": 75497472
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009869809428284856,
      "loss": 2.2558,
      "theoretical_loss": 4.993582837359012,
      "tokens_seen": 75563008
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009869608826479437,
      "loss": 2.4505,
      "theoretical_loss": 4.992962417248453,
      "tokens_seen": 75628544
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009869408224674023,
      "loss": 2.4398,
      "theoretical_loss": 4.992342684917482,
      "tokens_seen": 75694080
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009869207622868607,
      "loss": 2.2416,
      "theoretical_loss": 4.991723639009154,
      "tokens_seen": 75759616
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000986900702106319,
      "loss": 2.2164,
      "theoretical_loss": 4.991105278170377,
      "tokens_seen": 75825152
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009868806419257774,
      "loss": 2.5057,
      "theoretical_loss": 4.990487601051887,
      "tokens_seen": 75890688
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009868605817452357,
      "loss": 2.4679,
      "theoretical_loss": 4.989870606308244,
      "tokens_seen": 75956224
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000986840521564694,
      "loss": 2.3433,
      "theoretical_loss": 4.989254292597813,
      "tokens_seen": 76021760
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009868204613841525,
      "loss": 2.5069,
      "theoretical_loss": 4.9886386585827465,
      "tokens_seen": 76087296
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009868004012036108,
      "loss": 2.4961,
      "theoretical_loss": 4.988023702928979,
      "tokens_seen": 76152832
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009867803410230692,
      "loss": 2.3971,
      "theoretical_loss": 4.9874094243062075,
      "tokens_seen": 76218368
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009867602808425276,
      "loss": 2.4396,
      "theoretical_loss": 4.986795821387878,
      "tokens_seen": 76283904
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000986740220661986,
      "loss": 2.4312,
      "theoretical_loss": 4.986182892851171,
      "tokens_seen": 76349440
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009867201604814443,
      "loss": 2.3369,
      "theoretical_loss": 4.985570637376993,
      "tokens_seen": 76414976
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009867001003009026,
      "loss": 2.3453,
      "theoretical_loss": 4.984959053649959,
      "tokens_seen": 76480512
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000986680040120361,
      "loss": 2.3224,
      "theoretical_loss": 4.984348140358374,
      "tokens_seen": 76546048
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009866599799398196,
      "loss": 2.3089,
      "theoretical_loss": 4.983737896194229,
      "tokens_seen": 76611584
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009866399197592777,
      "loss": 2.3881,
      "theoretical_loss": 4.983128319853182,
      "tokens_seen": 76677120
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009866198595787363,
      "loss": 2.3237,
      "theoretical_loss": 4.982519410034545,
      "tokens_seen": 76742656
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009865997993981945,
      "loss": 2.2656,
      "theoretical_loss": 4.981911165441273,
      "tokens_seen": 76808192
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000986579739217653,
      "loss": 2.358,
      "theoretical_loss": 4.981303584779946,
      "tokens_seen": 76873728
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009865596790371114,
      "loss": 2.4148,
      "theoretical_loss": 4.980696666760761,
      "tokens_seen": 76939264
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009865396188565698,
      "loss": 2.3153,
      "theoretical_loss": 4.980090410097517,
      "tokens_seen": 77004800
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009865195586760281,
      "loss": 2.3638,
      "theoretical_loss": 4.979484813507599,
      "tokens_seen": 77070336
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009864994984954865,
      "loss": 2.3893,
      "theoretical_loss": 4.97887987571197,
      "tokens_seen": 77135872
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009864794383149448,
      "loss": 2.2725,
      "theoretical_loss": 4.9782755954351545,
      "tokens_seen": 77201408
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009864593781344032,
      "loss": 2.3585,
      "theoretical_loss": 4.977671971405225,
      "tokens_seen": 77266944
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009864393179538616,
      "loss": 2.3069,
      "theoretical_loss": 4.977069002353792,
      "tokens_seen": 77332480
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00098641925777332,
      "loss": 2.3565,
      "theoretical_loss": 4.976466687015991,
      "tokens_seen": 77398016
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009863991975927785,
      "loss": 2.375,
      "theoretical_loss": 4.975865024130467,
      "tokens_seen": 77463552
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009863791374122367,
      "loss": 2.2839,
      "theoretical_loss": 4.975264012439363,
      "tokens_seen": 77529088
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009863590772316952,
      "loss": 2.3887,
      "theoretical_loss": 4.974663650688306,
      "tokens_seen": 77594624
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009863390170511534,
      "loss": 2.3619,
      "theoretical_loss": 4.974063937626402,
      "tokens_seen": 77660160
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000986318956870612,
      "loss": 2.3161,
      "theoretical_loss": 4.973464872006213,
      "tokens_seen": 77725696
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009862988966900703,
      "loss": 2.2241,
      "theoretical_loss": 4.972866452583748,
      "tokens_seen": 77791232
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009862788365095287,
      "loss": 2.5561,
      "theoretical_loss": 4.972268678118454,
      "tokens_seen": 77856768
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000986258776328987,
      "loss": 2.2846,
      "theoretical_loss": 4.971671547373202,
      "tokens_seen": 77922304
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009862387161484454,
      "loss": 2.3102,
      "theoretical_loss": 4.971075059114272,
      "tokens_seen": 77987840
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009862186559679038,
      "loss": 2.2737,
      "theoretical_loss": 4.970479212111342,
      "tokens_seen": 78053376
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009861985957873621,
      "loss": 2.2125,
      "theoretical_loss": 4.969884005137479,
      "tokens_seen": 78118912
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009861785356068205,
      "loss": 2.4044,
      "theoretical_loss": 4.969289436969124,
      "tokens_seen": 78184448
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009861584754262789,
      "loss": 2.3445,
      "theoretical_loss": 4.968695506386077,
      "tokens_seen": 78249984
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009861384152457372,
      "loss": 2.1897,
      "theoretical_loss": 4.968102212171493,
      "tokens_seen": 78315520
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009861183550651956,
      "loss": 2.3495,
      "theoretical_loss": 4.967509553111862,
      "tokens_seen": 78381056
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000986098294884654,
      "loss": 2.197,
      "theoretical_loss": 4.966917527996999,
      "tokens_seen": 78446592
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009860782347041123,
      "loss": 2.3427,
      "theoretical_loss": 4.9663261356200366,
      "tokens_seen": 78512128
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009860581745235707,
      "loss": 2.339,
      "theoretical_loss": 4.965735374777409,
      "tokens_seen": 78577664
    },
    {
      "epoch": 0.02,
      "objective/train/advantage_avg": 0.4910324811935425,
      "objective/train/docs_used": 36308,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.3306944370269775,
      "objective/train/original_loss": 2.3306941986083984,
      "objective/train/theoretical_loss": 4.96514524426884,
      "objective/train/tokens_used": 99103200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24158093333244324,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503313541412354,
      "objective/train/weighted_lm_loss": 2.448256015777588,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.96514524426884,
      "tokens_seen": 78643200
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009860381143430292,
      "loss": 2.4381,
      "theoretical_loss": 4.96514524426884,
      "tokens_seen": 78643200
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009860180541624874,
      "loss": 2.333,
      "theoretical_loss": 4.964555742897334,
      "tokens_seen": 78708736
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000985997993981946,
      "loss": 2.45,
      "theoretical_loss": 4.963966869469162,
      "tokens_seen": 78774272
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009859779338014041,
      "loss": 2.2028,
      "theoretical_loss": 4.963378622793849,
      "tokens_seen": 78839808
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009859578736208627,
      "loss": 2.2797,
      "theoretical_loss": 4.962791001684167,
      "tokens_seen": 78905344
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000985937813440321,
      "loss": 2.2991,
      "theoretical_loss": 4.962204004956117,
      "tokens_seen": 78970880
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009859177532597794,
      "loss": 2.232,
      "theoretical_loss": 4.961617631428927,
      "tokens_seen": 79036416
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009858976930792378,
      "loss": 2.2345,
      "theoretical_loss": 4.961031879925029,
      "tokens_seen": 79101952
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009858776328986961,
      "loss": 2.3468,
      "theoretical_loss": 4.960446749270055,
      "tokens_seen": 79167488
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009858575727181545,
      "loss": 2.348,
      "theoretical_loss": 4.959862238292827,
      "tokens_seen": 79233024
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009858375125376129,
      "loss": 2.469,
      "theoretical_loss": 4.959278345825336,
      "tokens_seen": 79298560
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009858174523570712,
      "loss": 2.3083,
      "theoretical_loss": 4.958695070702744,
      "tokens_seen": 79364096
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009857973921765296,
      "loss": 2.3116,
      "theoretical_loss": 4.958112411763365,
      "tokens_seen": 79429632
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000985777331995988,
      "loss": 2.328,
      "theoretical_loss": 4.957530367848651,
      "tokens_seen": 79495168
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009857572718154463,
      "loss": 2.4212,
      "theoretical_loss": 4.95694893780319,
      "tokens_seen": 79560704
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009857372116349047,
      "loss": 2.325,
      "theoretical_loss": 4.956368120474689,
      "tokens_seen": 79626240
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000985717151454363,
      "loss": 2.3468,
      "theoretical_loss": 4.955787914713962,
      "tokens_seen": 79691776
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009856970912738214,
      "loss": 2.3578,
      "theoretical_loss": 4.955208319374924,
      "tokens_seen": 79757312
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00098567703109328,
      "loss": 2.4362,
      "theoretical_loss": 4.954629333314573,
      "tokens_seen": 79822848
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009856569709127381,
      "loss": 2.2681,
      "theoretical_loss": 4.954050955392988,
      "tokens_seen": 79888384
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009856369107321967,
      "loss": 2.2174,
      "theoretical_loss": 4.953473184473312,
      "tokens_seen": 79953920
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009856168505516549,
      "loss": 2.2807,
      "theoretical_loss": 4.952896019421743,
      "tokens_seen": 80019456
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009855967903711134,
      "loss": 2.329,
      "theoretical_loss": 4.952319459107521,
      "tokens_seen": 80084992
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009855767301905718,
      "loss": 2.2275,
      "theoretical_loss": 4.951743502402923,
      "tokens_seen": 80150528
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009855566700100302,
      "loss": 2.2448,
      "theoretical_loss": 4.951168148183246,
      "tokens_seen": 80216064
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009855366098294885,
      "loss": 2.3235,
      "theoretical_loss": 4.9505933953268,
      "tokens_seen": 80281600
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009855165496489469,
      "loss": 2.3358,
      "theoretical_loss": 4.9500192427149,
      "tokens_seen": 80347136
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009854964894684052,
      "loss": 2.2623,
      "theoretical_loss": 4.949445689231847,
      "tokens_seen": 80412672
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009854764292878636,
      "loss": 2.2786,
      "theoretical_loss": 4.948872733764926,
      "tokens_seen": 80478208
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000985456369107322,
      "loss": 2.1879,
      "theoretical_loss": 4.9483003752043935,
      "tokens_seen": 80543744
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009854363089267803,
      "loss": 2.1886,
      "theoretical_loss": 4.9477286124434645,
      "tokens_seen": 80609280
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009854162487462387,
      "loss": 2.3044,
      "theoretical_loss": 4.947157444378303,
      "tokens_seen": 80674816
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000985396188565697,
      "loss": 2.2717,
      "theoretical_loss": 4.946586869908014,
      "tokens_seen": 80740352
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009853761283851554,
      "loss": 2.268,
      "theoretical_loss": 4.946016887934633,
      "tokens_seen": 80805888
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009853560682046138,
      "loss": 2.37,
      "theoretical_loss": 4.945447497363112,
      "tokens_seen": 80871424
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009853360080240721,
      "loss": 2.3511,
      "theoretical_loss": 4.944878697101316,
      "tokens_seen": 80936960
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009853159478435307,
      "loss": 2.3928,
      "theoretical_loss": 4.944310486060004,
      "tokens_seen": 81002496
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009852958876629889,
      "loss": 2.2323,
      "theoretical_loss": 4.943742863152828,
      "tokens_seen": 81068032
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009852758274824474,
      "loss": 2.203,
      "theoretical_loss": 4.943175827296319,
      "tokens_seen": 81133568
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009852557673019056,
      "loss": 2.3613,
      "theoretical_loss": 4.942609377409873,
      "tokens_seen": 81199104
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009852357071213642,
      "loss": 2.3922,
      "theoretical_loss": 4.942043512415751,
      "tokens_seen": 81264640
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009852156469408225,
      "loss": 2.3077,
      "theoretical_loss": 4.94147823123906,
      "tokens_seen": 81330176
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000985195586760281,
      "loss": 2.4319,
      "theoretical_loss": 4.940913532807745,
      "tokens_seen": 81395712
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009851755265797393,
      "loss": 2.3457,
      "theoretical_loss": 4.940349416052586,
      "tokens_seen": 81461248
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009851554663991976,
      "loss": 2.3576,
      "theoretical_loss": 4.939785879907176,
      "tokens_seen": 81526784
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000985135406218656,
      "loss": 2.1788,
      "theoretical_loss": 4.939222923307925,
      "tokens_seen": 81592320
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009851153460381143,
      "loss": 2.3332,
      "theoretical_loss": 4.938660545194038,
      "tokens_seen": 81657856
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009850952858575727,
      "loss": 2.4885,
      "theoretical_loss": 4.938098744507517,
      "tokens_seen": 81723392
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000985075225677031,
      "loss": 2.4229,
      "theoretical_loss": 4.937537520193139,
      "tokens_seen": 81788928
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009850551654964894,
      "loss": 2.2958,
      "theoretical_loss": 4.936976871198457,
      "tokens_seen": 81854464
    },
    {
      "epoch": 0.02,
      "objective/train/advantage_avg": 0.48865529894828796,
      "objective/train/docs_used": 37569,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.141408681869507,
      "objective/train/original_loss": 2.141408920288086,
      "objective/train/theoretical_loss": 4.936416796473786,
      "objective/train/tokens_used": 102380000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24008464813232422,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500857830047607,
      "objective/train/weighted_lm_loss": 2.2487919330596924,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.936416796473786,
      "tokens_seen": 81920000
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009850351053159478,
      "loss": 2.3225,
      "theoretical_loss": 4.936416796473786,
      "tokens_seen": 81920000
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009850150451354062,
      "loss": 2.3456,
      "theoretical_loss": 4.935857294972193,
      "tokens_seen": 81985536
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009849949849548645,
      "loss": 2.1802,
      "theoretical_loss": 4.93529836564949,
      "tokens_seen": 82051072
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000984974924774323,
      "loss": 2.3381,
      "theoretical_loss": 4.934740007464219,
      "tokens_seen": 82116608
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009849548645937815,
      "loss": 2.3167,
      "theoretical_loss": 4.934182219377654,
      "tokens_seen": 82182144
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009849348044132398,
      "loss": 2.2076,
      "theoretical_loss": 4.9336250003537785,
      "tokens_seen": 82247680
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009849147442326982,
      "loss": 2.2008,
      "theoretical_loss": 4.933068349359283,
      "tokens_seen": 82313216
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0009848946840521565,
      "loss": 2.255,
      "theoretical_loss": 4.93251226536356,
      "tokens_seen": 82378752
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000984874623871615,
      "loss": 2.3702,
      "theoretical_loss": 4.931956747338681,
      "tokens_seen": 82444288
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009848545636910733,
      "loss": 2.2657,
      "theoretical_loss": 4.931401794259408,
      "tokens_seen": 82509824
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009848345035105316,
      "loss": 2.3108,
      "theoretical_loss": 4.93084740510316,
      "tokens_seen": 82575360
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00098481444332999,
      "loss": 2.289,
      "theoretical_loss": 4.930293578850026,
      "tokens_seen": 82640896
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009847943831494484,
      "loss": 2.2171,
      "theoretical_loss": 4.9297403144827445,
      "tokens_seen": 82706432
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009847743229689067,
      "loss": 2.2979,
      "theoretical_loss": 4.929187610986693,
      "tokens_seen": 82771968
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000984754262788365,
      "loss": 2.2295,
      "theoretical_loss": 4.928635467349885,
      "tokens_seen": 82837504
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009847342026078234,
      "loss": 2.3672,
      "theoretical_loss": 4.9280838825629605,
      "tokens_seen": 82903040
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009847141424272818,
      "loss": 2.3961,
      "theoretical_loss": 4.927532855619173,
      "tokens_seen": 82968576
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009846940822467404,
      "loss": 2.2037,
      "theoretical_loss": 4.926982385514384,
      "tokens_seen": 83034112
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009846740220661985,
      "loss": 2.3054,
      "theoretical_loss": 4.92643247124705,
      "tokens_seen": 83099648
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009846539618856571,
      "loss": 2.2816,
      "theoretical_loss": 4.925883111818225,
      "tokens_seen": 83165184
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009846339017051153,
      "loss": 2.2191,
      "theoretical_loss": 4.925334306231532,
      "tokens_seen": 83230720
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009846138415245738,
      "loss": 2.36,
      "theoretical_loss": 4.924786053493177,
      "tokens_seen": 83296256
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009845937813440322,
      "loss": 2.3395,
      "theoretical_loss": 4.924238352611924,
      "tokens_seen": 83361792
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009845737211634906,
      "loss": 2.3407,
      "theoretical_loss": 4.923691202599091,
      "tokens_seen": 83427328
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000984553660982949,
      "loss": 2.0856,
      "theoretical_loss": 4.923144602468545,
      "tokens_seen": 83492864
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009845336008024073,
      "loss": 2.2128,
      "theoretical_loss": 4.922598551236689,
      "tokens_seen": 83558400
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009845135406218656,
      "loss": 2.4467,
      "theoretical_loss": 4.922053047922455,
      "tokens_seen": 83623936
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000984493480441324,
      "loss": 2.266,
      "theoretical_loss": 4.921508091547297,
      "tokens_seen": 83689472
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009844734202607824,
      "loss": 2.2651,
      "theoretical_loss": 4.920963681135177,
      "tokens_seen": 83755008
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009844533600802407,
      "loss": 2.2465,
      "theoretical_loss": 4.9204198157125685,
      "tokens_seen": 83820544
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000984433299899699,
      "loss": 2.2151,
      "theoretical_loss": 4.919876494308432,
      "tokens_seen": 83886080
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009844132397191575,
      "loss": 2.3275,
      "theoretical_loss": 4.91933371595422,
      "tokens_seen": 83951616
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009843931795386158,
      "loss": 2.3892,
      "theoretical_loss": 4.918791479683863,
      "tokens_seen": 84017152
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009843731193580742,
      "loss": 2.31,
      "theoretical_loss": 4.91824978453376,
      "tokens_seen": 84082688
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009843530591775325,
      "loss": 2.3452,
      "theoretical_loss": 4.917708629542775,
      "tokens_seen": 84148224
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009843329989969911,
      "loss": 2.2599,
      "theoretical_loss": 4.917168013752225,
      "tokens_seen": 84213760
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009843129388164493,
      "loss": 2.2059,
      "theoretical_loss": 4.916627936205872,
      "tokens_seen": 84279296
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009842928786359078,
      "loss": 2.2837,
      "theoretical_loss": 4.916088395949915,
      "tokens_seen": 84344832
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000984272818455366,
      "loss": 2.4008,
      "theoretical_loss": 4.915549392032985,
      "tokens_seen": 84410368
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009842527582748246,
      "loss": 2.4535,
      "theoretical_loss": 4.915010923506134,
      "tokens_seen": 84475904
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000984232698094283,
      "loss": 2.3609,
      "theoretical_loss": 4.914472989422826,
      "tokens_seen": 84541440
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009842126379137413,
      "loss": 2.1934,
      "theoretical_loss": 4.913935588838932,
      "tokens_seen": 84606976
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009841925777331997,
      "loss": 2.3023,
      "theoretical_loss": 4.913398720812719,
      "tokens_seen": 84672512
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000984172517552658,
      "loss": 2.2574,
      "theoretical_loss": 4.9128623844048445,
      "tokens_seen": 84738048
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009841524573721164,
      "loss": 2.2642,
      "theoretical_loss": 4.91232657867835,
      "tokens_seen": 84803584
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009841323971915747,
      "loss": 2.2327,
      "theoretical_loss": 4.911791302698646,
      "tokens_seen": 84869120
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000984112337011033,
      "loss": 2.3648,
      "theoretical_loss": 4.9112565555335115,
      "tokens_seen": 84934656
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009840922768304915,
      "loss": 2.3775,
      "theoretical_loss": 4.910722336253086,
      "tokens_seen": 85000192
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009840722166499498,
      "loss": 2.2554,
      "theoretical_loss": 4.910188643929853,
      "tokens_seen": 85065728
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009840521564694082,
      "loss": 2.386,
      "theoretical_loss": 4.909655477638648,
      "tokens_seen": 85131264
    },
    {
      "epoch": 0.03,
      "objective/train/advantage_avg": 0.4955958127975464,
      "objective/train/docs_used": 38678,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.1719961166381836,
      "objective/train/original_loss": 2.1719963550567627,
      "objective/train/theoretical_loss": 4.909122836456632,
      "objective/train/tokens_used": 105656800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24580909311771393,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508092641830444,
      "objective/train/weighted_lm_loss": 2.282179594039917,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 4.909122836456632,
      "tokens_seen": 85196800
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009840320962888666,
      "loss": 2.2857,
      "theoretical_loss": 4.909122836456632,
      "tokens_seen": 85196800
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000984012036108325,
      "loss": 2.243,
      "theoretical_loss": 4.908590719463298,
      "tokens_seen": 85262336
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009839919759277833,
      "loss": 2.3113,
      "theoretical_loss": 4.9080591257404595,
      "tokens_seen": 85327872
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009839719157472419,
      "loss": 2.2506,
      "theoretical_loss": 4.907528054372239,
      "tokens_seen": 85393408
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009839518555667,
      "loss": 2.1794,
      "theoretical_loss": 4.906997504445066,
      "tokens_seen": 85458944
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009839317953861586,
      "loss": 2.2323,
      "theoretical_loss": 4.906467475047664,
      "tokens_seen": 85524480
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009839117352056167,
      "loss": 2.4132,
      "theoretical_loss": 4.905937965271047,
      "tokens_seen": 85590016
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009838916750250753,
      "loss": 2.314,
      "theoretical_loss": 4.905408974208514,
      "tokens_seen": 85655552
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009838716148445337,
      "loss": 2.3108,
      "theoretical_loss": 4.904880500955633,
      "tokens_seen": 85721088
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000983851554663992,
      "loss": 2.1642,
      "theoretical_loss": 4.904352544610242,
      "tokens_seen": 85786624
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009838314944834504,
      "loss": 2.354,
      "theoretical_loss": 4.903825104272437,
      "tokens_seen": 85852160
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009838114343029088,
      "loss": 2.2949,
      "theoretical_loss": 4.903298179044569,
      "tokens_seen": 85917696
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009837913741223671,
      "loss": 2.2796,
      "theoretical_loss": 4.90277176803123,
      "tokens_seen": 85983232
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009837713139418255,
      "loss": 2.259,
      "theoretical_loss": 4.902245870339252,
      "tokens_seen": 86048768
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009837512537612838,
      "loss": 2.3348,
      "theoretical_loss": 4.9017204850776945,
      "tokens_seen": 86114304
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009837311935807422,
      "loss": 2.3361,
      "theoretical_loss": 4.901195611357844,
      "tokens_seen": 86179840
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009837111334002006,
      "loss": 2.1493,
      "theoretical_loss": 4.9006712482931984,
      "tokens_seen": 86245376
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000983691073219659,
      "loss": 2.3035,
      "theoretical_loss": 4.9001473949994665,
      "tokens_seen": 86310912
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009836710130391173,
      "loss": 2.1573,
      "theoretical_loss": 4.899624050594557,
      "tokens_seen": 86376448
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009836509528585757,
      "loss": 2.3964,
      "theoretical_loss": 4.899101214198575,
      "tokens_seen": 86441984
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000983630892678034,
      "loss": 2.1748,
      "theoretical_loss": 4.89857888493381,
      "tokens_seen": 86507520
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009836108324974926,
      "loss": 2.2292,
      "theoretical_loss": 4.898057061924732,
      "tokens_seen": 86573056
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009835907723169507,
      "loss": 2.3679,
      "theoretical_loss": 4.897535744297986,
      "tokens_seen": 86638592
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009835707121364093,
      "loss": 2.3989,
      "theoretical_loss": 4.89701493118238,
      "tokens_seen": 86704128
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009835506519558675,
      "loss": 2.1716,
      "theoretical_loss": 4.896494621708882,
      "tokens_seen": 86769664
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000983530591775326,
      "loss": 2.2875,
      "theoretical_loss": 4.895974815010612,
      "tokens_seen": 86835200
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009835105315947844,
      "loss": 2.2411,
      "theoretical_loss": 4.895455510222835,
      "tokens_seen": 86900736
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009834904714142428,
      "loss": 2.2779,
      "theoretical_loss": 4.8949367064829525,
      "tokens_seen": 86966272
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009834704112337011,
      "loss": 2.2055,
      "theoretical_loss": 4.8944184029305,
      "tokens_seen": 87031808
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009834503510531595,
      "loss": 2.3822,
      "theoretical_loss": 4.893900598707134,
      "tokens_seen": 87097344
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009834302908726179,
      "loss": 2.1673,
      "theoretical_loss": 4.893383292956631,
      "tokens_seen": 87162880
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009834102306920762,
      "loss": 2.3037,
      "theoretical_loss": 4.892866484824879,
      "tokens_seen": 87228416
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009833901705115346,
      "loss": 2.1902,
      "theoretical_loss": 4.892350173459863,
      "tokens_seen": 87293952
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000983370110330993,
      "loss": 2.2473,
      "theoretical_loss": 4.891834358011673,
      "tokens_seen": 87359488
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009833500501504515,
      "loss": 2.1875,
      "theoretical_loss": 4.891319037632486,
      "tokens_seen": 87425024
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009833299899699097,
      "loss": 2.2182,
      "theoretical_loss": 4.890804211476562,
      "tokens_seen": 87490560
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009833099297893682,
      "loss": 2.2744,
      "theoretical_loss": 4.890289878700239,
      "tokens_seen": 87556096
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009832898696088264,
      "loss": 2.202,
      "theoretical_loss": 4.889776038461924,
      "tokens_seen": 87621632
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000983269809428285,
      "loss": 2.203,
      "theoretical_loss": 4.88926268992209,
      "tokens_seen": 87687168
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009832497492477433,
      "loss": 2.216,
      "theoretical_loss": 4.888749832243265,
      "tokens_seen": 87752704
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009832296890672017,
      "loss": 2.2196,
      "theoretical_loss": 4.888237464590028,
      "tokens_seen": 87818240
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00098320962888666,
      "loss": 2.3488,
      "theoretical_loss": 4.887725586129003,
      "tokens_seen": 87883776
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009831895687061184,
      "loss": 2.2586,
      "theoretical_loss": 4.887214196028848,
      "tokens_seen": 87949312
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009831695085255768,
      "loss": 2.2706,
      "theoretical_loss": 4.886703293460259,
      "tokens_seen": 88014848
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009831494483450351,
      "loss": 2.2567,
      "theoretical_loss": 4.8861928775959464,
      "tokens_seen": 88080384
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009831293881644935,
      "loss": 2.2866,
      "theoretical_loss": 4.885682947610648,
      "tokens_seen": 88145920
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009831093279839519,
      "loss": 2.2333,
      "theoretical_loss": 4.885173502681107,
      "tokens_seen": 88211456
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009830892678034102,
      "loss": 2.2126,
      "theoretical_loss": 4.884664541986075,
      "tokens_seen": 88276992
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009830692076228686,
      "loss": 2.3505,
      "theoretical_loss": 4.884156064706302,
      "tokens_seen": 88342528
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000983049147442327,
      "loss": 2.1561,
      "theoretical_loss": 4.883648070024529,
      "tokens_seen": 88408064
    },
    {
      "epoch": 0.03,
      "objective/train/advantage_avg": 0.49346664547920227,
      "objective/train/docs_used": 39491,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.0473103523254395,
      "objective/train/original_loss": 2.0473103523254395,
      "objective/train/theoretical_loss": 4.8831405571254844,
      "objective/train/tokens_used": 108933600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24416816234588623,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505878925323486,
      "objective/train/weighted_lm_loss": 2.150813102722168,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.8831405571254844,
      "tokens_seen": 88473600
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009830290872617853,
      "loss": 2.2569,
      "theoretical_loss": 4.8831405571254844,
      "tokens_seen": 88473600
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009830090270812437,
      "loss": 2.2434,
      "theoretical_loss": 4.8826335251958755,
      "tokens_seen": 88539136
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009829889669007023,
      "loss": 2.2955,
      "theoretical_loss": 4.882126973424384,
      "tokens_seen": 88604672
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009829689067201604,
      "loss": 2.3253,
      "theoretical_loss": 4.881620901001661,
      "tokens_seen": 88670208
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000982948846539619,
      "loss": 2.2168,
      "theoretical_loss": 4.881115307120314,
      "tokens_seen": 88735744
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009829287863590771,
      "loss": 2.155,
      "theoretical_loss": 4.88061019097491,
      "tokens_seen": 88801280
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009829087261785357,
      "loss": 2.3044,
      "theoretical_loss": 4.880105551761961,
      "tokens_seen": 88866816
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000982888665997994,
      "loss": 2.2709,
      "theoretical_loss": 4.879601388679923,
      "tokens_seen": 88932352
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009828686058174524,
      "loss": 2.2732,
      "theoretical_loss": 4.879097700929192,
      "tokens_seen": 88997888
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009828485456369108,
      "loss": 2.1625,
      "theoretical_loss": 4.878594487712088,
      "tokens_seen": 89063424
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009828284854563692,
      "loss": 2.4076,
      "theoretical_loss": 4.87809174823286,
      "tokens_seen": 89128960
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009828084252758275,
      "loss": 2.3207,
      "theoretical_loss": 4.877589481697674,
      "tokens_seen": 89194496
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009827883650952859,
      "loss": 2.1592,
      "theoretical_loss": 4.877087687314608,
      "tokens_seen": 89260032
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009827683049147442,
      "loss": 2.1353,
      "theoretical_loss": 4.876586364293646,
      "tokens_seen": 89325568
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009827482447342026,
      "loss": 2.2606,
      "theoretical_loss": 4.876085511846673,
      "tokens_seen": 89391104
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000982728184553661,
      "loss": 2.2478,
      "theoretical_loss": 4.875585129187467,
      "tokens_seen": 89456640
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009827081243731193,
      "loss": 2.279,
      "theoretical_loss": 4.875085215531697,
      "tokens_seen": 89522176
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009826880641925777,
      "loss": 2.2824,
      "theoretical_loss": 4.87458577009691,
      "tokens_seen": 89587712
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000982668004012036,
      "loss": 2.2527,
      "theoretical_loss": 4.874086792102535,
      "tokens_seen": 89653248
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009826479438314944,
      "loss": 2.2688,
      "theoretical_loss": 4.873588280769868,
      "tokens_seen": 89718784
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000982627883650953,
      "loss": 2.0824,
      "theoretical_loss": 4.873090235322069,
      "tokens_seen": 89784320
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009826078234704111,
      "loss": 2.0821,
      "theoretical_loss": 4.87259265498416,
      "tokens_seen": 89849856
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009825877632898697,
      "loss": 2.2779,
      "theoretical_loss": 4.872095538983015,
      "tokens_seen": 89915392
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009825677031093279,
      "loss": 2.2936,
      "theoretical_loss": 4.871598886547356,
      "tokens_seen": 89980928
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009825476429287864,
      "loss": 2.2121,
      "theoretical_loss": 4.871102696907745,
      "tokens_seen": 90046464
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009825275827482448,
      "loss": 2.2357,
      "theoretical_loss": 4.870606969296581,
      "tokens_seen": 90112000
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009825075225677032,
      "loss": 2.5256,
      "theoretical_loss": 4.870111702948094,
      "tokens_seen": 90177536
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009824874623871615,
      "loss": 2.2016,
      "theoretical_loss": 4.869616897098336,
      "tokens_seen": 90243072
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00098246740220662,
      "loss": 2.1991,
      "theoretical_loss": 4.869122550985182,
      "tokens_seen": 90308608
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009824473420260783,
      "loss": 2.3257,
      "theoretical_loss": 4.868628663848316,
      "tokens_seen": 90374144
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009824272818455366,
      "loss": 2.3091,
      "theoretical_loss": 4.868135234929232,
      "tokens_seen": 90439680
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000982407221664995,
      "loss": 2.3082,
      "theoretical_loss": 4.867642263471224,
      "tokens_seen": 90505216
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009823871614844533,
      "loss": 2.1111,
      "theoretical_loss": 4.867149748719386,
      "tokens_seen": 90570752
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009823671013039117,
      "loss": 2.3432,
      "theoretical_loss": 4.866657689920601,
      "tokens_seen": 90636288
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00098234704112337,
      "loss": 2.1497,
      "theoretical_loss": 4.866166086323535,
      "tokens_seen": 90701824
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009823269809428284,
      "loss": 2.1864,
      "theoretical_loss": 4.865674937178635,
      "tokens_seen": 90767360
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009823069207622868,
      "loss": 2.1773,
      "theoretical_loss": 4.865184241738126,
      "tokens_seen": 90832896
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009822868605817452,
      "loss": 2.2805,
      "theoretical_loss": 4.864693999255998,
      "tokens_seen": 90898432
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009822668004012037,
      "loss": 2.3565,
      "theoretical_loss": 4.864204208988003,
      "tokens_seen": 90963968
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009822467402206619,
      "loss": 2.1671,
      "theoretical_loss": 4.8637148701916555,
      "tokens_seen": 91029504
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009822266800401205,
      "loss": 2.2914,
      "theoretical_loss": 4.863225982126219,
      "tokens_seen": 91095040
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009822066198595786,
      "loss": 2.2608,
      "theoretical_loss": 4.862737544052706,
      "tokens_seen": 91160576
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009821865596790372,
      "loss": 2.1416,
      "theoretical_loss": 4.86224955523387,
      "tokens_seen": 91226112
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009821664994984955,
      "loss": 2.3752,
      "theoretical_loss": 4.8617620149342,
      "tokens_seen": 91291648
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000982146439317954,
      "loss": 2.2634,
      "theoretical_loss": 4.861274922419918,
      "tokens_seen": 91357184
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009821263791374123,
      "loss": 2.322,
      "theoretical_loss": 4.86078827695897,
      "tokens_seen": 91422720
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009821063189568706,
      "loss": 2.2562,
      "theoretical_loss": 4.860302077821023,
      "tokens_seen": 91488256
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000982086258776329,
      "loss": 2.1679,
      "theoretical_loss": 4.859816324277461,
      "tokens_seen": 91553792
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009820661985957874,
      "loss": 2.2282,
      "theoretical_loss": 4.859331015601375,
      "tokens_seen": 91619328
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009820461384152457,
      "loss": 2.1868,
      "theoretical_loss": 4.858846151067562,
      "tokens_seen": 91684864
    },
    {
      "epoch": 0.03,
      "objective/train/advantage_avg": 0.49189305305480957,
      "objective/train/docs_used": 40748,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.9773281812667847,
      "objective/train/original_loss": 1.9773280620574951,
      "objective/train/theoretical_loss": 4.858361729952518,
      "objective/train/tokens_used": 112210400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24240271747112274,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504215955734253,
      "objective/train/weighted_lm_loss": 2.0769896507263184,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.858361729952518,
      "tokens_seen": 91750400
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000982026078234704,
      "loss": 2.2005,
      "theoretical_loss": 4.858361729952518,
      "tokens_seen": 91750400
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009820060180541624,
      "loss": 2.199,
      "theoretical_loss": 4.857877751534435,
      "tokens_seen": 91815936
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009819859578736208,
      "loss": 2.2238,
      "theoretical_loss": 4.857394215093191,
      "tokens_seen": 91881472
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009819658976930794,
      "loss": 2.2805,
      "theoretical_loss": 4.856911119910352,
      "tokens_seen": 91947008
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009819458375125375,
      "loss": 2.1968,
      "theoretical_loss": 4.856428465269159,
      "tokens_seen": 92012544
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009819257773319961,
      "loss": 2.1694,
      "theoretical_loss": 4.85594625045453,
      "tokens_seen": 92078080
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009819057171514545,
      "loss": 2.2446,
      "theoretical_loss": 4.855464474753049,
      "tokens_seen": 92143616
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009818856569709128,
      "loss": 2.1264,
      "theoretical_loss": 4.854983137452964,
      "tokens_seen": 92209152
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009818655967903712,
      "loss": 2.2233,
      "theoretical_loss": 4.8545022378441836,
      "tokens_seen": 92274688
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009818455366098296,
      "loss": 2.3523,
      "theoretical_loss": 4.854021775218266,
      "tokens_seen": 92340224
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000981825476429288,
      "loss": 2.2655,
      "theoretical_loss": 4.853541748868423,
      "tokens_seen": 92405760
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009818054162487463,
      "loss": 2.3512,
      "theoretical_loss": 4.853062158089504,
      "tokens_seen": 92471296
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009817853560682046,
      "loss": 2.1191,
      "theoretical_loss": 4.852583002178001,
      "tokens_seen": 92536832
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000981765295887663,
      "loss": 2.1088,
      "theoretical_loss": 4.852104280432039,
      "tokens_seen": 92602368
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009817452357071214,
      "loss": 2.2118,
      "theoretical_loss": 4.85162599215137,
      "tokens_seen": 92667904
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009817251755265797,
      "loss": 2.256,
      "theoretical_loss": 4.85114813663737,
      "tokens_seen": 92733440
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000981705115346038,
      "loss": 2.2957,
      "theoretical_loss": 4.8506707131930344,
      "tokens_seen": 92798976
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009816850551654965,
      "loss": 2.0396,
      "theoretical_loss": 4.850193721122972,
      "tokens_seen": 92864512
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009816649949849548,
      "loss": 2.2755,
      "theoretical_loss": 4.849717159733401,
      "tokens_seen": 92930048
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009816449348044134,
      "loss": 2.2867,
      "theoretical_loss": 4.849241028332142,
      "tokens_seen": 92995584
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009816248746238715,
      "loss": 2.2158,
      "theoretical_loss": 4.848765326228618,
      "tokens_seen": 93061120
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009816048144433301,
      "loss": 2.1635,
      "theoretical_loss": 4.848290052733843,
      "tokens_seen": 93126656
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009815847542627883,
      "loss": 2.186,
      "theoretical_loss": 4.847815207160422,
      "tokens_seen": 93192192
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009815646940822468,
      "loss": 2.2273,
      "theoretical_loss": 4.847340788822546,
      "tokens_seen": 93257728
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009815446339017052,
      "loss": 2.1433,
      "theoretical_loss": 4.846866797035984,
      "tokens_seen": 93323264
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009815245737211636,
      "loss": 2.2482,
      "theoretical_loss": 4.846393231118083,
      "tokens_seen": 93388800
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000981504513540622,
      "loss": 2.2104,
      "theoretical_loss": 4.845920090387757,
      "tokens_seen": 93454336
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009814844533600803,
      "loss": 2.1955,
      "theoretical_loss": 4.84544737416549,
      "tokens_seen": 93519872
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009814643931795387,
      "loss": 2.2442,
      "theoretical_loss": 4.844975081773322,
      "tokens_seen": 93585408
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000981444332998997,
      "loss": 2.1941,
      "theoretical_loss": 4.844503212534856,
      "tokens_seen": 93650944
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009814242728184554,
      "loss": 2.1822,
      "theoretical_loss": 4.844031765775242,
      "tokens_seen": 93716480
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009814042126379137,
      "loss": 2.1861,
      "theoretical_loss": 4.843560740821177,
      "tokens_seen": 93782016
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000981384152457372,
      "loss": 2.3743,
      "theoretical_loss": 4.843090137000904,
      "tokens_seen": 93847552
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009813640922768305,
      "loss": 2.0544,
      "theoretical_loss": 4.842619953644203,
      "tokens_seen": 93913088
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009813440320962888,
      "loss": 2.3207,
      "theoretical_loss": 4.842150190082384,
      "tokens_seen": 93978624
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009813239719157472,
      "loss": 2.0991,
      "theoretical_loss": 4.841680845648289,
      "tokens_seen": 94044160
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009813039117352056,
      "loss": 2.163,
      "theoretical_loss": 4.841211919676287,
      "tokens_seen": 94109696
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009812838515546641,
      "loss": 2.1717,
      "theoretical_loss": 4.84074341150226,
      "tokens_seen": 94175232
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009812637913741223,
      "loss": 2.1603,
      "theoretical_loss": 4.840275320463608,
      "tokens_seen": 94240768
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009812437311935809,
      "loss": 2.2077,
      "theoretical_loss": 4.839807645899245,
      "tokens_seen": 94306304
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000981223671013039,
      "loss": 2.0532,
      "theoretical_loss": 4.839340387149586,
      "tokens_seen": 94371840
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009812036108324976,
      "loss": 2.1507,
      "theoretical_loss": 4.838873543556552,
      "tokens_seen": 94437376
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000981183550651956,
      "loss": 2.1606,
      "theoretical_loss": 4.838407114463558,
      "tokens_seen": 94502912
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009811634904714143,
      "loss": 2.2777,
      "theoretical_loss": 4.837941099215515,
      "tokens_seen": 94568448
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009811434302908727,
      "loss": 2.1049,
      "theoretical_loss": 4.837475497158817,
      "tokens_seen": 94633984
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000981123370110331,
      "loss": 2.2131,
      "theoretical_loss": 4.83701030764135,
      "tokens_seen": 94699520
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009811033099297894,
      "loss": 2.0608,
      "theoretical_loss": 4.836545530012472,
      "tokens_seen": 94765056
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009810832497492478,
      "loss": 2.1397,
      "theoretical_loss": 4.83608116362302,
      "tokens_seen": 94830592
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009810631895687061,
      "loss": 2.1879,
      "theoretical_loss": 4.835617207825303,
      "tokens_seen": 94896128
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009810431293881645,
      "loss": 2.2057,
      "theoretical_loss": 4.835153661973092,
      "tokens_seen": 94961664
    },
    {
      "epoch": 0.03,
      "objective/train/advantage_avg": 0.48443371057510376,
      "objective/train/docs_used": 42039,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8953099250793457,
      "objective/train/original_loss": 1.8953099250793457,
      "objective/train/theoretical_loss": 4.8346905254216255,
      "objective/train/tokens_used": 115487200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23641405999660492,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0496450662612915,
      "objective/train/weighted_lm_loss": 1.9899815320968628,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.8346905254216255,
      "tokens_seen": 95027200
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009810230692076228,
      "loss": 2.0976,
      "theoretical_loss": 4.8346905254216255,
      "tokens_seen": 95027200
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009810030090270812,
      "loss": 2.1386,
      "theoretical_loss": 4.834227797527595,
      "tokens_seen": 95092736
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009809829488465396,
      "loss": 2.1371,
      "theoretical_loss": 4.83376547764915,
      "tokens_seen": 95158272
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000980962888665998,
      "loss": 2.1342,
      "theoretical_loss": 4.833303565145885,
      "tokens_seen": 95223808
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009809428284854563,
      "loss": 2.2161,
      "theoretical_loss": 4.832842059378843,
      "tokens_seen": 95289344
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009809227683049149,
      "loss": 2.1805,
      "theoretical_loss": 4.832380959710505,
      "tokens_seen": 95354880
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000980902708124373,
      "loss": 2.1964,
      "theoretical_loss": 4.831920265504792,
      "tokens_seen": 95420416
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009808826479438316,
      "loss": 2.1742,
      "theoretical_loss": 4.831459976127054,
      "tokens_seen": 95485952
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009808625877632897,
      "loss": 2.2361,
      "theoretical_loss": 4.831000090944068,
      "tokens_seen": 95551488
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009808425275827483,
      "loss": 2.2147,
      "theoretical_loss": 4.830540609324041,
      "tokens_seen": 95617024
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009808224674022067,
      "loss": 2.2732,
      "theoretical_loss": 4.830081530636594,
      "tokens_seen": 95682560
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000980802407221665,
      "loss": 2.1651,
      "theoretical_loss": 4.829622854252765,
      "tokens_seen": 95748096
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009807823470411234,
      "loss": 2.2456,
      "theoretical_loss": 4.829164579545006,
      "tokens_seen": 95813632
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009807622868605818,
      "loss": 2.1658,
      "theoretical_loss": 4.828706705887171,
      "tokens_seen": 95879168
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009807422266800401,
      "loss": 2.2252,
      "theoretical_loss": 4.8282492326545245,
      "tokens_seen": 95944704
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009807221664994985,
      "loss": 2.2021,
      "theoretical_loss": 4.827792159223722,
      "tokens_seen": 96010240
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009807021063189569,
      "loss": 2.2582,
      "theoretical_loss": 4.82733548497282,
      "tokens_seen": 96075776
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009806820461384152,
      "loss": 2.0183,
      "theoretical_loss": 4.8268792092812625,
      "tokens_seen": 96141312
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009806619859578736,
      "loss": 2.0873,
      "theoretical_loss": 4.826423331529884,
      "tokens_seen": 96206848
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000980641925777332,
      "loss": 2.1451,
      "theoretical_loss": 4.825967851100897,
      "tokens_seen": 96272384
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009806218655967903,
      "loss": 2.3053,
      "theoretical_loss": 4.825512767377897,
      "tokens_seen": 96337920
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009806018054162487,
      "loss": 2.149,
      "theoretical_loss": 4.825058079745853,
      "tokens_seen": 96403456
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000980581745235707,
      "loss": 2.1526,
      "theoretical_loss": 4.824603787591102,
      "tokens_seen": 96468992
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009805616850551656,
      "loss": 2.1577,
      "theoretical_loss": 4.824149890301353,
      "tokens_seen": 96534528
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009805416248746238,
      "loss": 2.1158,
      "theoretical_loss": 4.823696387265674,
      "tokens_seen": 96600064
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009805215646940823,
      "loss": 2.2835,
      "theoretical_loss": 4.823243277874491,
      "tokens_seen": 96665600
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009805015045135407,
      "loss": 2.2474,
      "theoretical_loss": 4.822790561519591,
      "tokens_seen": 96731136
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000980481444332999,
      "loss": 2.1669,
      "theoretical_loss": 4.822338237594105,
      "tokens_seen": 96796672
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009804613841524574,
      "loss": 2.1423,
      "theoretical_loss": 4.821886305492513,
      "tokens_seen": 96862208
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009804413239719158,
      "loss": 2.2884,
      "theoretical_loss": 4.8214347646106415,
      "tokens_seen": 96927744
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009804212637913741,
      "loss": 2.162,
      "theoretical_loss": 4.8209836143456535,
      "tokens_seen": 96993280
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009804012036108325,
      "loss": 2.0766,
      "theoretical_loss": 4.820532854096047,
      "tokens_seen": 97058816
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009803811434302909,
      "loss": 2.2347,
      "theoretical_loss": 4.820082483261652,
      "tokens_seen": 97124352
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009803610832497492,
      "loss": 2.1397,
      "theoretical_loss": 4.81963250124363,
      "tokens_seen": 97189888
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009803410230692076,
      "loss": 2.0618,
      "theoretical_loss": 4.81918290744446,
      "tokens_seen": 97255424
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000980320962888666,
      "loss": 2.0536,
      "theoretical_loss": 4.818733701267945,
      "tokens_seen": 97320960
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009803009027081245,
      "loss": 2.2106,
      "theoretical_loss": 4.818284882119203,
      "tokens_seen": 97386496
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009802808425275827,
      "loss": 2.3802,
      "theoretical_loss": 4.817836449404666,
      "tokens_seen": 97452032
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009802607823470413,
      "loss": 2.2977,
      "theoretical_loss": 4.817388402532074,
      "tokens_seen": 97517568
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009802407221664994,
      "loss": 2.316,
      "theoretical_loss": 4.81694074091047,
      "tokens_seen": 97583104
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000980220661985958,
      "loss": 2.4769,
      "theoretical_loss": 4.816493463950202,
      "tokens_seen": 97648640
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009802006018054163,
      "loss": 2.2114,
      "theoretical_loss": 4.816046571062912,
      "tokens_seen": 97714176
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009801805416248747,
      "loss": 2.2655,
      "theoretical_loss": 4.815600061661536,
      "tokens_seen": 97779712
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000980160481444333,
      "loss": 2.0975,
      "theoretical_loss": 4.815153935160305,
      "tokens_seen": 97845248
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009801404212637914,
      "loss": 2.1909,
      "theoretical_loss": 4.814708190974729,
      "tokens_seen": 97910784
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009801203610832498,
      "loss": 2.2142,
      "theoretical_loss": 4.8142628285216045,
      "tokens_seen": 97976320
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009801003009027082,
      "loss": 2.1922,
      "theoretical_loss": 4.813817847219008,
      "tokens_seen": 98041856
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009800802407221665,
      "loss": 2.1104,
      "theoretical_loss": 4.81337324648629,
      "tokens_seen": 98107392
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009800601805416249,
      "loss": 2.0705,
      "theoretical_loss": 4.812929025744072,
      "tokens_seen": 98172928
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009800401203610832,
      "loss": 2.0202,
      "theoretical_loss": 4.812485184414244,
      "tokens_seen": 98238464
    },
    {
      "epoch": 0.03,
      "objective/train/advantage_avg": 0.4935273230075836,
      "objective/train/docs_used": 43181,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.0869107246398926,
      "objective/train/original_loss": 2.0869107246398926,
      "objective/train/theoretical_loss": 4.812041721919962,
      "objective/train/tokens_used": 118764000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24405071139335632,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050593376159668,
      "objective/train/weighted_lm_loss": 2.192361831665039,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.812041721919962,
      "tokens_seen": 98304000
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009800200601805416,
      "loss": 2.1919,
      "theoretical_loss": 4.812041721919962,
      "tokens_seen": 98304000
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00098,
      "loss": 2.1269,
      "theoretical_loss": 4.8115986376856394,
      "tokens_seen": 98369536
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009799799398194583,
      "loss": 2.1634,
      "theoretical_loss": 4.81115593113695,
      "tokens_seen": 98435072
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009799598796389167,
      "loss": 2.0142,
      "theoretical_loss": 4.8107136017008205,
      "tokens_seen": 98500608
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009799398194583753,
      "loss": 2.0447,
      "theoretical_loss": 4.810271648805427,
      "tokens_seen": 98566144
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009799197592778334,
      "loss": 2.1669,
      "theoretical_loss": 4.809830071880194,
      "tokens_seen": 98631680
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979899699097292,
      "loss": 2.2777,
      "theoretical_loss": 4.809388870355785,
      "tokens_seen": 98697216
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009798796389167501,
      "loss": 2.1788,
      "theoretical_loss": 4.808948043664107,
      "tokens_seen": 98762752
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009798595787362087,
      "loss": 2.1686,
      "theoretical_loss": 4.8085075912383015,
      "tokens_seen": 98828288
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979839518555667,
      "loss": 2.0359,
      "theoretical_loss": 4.808067512512741,
      "tokens_seen": 98893824
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009798194583751254,
      "loss": 2.0967,
      "theoretical_loss": 4.807627806923029,
      "tokens_seen": 98959360
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009797993981945838,
      "loss": 2.2084,
      "theoretical_loss": 4.807188473905994,
      "tokens_seen": 99024896
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009797793380140422,
      "loss": 2.267,
      "theoretical_loss": 4.806749512899687,
      "tokens_seen": 99090432
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009797592778335005,
      "loss": 2.1556,
      "theoretical_loss": 4.806310923343375,
      "tokens_seen": 99155968
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979739217652959,
      "loss": 2.1009,
      "theoretical_loss": 4.805872704677542,
      "tokens_seen": 99221504
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009797191574724173,
      "loss": 2.1948,
      "theoretical_loss": 4.805434856343886,
      "tokens_seen": 99287040
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009796990972918756,
      "loss": 2.2327,
      "theoretical_loss": 4.804997377785307,
      "tokens_seen": 99352576
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979679037111334,
      "loss": 2.2464,
      "theoretical_loss": 4.8045602684459165,
      "tokens_seen": 99418112
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009796589769307923,
      "loss": 2.168,
      "theoretical_loss": 4.804123527771022,
      "tokens_seen": 99483648
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009796389167502507,
      "loss": 2.1951,
      "theoretical_loss": 4.803687155207134,
      "tokens_seen": 99549184
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979618856569709,
      "loss": 2.1594,
      "theoretical_loss": 4.8032511502019535,
      "tokens_seen": 99614720
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009795987963891674,
      "loss": 2.1376,
      "theoretical_loss": 4.802815512204375,
      "tokens_seen": 99680256
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979578736208626,
      "loss": 2.1674,
      "theoretical_loss": 4.80238024066448,
      "tokens_seen": 99745792
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009795586760280842,
      "loss": 2.0653,
      "theoretical_loss": 4.801945335033535,
      "tokens_seen": 99811328
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009795386158475427,
      "loss": 2.3154,
      "theoretical_loss": 4.801510794763988,
      "tokens_seen": 99876864
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979518555667001,
      "loss": 2.0517,
      "theoretical_loss": 4.801076619309466,
      "tokens_seen": 99942400
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009794984954864595,
      "loss": 2.0963,
      "theoretical_loss": 4.800642808124767,
      "tokens_seen": 100007936
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009794784353059178,
      "loss": 2.3449,
      "theoretical_loss": 4.800209360665864,
      "tokens_seen": 100073472
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009794583751253762,
      "loss": 2.1891,
      "theoretical_loss": 4.799776276389897,
      "tokens_seen": 100139008
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009794383149448345,
      "loss": 2.183,
      "theoretical_loss": 4.79934355475517,
      "tokens_seen": 100204544
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979418254764293,
      "loss": 2.1676,
      "theoretical_loss": 4.79891119522115,
      "tokens_seen": 100270080
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009793981945837513,
      "loss": 2.1522,
      "theoretical_loss": 4.798479197248461,
      "tokens_seen": 100335616
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009793781344032096,
      "loss": 2.364,
      "theoretical_loss": 4.798047560298882,
      "tokens_seen": 100401152
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979358074222668,
      "loss": 2.2696,
      "theoretical_loss": 4.797616283835346,
      "tokens_seen": 100466688
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009793380140421264,
      "loss": 2.0181,
      "theoretical_loss": 4.797185367321932,
      "tokens_seen": 100532224
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009793179538615847,
      "loss": 2.1756,
      "theoretical_loss": 4.796754810223866,
      "tokens_seen": 100597760
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979297893681043,
      "loss": 2.0873,
      "theoretical_loss": 4.796324612007515,
      "tokens_seen": 100663296
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009792778335005014,
      "loss": 2.1305,
      "theoretical_loss": 4.795894772140386,
      "tokens_seen": 100728832
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009792577733199598,
      "loss": 2.2556,
      "theoretical_loss": 4.795465290091123,
      "tokens_seen": 100794368
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009792377131394182,
      "loss": 2.0559,
      "theoretical_loss": 4.795036165329499,
      "tokens_seen": 100859904
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009792176529588767,
      "loss": 2.1793,
      "theoretical_loss": 4.794607397326421,
      "tokens_seen": 100925440
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979197592778335,
      "loss": 2.1214,
      "theoretical_loss": 4.794178985553919,
      "tokens_seen": 100990976
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009791775325977935,
      "loss": 2.2544,
      "theoretical_loss": 4.793750929485149,
      "tokens_seen": 101056512
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009791574724172518,
      "loss": 2.3256,
      "theoretical_loss": 4.793323228594385,
      "tokens_seen": 101122048
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009791374122367102,
      "loss": 2.087,
      "theoretical_loss": 4.792895882357019,
      "tokens_seen": 101187584
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009791173520561686,
      "loss": 2.0553,
      "theoretical_loss": 4.792468890249557,
      "tokens_seen": 101253120
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979097291875627,
      "loss": 2.1112,
      "theoretical_loss": 4.792042251749617,
      "tokens_seen": 101318656
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009790772316950853,
      "loss": 2.1636,
      "theoretical_loss": 4.791615966335922,
      "tokens_seen": 101384192
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009790571715145436,
      "loss": 2.1967,
      "theoretical_loss": 4.791190033488302,
      "tokens_seen": 101449728
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000979037111334002,
      "loss": 2.1536,
      "theoretical_loss": 4.790764452687689,
      "tokens_seen": 101515264
    },
    {
      "epoch": 0.03,
      "objective/train/advantage_avg": 0.49196192622184753,
      "objective/train/docs_used": 44310,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.0242908000946045,
      "objective/train/original_loss": 2.0242908000946045,
      "objective/train/theoretical_loss": 4.790339223416113,
      "objective/train/tokens_used": 122040800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2426905632019043,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504299402236938,
      "objective/train/weighted_lm_loss": 2.1261332035064697,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.790339223416113,
      "tokens_seen": 101580800
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009790170511534604,
      "loss": 2.0924,
      "theoretical_loss": 4.790339223416113,
      "tokens_seen": 101580800
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009789969909729187,
      "loss": 2.1643,
      "theoretical_loss": 4.7899143451566975,
      "tokens_seen": 101646336
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000978976930792377,
      "loss": 2.1168,
      "theoretical_loss": 4.7894898173936635,
      "tokens_seen": 101711872
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009789568706118357,
      "loss": 2.0482,
      "theoretical_loss": 4.789065639612316,
      "tokens_seen": 101777408
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009789368104312938,
      "loss": 2.17,
      "theoretical_loss": 4.78864181129905,
      "tokens_seen": 101842944
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009789167502507524,
      "loss": 2.0711,
      "theoretical_loss": 4.788218331941344,
      "tokens_seen": 101908480
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009788966900702105,
      "loss": 2.0941,
      "theoretical_loss": 4.787795201027757,
      "tokens_seen": 101974016
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009788766298896691,
      "loss": 2.149,
      "theoretical_loss": 4.787372418047926,
      "tokens_seen": 102039552
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009788565697091275,
      "loss": 2.2792,
      "theoretical_loss": 4.786949982492559,
      "tokens_seen": 102105088
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009788365095285858,
      "loss": 2.1955,
      "theoretical_loss": 4.786527893853441,
      "tokens_seen": 102170624
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009788164493480442,
      "loss": 2.046,
      "theoretical_loss": 4.786106151623423,
      "tokens_seen": 102236160
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009787963891675026,
      "loss": 2.0414,
      "theoretical_loss": 4.785684755296422,
      "tokens_seen": 102301696
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000978776328986961,
      "loss": 2.0363,
      "theoretical_loss": 4.785263704367418,
      "tokens_seen": 102367232
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009787562688064193,
      "loss": 2.0321,
      "theoretical_loss": 4.7848429983324525,
      "tokens_seen": 102432768
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009787362086258777,
      "loss": 2.1357,
      "theoretical_loss": 4.784422636688622,
      "tokens_seen": 102498304
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000978716148445336,
      "loss": 2.168,
      "theoretical_loss": 4.784002618934078,
      "tokens_seen": 102563840
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009786960882647944,
      "loss": 2.0982,
      "theoretical_loss": 4.783582944568026,
      "tokens_seen": 102629376
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009786760280842527,
      "loss": 2.145,
      "theoretical_loss": 4.783163613090714,
      "tokens_seen": 102694912
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000978655967903711,
      "loss": 2.0991,
      "theoretical_loss": 4.782744624003442,
      "tokens_seen": 102760448
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009786359077231695,
      "loss": 2.2595,
      "theoretical_loss": 4.782325976808549,
      "tokens_seen": 102825984
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009786158475426278,
      "loss": 2.2138,
      "theoretical_loss": 4.7819076710094155,
      "tokens_seen": 102891520
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009785957873620864,
      "loss": 2.1458,
      "theoretical_loss": 4.781489706110458,
      "tokens_seen": 102957056
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009785757271815446,
      "loss": 2.0066,
      "theoretical_loss": 4.781072081617127,
      "tokens_seen": 103022592
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009785556670010031,
      "loss": 2.1166,
      "theoretical_loss": 4.780654797035907,
      "tokens_seen": 103088128
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009785356068204615,
      "loss": 2.1703,
      "theoretical_loss": 4.780237851874308,
      "tokens_seen": 103153664
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009785155466399199,
      "loss": 2.1396,
      "theoretical_loss": 4.779821245640869,
      "tokens_seen": 103219200
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009784954864593782,
      "loss": 2.1718,
      "theoretical_loss": 4.779404977845148,
      "tokens_seen": 103284736
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009784754262788366,
      "loss": 2.1923,
      "theoretical_loss": 4.778989047997728,
      "tokens_seen": 103350272
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000978455366098295,
      "loss": 2.0623,
      "theoretical_loss": 4.778573455610207,
      "tokens_seen": 103415808
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009784353059177533,
      "loss": 2.1395,
      "theoretical_loss": 4.778158200195196,
      "tokens_seen": 103481344
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009784152457372117,
      "loss": 2.007,
      "theoretical_loss": 4.777743281266321,
      "tokens_seen": 103546880
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00097839518555667,
      "loss": 2.0959,
      "theoretical_loss": 4.777328698338218,
      "tokens_seen": 103612416
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009783751253761284,
      "loss": 2.146,
      "theoretical_loss": 4.776914450926526,
      "tokens_seen": 103677952
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009783550651955868,
      "loss": 2.1683,
      "theoretical_loss": 4.776500538547891,
      "tokens_seen": 103743488
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009783350050150451,
      "loss": 2.1065,
      "theoretical_loss": 4.776086960719956,
      "tokens_seen": 103809024
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009783149448345035,
      "loss": 2.0547,
      "theoretical_loss": 4.775673716961369,
      "tokens_seen": 103874560
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009782948846539618,
      "loss": 2.2282,
      "theoretical_loss": 4.775260806791767,
      "tokens_seen": 103940096
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009782748244734202,
      "loss": 2.2749,
      "theoretical_loss": 4.774848229731784,
      "tokens_seen": 104005632
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009782547642928786,
      "loss": 2.2745,
      "theoretical_loss": 4.774435985303043,
      "tokens_seen": 104071168
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009782347041123371,
      "loss": 2.1922,
      "theoretical_loss": 4.774024073028156,
      "tokens_seen": 104136704
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009782146439317953,
      "loss": 1.9764,
      "theoretical_loss": 4.7736124924307175,
      "tokens_seen": 104202240
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009781945837512539,
      "loss": 1.9248,
      "theoretical_loss": 4.773201243035307,
      "tokens_seen": 104267776
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009781745235707122,
      "loss": 2.1946,
      "theoretical_loss": 4.772790324367482,
      "tokens_seen": 104333312
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009781544633901706,
      "loss": 2.1539,
      "theoretical_loss": 4.772379735953779,
      "tokens_seen": 104398848
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000978134403209629,
      "loss": 2.1859,
      "theoretical_loss": 4.771969477321704,
      "tokens_seen": 104464384
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009781143430290873,
      "loss": 2.0899,
      "theoretical_loss": 4.771559547999743,
      "tokens_seen": 104529920
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009780942828485457,
      "loss": 2.1757,
      "theoretical_loss": 4.771149947517346,
      "tokens_seen": 104595456
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000978074222668004,
      "loss": 2.159,
      "theoretical_loss": 4.770740675404928,
      "tokens_seen": 104660992
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009780541624874624,
      "loss": 2.044,
      "theoretical_loss": 4.770331731193872,
      "tokens_seen": 104726528
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009780341023069208,
      "loss": 2.0043,
      "theoretical_loss": 4.769923114416522,
      "tokens_seen": 104792064
    },
    {
      "epoch": 0.03,
      "objective/train/advantage_avg": 0.49358728528022766,
      "objective/train/docs_used": 45444,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.1862094402313232,
      "objective/train/original_loss": 2.1862096786499023,
      "objective/train/theoretical_loss": 4.769514824606178,
      "objective/train/tokens_used": 125317600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2443910837173462,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506013631820679,
      "objective/train/weighted_lm_loss": 2.2968454360961914,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.769514824606178,
      "tokens_seen": 104857600
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009780140421263791,
      "loss": 2.1955,
      "theoretical_loss": 4.769514824606178,
      "tokens_seen": 104857600
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009779939819458375,
      "loss": 2.0942,
      "theoretical_loss": 4.7691068612971,
      "tokens_seen": 104923136
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009779739217652959,
      "loss": 1.9648,
      "theoretical_loss": 4.768699224024498,
      "tokens_seen": 104988672
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009779538615847542,
      "loss": 2.12,
      "theoretical_loss": 4.768291912324538,
      "tokens_seen": 105054208
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009779338014042126,
      "loss": 2.025,
      "theoretical_loss": 4.7678849257343305,
      "tokens_seen": 105119744
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000977913741223671,
      "loss": 2.1114,
      "theoretical_loss": 4.767478263791933,
      "tokens_seen": 105185280
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009778936810431293,
      "loss": 2.0773,
      "theoretical_loss": 4.7670719260363486,
      "tokens_seen": 105250816
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009778736208625879,
      "loss": 2.0793,
      "theoretical_loss": 4.76666591200752,
      "tokens_seen": 105316352
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000977853560682046,
      "loss": 2.0647,
      "theoretical_loss": 4.766260221246329,
      "tokens_seen": 105381888
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009778335005015046,
      "loss": 2.0955,
      "theoretical_loss": 4.765854853294592,
      "tokens_seen": 105447424
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000977813440320963,
      "loss": 2.0974,
      "theoretical_loss": 4.765449807695062,
      "tokens_seen": 105512960
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009777933801404213,
      "loss": 2.1978,
      "theoretical_loss": 4.765045083991422,
      "tokens_seen": 105578496
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009777733199598797,
      "loss": 2.0725,
      "theoretical_loss": 4.764640681728281,
      "tokens_seen": 105644032
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000977753259779338,
      "loss": 2.3367,
      "theoretical_loss": 4.764236600451178,
      "tokens_seen": 105709568
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009777331995987964,
      "loss": 2.19,
      "theoretical_loss": 4.763832839706574,
      "tokens_seen": 105775104
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009777131394182548,
      "loss": 2.0348,
      "theoretical_loss": 4.763429399041852,
      "tokens_seen": 105840640
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009776930792377131,
      "loss": 2.025,
      "theoretical_loss": 4.763026278005314,
      "tokens_seen": 105906176
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009776730190571715,
      "loss": 2.2138,
      "theoretical_loss": 4.762623476146176,
      "tokens_seen": 105971712
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009776529588766299,
      "loss": 2.0778,
      "theoretical_loss": 4.762220993014572,
      "tokens_seen": 106037248
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009776328986960882,
      "loss": 2.0225,
      "theoretical_loss": 4.761818828161545,
      "tokens_seen": 106102784
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009776128385155466,
      "loss": 2.1458,
      "theoretical_loss": 4.761416981139046,
      "tokens_seen": 106168320
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000977592778335005,
      "loss": 2.018,
      "theoretical_loss": 4.7610154514999365,
      "tokens_seen": 106233856
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009775727181544633,
      "loss": 2.1109,
      "theoretical_loss": 4.760614238797981,
      "tokens_seen": 106299392
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000977552657973922,
      "loss": 2.1624,
      "theoretical_loss": 4.760213342587843,
      "tokens_seen": 106364928
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00097753259779338,
      "loss": 2.1445,
      "theoretical_loss": 4.75981276242509,
      "tokens_seen": 106430464
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009775125376128386,
      "loss": 2.0961,
      "theoretical_loss": 4.759412497866186,
      "tokens_seen": 106496000
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009774924774322968,
      "loss": 2.1643,
      "theoretical_loss": 4.759012548468487,
      "tokens_seen": 106561536
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009774724172517553,
      "loss": 2.0578,
      "theoretical_loss": 4.758612913790243,
      "tokens_seen": 106627072
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009774523570712137,
      "loss": 2.0945,
      "theoretical_loss": 4.758213593390595,
      "tokens_seen": 106692608
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000977432296890672,
      "loss": 2.0254,
      "theoretical_loss": 4.757814586829572,
      "tokens_seen": 106758144
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009774122367101304,
      "loss": 2.1265,
      "theoretical_loss": 4.757415893668089,
      "tokens_seen": 106823680
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009773921765295888,
      "loss": 2.1485,
      "theoretical_loss": 4.75701751346794,
      "tokens_seen": 106889216
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009773721163490472,
      "loss": 2.0516,
      "theoretical_loss": 4.756619445791808,
      "tokens_seen": 106954752
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009773520561685055,
      "loss": 2.1517,
      "theoretical_loss": 4.756221690203244,
      "tokens_seen": 107020288
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009773319959879639,
      "loss": 2.2329,
      "theoretical_loss": 4.755824246266685,
      "tokens_seen": 107085824
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009773119358074222,
      "loss": 2.1316,
      "theoretical_loss": 4.755427113547437,
      "tokens_seen": 107151360
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009772918756268806,
      "loss": 2.0784,
      "theoretical_loss": 4.755030291611678,
      "tokens_seen": 107216896
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000977271815446339,
      "loss": 2.1254,
      "theoretical_loss": 4.754633780026456,
      "tokens_seen": 107282432
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009772517552657975,
      "loss": 2.122,
      "theoretical_loss": 4.754237578359687,
      "tokens_seen": 107347968
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009772316950852557,
      "loss": 2.093,
      "theoretical_loss": 4.753841686180152,
      "tokens_seen": 107413504
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009772116349047143,
      "loss": 2.1048,
      "theoretical_loss": 4.753446103057492,
      "tokens_seen": 107479040
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009771915747241726,
      "loss": 2.1523,
      "theoretical_loss": 4.753050828562211,
      "tokens_seen": 107544576
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000977171514543631,
      "loss": 2.1161,
      "theoretical_loss": 4.752655862265669,
      "tokens_seen": 107610112
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009771514543630894,
      "loss": 2.0962,
      "theoretical_loss": 4.752261203740085,
      "tokens_seen": 107675648
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009771313941825477,
      "loss": 2.1709,
      "theoretical_loss": 4.751866852558529,
      "tokens_seen": 107741184
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000977111334002006,
      "loss": 2.0602,
      "theoretical_loss": 4.751472808294924,
      "tokens_seen": 107806720
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009770912738214644,
      "loss": 2.2369,
      "theoretical_loss": 4.751079070524041,
      "tokens_seen": 107872256
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009770712136409228,
      "loss": 2.185,
      "theoretical_loss": 4.750685638821498,
      "tokens_seen": 107937792
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009770511534603812,
      "loss": 2.0072,
      "theoretical_loss": 4.7502925127637585,
      "tokens_seen": 108003328
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009770310932798395,
      "loss": 2.132,
      "theoretical_loss": 4.749899691928128,
      "tokens_seen": 108068864
    },
    {
      "epoch": 0.03,
      "objective/train/advantage_avg": 0.49381500482559204,
      "objective/train/docs_used": 46682,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.180783271789551,
      "objective/train/original_loss": 2.180783748626709,
      "objective/train/theoretical_loss": 4.749507175892756,
      "objective/train/tokens_used": 128594400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24419686198234558,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506229400634766,
      "objective/train/weighted_lm_loss": 2.291189432144165,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.749507175892756,
      "tokens_seen": 108134400
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000977011033099298,
      "loss": 2.1949,
      "theoretical_loss": 4.749507175892756,
      "tokens_seen": 108134400
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009769909729187563,
      "loss": 2.1011,
      "theoretical_loss": 4.7491149642366235,
      "tokens_seen": 108199936
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009769709127382146,
      "loss": 2.0941,
      "theoretical_loss": 4.7487230565395535,
      "tokens_seen": 108265472
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000976950852557673,
      "loss": 2.1314,
      "theoretical_loss": 4.7483314523822004,
      "tokens_seen": 108331008
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009769307923771313,
      "loss": 2.2006,
      "theoretical_loss": 4.74794015134605,
      "tokens_seen": 108396544
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009769107321965897,
      "loss": 2.0744,
      "theoretical_loss": 4.74754915301342,
      "tokens_seen": 108462080
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009768906720160483,
      "loss": 2.047,
      "theoretical_loss": 4.747158456967452,
      "tokens_seen": 108527616
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009768706118355064,
      "loss": 2.1796,
      "theoretical_loss": 4.746768062792116,
      "tokens_seen": 108593152
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000976850551654965,
      "loss": 2.0949,
      "theoretical_loss": 4.746377970072205,
      "tokens_seen": 108658688
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009768304914744234,
      "loss": 2.1516,
      "theoretical_loss": 4.745988178393331,
      "tokens_seen": 108724224
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009768104312938817,
      "loss": 2.1688,
      "theoretical_loss": 4.745598687341927,
      "tokens_seen": 108789760
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00097679037111334,
      "loss": 2.1327,
      "theoretical_loss": 4.745209496505241,
      "tokens_seen": 108855296
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009767703109327985,
      "loss": 2.1034,
      "theoretical_loss": 4.744820605471338,
      "tokens_seen": 108920832
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009767502507522568,
      "loss": 2.1114,
      "theoretical_loss": 4.744432013829094,
      "tokens_seen": 108986368
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009767301905717152,
      "loss": 2.1984,
      "theoretical_loss": 4.744043721168196,
      "tokens_seen": 109051904
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009767101303911735,
      "loss": 2.0895,
      "theoretical_loss": 4.7436557270791395,
      "tokens_seen": 109117440
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000976690070210632,
      "loss": 2.0292,
      "theoretical_loss": 4.743268031153226,
      "tokens_seen": 109182976
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009766700100300903,
      "loss": 2.0267,
      "theoretical_loss": 4.742880632982564,
      "tokens_seen": 109248512
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009766499498495486,
      "loss": 2.1127,
      "theoretical_loss": 4.74249353216006,
      "tokens_seen": 109314048
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000976629889669007,
      "loss": 1.9842,
      "theoretical_loss": 4.742106728279422,
      "tokens_seen": 109379584
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009766098294884654,
      "loss": 2.1229,
      "theoretical_loss": 4.741720220935159,
      "tokens_seen": 109445120
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009765897693079237,
      "loss": 2.0699,
      "theoretical_loss": 4.741334009722573,
      "tokens_seen": 109510656
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009765697091273823,
      "loss": 2.1166,
      "theoretical_loss": 4.740948094237761,
      "tokens_seen": 109576192
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009765496489468406,
      "loss": 2.1423,
      "theoretical_loss": 4.7405624740776116,
      "tokens_seen": 109641728
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009765295887662989,
      "loss": 2.1827,
      "theoretical_loss": 4.7401771488398055,
      "tokens_seen": 109707264
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009765095285857573,
      "loss": 2.1094,
      "theoretical_loss": 4.739792118122809,
      "tokens_seen": 109772800
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009764894684052156,
      "loss": 2.0291,
      "theoretical_loss": 4.739407381525874,
      "tokens_seen": 109838336
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000976469408224674,
      "loss": 2.1474,
      "theoretical_loss": 4.73902293864904,
      "tokens_seen": 109903872
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009764493480441325,
      "loss": 2.1552,
      "theoretical_loss": 4.738638789093124,
      "tokens_seen": 109969408
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009764292878635907,
      "loss": 2.0372,
      "theoretical_loss": 4.738254932459727,
      "tokens_seen": 110034944
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009764092276830492,
      "loss": 2.1613,
      "theoretical_loss": 4.7378713683512235,
      "tokens_seen": 110100480
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009763891675025074,
      "loss": 2.0998,
      "theoretical_loss": 4.737488096370768,
      "tokens_seen": 110166016
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009763691073219659,
      "loss": 2.0198,
      "theoretical_loss": 4.737105116122288,
      "tokens_seen": 110231552
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009763490471414243,
      "loss": 2.1154,
      "theoretical_loss": 4.736722427210483,
      "tokens_seen": 110297088
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009763289869608826,
      "loss": 2.2146,
      "theoretical_loss": 4.7363400292408215,
      "tokens_seen": 110362624
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000976308926780341,
      "loss": 2.1661,
      "theoretical_loss": 4.735957921819542,
      "tokens_seen": 110428160
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009762888665997995,
      "loss": 2.1288,
      "theoretical_loss": 4.735576104553646,
      "tokens_seen": 110493696
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009762688064192577,
      "loss": 2.0877,
      "theoretical_loss": 4.735194577050903,
      "tokens_seen": 110559232
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009762487462387162,
      "loss": 2.2459,
      "theoretical_loss": 4.734813338919842,
      "tokens_seen": 110624768
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009762286860581745,
      "loss": 2.0061,
      "theoretical_loss": 4.734432389769756,
      "tokens_seen": 110690304
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009762086258776329,
      "loss": 2.0924,
      "theoretical_loss": 4.734051729210693,
      "tokens_seen": 110755840
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009761885656970913,
      "loss": 2.0756,
      "theoretical_loss": 4.733671356853456,
      "tokens_seen": 110821376
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009761685055165496,
      "loss": 2.111,
      "theoretical_loss": 4.733291272309609,
      "tokens_seen": 110886912
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000976148445336008,
      "loss": 2.1974,
      "theoretical_loss": 4.732911475191464,
      "tokens_seen": 110952448
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009761283851554664,
      "loss": 1.9964,
      "theoretical_loss": 4.732531965112083,
      "tokens_seen": 111017984
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009761083249749247,
      "loss": 2.1016,
      "theoretical_loss": 4.732152741685281,
      "tokens_seen": 111083520
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009760882647943832,
      "loss": 2.0433,
      "theoretical_loss": 4.731773804525616,
      "tokens_seen": 111149056
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009760682046138415,
      "loss": 2.0489,
      "theoretical_loss": 4.731395153248394,
      "tokens_seen": 111214592
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009760481444332999,
      "loss": 1.9634,
      "theoretical_loss": 4.731016787469663,
      "tokens_seen": 111280128
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009760280842527583,
      "loss": 2.1297,
      "theoretical_loss": 4.730638706806213,
      "tokens_seen": 111345664
    },
    {
      "epoch": 0.03,
      "objective/train/advantage_avg": 0.49222612380981445,
      "objective/train/docs_used": 47788,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.9619340896606445,
      "objective/train/original_loss": 1.961933970451355,
      "objective/train/theoretical_loss": 4.73026091087557,
      "objective/train/tokens_used": 131871200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24286691844463348,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050457239151001,
      "objective/train/weighted_lm_loss": 2.060546636581421,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.73026091087557,
      "tokens_seen": 111411200
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009760080240722167,
      "loss": 2.0969,
      "theoretical_loss": 4.73026091087557,
      "tokens_seen": 111411200
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000975987963891675,
      "loss": 2.1887,
      "theoretical_loss": 4.729883399296005,
      "tokens_seen": 111476736
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009759679037111334,
      "loss": 2.049,
      "theoretical_loss": 4.729506171686518,
      "tokens_seen": 111542272
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009759478435305919,
      "loss": 2.0852,
      "theoretical_loss": 4.729129227666847,
      "tokens_seen": 111607808
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009759277833500502,
      "loss": 2.1017,
      "theoretical_loss": 4.728752566857459,
      "tokens_seen": 111673344
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009759077231695086,
      "loss": 2.1405,
      "theoretical_loss": 4.728376188879554,
      "tokens_seen": 111738880
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009758876629889669,
      "loss": 2.1897,
      "theoretical_loss": 4.7280000933550586,
      "tokens_seen": 111804416
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009758676028084253,
      "loss": 2.1838,
      "theoretical_loss": 4.727624279906628,
      "tokens_seen": 111869952
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009758475426278837,
      "loss": 2.047,
      "theoretical_loss": 4.72724874815764,
      "tokens_seen": 111935488
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009758274824473421,
      "loss": 2.0746,
      "theoretical_loss": 4.726873497732196,
      "tokens_seen": 112001024
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009758074222668004,
      "loss": 2.2697,
      "theoretical_loss": 4.726498528255121,
      "tokens_seen": 112066560
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009757873620862589,
      "loss": 2.1296,
      "theoretical_loss": 4.7261238393519545,
      "tokens_seen": 112132096
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009757673019057171,
      "loss": 2.1274,
      "theoretical_loss": 4.725749430648958,
      "tokens_seen": 112197632
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009757472417251756,
      "loss": 2.0796,
      "theoretical_loss": 4.725375301773107,
      "tokens_seen": 112263168
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009757271815446339,
      "loss": 2.0802,
      "theoretical_loss": 4.725001452352091,
      "tokens_seen": 112328704
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009757071213640923,
      "loss": 2.1455,
      "theoretical_loss": 4.724627882014311,
      "tokens_seen": 112394240
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009756870611835507,
      "loss": 2.0957,
      "theoretical_loss": 4.724254590388881,
      "tokens_seen": 112459776
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009756670010030091,
      "loss": 1.9454,
      "theoretical_loss": 4.723881577105621,
      "tokens_seen": 112525312
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009756469408224674,
      "loss": 2.157,
      "theoretical_loss": 4.723508841795058,
      "tokens_seen": 112590848
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009756268806419259,
      "loss": 2.1352,
      "theoretical_loss": 4.723136384088427,
      "tokens_seen": 112656384
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009756068204613841,
      "loss": 2.078,
      "theoretical_loss": 4.722764203617663,
      "tokens_seen": 112721920
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009755867602808426,
      "loss": 2.1686,
      "theoretical_loss": 4.722392300015404,
      "tokens_seen": 112787456
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000975566700100301,
      "loss": 2.1921,
      "theoretical_loss": 4.7220206729149865,
      "tokens_seen": 112852992
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009755466399197593,
      "loss": 2.0499,
      "theoretical_loss": 4.72164932195045,
      "tokens_seen": 112918528
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009755265797392177,
      "loss": 2.1789,
      "theoretical_loss": 4.7212782467565235,
      "tokens_seen": 112984064
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000975506519558676,
      "loss": 2.0737,
      "theoretical_loss": 4.720907446968635,
      "tokens_seen": 113049600
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009754864593781344,
      "loss": 2.0673,
      "theoretical_loss": 4.720536922222907,
      "tokens_seen": 113115136
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009754663991975929,
      "loss": 2.1072,
      "theoretical_loss": 4.720166672156147,
      "tokens_seen": 113180672
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009754463390170511,
      "loss": 2.0866,
      "theoretical_loss": 4.719796696405858,
      "tokens_seen": 113246208
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009754262788365096,
      "loss": 2.1345,
      "theoretical_loss": 4.719426994610227,
      "tokens_seen": 113311744
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009754062186559678,
      "loss": 2.0008,
      "theoretical_loss": 4.719057566408132,
      "tokens_seen": 113377280
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009753861584754263,
      "loss": 2.0752,
      "theoretical_loss": 4.718688411439129,
      "tokens_seen": 113442816
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009753660982948847,
      "loss": 2.0776,
      "theoretical_loss": 4.718319529343462,
      "tokens_seen": 113508352
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000975346038114343,
      "loss": 2.1316,
      "theoretical_loss": 4.7179509197620515,
      "tokens_seen": 113573888
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009753259779338014,
      "loss": 2.082,
      "theoretical_loss": 4.717582582336503,
      "tokens_seen": 113639424
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009753059177532599,
      "loss": 1.9908,
      "theoretical_loss": 4.7172145167090935,
      "tokens_seen": 113704960
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009752858575727181,
      "loss": 2.1448,
      "theoretical_loss": 4.716846722522781,
      "tokens_seen": 113770496
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009752657973921766,
      "loss": 2.0554,
      "theoretical_loss": 4.716479199421195,
      "tokens_seen": 113836032
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009752457372116349,
      "loss": 1.9584,
      "theoretical_loss": 4.716111947048638,
      "tokens_seen": 113901568
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009752256770310933,
      "loss": 2.0346,
      "theoretical_loss": 4.715744965050087,
      "tokens_seen": 113967104
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009752056168505517,
      "loss": 2.1214,
      "theoretical_loss": 4.715378253071181,
      "tokens_seen": 114032640
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00097518555667001,
      "loss": 2.01,
      "theoretical_loss": 4.715011810758236,
      "tokens_seen": 114098176
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009751654964894684,
      "loss": 2.0735,
      "theoretical_loss": 4.714645637758226,
      "tokens_seen": 114163712
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009751454363089268,
      "loss": 2.0352,
      "theoretical_loss": 4.714279733718793,
      "tokens_seen": 114229248
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009751253761283851,
      "loss": 2.1201,
      "theoretical_loss": 4.713914098288242,
      "tokens_seen": 114294784
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009751053159478436,
      "loss": 2.0151,
      "theoretical_loss": 4.71354873111554,
      "tokens_seen": 114360320
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009750852557673019,
      "loss": 1.9781,
      "theoretical_loss": 4.7131836318503115,
      "tokens_seen": 114425856
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009750651955867603,
      "loss": 2.0794,
      "theoretical_loss": 4.7128188001428395,
      "tokens_seen": 114491392
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009750451354062187,
      "loss": 2.075,
      "theoretical_loss": 4.712454235644064,
      "tokens_seen": 114556928
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009750250752256771,
      "loss": 2.1529,
      "theoretical_loss": 4.71208993800558,
      "tokens_seen": 114622464
    },
    {
      "epoch": 0.03,
      "objective/train/advantage_avg": 0.49528729915618896,
      "objective/train/docs_used": 48994,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.978676676750183,
      "objective/train/original_loss": 1.9786767959594727,
      "objective/train/theoretical_loss": 4.711725906879634,
      "objective/train/tokens_used": 135148000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24554309248924255,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507770776748657,
      "objective/train/weighted_lm_loss": 2.0792107582092285,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.711725906879634,
      "tokens_seen": 114688000
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009750050150451354,
      "loss": 2.0281,
      "theoretical_loss": 4.711725906879634,
      "tokens_seen": 114688000
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009749849548645938,
      "loss": 2.0641,
      "theoretical_loss": 4.7113621419191265,
      "tokens_seen": 114753536
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009749648946840521,
      "loss": 2.1124,
      "theoretical_loss": 4.710998642777606,
      "tokens_seen": 114819072
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009749448345035106,
      "loss": 2.0097,
      "theoretical_loss": 4.710635409109269,
      "tokens_seen": 114884608
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009749247743229689,
      "loss": 2.2112,
      "theoretical_loss": 4.710272440568962,
      "tokens_seen": 114950144
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009749047141424273,
      "loss": 2.1056,
      "theoretical_loss": 4.709909736812172,
      "tokens_seen": 115015680
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009748846539618856,
      "loss": 2.2072,
      "theoretical_loss": 4.709547297495034,
      "tokens_seen": 115081216
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009748645937813441,
      "loss": 2.1081,
      "theoretical_loss": 4.709185122274322,
      "tokens_seen": 115146752
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009748445336008024,
      "loss": 2.0524,
      "theoretical_loss": 4.708823210807452,
      "tokens_seen": 115212288
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009748244734202608,
      "loss": 1.8597,
      "theoretical_loss": 4.708461562752479,
      "tokens_seen": 115277824
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009748044132397191,
      "loss": 2.0625,
      "theoretical_loss": 4.708100177768094,
      "tokens_seen": 115343360
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009747843530591775,
      "loss": 2.0811,
      "theoretical_loss": 4.707739055513627,
      "tokens_seen": 115408896
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0009747642928786359,
      "loss": 2.183,
      "theoretical_loss": 4.707378195649038,
      "tokens_seen": 115474432
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009747442326980943,
      "loss": 1.9737,
      "theoretical_loss": 4.7070175978349225,
      "tokens_seen": 115539968
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009747241725175526,
      "loss": 2.1049,
      "theoretical_loss": 4.7066572617325075,
      "tokens_seen": 115605504
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009747041123370111,
      "loss": 2.0388,
      "theoretical_loss": 4.706297187003647,
      "tokens_seen": 115671040
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009746840521564694,
      "loss": 2.0434,
      "theoretical_loss": 4.705937373310828,
      "tokens_seen": 115736576
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009746639919759278,
      "loss": 2.2422,
      "theoretical_loss": 4.705577820317158,
      "tokens_seen": 115802112
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009746439317953862,
      "loss": 2.1174,
      "theoretical_loss": 4.705218527686375,
      "tokens_seen": 115867648
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009746238716148445,
      "loss": 2.1772,
      "theoretical_loss": 4.704859495082837,
      "tokens_seen": 115933184
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009746038114343029,
      "loss": 2.1444,
      "theoretical_loss": 4.704500722171526,
      "tokens_seen": 115998720
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009745837512537614,
      "loss": 2.2104,
      "theoretical_loss": 4.704142208618045,
      "tokens_seen": 116064256
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009745636910732196,
      "loss": 2.0155,
      "theoretical_loss": 4.703783954088612,
      "tokens_seen": 116129792
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009745436308926781,
      "loss": 2.3143,
      "theoretical_loss": 4.703425958250071,
      "tokens_seen": 116195328
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009745235707121363,
      "loss": 2.1153,
      "theoretical_loss": 4.703068220769872,
      "tokens_seen": 116260864
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009745035105315948,
      "loss": 2.0486,
      "theoretical_loss": 4.702710741316087,
      "tokens_seen": 116326400
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009744834503510532,
      "loss": 2.1845,
      "theoretical_loss": 4.702353519557398,
      "tokens_seen": 116391936
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009744633901705115,
      "loss": 2.2034,
      "theoretical_loss": 4.701996555163101,
      "tokens_seen": 116457472
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00097444332998997,
      "loss": 2.0476,
      "theoretical_loss": 4.701639847803098,
      "tokens_seen": 116523008
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009744232698094282,
      "loss": 2.1367,
      "theoretical_loss": 4.701283397147904,
      "tokens_seen": 116588544
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009744032096288867,
      "loss": 2.0433,
      "theoretical_loss": 4.700927202868639,
      "tokens_seen": 116654080
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009743831494483451,
      "loss": 2.0795,
      "theoretical_loss": 4.700571264637029,
      "tokens_seen": 116719616
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009743630892678034,
      "loss": 1.985,
      "theoretical_loss": 4.700215582125406,
      "tokens_seen": 116785152
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009743430290872618,
      "loss": 2.0779,
      "theoretical_loss": 4.699860155006702,
      "tokens_seen": 116850688
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009743229689067203,
      "loss": 2.0759,
      "theoretical_loss": 4.699504982954452,
      "tokens_seen": 116916224
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009743029087261785,
      "loss": 2.1299,
      "theoretical_loss": 4.699150065642794,
      "tokens_seen": 116981760
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000974282848545637,
      "loss": 1.9488,
      "theoretical_loss": 4.698795402746457,
      "tokens_seen": 117047296
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009742627883650953,
      "loss": 2.0521,
      "theoretical_loss": 4.6984409939407765,
      "tokens_seen": 117112832
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009742427281845537,
      "loss": 2.3108,
      "theoretical_loss": 4.698086838901676,
      "tokens_seen": 117178368
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009742226680040121,
      "loss": 2.0137,
      "theoretical_loss": 4.697732937305678,
      "tokens_seen": 117243904
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009742026078234705,
      "loss": 2.1527,
      "theoretical_loss": 4.697379288829895,
      "tokens_seen": 117309440
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009741825476429288,
      "loss": 1.9996,
      "theoretical_loss": 4.697025893152032,
      "tokens_seen": 117374976
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009741624874623872,
      "loss": 2.1059,
      "theoretical_loss": 4.696672749950385,
      "tokens_seen": 117440512
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009741424272818455,
      "loss": 1.938,
      "theoretical_loss": 4.696319858903838,
      "tokens_seen": 117506048
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000974122367101304,
      "loss": 2.1274,
      "theoretical_loss": 4.69596721969186,
      "tokens_seen": 117571584
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009741023069207623,
      "loss": 2.1286,
      "theoretical_loss": 4.69561483199451,
      "tokens_seen": 117637120
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009740822467402207,
      "loss": 2.0024,
      "theoretical_loss": 4.695262695492428,
      "tokens_seen": 117702656
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009740621865596791,
      "loss": 1.99,
      "theoretical_loss": 4.694910809866839,
      "tokens_seen": 117768192
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009740421263791375,
      "loss": 2.1947,
      "theoretical_loss": 4.694559174799549,
      "tokens_seen": 117833728
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009740220661985958,
      "loss": 2.0581,
      "theoretical_loss": 4.694207789972943,
      "tokens_seen": 117899264
    },
    {
      "epoch": 0.04,
      "objective/train/advantage_avg": 0.4920607805252075,
      "objective/train/docs_used": 50256,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.054048776626587,
      "objective/train/original_loss": 2.054048776626587,
      "objective/train/theoretical_loss": 4.693856655069986,
      "objective/train/tokens_used": 138424800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2430429458618164,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504416227340698,
      "objective/train/weighted_lm_loss": 2.157548666000366,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 4.693856655069986,
      "tokens_seen": 117964800
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009740020060180542,
      "loss": 2.1324,
      "theoretical_loss": 4.693856655069986,
      "tokens_seen": 117964800
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009739819458375125,
      "loss": 2.1721,
      "theoretical_loss": 4.693505769774223,
      "tokens_seen": 118030336
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000973961885656971,
      "loss": 2.1294,
      "theoretical_loss": 4.6931551337697694,
      "tokens_seen": 118095872
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009739418254764293,
      "loss": 2.0215,
      "theoretical_loss": 4.692804746741322,
      "tokens_seen": 118161408
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009739217652958877,
      "loss": 2.1654,
      "theoretical_loss": 4.692454608374145,
      "tokens_seen": 118226944
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000973901705115346,
      "loss": 2.1852,
      "theoretical_loss": 4.692104718354077,
      "tokens_seen": 118292480
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009738816449348045,
      "loss": 1.9563,
      "theoretical_loss": 4.69175507636753,
      "tokens_seen": 118358016
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009738615847542628,
      "loss": 2.0985,
      "theoretical_loss": 4.691405682101482,
      "tokens_seen": 118423552
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009738415245737212,
      "loss": 2.0069,
      "theoretical_loss": 4.6910565352434785,
      "tokens_seen": 118489088
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009738214643931796,
      "loss": 2.1439,
      "theoretical_loss": 4.690707635481634,
      "tokens_seen": 118554624
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009738014042126379,
      "loss": 2.0671,
      "theoretical_loss": 4.690358982504628,
      "tokens_seen": 118620160
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009737813440320963,
      "loss": 1.9935,
      "theoretical_loss": 4.690010576001702,
      "tokens_seen": 118685696
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009737612838515547,
      "loss": 2.0938,
      "theoretical_loss": 4.689662415662662,
      "tokens_seen": 118751232
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000973741223671013,
      "loss": 2.0523,
      "theoretical_loss": 4.689314501177874,
      "tokens_seen": 118816768
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009737211634904715,
      "loss": 2.0498,
      "theoretical_loss": 4.6889668322382665,
      "tokens_seen": 118882304
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009737011033099298,
      "loss": 2.025,
      "theoretical_loss": 4.688619408535323,
      "tokens_seen": 118947840
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009736810431293882,
      "loss": 2.1013,
      "theoretical_loss": 4.688272229761087,
      "tokens_seen": 119013376
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009736609829488466,
      "loss": 2.1606,
      "theoretical_loss": 4.687925295608157,
      "tokens_seen": 119078912
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009736409227683049,
      "loss": 2.1686,
      "theoretical_loss": 4.687578605769687,
      "tokens_seen": 119144448
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009736208625877633,
      "loss": 2.0253,
      "theoretical_loss": 4.6872321599393825,
      "tokens_seen": 119209984
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009736008024072218,
      "loss": 2.0954,
      "theoretical_loss": 4.686885957811503,
      "tokens_seen": 119275520
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00097358074222668,
      "loss": 1.9549,
      "theoretical_loss": 4.6865399990808605,
      "tokens_seen": 119341056
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009735606820461385,
      "loss": 1.9711,
      "theoretical_loss": 4.686194283442811,
      "tokens_seen": 119406592
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009735406218655967,
      "loss": 2.1033,
      "theoretical_loss": 4.685848810593263,
      "tokens_seen": 119472128
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009735205616850552,
      "loss": 1.9815,
      "theoretical_loss": 4.685503580228671,
      "tokens_seen": 119537664
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009735005015045136,
      "loss": 2.1605,
      "theoretical_loss": 4.685158592046037,
      "tokens_seen": 119603200
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009734804413239719,
      "loss": 1.963,
      "theoretical_loss": 4.684813845742902,
      "tokens_seen": 119668736
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009734603811434303,
      "loss": 2.0143,
      "theoretical_loss": 4.684469341017357,
      "tokens_seen": 119734272
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009734403209628886,
      "loss": 2.0491,
      "theoretical_loss": 4.684125077568028,
      "tokens_seen": 119799808
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000973420260782347,
      "loss": 2.2728,
      "theoretical_loss": 4.683781055094087,
      "tokens_seen": 119865344
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009734002006018055,
      "loss": 2.0802,
      "theoretical_loss": 4.683437273295245,
      "tokens_seen": 119930880
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009733801404212637,
      "loss": 2.1258,
      "theoretical_loss": 4.683093731871746,
      "tokens_seen": 119996416
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009733600802407222,
      "loss": 1.933,
      "theoretical_loss": 4.682750430524376,
      "tokens_seen": 120061952
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009733400200601806,
      "loss": 2.0386,
      "theoretical_loss": 4.682407368954454,
      "tokens_seen": 120127488
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009733199598796389,
      "loss": 2.1227,
      "theoretical_loss": 4.682064546863834,
      "tokens_seen": 120193024
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009732998996990973,
      "loss": 2.0572,
      "theoretical_loss": 4.681721963954903,
      "tokens_seen": 120258560
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009732798395185557,
      "loss": 2.2008,
      "theoretical_loss": 4.6813796199305795,
      "tokens_seen": 120324096
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000973259779338014,
      "loss": 2.0648,
      "theoretical_loss": 4.681037514494313,
      "tokens_seen": 120389632
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009732397191574725,
      "loss": 2.1425,
      "theoretical_loss": 4.680695647350082,
      "tokens_seen": 120455168
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009732196589769307,
      "loss": 2.1755,
      "theoretical_loss": 4.680354018202394,
      "tokens_seen": 120520704
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009731995987963892,
      "loss": 2.08,
      "theoretical_loss": 4.68001262675628,
      "tokens_seen": 120586240
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009731795386158475,
      "loss": 1.8476,
      "theoretical_loss": 4.6796714727173025,
      "tokens_seen": 120651776
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009731594784353059,
      "loss": 2.1225,
      "theoretical_loss": 4.679330555791543,
      "tokens_seen": 120717312
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009731394182547643,
      "loss": 2.1533,
      "theoretical_loss": 4.678989875685607,
      "tokens_seen": 120782848
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009731193580742227,
      "loss": 2.0241,
      "theoretical_loss": 4.678649432106627,
      "tokens_seen": 120848384
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000973099297893681,
      "loss": 1.9959,
      "theoretical_loss": 4.678309224762249,
      "tokens_seen": 120913920
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009730792377131395,
      "loss": 2.239,
      "theoretical_loss": 4.677969253360644,
      "tokens_seen": 120979456
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009730591775325977,
      "loss": 2.1644,
      "theoretical_loss": 4.677629517610498,
      "tokens_seen": 121044992
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009730391173520562,
      "loss": 1.9964,
      "theoretical_loss": 4.677290017221017,
      "tokens_seen": 121110528
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009730190571715145,
      "loss": 2.1193,
      "theoretical_loss": 4.67695075190192,
      "tokens_seen": 121176064
    },
    {
      "epoch": 0.04,
      "objective/train/advantage_avg": 0.4931885600090027,
      "objective/train/docs_used": 51395,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.00555157661438,
      "objective/train/original_loss": 2.00555157661438,
      "objective/train/theoretical_loss": 4.676611721363443,
      "objective/train/tokens_used": 141701600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24360352754592896,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050557255744934,
      "objective/train/weighted_lm_loss": 2.1069064140319824,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.676611721363443,
      "tokens_seen": 121241600
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009729989969909729,
      "loss": 2.107,
      "theoretical_loss": 4.676611721363443,
      "tokens_seen": 121241600
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009729789368104313,
      "loss": 1.9974,
      "theoretical_loss": 4.6762729253163355,
      "tokens_seen": 121307136
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009729588766298897,
      "loss": 2.0427,
      "theoretical_loss": 4.675934363471857,
      "tokens_seen": 121372672
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009729388164493481,
      "loss": 1.9648,
      "theoretical_loss": 4.675596035541782,
      "tokens_seen": 121438208
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009729187562688064,
      "loss": 2.0273,
      "theoretical_loss": 4.675257941238391,
      "tokens_seen": 121503744
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009728986960882649,
      "loss": 2.041,
      "theoretical_loss": 4.674920080274475,
      "tokens_seen": 121569280
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009728786359077232,
      "loss": 2.0907,
      "theoretical_loss": 4.674582452363334,
      "tokens_seen": 121634816
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009728585757271816,
      "loss": 2.0461,
      "theoretical_loss": 4.674245057218775,
      "tokens_seen": 121700352
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00097283851554664,
      "loss": 2.1216,
      "theoretical_loss": 4.673907894555105,
      "tokens_seen": 121765888
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009728184553660983,
      "loss": 1.9807,
      "theoretical_loss": 4.673570964087141,
      "tokens_seen": 121831424
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009727983951855567,
      "loss": 2.1125,
      "theoretical_loss": 4.673234265530201,
      "tokens_seen": 121896960
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009727783350050151,
      "loss": 1.9752,
      "theoretical_loss": 4.672897798600104,
      "tokens_seen": 121962496
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009727582748244734,
      "loss": 2.0491,
      "theoretical_loss": 4.672561563013172,
      "tokens_seen": 122028032
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009727382146439319,
      "loss": 2.019,
      "theoretical_loss": 4.672225558486223,
      "tokens_seen": 122093568
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009727181544633902,
      "loss": 2.064,
      "theoretical_loss": 4.671889784736576,
      "tokens_seen": 122159104
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009726980942828486,
      "loss": 2.0167,
      "theoretical_loss": 4.671554241482047,
      "tokens_seen": 122224640
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000972678034102307,
      "loss": 2.2564,
      "theoretical_loss": 4.671218928440947,
      "tokens_seen": 122290176
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009726579739217653,
      "loss": 2.0625,
      "theoretical_loss": 4.670883845332084,
      "tokens_seen": 122355712
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009726379137412237,
      "loss": 1.9611,
      "theoretical_loss": 4.670548991874758,
      "tokens_seen": 122421248
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009726178535606822,
      "loss": 1.9603,
      "theoretical_loss": 4.670214367788761,
      "tokens_seen": 122486784
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009725977933801404,
      "loss": 2.095,
      "theoretical_loss": 4.66987997279438,
      "tokens_seen": 122552320
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009725777331995989,
      "loss": 1.9932,
      "theoretical_loss": 4.669545806612389,
      "tokens_seen": 122617856
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009725576730190571,
      "loss": 2.0098,
      "theoretical_loss": 4.669211868964052,
      "tokens_seen": 122683392
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009725376128385156,
      "loss": 2.0485,
      "theoretical_loss": 4.668878159571123,
      "tokens_seen": 122748928
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000972517552657974,
      "loss": 2.0402,
      "theoretical_loss": 4.6685446781558415,
      "tokens_seen": 122814464
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009724974924774323,
      "loss": 2.0854,
      "theoretical_loss": 4.668211424440934,
      "tokens_seen": 122880000
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009724774322968907,
      "loss": 2.1726,
      "theoretical_loss": 4.66787839814961,
      "tokens_seen": 122945536
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000972457372116349,
      "loss": 2.0551,
      "theoretical_loss": 4.667545599005566,
      "tokens_seen": 123011072
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009724373119358074,
      "loss": 2.0483,
      "theoretical_loss": 4.667213026732977,
      "tokens_seen": 123076608
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009724172517552659,
      "loss": 1.9799,
      "theoretical_loss": 4.6668806810565036,
      "tokens_seen": 123142144
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009723971915747241,
      "loss": 2.1432,
      "theoretical_loss": 4.666548561701285,
      "tokens_seen": 123207680
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009723771313941826,
      "loss": 2.091,
      "theoretical_loss": 4.666216668392939,
      "tokens_seen": 123273216
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000972357071213641,
      "loss": 2.0325,
      "theoretical_loss": 4.665885000857562,
      "tokens_seen": 123338752
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009723370110330993,
      "loss": 2.0689,
      "theoretical_loss": 4.665553558821729,
      "tokens_seen": 123404288
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009723169508525577,
      "loss": 2.125,
      "theoretical_loss": 4.665222342012491,
      "tokens_seen": 123469824
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009722968906720161,
      "loss": 1.9438,
      "theoretical_loss": 4.664891350157374,
      "tokens_seen": 123535360
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009722768304914744,
      "loss": 2.0849,
      "theoretical_loss": 4.664560582984374,
      "tokens_seen": 123600896
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009722567703109329,
      "loss": 1.9947,
      "theoretical_loss": 4.664230040221966,
      "tokens_seen": 123666432
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009722367101303911,
      "loss": 2.2468,
      "theoretical_loss": 4.663899721599093,
      "tokens_seen": 123731968
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009722166499498496,
      "loss": 1.9996,
      "theoretical_loss": 4.66356962684517,
      "tokens_seen": 123797504
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009721965897693079,
      "loss": 2.03,
      "theoretical_loss": 4.663239755690082,
      "tokens_seen": 123863040
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009721765295887663,
      "loss": 2.0711,
      "theoretical_loss": 4.66291010786418,
      "tokens_seen": 123928576
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009721564694082247,
      "loss": 1.9784,
      "theoretical_loss": 4.6625806830982865,
      "tokens_seen": 123994112
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009721364092276831,
      "loss": 1.9567,
      "theoretical_loss": 4.662251481123688,
      "tokens_seen": 124059648
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009721163490471414,
      "loss": 2.0798,
      "theoretical_loss": 4.661922501672136,
      "tokens_seen": 124125184
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009720962888665999,
      "loss": 2.0622,
      "theoretical_loss": 4.66159374447585,
      "tokens_seen": 124190720
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009720762286860581,
      "loss": 2.0201,
      "theoretical_loss": 4.661265209267507,
      "tokens_seen": 124256256
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009720561685055166,
      "loss": 1.9692,
      "theoretical_loss": 4.660936895780253,
      "tokens_seen": 124321792
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009720361083249749,
      "loss": 2.1423,
      "theoretical_loss": 4.660608803747688,
      "tokens_seen": 124387328
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009720160481444333,
      "loss": 2.0592,
      "theoretical_loss": 4.660280932903879,
      "tokens_seen": 124452864
    },
    {
      "epoch": 0.04,
      "objective/train/advantage_avg": 0.4943527281284332,
      "objective/train/docs_used": 52572,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8934916257858276,
      "objective/train/original_loss": 1.8934917449951172,
      "objective/train/theoretical_loss": 4.659953282983348,
      "objective/train/tokens_used": 144978400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2446935772895813,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050679326057434,
      "objective/train/weighted_lm_loss": 1.9894511699676514,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.659953282983348,
      "tokens_seen": 124518400
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009719959879638917,
      "loss": 1.9832,
      "theoretical_loss": 4.659953282983348,
      "tokens_seen": 124518400
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009719759277833501,
      "loss": 2.0029,
      "theoretical_loss": 4.659625853721076,
      "tokens_seen": 124583936
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009719558676028084,
      "loss": 2.167,
      "theoretical_loss": 4.659298644852502,
      "tokens_seen": 124649472
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009719358074222668,
      "loss": 2.0752,
      "theoretical_loss": 4.65897165611352,
      "tokens_seen": 124715008
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009719157472417252,
      "loss": 2.1872,
      "theoretical_loss": 4.658644887240481,
      "tokens_seen": 124780544
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009718956870611836,
      "loss": 2.0446,
      "theoretical_loss": 4.6583183379701865,
      "tokens_seen": 124846080
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009718756268806419,
      "loss": 2.0463,
      "theoretical_loss": 4.657992008039894,
      "tokens_seen": 124911616
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009718555667001004,
      "loss": 2.1342,
      "theoretical_loss": 4.657665897187313,
      "tokens_seen": 124977152
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009718355065195586,
      "loss": 2.1419,
      "theoretical_loss": 4.657340005150602,
      "tokens_seen": 125042688
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009718154463390171,
      "loss": 2.1212,
      "theoretical_loss": 4.657014331668371,
      "tokens_seen": 125108224
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009717953861584754,
      "loss": 2.106,
      "theoretical_loss": 4.656688876479679,
      "tokens_seen": 125173760
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009717753259779338,
      "loss": 2.0815,
      "theoretical_loss": 4.656363639324032,
      "tokens_seen": 125239296
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009717552657973922,
      "loss": 2.0729,
      "theoretical_loss": 4.656038619941382,
      "tokens_seen": 125304832
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009717352056168506,
      "loss": 1.8985,
      "theoretical_loss": 4.655713818072132,
      "tokens_seen": 125370368
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009717151454363089,
      "loss": 2.0584,
      "theoretical_loss": 4.655389233457123,
      "tokens_seen": 125435904
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009716950852557674,
      "loss": 2.067,
      "theoretical_loss": 4.6550648658376454,
      "tokens_seen": 125501440
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009716750250752256,
      "loss": 2.0179,
      "theoretical_loss": 4.654740714955429,
      "tokens_seen": 125566976
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009716549648946841,
      "loss": 2.0203,
      "theoretical_loss": 4.654416780552648,
      "tokens_seen": 125632512
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009716349047141424,
      "loss": 2.1031,
      "theoretical_loss": 4.654093062371917,
      "tokens_seen": 125698048
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009716148445336008,
      "loss": 2.1167,
      "theoretical_loss": 4.65376956015629,
      "tokens_seen": 125763584
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009715947843530592,
      "loss": 2.0359,
      "theoretical_loss": 4.653446273649259,
      "tokens_seen": 125829120
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009715747241725175,
      "loss": 2.0423,
      "theoretical_loss": 4.653123202594757,
      "tokens_seen": 125894656
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009715546639919759,
      "loss": 2.1098,
      "theoretical_loss": 4.652800346737152,
      "tokens_seen": 125960192
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009715346038114344,
      "loss": 2.062,
      "theoretical_loss": 4.652477705821248,
      "tokens_seen": 126025728
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009715145436308926,
      "loss": 1.9808,
      "theoretical_loss": 4.652155279592286,
      "tokens_seen": 126091264
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009714944834503511,
      "loss": 2.0248,
      "theoretical_loss": 4.651833067795939,
      "tokens_seen": 126156800
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009714744232698093,
      "loss": 1.9461,
      "theoretical_loss": 4.651511070178314,
      "tokens_seen": 126222336
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009714543630892678,
      "loss": 2.1108,
      "theoretical_loss": 4.65118928648595,
      "tokens_seen": 126287872
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009714343029087263,
      "loss": 2.0036,
      "theoretical_loss": 4.650867716465819,
      "tokens_seen": 126353408
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009714142427281845,
      "loss": 1.8282,
      "theoretical_loss": 4.65054635986532,
      "tokens_seen": 126418944
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000971394182547643,
      "loss": 2.0785,
      "theoretical_loss": 4.650225216432285,
      "tokens_seen": 126484480
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009713741223671014,
      "loss": 2.1061,
      "theoretical_loss": 4.649904285914971,
      "tokens_seen": 126550016
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009713540621865597,
      "loss": 2.1424,
      "theoretical_loss": 4.6495835680620665,
      "tokens_seen": 126615552
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009713340020060181,
      "loss": 2.0539,
      "theoretical_loss": 4.649263062622682,
      "tokens_seen": 126681088
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009713139418254765,
      "loss": 1.9926,
      "theoretical_loss": 4.648942769346356,
      "tokens_seen": 126746624
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009712938816449348,
      "loss": 2.0468,
      "theoretical_loss": 4.648622687983053,
      "tokens_seen": 126812160
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009712738214643933,
      "loss": 1.9613,
      "theoretical_loss": 4.648302818283158,
      "tokens_seen": 126877696
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009712537612838515,
      "loss": 2.1172,
      "theoretical_loss": 4.64798315999748,
      "tokens_seen": 126943232
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00097123370110331,
      "loss": 2.1304,
      "theoretical_loss": 4.647663712877252,
      "tokens_seen": 127008768
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009712136409227683,
      "loss": 2.0174,
      "theoretical_loss": 4.6473444766741245,
      "tokens_seen": 127074304
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009711935807422267,
      "loss": 2.11,
      "theoretical_loss": 4.64702545114017,
      "tokens_seen": 127139840
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009711735205616851,
      "loss": 2.0202,
      "theoretical_loss": 4.646706636027881,
      "tokens_seen": 127205376
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009711534603811435,
      "loss": 2.0189,
      "theoretical_loss": 4.646388031090163,
      "tokens_seen": 127270912
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009711334002006018,
      "loss": 2.0739,
      "theoretical_loss": 4.646069636080346,
      "tokens_seen": 127336448
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009711133400200603,
      "loss": 2.0184,
      "theoretical_loss": 4.645751450752172,
      "tokens_seen": 127401984
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009710932798395185,
      "loss": 1.9785,
      "theoretical_loss": 4.645433474859798,
      "tokens_seen": 127467520
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000971073219658977,
      "loss": 1.9877,
      "theoretical_loss": 4.645115708157796,
      "tokens_seen": 127533056
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009710531594784353,
      "loss": 1.9167,
      "theoretical_loss": 4.6447981504011535,
      "tokens_seen": 127598592
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009710330992978937,
      "loss": 2.11,
      "theoretical_loss": 4.644480801345268,
      "tokens_seen": 127664128
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009710130391173521,
      "loss": 2.0886,
      "theoretical_loss": 4.644163660745951,
      "tokens_seen": 127729664
    },
    {
      "epoch": 0.04,
      "objective/train/advantage_avg": 0.49203452467918396,
      "objective/train/docs_used": 53741,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.9939597845077515,
      "objective/train/original_loss": 1.993959665298462,
      "objective/train/theoretical_loss": 4.6438467283594225,
      "objective/train/tokens_used": 148255200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2424752116203308,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050436019897461,
      "objective/train/weighted_lm_loss": 2.0946204662323,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.6438467283594225,
      "tokens_seen": 127795200
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009709929789368105,
      "loss": 2.1437,
      "theoretical_loss": 4.6438467283594225,
      "tokens_seen": 127795200
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009709729187562688,
      "loss": 2.1764,
      "theoretical_loss": 4.643530003942313,
      "tokens_seen": 127860736
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009709528585757272,
      "loss": 1.9627,
      "theoretical_loss": 4.643213487251664,
      "tokens_seen": 127926272
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009709327983951856,
      "loss": 1.9454,
      "theoretical_loss": 4.642897178044924,
      "tokens_seen": 127991808
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000970912738214644,
      "loss": 1.9753,
      "theoretical_loss": 4.6425810760799475,
      "tokens_seen": 128057344
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009708926780341023,
      "loss": 2.1277,
      "theoretical_loss": 4.642265181114996,
      "tokens_seen": 128122880
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009708726178535608,
      "loss": 1.9976,
      "theoretical_loss": 4.641949492908737,
      "tokens_seen": 128188416
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000970852557673019,
      "loss": 2.0455,
      "theoretical_loss": 4.641634011220242,
      "tokens_seen": 128253952
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009708324974924775,
      "loss": 2.1228,
      "theoretical_loss": 4.641318735808988,
      "tokens_seen": 128319488
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009708124373119358,
      "loss": 1.9554,
      "theoretical_loss": 4.641003666434852,
      "tokens_seen": 128385024
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009707923771313942,
      "loss": 2.0163,
      "theoretical_loss": 4.640688802858113,
      "tokens_seen": 128450560
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009707723169508526,
      "loss": 2.0832,
      "theoretical_loss": 4.640374144839454,
      "tokens_seen": 128516096
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000970752256770311,
      "loss": 2.1972,
      "theoretical_loss": 4.640059692139955,
      "tokens_seen": 128581632
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009707321965897693,
      "loss": 1.9591,
      "theoretical_loss": 4.639745444521098,
      "tokens_seen": 128647168
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009707121364092278,
      "loss": 2.0579,
      "theoretical_loss": 4.63943140174476,
      "tokens_seen": 128712704
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000970692076228686,
      "loss": 2.126,
      "theoretical_loss": 4.63911756357322,
      "tokens_seen": 128778240
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009706720160481445,
      "loss": 2.0433,
      "theoretical_loss": 4.63880392976915,
      "tokens_seen": 128843776
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009706519558676028,
      "loss": 1.974,
      "theoretical_loss": 4.638490500095619,
      "tokens_seen": 128909312
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009706318956870612,
      "loss": 2.0815,
      "theoretical_loss": 4.6381772743160905,
      "tokens_seen": 128974848
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009706118355065196,
      "loss": 2.1851,
      "theoretical_loss": 4.637864252194423,
      "tokens_seen": 129040384
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009705917753259779,
      "loss": 2.0548,
      "theoretical_loss": 4.637551433494869,
      "tokens_seen": 129105920
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009705717151454363,
      "loss": 2.0605,
      "theoretical_loss": 4.637238817982071,
      "tokens_seen": 129171456
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009705516549648948,
      "loss": 1.9794,
      "theoretical_loss": 4.636926405421065,
      "tokens_seen": 129236992
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000970531594784353,
      "loss": 2.1056,
      "theoretical_loss": 4.636614195577276,
      "tokens_seen": 129302528
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009705115346038115,
      "loss": 1.8769,
      "theoretical_loss": 4.63630218821652,
      "tokens_seen": 129368064
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009704914744232697,
      "loss": 2.0787,
      "theoretical_loss": 4.635990383105003,
      "tokens_seen": 129433600
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009704714142427282,
      "loss": 1.9702,
      "theoretical_loss": 4.635678780009318,
      "tokens_seen": 129499136
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009704513540621866,
      "loss": 2.0665,
      "theoretical_loss": 4.6353673786964436,
      "tokens_seen": 129564672
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009704312938816449,
      "loss": 1.9906,
      "theoretical_loss": 4.635056178933747,
      "tokens_seen": 129630208
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009704112337011033,
      "loss": 2.0691,
      "theoretical_loss": 4.6347451804889825,
      "tokens_seen": 129695744
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009703911735205618,
      "loss": 1.9572,
      "theoretical_loss": 4.634434383130284,
      "tokens_seen": 129761280
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00097037111334002,
      "loss": 2.1197,
      "theoretical_loss": 4.634123786626175,
      "tokens_seen": 129826816
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009703510531594785,
      "loss": 2.0133,
      "theoretical_loss": 4.633813390745559,
      "tokens_seen": 129892352
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009703309929789367,
      "loss": 2.1973,
      "theoretical_loss": 4.633503195257724,
      "tokens_seen": 129957888
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009703109327983952,
      "loss": 2.1009,
      "theoretical_loss": 4.633193199932336,
      "tokens_seen": 130023424
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009702908726178536,
      "loss": 2.1452,
      "theoretical_loss": 4.632883404539445,
      "tokens_seen": 130088960
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009702708124373119,
      "loss": 2.0481,
      "theoretical_loss": 4.632573808849479,
      "tokens_seen": 130154496
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009702507522567703,
      "loss": 2.0503,
      "theoretical_loss": 4.632264412633246,
      "tokens_seen": 130220032
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009702306920762287,
      "loss": 2.0984,
      "theoretical_loss": 4.631955215661932,
      "tokens_seen": 130285568
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000970210631895687,
      "loss": 2.1714,
      "theoretical_loss": 4.6316462177071,
      "tokens_seen": 130351104
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009701905717151455,
      "loss": 2.1332,
      "theoretical_loss": 4.63133741854069,
      "tokens_seen": 130416640
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009701705115346038,
      "loss": 2.0043,
      "theoretical_loss": 4.631028817935018,
      "tokens_seen": 130482176
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009701504513540622,
      "loss": 1.9436,
      "theoretical_loss": 4.630720415662774,
      "tokens_seen": 130547712
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009701303911735205,
      "loss": 1.9561,
      "theoretical_loss": 4.630412211497022,
      "tokens_seen": 130613248
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000970110330992979,
      "loss": 1.8725,
      "theoretical_loss": 4.630104205211202,
      "tokens_seen": 130678784
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009700902708124373,
      "loss": 2.0146,
      "theoretical_loss": 4.629796396579124,
      "tokens_seen": 130744320
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009700702106318957,
      "loss": 1.9803,
      "theoretical_loss": 4.629488785374969,
      "tokens_seen": 130809856
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000970050150451354,
      "loss": 2.1308,
      "theoretical_loss": 4.629181371373292,
      "tokens_seen": 130875392
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009700300902708125,
      "loss": 2.0169,
      "theoretical_loss": 4.628874154349017,
      "tokens_seen": 130940928
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009700100300902708,
      "loss": 1.9983,
      "theoretical_loss": 4.628567134077436,
      "tokens_seen": 131006464
    },
    {
      "epoch": 0.04,
      "objective/train/advantage_avg": 0.49247100949287415,
      "objective/train/docs_used": 54897,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.9558112621307373,
      "objective/train/original_loss": 1.9558112621307373,
      "objective/train/theoretical_loss": 4.628260310334209,
      "objective/train/tokens_used": 151532000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24331486225128174,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05048406124115,
      "objective/train/weighted_lm_loss": 2.0544002056121826,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 4.628260310334209,
      "tokens_seen": 131072000
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009699899699097292,
      "loss": 1.9695,
      "theoretical_loss": 4.628260310334209,
      "tokens_seen": 131072000
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009699699097291876,
      "loss": 2.2511,
      "theoretical_loss": 4.6279536828953685,
      "tokens_seen": 131137536
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000969949849548646,
      "loss": 1.9499,
      "theoretical_loss": 4.627647251537309,
      "tokens_seen": 131203072
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009699297893681044,
      "loss": 1.9427,
      "theoretical_loss": 4.627341016036794,
      "tokens_seen": 131268608
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009699097291875627,
      "loss": 2.0919,
      "theoretical_loss": 4.6270349761709495,
      "tokens_seen": 131334144
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009698896690070212,
      "loss": 1.9728,
      "theoretical_loss": 4.626729131717271,
      "tokens_seen": 131399680
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009698696088264794,
      "loss": 1.9052,
      "theoretical_loss": 4.626423482453612,
      "tokens_seen": 131465216
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009698495486459379,
      "loss": 2.1468,
      "theoretical_loss": 4.626118028158196,
      "tokens_seen": 131530752
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009698294884653962,
      "loss": 1.9891,
      "theoretical_loss": 4.625812768609601,
      "tokens_seen": 131596288
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009698094282848546,
      "loss": 1.9941,
      "theoretical_loss": 4.6255077035867735,
      "tokens_seen": 131661824
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000969789368104313,
      "loss": 2.2123,
      "theoretical_loss": 4.625202832869016,
      "tokens_seen": 131727360
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009697693079237714,
      "loss": 2.0563,
      "theoretical_loss": 4.624898156235994,
      "tokens_seen": 131792896
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009697492477432297,
      "loss": 2.0261,
      "theoretical_loss": 4.62459367346773,
      "tokens_seen": 131858432
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009697291875626882,
      "loss": 2.0436,
      "theoretical_loss": 4.624289384344609,
      "tokens_seen": 131923968
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009697091273821464,
      "loss": 2.1076,
      "theoretical_loss": 4.623985288647368,
      "tokens_seen": 131989504
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009696890672016049,
      "loss": 1.9861,
      "theoretical_loss": 4.623681386157104,
      "tokens_seen": 132055040
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009696690070210632,
      "loss": 2.0016,
      "theoretical_loss": 4.623377676655271,
      "tokens_seen": 132120576
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009696489468405216,
      "loss": 2.0137,
      "theoretical_loss": 4.623074159923678,
      "tokens_seen": 132186112
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00096962888665998,
      "loss": 2.0859,
      "theoretical_loss": 4.622770835744487,
      "tokens_seen": 132251648
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009696088264794383,
      "loss": 2.0853,
      "theoretical_loss": 4.622467703900216,
      "tokens_seen": 132317184
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009695887662988967,
      "loss": 2.1195,
      "theoretical_loss": 4.622164764173735,
      "tokens_seen": 132382720
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009695687061183552,
      "loss": 2.0064,
      "theoretical_loss": 4.621862016348267,
      "tokens_seen": 132448256
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009695486459378134,
      "loss": 2.1665,
      "theoretical_loss": 4.621559460207386,
      "tokens_seen": 132513792
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009695285857572719,
      "loss": 2.141,
      "theoretical_loss": 4.621257095535019,
      "tokens_seen": 132579328
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009695085255767301,
      "loss": 2.0802,
      "theoretical_loss": 4.62095492211544,
      "tokens_seen": 132644864
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009694884653961886,
      "loss": 2.1012,
      "theoretical_loss": 4.620652939733276,
      "tokens_seen": 132710400
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000969468405215647,
      "loss": 2.2375,
      "theoretical_loss": 4.6203511481735,
      "tokens_seen": 132775936
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009694483450351053,
      "loss": 2.2068,
      "theoretical_loss": 4.6200495472214325,
      "tokens_seen": 132841472
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009694282848545637,
      "loss": 2.1315,
      "theoretical_loss": 4.6197481366627455,
      "tokens_seen": 132907008
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009694082246740222,
      "loss": 2.1076,
      "theoretical_loss": 4.6194469162834535,
      "tokens_seen": 132972544
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009693881644934804,
      "loss": 2.0728,
      "theoretical_loss": 4.6191458858699175,
      "tokens_seen": 133038080
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009693681043129389,
      "loss": 2.0247,
      "theoretical_loss": 4.618845045208845,
      "tokens_seen": 133103616
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009693480441323971,
      "loss": 2.0239,
      "theoretical_loss": 4.618544394087287,
      "tokens_seen": 133169152
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009693279839518556,
      "loss": 2.0453,
      "theoretical_loss": 4.618243932292639,
      "tokens_seen": 133234688
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000969307923771314,
      "loss": 2.0592,
      "theoretical_loss": 4.617943659612637,
      "tokens_seen": 133300224
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009692878635907723,
      "loss": 1.944,
      "theoretical_loss": 4.617643575835361,
      "tokens_seen": 133365760
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009692678034102307,
      "loss": 1.8841,
      "theoretical_loss": 4.617343680749233,
      "tokens_seen": 133431296
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009692477432296891,
      "loss": 2.0001,
      "theoretical_loss": 4.617043974143014,
      "tokens_seen": 133496832
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009692276830491474,
      "loss": 1.9382,
      "theoretical_loss": 4.616744455805806,
      "tokens_seen": 133562368
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009692076228686059,
      "loss": 2.1043,
      "theoretical_loss": 4.616445125527051,
      "tokens_seen": 133627904
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009691875626880642,
      "loss": 1.9803,
      "theoretical_loss": 4.61614598309653,
      "tokens_seen": 133693440
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009691675025075226,
      "loss": 2.0213,
      "theoretical_loss": 4.6158470283043584,
      "tokens_seen": 133758976
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009691474423269809,
      "loss": 1.9767,
      "theoretical_loss": 4.615548260940992,
      "tokens_seen": 133824512
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009691273821464394,
      "loss": 2.1192,
      "theoretical_loss": 4.615249680797223,
      "tokens_seen": 133890048
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009691073219658977,
      "loss": 2.1185,
      "theoretical_loss": 4.614951287664179,
      "tokens_seen": 133955584
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009690872617853561,
      "loss": 2.1068,
      "theoretical_loss": 4.614653081333321,
      "tokens_seen": 134021120
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009690672016048144,
      "loss": 2.0312,
      "theoretical_loss": 4.614355061596447,
      "tokens_seen": 134086656
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009690471414242729,
      "loss": 2.1178,
      "theoretical_loss": 4.614057228245686,
      "tokens_seen": 134152192
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009690270812437312,
      "loss": 2.1038,
      "theoretical_loss": 4.613759581073502,
      "tokens_seen": 134217728
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009690070210631896,
      "loss": 1.9887,
      "theoretical_loss": 4.61346211987269,
      "tokens_seen": 134283264
    },
    {
      "epoch": 0.04,
      "objective/train/advantage_avg": 0.49365389347076416,
      "objective/train/docs_used": 56110,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.9194551706314087,
      "objective/train/original_loss": 1.9194552898406982,
      "objective/train/theoretical_loss": 4.613164844436379,
      "objective/train/tokens_used": 154808800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24395570158958435,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506056547164917,
      "objective/train/weighted_lm_loss": 2.0164098739624023,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 4.613164844436379,
      "tokens_seen": 134348800
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009689869608826479,
      "loss": 2.1666,
      "theoretical_loss": 4.613164844436379,
      "tokens_seen": 134348800
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009689669007021064,
      "loss": 2.1292,
      "theoretical_loss": 4.6128677545580254,
      "tokens_seen": 134414336
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009689468405215647,
      "loss": 2.0502,
      "theoretical_loss": 4.612570850031418,
      "tokens_seen": 134479872
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009689267803410231,
      "loss": 2.0157,
      "theoretical_loss": 4.612274130650675,
      "tokens_seen": 134545408
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009689067201604814,
      "loss": 1.9991,
      "theoretical_loss": 4.611977596210242,
      "tokens_seen": 134610944
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009688866599799398,
      "loss": 1.9506,
      "theoretical_loss": 4.6116812465048955,
      "tokens_seen": 134676480
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009688665997993982,
      "loss": 2.0571,
      "theoretical_loss": 4.611385081329736,
      "tokens_seen": 134742016
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009688465396188566,
      "loss": 2.0394,
      "theoretical_loss": 4.611089100480193,
      "tokens_seen": 134807552
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009688264794383149,
      "loss": 1.8767,
      "theoretical_loss": 4.610793303752023,
      "tokens_seen": 134873088
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009688064192577734,
      "loss": 2.0354,
      "theoretical_loss": 4.610497690941305,
      "tokens_seen": 134938624
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009687863590772317,
      "loss": 2.1914,
      "theoretical_loss": 4.610202261844444,
      "tokens_seen": 135004160
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009687662988966901,
      "loss": 2.0961,
      "theoretical_loss": 4.609907016258171,
      "tokens_seen": 135069696
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009687462387161484,
      "loss": 2.1632,
      "theoretical_loss": 4.609611953979537,
      "tokens_seen": 135135232
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009687261785356068,
      "loss": 1.9911,
      "theoretical_loss": 4.609317074805919,
      "tokens_seen": 135200768
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009687061183550652,
      "loss": 1.9436,
      "theoretical_loss": 4.6090223785350135,
      "tokens_seen": 135266304
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009686860581745236,
      "loss": 1.974,
      "theoretical_loss": 4.608727864964839,
      "tokens_seen": 135331840
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009686659979939819,
      "loss": 1.9855,
      "theoretical_loss": 4.608433533893738,
      "tokens_seen": 135397376
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009686459378134404,
      "loss": 2.0362,
      "theoretical_loss": 4.608139385120367,
      "tokens_seen": 135462912
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009686258776328986,
      "loss": 2.0603,
      "theoretical_loss": 4.607845418443706,
      "tokens_seen": 135528448
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009686058174523571,
      "loss": 1.95,
      "theoretical_loss": 4.6075516336630535,
      "tokens_seen": 135593984
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009685857572718155,
      "loss": 2.0118,
      "theoretical_loss": 4.607258030578024,
      "tokens_seen": 135659520
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009685656970912738,
      "loss": 2.0902,
      "theoretical_loss": 4.606964608988553,
      "tokens_seen": 135725056
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009685456369107322,
      "loss": 2.266,
      "theoretical_loss": 4.606671368694888,
      "tokens_seen": 135790592
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009685255767301905,
      "loss": 1.9943,
      "theoretical_loss": 4.6063783094975985,
      "tokens_seen": 135856128
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009685055165496489,
      "loss": 2.0378,
      "theoretical_loss": 4.606085431197563,
      "tokens_seen": 135921664
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009684854563691074,
      "loss": 2.0629,
      "theoretical_loss": 4.6057927335959805,
      "tokens_seen": 135987200
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009684653961885657,
      "loss": 2.0653,
      "theoretical_loss": 4.6055002164943595,
      "tokens_seen": 136052736
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009684453360080241,
      "loss": 2.1102,
      "theoretical_loss": 4.605207879694525,
      "tokens_seen": 136118272
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009684252758274826,
      "loss": 1.9009,
      "theoretical_loss": 4.604915722998614,
      "tokens_seen": 136183808
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009684052156469408,
      "loss": 1.9044,
      "theoretical_loss": 4.604623746209077,
      "tokens_seen": 136249344
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009683851554663993,
      "loss": 2.0011,
      "theoretical_loss": 4.604331949128672,
      "tokens_seen": 136314880
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009683650952858575,
      "loss": 2.173,
      "theoretical_loss": 4.604040331560473,
      "tokens_seen": 136380416
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000968345035105316,
      "loss": 2.0807,
      "theoretical_loss": 4.603748893307861,
      "tokens_seen": 136445952
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009683249749247744,
      "loss": 2.007,
      "theoretical_loss": 4.603457634174527,
      "tokens_seen": 136511488
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009683049147442327,
      "loss": 2.0617,
      "theoretical_loss": 4.603166553964474,
      "tokens_seen": 136577024
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009682848545636911,
      "loss": 2.2422,
      "theoretical_loss": 4.602875652482009,
      "tokens_seen": 136642560
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009682647943831495,
      "loss": 1.9923,
      "theoretical_loss": 4.60258492953175,
      "tokens_seen": 136708096
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009682447342026078,
      "loss": 2.0762,
      "theoretical_loss": 4.60229438491862,
      "tokens_seen": 136773632
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009682246740220663,
      "loss": 2.1774,
      "theoretical_loss": 4.60200401844785,
      "tokens_seen": 136839168
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009682046138415246,
      "loss": 1.9386,
      "theoretical_loss": 4.601713829924977,
      "tokens_seen": 136904704
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000968184553660983,
      "loss": 2.037,
      "theoretical_loss": 4.601423819155842,
      "tokens_seen": 136970240
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009681644934804413,
      "loss": 1.9515,
      "theoretical_loss": 4.601133985946591,
      "tokens_seen": 137035776
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009681444332998998,
      "loss": 2.0307,
      "theoretical_loss": 4.6008443301036746,
      "tokens_seen": 137101312
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009681243731193581,
      "loss": 1.9943,
      "theoretical_loss": 4.600554851433845,
      "tokens_seen": 137166848
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009681043129388165,
      "loss": 2.1094,
      "theoretical_loss": 4.60026554974416,
      "tokens_seen": 137232384
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009680842527582748,
      "loss": 2.0925,
      "theoretical_loss": 4.599976424841977,
      "tokens_seen": 137297920
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009680641925777333,
      "loss": 2.0283,
      "theoretical_loss": 4.5996874765349585,
      "tokens_seen": 137363456
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009680441323971916,
      "loss": 2.0397,
      "theoretical_loss": 4.599398704631062,
      "tokens_seen": 137428992
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00096802407221665,
      "loss": 1.9583,
      "theoretical_loss": 4.599110108938549,
      "tokens_seen": 137494528
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009680040120361083,
      "loss": 1.9251,
      "theoretical_loss": 4.598821689265982,
      "tokens_seen": 137560064
    },
    {
      "epoch": 0.04,
      "objective/train/advantage_avg": 0.49249258637428284,
      "objective/train/docs_used": 57180,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.132284641265869,
      "objective/train/original_loss": 2.1322851181030273,
      "objective/train/theoretical_loss": 4.598533445422221,
      "objective/train/tokens_used": 158085600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24288922548294067,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05048406124115,
      "objective/train/weighted_lm_loss": 2.2398180961608887,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 4.598533445422221,
      "tokens_seen": 137625600
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009679839518555668,
      "loss": 2.1706,
      "theoretical_loss": 4.598533445422221,
      "tokens_seen": 137625600
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009679638916750251,
      "loss": 1.9657,
      "theoretical_loss": 4.598245377216422,
      "tokens_seen": 137691136
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009679438314944835,
      "loss": 1.9385,
      "theoretical_loss": 4.5979574844580435,
      "tokens_seen": 137756672
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009679237713139418,
      "loss": 2.0471,
      "theoretical_loss": 4.597669766956837,
      "tokens_seen": 137822208
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009679037111334002,
      "loss": 2.1599,
      "theoretical_loss": 4.597382224522855,
      "tokens_seen": 137887744
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009678836509528586,
      "loss": 1.9596,
      "theoretical_loss": 4.597094856966441,
      "tokens_seen": 137953280
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000967863590772317,
      "loss": 1.9701,
      "theoretical_loss": 4.596807664098238,
      "tokens_seen": 138018816
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009678435305917753,
      "loss": 2.0581,
      "theoretical_loss": 4.596520645729181,
      "tokens_seen": 138084352
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009678234704112338,
      "loss": 2.1893,
      "theoretical_loss": 4.596233801670502,
      "tokens_seen": 138149888
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009678034102306921,
      "loss": 1.8744,
      "theoretical_loss": 4.595947131733725,
      "tokens_seen": 138215424
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009677833500501505,
      "loss": 2.1235,
      "theoretical_loss": 4.595660635730665,
      "tokens_seen": 138280960
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009677632898696088,
      "loss": 2.1337,
      "theoretical_loss": 4.595374313473435,
      "tokens_seen": 138346496
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009677432296890672,
      "loss": 1.9614,
      "theoretical_loss": 4.595088164774435,
      "tokens_seen": 138412032
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009677231695085256,
      "loss": 1.9998,
      "theoretical_loss": 4.594802189446356,
      "tokens_seen": 138477568
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000967703109327984,
      "loss": 2.0338,
      "theoretical_loss": 4.594516387302184,
      "tokens_seen": 138543104
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009676830491474423,
      "loss": 2.118,
      "theoretical_loss": 4.5942307581551916,
      "tokens_seen": 138608640
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009676629889669008,
      "loss": 2.167,
      "theoretical_loss": 4.593945301818941,
      "tokens_seen": 138674176
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000967642928786359,
      "loss": 1.9812,
      "theoretical_loss": 4.593660018107286,
      "tokens_seen": 138739712
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009676228686058175,
      "loss": 1.9162,
      "theoretical_loss": 4.593374906834365,
      "tokens_seen": 138805248
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009676028084252759,
      "loss": 1.8815,
      "theoretical_loss": 4.593089967814606,
      "tokens_seen": 138870784
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009675827482447342,
      "loss": 2.0387,
      "theoretical_loss": 4.592805200862726,
      "tokens_seen": 138936320
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009675626880641926,
      "loss": 2.0424,
      "theoretical_loss": 4.592520605793725,
      "tokens_seen": 139001856
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009675426278836509,
      "loss": 2.0179,
      "theoretical_loss": 4.592236182422891,
      "tokens_seen": 139067392
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009675225677031093,
      "loss": 1.9477,
      "theoretical_loss": 4.591951930565798,
      "tokens_seen": 139132928
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009675025075225678,
      "loss": 1.9914,
      "theoretical_loss": 4.591667850038302,
      "tokens_seen": 139198464
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000967482447342026,
      "loss": 2.1217,
      "theoretical_loss": 4.591383940656547,
      "tokens_seen": 139264000
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009674623871614845,
      "loss": 2.0834,
      "theoretical_loss": 4.59110020223696,
      "tokens_seen": 139329536
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009674423269809429,
      "loss": 2.0082,
      "theoretical_loss": 4.590816634596247,
      "tokens_seen": 139395072
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009674222668004012,
      "loss": 1.883,
      "theoretical_loss": 4.590533237551401,
      "tokens_seen": 139460608
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009674022066198596,
      "loss": 2.0717,
      "theoretical_loss": 4.590250010919696,
      "tokens_seen": 139526144
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000967382146439318,
      "loss": 1.941,
      "theoretical_loss": 4.5899669545186885,
      "tokens_seen": 139591680
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009673620862587763,
      "loss": 2.1246,
      "theoretical_loss": 4.589684068166212,
      "tokens_seen": 139657216
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009673420260782348,
      "loss": 1.9705,
      "theoretical_loss": 4.589401351680385,
      "tokens_seen": 139722752
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000967321965897693,
      "loss": 2.0961,
      "theoretical_loss": 4.5891188048796,
      "tokens_seen": 139788288
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009673019057171515,
      "loss": 2.0898,
      "theoretical_loss": 4.588836427582536,
      "tokens_seen": 139853824
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009672818455366098,
      "loss": 1.9653,
      "theoretical_loss": 4.588554219608145,
      "tokens_seen": 139919360
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009672617853560682,
      "loss": 2.0289,
      "theoretical_loss": 4.588272180775659,
      "tokens_seen": 139984896
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009672417251755266,
      "loss": 2.0156,
      "theoretical_loss": 4.587990310904587,
      "tokens_seen": 140050432
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000967221664994985,
      "loss": 1.8817,
      "theoretical_loss": 4.587708609814715,
      "tokens_seen": 140115968
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009672016048144433,
      "loss": 2.0011,
      "theoretical_loss": 4.587427077326108,
      "tokens_seen": 140181504
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009671815446339017,
      "loss": 2.0467,
      "theoretical_loss": 4.587145713259102,
      "tokens_seen": 140247040
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00096716148445336,
      "loss": 1.9799,
      "theoretical_loss": 4.586864517434312,
      "tokens_seen": 140312576
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009671414242728185,
      "loss": 2.0688,
      "theoretical_loss": 4.586583489672625,
      "tokens_seen": 140378112
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009671213640922768,
      "loss": 1.9412,
      "theoretical_loss": 4.586302629795206,
      "tokens_seen": 140443648
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009671013039117352,
      "loss": 1.9925,
      "theoretical_loss": 4.5860219376234905,
      "tokens_seen": 140509184
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009670812437311936,
      "loss": 2.0053,
      "theoretical_loss": 4.585741412979187,
      "tokens_seen": 140574720
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000967061183550652,
      "loss": 1.9645,
      "theoretical_loss": 4.585461055684279,
      "tokens_seen": 140640256
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009670411233701103,
      "loss": 2.204,
      "theoretical_loss": 4.58518086556102,
      "tokens_seen": 140705792
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009670210631895687,
      "loss": 1.9912,
      "theoretical_loss": 4.584900842431934,
      "tokens_seen": 140771328
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000967001003009027,
      "loss": 2.0303,
      "theoretical_loss": 4.58462098611982,
      "tokens_seen": 140836864
    },
    {
      "epoch": 0.04,
      "objective/train/advantage_avg": 0.49588945508003235,
      "objective/train/docs_used": 58312,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7948875427246094,
      "objective/train/original_loss": 1.7948875427246094,
      "objective/train/theoretical_loss": 4.584341296447742,
      "objective/train/tokens_used": 161362400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24614021182060242,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508403778076172,
      "objective/train/weighted_lm_loss": 1.8861085176467896,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.584341296447742,
      "tokens_seen": 140902400
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009669809428284855,
      "loss": 1.8735,
      "theoretical_loss": 4.584341296447742,
      "tokens_seen": 140902400
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009669608826479439,
      "loss": 2.0184,
      "theoretical_loss": 4.584061773239039,
      "tokens_seen": 140967936
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009669408224674022,
      "loss": 1.9653,
      "theoretical_loss": 4.583782416317316,
      "tokens_seen": 141033472
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009669207622868606,
      "loss": 1.8443,
      "theoretical_loss": 4.583503225506448,
      "tokens_seen": 141099008
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000966900702106319,
      "loss": 2.1068,
      "theoretical_loss": 4.583224200630577,
      "tokens_seen": 141164544
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009668806419257774,
      "loss": 2.0342,
      "theoretical_loss": 4.582945341514114,
      "tokens_seen": 141230080
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009668605817452357,
      "loss": 2.0212,
      "theoretical_loss": 4.582666647981739,
      "tokens_seen": 141295616
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009668405215646942,
      "loss": 2.0599,
      "theoretical_loss": 4.582388119858393,
      "tokens_seen": 141361152
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009668204613841525,
      "loss": 2.0357,
      "theoretical_loss": 4.582109756969289,
      "tokens_seen": 141426688
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009668004012036109,
      "loss": 1.7923,
      "theoretical_loss": 4.581831559139902,
      "tokens_seen": 141492224
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009667803410230693,
      "loss": 1.9931,
      "theoretical_loss": 4.581553526195974,
      "tokens_seen": 141557760
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009667602808425276,
      "loss": 2.0577,
      "theoretical_loss": 4.581275657963511,
      "tokens_seen": 141623296
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000966740220661986,
      "loss": 2.0293,
      "theoretical_loss": 4.580997954268781,
      "tokens_seen": 141688832
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009667201604814444,
      "loss": 1.9801,
      "theoretical_loss": 4.5807204149383205,
      "tokens_seen": 141754368
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009667001003009027,
      "loss": 2.028,
      "theoretical_loss": 4.580443039798922,
      "tokens_seen": 141819904
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009666800401203612,
      "loss": 2.027,
      "theoretical_loss": 4.580165828677647,
      "tokens_seen": 141885440
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009666599799398194,
      "loss": 2.0958,
      "theoretical_loss": 4.579888781401816,
      "tokens_seen": 141950976
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009666399197592779,
      "loss": 2.0398,
      "theoretical_loss": 4.579611897799008,
      "tokens_seen": 142016512
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009666198595787363,
      "loss": 1.9667,
      "theoretical_loss": 4.57933517769707,
      "tokens_seen": 142082048
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009665997993981946,
      "loss": 2.0251,
      "theoretical_loss": 4.579058620924103,
      "tokens_seen": 142147584
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000966579739217653,
      "loss": 1.9583,
      "theoretical_loss": 4.578782227308471,
      "tokens_seen": 142213120
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009665596790371113,
      "loss": 2.063,
      "theoretical_loss": 4.578505996678796,
      "tokens_seen": 142278656
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009665396188565697,
      "loss": 2.1415,
      "theoretical_loss": 4.578229928863959,
      "tokens_seen": 142344192
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009665195586760282,
      "loss": 2.0813,
      "theoretical_loss": 4.577954023693101,
      "tokens_seen": 142409728
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009664994984954864,
      "loss": 1.9436,
      "theoretical_loss": 4.577678280995619,
      "tokens_seen": 142475264
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009664794383149449,
      "loss": 1.9627,
      "theoretical_loss": 4.577402700601168,
      "tokens_seen": 142540800
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009664593781344033,
      "loss": 2.0339,
      "theoretical_loss": 4.57712728233966,
      "tokens_seen": 142606336
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009664393179538616,
      "loss": 1.9898,
      "theoretical_loss": 4.576852026041261,
      "tokens_seen": 142671872
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00096641925777332,
      "loss": 1.9128,
      "theoretical_loss": 4.5765769315363976,
      "tokens_seen": 142737408
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009663991975927783,
      "loss": 2.0191,
      "theoretical_loss": 4.576301998655747,
      "tokens_seen": 142802944
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009663791374122367,
      "loss": 1.9703,
      "theoretical_loss": 4.576027227230245,
      "tokens_seen": 142868480
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009663590772316952,
      "loss": 2.0606,
      "theoretical_loss": 4.5757526170910765,
      "tokens_seen": 142934016
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009663390170511534,
      "loss": 2.0374,
      "theoretical_loss": 4.575478168069687,
      "tokens_seen": 142999552
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009663189568706119,
      "loss": 2.121,
      "theoretical_loss": 4.575203879997771,
      "tokens_seen": 143065088
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009662988966900702,
      "loss": 2.0917,
      "theoretical_loss": 4.574929752707274,
      "tokens_seen": 143130624
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009662788365095286,
      "loss": 1.9206,
      "theoretical_loss": 4.5746557860304,
      "tokens_seen": 143196160
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000966258776328987,
      "loss": 2.0839,
      "theoretical_loss": 4.574381979799599,
      "tokens_seen": 143261696
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009662387161484454,
      "loss": 2.0296,
      "theoretical_loss": 4.574108333847575,
      "tokens_seen": 143327232
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009662186559679037,
      "loss": 1.9963,
      "theoretical_loss": 4.573834848007284,
      "tokens_seen": 143392768
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009661985957873621,
      "loss": 2.0374,
      "theoretical_loss": 4.573561522111927,
      "tokens_seen": 143458304
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009661785356068204,
      "loss": 2.0905,
      "theoretical_loss": 4.573288355994963,
      "tokens_seen": 143523840
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009661584754262789,
      "loss": 1.8426,
      "theoretical_loss": 4.573015349490094,
      "tokens_seen": 143589376
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009661384152457372,
      "loss": 2.0254,
      "theoretical_loss": 4.572742502431272,
      "tokens_seen": 143654912
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009661183550651956,
      "loss": 2.0699,
      "theoretical_loss": 4.572469814652698,
      "tokens_seen": 143720448
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000966098294884654,
      "loss": 2.0789,
      "theoretical_loss": 4.572197285988824,
      "tokens_seen": 143785984
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009660782347041124,
      "loss": 1.925,
      "theoretical_loss": 4.571924916274344,
      "tokens_seen": 143851520
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009660581745235707,
      "loss": 1.8722,
      "theoretical_loss": 4.571652705344202,
      "tokens_seen": 143917056
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009660381143430291,
      "loss": 2.0093,
      "theoretical_loss": 4.571380653033588,
      "tokens_seen": 143982592
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009660180541624874,
      "loss": 2.0186,
      "theoretical_loss": 4.571108759177939,
      "tokens_seen": 144048128
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009659979939819459,
      "loss": 2.0186,
      "theoretical_loss": 4.570837023612935,
      "tokens_seen": 144113664
    },
    {
      "epoch": 0.04,
      "objective/train/advantage_avg": 0.4903619587421417,
      "objective/train/docs_used": 59583,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8303672075271606,
      "objective/train/original_loss": 1.830367088317871,
      "objective/train/theoretical_loss": 4.570565446174504,
      "objective/train/tokens_used": 164639200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2413649708032608,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502631664276123,
      "objective/train/weighted_lm_loss": 1.9222012758255005,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.570565446174504,
      "tokens_seen": 144179200
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009659779338014042,
      "loss": 2.0127,
      "theoretical_loss": 4.570565446174504,
      "tokens_seen": 144179200
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009659578736208626,
      "loss": 1.9161,
      "theoretical_loss": 4.570294026698815,
      "tokens_seen": 144244736
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009659378134403209,
      "loss": 2.0468,
      "theoretical_loss": 4.570022765022285,
      "tokens_seen": 144310272
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009659177532597794,
      "loss": 2.0074,
      "theoretical_loss": 4.569751660981572,
      "tokens_seen": 144375808
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009658976930792377,
      "loss": 1.9636,
      "theoretical_loss": 4.569480714413578,
      "tokens_seen": 144441344
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009658776328986961,
      "loss": 2.0375,
      "theoretical_loss": 4.569209925155445,
      "tokens_seen": 144506880
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009658575727181545,
      "loss": 1.9634,
      "theoretical_loss": 4.568939293044564,
      "tokens_seen": 144572416
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009658375125376129,
      "loss": 2.0068,
      "theoretical_loss": 4.56866881791856,
      "tokens_seen": 144637952
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009658174523570712,
      "loss": 2.078,
      "theoretical_loss": 4.568398499615305,
      "tokens_seen": 144703488
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009657973921765297,
      "loss": 2.0601,
      "theoretical_loss": 4.568128337972907,
      "tokens_seen": 144769024
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009657773319959879,
      "loss": 2.0646,
      "theoretical_loss": 4.567858332829719,
      "tokens_seen": 144834560
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009657572718154464,
      "loss": 1.9944,
      "theoretical_loss": 4.567588484024331,
      "tokens_seen": 144900096
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009657372116349047,
      "loss": 2.0121,
      "theoretical_loss": 4.56731879139557,
      "tokens_seen": 144965632
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009657171514543631,
      "loss": 2.0377,
      "theoretical_loss": 4.567049254782509,
      "tokens_seen": 145031168
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009656970912738215,
      "loss": 1.9483,
      "theoretical_loss": 4.566779874024453,
      "tokens_seen": 145096704
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009656770310932798,
      "loss": 1.8921,
      "theoretical_loss": 4.566510648960948,
      "tokens_seen": 145162240
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009656569709127382,
      "loss": 1.9893,
      "theoretical_loss": 4.566241579431776,
      "tokens_seen": 145227776
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009656369107321967,
      "loss": 1.9413,
      "theoretical_loss": 4.565972665276957,
      "tokens_seen": 145293312
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009656168505516549,
      "loss": 2.1933,
      "theoretical_loss": 4.56570390633675,
      "tokens_seen": 145358848
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009655967903711134,
      "loss": 1.882,
      "theoretical_loss": 4.565435302451645,
      "tokens_seen": 145424384
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009655767301905716,
      "loss": 2.0421,
      "theoretical_loss": 4.565166853462371,
      "tokens_seen": 145489920
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009655566700100301,
      "loss": 2.0148,
      "theoretical_loss": 4.56489855920989,
      "tokens_seen": 145555456
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009655366098294885,
      "loss": 2.1477,
      "theoretical_loss": 4.564630419535405,
      "tokens_seen": 145620992
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009655165496489468,
      "loss": 1.9786,
      "theoretical_loss": 4.564362434280344,
      "tokens_seen": 145686528
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009654964894684052,
      "loss": 2.0122,
      "theoretical_loss": 4.564094603286375,
      "tokens_seen": 145752064
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009654764292878637,
      "loss": 2.0539,
      "theoretical_loss": 4.5638269263954,
      "tokens_seen": 145817600
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000965456369107322,
      "loss": 2.0813,
      "theoretical_loss": 4.563559403449551,
      "tokens_seen": 145883136
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009654363089267804,
      "loss": 1.826,
      "theoretical_loss": 4.563292034291193,
      "tokens_seen": 145948672
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009654162487462387,
      "loss": 1.9389,
      "theoretical_loss": 4.5630248187629245,
      "tokens_seen": 146014208
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009653961885656971,
      "loss": 1.9886,
      "theoretical_loss": 4.562757756707574,
      "tokens_seen": 146079744
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009653761283851556,
      "loss": 2.0087,
      "theoretical_loss": 4.5624908479682045,
      "tokens_seen": 146145280
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009653560682046138,
      "loss": 1.9831,
      "theoretical_loss": 4.562224092388105,
      "tokens_seen": 146210816
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009653360080240723,
      "loss": 1.9913,
      "theoretical_loss": 4.561957489810798,
      "tokens_seen": 146276352
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009653159478435306,
      "loss": 1.9859,
      "theoretical_loss": 4.561691040080035,
      "tokens_seen": 146341888
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000965295887662989,
      "loss": 1.9011,
      "theoretical_loss": 4.561424743039797,
      "tokens_seen": 146407424
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009652758274824474,
      "loss": 2.0157,
      "theoretical_loss": 4.561158598534295,
      "tokens_seen": 146472960
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009652557673019058,
      "loss": 1.9363,
      "theoretical_loss": 4.5608926064079665,
      "tokens_seen": 146538496
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009652357071213641,
      "loss": 2.0811,
      "theoretical_loss": 4.560626766505479,
      "tokens_seen": 146604032
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009652156469408225,
      "loss": 1.9238,
      "theoretical_loss": 4.5603610786717255,
      "tokens_seen": 146669568
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009651955867602808,
      "loss": 2.1267,
      "theoretical_loss": 4.56009554275183,
      "tokens_seen": 146735104
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009651755265797393,
      "loss": 1.9993,
      "theoretical_loss": 4.559830158591139,
      "tokens_seen": 146800640
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009651554663991976,
      "loss": 1.9902,
      "theoretical_loss": 4.559564926035229,
      "tokens_seen": 146866176
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000965135406218656,
      "loss": 1.8406,
      "theoretical_loss": 4.5592998449299,
      "tokens_seen": 146931712
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009651153460381144,
      "loss": 1.9684,
      "theoretical_loss": 4.559034915121179,
      "tokens_seen": 146997248
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009650952858575728,
      "loss": 1.9249,
      "theoretical_loss": 4.558770136455316,
      "tokens_seen": 147062784
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009650752256770311,
      "loss": 2.013,
      "theoretical_loss": 4.55850550877879,
      "tokens_seen": 147128320
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009650551654964895,
      "loss": 2.001,
      "theoretical_loss": 4.558241031938298,
      "tokens_seen": 147193856
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009650351053159478,
      "loss": 1.8982,
      "theoretical_loss": 4.557976705780766,
      "tokens_seen": 147259392
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009650150451354063,
      "loss": 1.9986,
      "theoretical_loss": 4.557712530153342,
      "tokens_seen": 147324928
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009649949849548646,
      "loss": 1.932,
      "theoretical_loss": 4.557448504903396,
      "tokens_seen": 147390464
    },
    {
      "epoch": 0.04,
      "objective/train/advantage_avg": 0.493988960981369,
      "objective/train/docs_used": 60747,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7616697549819946,
      "objective/train/original_loss": 1.761669635772705,
      "objective/train/theoretical_loss": 4.557184629878522,
      "objective/train/tokens_used": 167916000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2443990260362625,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050641417503357,
      "objective/train/weighted_lm_loss": 1.8508374691009521,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 4.557184629878522,
      "tokens_seen": 147456000
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000964974924774323,
      "loss": 1.8862,
      "theoretical_loss": 4.557184629878522,
      "tokens_seen": 147456000
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009649548645937813,
      "loss": 1.9551,
      "theoretical_loss": 4.556920904926534,
      "tokens_seen": 147521536
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009649348044132398,
      "loss": 1.9787,
      "theoretical_loss": 4.556657329895469,
      "tokens_seen": 147587072
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009649147442326981,
      "loss": 2.0888,
      "theoretical_loss": 4.556393904633584,
      "tokens_seen": 147652608
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009648946840521565,
      "loss": 1.9027,
      "theoretical_loss": 4.5561306289893615,
      "tokens_seen": 147718144
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009648746238716149,
      "loss": 1.9725,
      "theoretical_loss": 4.555867502811498,
      "tokens_seen": 147783680
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009648545636910733,
      "loss": 2.0147,
      "theoretical_loss": 4.5556045259489135,
      "tokens_seen": 147849216
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009648345035105316,
      "loss": 2.0284,
      "theoretical_loss": 4.555341698250748,
      "tokens_seen": 147914752
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00096481444332999,
      "loss": 2.0678,
      "theoretical_loss": 4.555079019566359,
      "tokens_seen": 147980288
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009647943831494483,
      "loss": 1.9821,
      "theoretical_loss": 4.554816489745324,
      "tokens_seen": 148045824
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009647743229689068,
      "loss": 1.9895,
      "theoretical_loss": 4.554554108637437,
      "tokens_seen": 148111360
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009647542627883651,
      "loss": 2.0104,
      "theoretical_loss": 4.554291876092713,
      "tokens_seen": 148176896
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009647342026078235,
      "loss": 1.8984,
      "theoretical_loss": 4.554029791961382,
      "tokens_seen": 148242432
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009647141424272819,
      "loss": 1.936,
      "theoretical_loss": 4.5537678560938915,
      "tokens_seen": 148307968
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009646940822467402,
      "loss": 2.0408,
      "theoretical_loss": 4.553506068340907,
      "tokens_seen": 148373504
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0009646740220661986,
      "loss": 1.88,
      "theoretical_loss": 4.55324442855331,
      "tokens_seen": 148439040
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009646539618856571,
      "loss": 1.8749,
      "theoretical_loss": 4.552982936582194,
      "tokens_seen": 148504576
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009646339017051153,
      "loss": 2.0451,
      "theoretical_loss": 4.552721592278875,
      "tokens_seen": 148570112
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009646138415245738,
      "loss": 1.9803,
      "theoretical_loss": 4.552460395494878,
      "tokens_seen": 148635648
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000964593781344032,
      "loss": 1.9701,
      "theoretical_loss": 4.552199346081946,
      "tokens_seen": 148701184
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009645737211634905,
      "loss": 2.0254,
      "theoretical_loss": 4.551938443892035,
      "tokens_seen": 148766720
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009645536609829489,
      "loss": 1.8414,
      "theoretical_loss": 4.551677688777316,
      "tokens_seen": 148832256
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009645336008024072,
      "loss": 2.0927,
      "theoretical_loss": 4.55141708059017,
      "tokens_seen": 148897792
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009645135406218656,
      "loss": 1.9209,
      "theoretical_loss": 4.551156619183196,
      "tokens_seen": 148963328
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009644934804413241,
      "loss": 1.7724,
      "theoretical_loss": 4.5508963044092035,
      "tokens_seen": 149028864
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009644734202607823,
      "loss": 2.0814,
      "theoretical_loss": 4.550636136121211,
      "tokens_seen": 149094400
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009644533600802408,
      "loss": 1.8497,
      "theoretical_loss": 4.5503761141724555,
      "tokens_seen": 149159936
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000964433299899699,
      "loss": 1.8274,
      "theoretical_loss": 4.550116238416379,
      "tokens_seen": 149225472
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009644132397191575,
      "loss": 1.9875,
      "theoretical_loss": 4.54985650870664,
      "tokens_seen": 149291008
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009643931795386159,
      "loss": 1.929,
      "theoretical_loss": 4.549596924897102,
      "tokens_seen": 149356544
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009643731193580742,
      "loss": 1.9954,
      "theoretical_loss": 4.549337486841843,
      "tokens_seen": 149422080
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009643530591775326,
      "loss": 1.8849,
      "theoretical_loss": 4.54907819439515,
      "tokens_seen": 149487616
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000964332998996991,
      "loss": 1.9833,
      "theoretical_loss": 4.54881904741152,
      "tokens_seen": 149553152
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009643129388164493,
      "loss": 1.8151,
      "theoretical_loss": 4.548560045745656,
      "tokens_seen": 149618688
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009642928786359078,
      "loss": 2.1873,
      "theoretical_loss": 4.548301189252473,
      "tokens_seen": 149684224
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000964272818455366,
      "loss": 1.9426,
      "theoretical_loss": 4.548042477787093,
      "tokens_seen": 149749760
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009642527582748245,
      "loss": 2.0007,
      "theoretical_loss": 4.547783911204846,
      "tokens_seen": 149815296
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009642326980942828,
      "loss": 2.0073,
      "theoretical_loss": 4.54752548936127,
      "tokens_seen": 149880832
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009642126379137412,
      "loss": 1.9919,
      "theoretical_loss": 4.54726721211211,
      "tokens_seen": 149946368
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009641925777331996,
      "loss": 1.9581,
      "theoretical_loss": 4.547009079313316,
      "tokens_seen": 150011904
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000964172517552658,
      "loss": 1.9835,
      "theoretical_loss": 4.546751090821047,
      "tokens_seen": 150077440
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009641524573721163,
      "loss": 1.9467,
      "theoretical_loss": 4.546493246491666,
      "tokens_seen": 150142976
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009641323971915748,
      "loss": 1.88,
      "theoretical_loss": 4.546235546181743,
      "tokens_seen": 150208512
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009641123370110331,
      "loss": 2.0288,
      "theoretical_loss": 4.545977989748052,
      "tokens_seen": 150274048
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009640922768304915,
      "loss": 1.8078,
      "theoretical_loss": 4.5457205770475735,
      "tokens_seen": 150339584
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009640722166499498,
      "loss": 1.8783,
      "theoretical_loss": 4.5454633079374895,
      "tokens_seen": 150405120
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009640521564694082,
      "loss": 1.853,
      "theoretical_loss": 4.545206182275189,
      "tokens_seen": 150470656
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009640320962888666,
      "loss": 1.8686,
      "theoretical_loss": 4.544949199918264,
      "tokens_seen": 150536192
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000964012036108325,
      "loss": 1.9122,
      "theoretical_loss": 4.544692360724509,
      "tokens_seen": 150601728
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009639919759277833,
      "loss": 1.8843,
      "theoretical_loss": 4.544435664551921,
      "tokens_seen": 150667264
    },
    {
      "epoch": 0.05,
      "objective/train/advantage_avg": 0.49405166506767273,
      "objective/train/docs_used": 61752,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8966041803359985,
      "objective/train/original_loss": 1.8966041803359985,
      "objective/train/theoretical_loss": 4.5441791112587016,
      "objective/train/tokens_used": 171192800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2443924993276596,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506476163864136,
      "objective/train/weighted_lm_loss": 1.9925545454025269,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 4.5441791112587016,
      "tokens_seen": 150732800
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009639719157472417,
      "loss": 1.9816,
      "theoretical_loss": 4.5441791112587016,
      "tokens_seen": 150732800
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009639518555667002,
      "loss": 2.1063,
      "theoretical_loss": 4.543922700703251,
      "tokens_seen": 150798336
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009639317953861585,
      "loss": 2.0471,
      "theoretical_loss": 4.5436664327441765,
      "tokens_seen": 150863872
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009639117352056169,
      "loss": 2.0619,
      "theoretical_loss": 4.543410307240281,
      "tokens_seen": 150929408
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009638916750250753,
      "loss": 1.9035,
      "theoretical_loss": 4.5431543240505725,
      "tokens_seen": 150994944
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009638716148445337,
      "loss": 1.9168,
      "theoretical_loss": 4.542898483034259,
      "tokens_seen": 151060480
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000963851554663992,
      "loss": 1.9746,
      "theoretical_loss": 4.542642784050747,
      "tokens_seen": 151126016
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009638314944834505,
      "loss": 2.0691,
      "theoretical_loss": 4.542387226959644,
      "tokens_seen": 151191552
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009638114343029087,
      "loss": 1.9151,
      "theoretical_loss": 4.5421318116207585,
      "tokens_seen": 151257088
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009637913741223672,
      "loss": 1.8876,
      "theoretical_loss": 4.541876537894095,
      "tokens_seen": 151322624
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009637713139418255,
      "loss": 1.8877,
      "theoretical_loss": 4.54162140563986,
      "tokens_seen": 151388160
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009637512537612839,
      "loss": 1.8514,
      "theoretical_loss": 4.541366414718456,
      "tokens_seen": 151453696
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009637311935807423,
      "loss": 1.9069,
      "theoretical_loss": 4.541111564990485,
      "tokens_seen": 151519232
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009637111334002006,
      "loss": 1.9673,
      "theoretical_loss": 4.540856856316747,
      "tokens_seen": 151584768
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000963691073219659,
      "loss": 1.8871,
      "theoretical_loss": 4.540602288558238,
      "tokens_seen": 151650304
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009636710130391175,
      "loss": 1.8173,
      "theoretical_loss": 4.5403478615761514,
      "tokens_seen": 151715840
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009636509528585757,
      "loss": 2.0094,
      "theoretical_loss": 4.540093575231879,
      "tokens_seen": 151781376
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009636308926780342,
      "loss": 1.8458,
      "theoretical_loss": 4.539839429387005,
      "tokens_seen": 151846912
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009636108324974924,
      "loss": 1.9317,
      "theoretical_loss": 4.539585423903313,
      "tokens_seen": 151912448
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009635907723169509,
      "loss": 1.9469,
      "theoretical_loss": 4.539331558642782,
      "tokens_seen": 151977984
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009635707121364093,
      "loss": 1.8159,
      "theoretical_loss": 4.539077833467582,
      "tokens_seen": 152043520
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009635506519558676,
      "loss": 2.0562,
      "theoretical_loss": 4.5388242482400845,
      "tokens_seen": 152109056
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000963530591775326,
      "loss": 1.9007,
      "theoretical_loss": 4.53857080282285,
      "tokens_seen": 152174592
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009635105315947845,
      "loss": 1.8905,
      "theoretical_loss": 4.538317497078634,
      "tokens_seen": 152240128
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009634904714142427,
      "loss": 1.8761,
      "theoretical_loss": 4.538064330870389,
      "tokens_seen": 152305664
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009634704112337012,
      "loss": 2.1206,
      "theoretical_loss": 4.537811304061257,
      "tokens_seen": 152371200
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009634503510531594,
      "loss": 1.8516,
      "theoretical_loss": 4.537558416514575,
      "tokens_seen": 152436736
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009634302908726179,
      "loss": 1.955,
      "theoretical_loss": 4.537305668093871,
      "tokens_seen": 152502272
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009634102306920763,
      "loss": 2.0363,
      "theoretical_loss": 4.537053058662869,
      "tokens_seen": 152567808
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009633901705115346,
      "loss": 2.0132,
      "theoretical_loss": 4.53680058808548,
      "tokens_seen": 152633344
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000963370110330993,
      "loss": 1.9282,
      "theoretical_loss": 4.53654825622581,
      "tokens_seen": 152698880
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009633500501504514,
      "loss": 1.9805,
      "theoretical_loss": 4.536296062948156,
      "tokens_seen": 152764416
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009633299899699097,
      "loss": 1.8875,
      "theoretical_loss": 4.536044008117005,
      "tokens_seen": 152829952
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009633099297893682,
      "loss": 2.0565,
      "theoretical_loss": 4.535792091597035,
      "tokens_seen": 152895488
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009632898696088264,
      "loss": 2.0856,
      "theoretical_loss": 4.535540313253112,
      "tokens_seen": 152961024
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009632698094282849,
      "loss": 1.9599,
      "theoretical_loss": 4.535288672950296,
      "tokens_seen": 153026560
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009632497492477432,
      "loss": 1.9393,
      "theoretical_loss": 4.535037170553833,
      "tokens_seen": 153092096
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009632296890672016,
      "loss": 1.9247,
      "theoretical_loss": 4.534785805929161,
      "tokens_seen": 153157632
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00096320962888666,
      "loss": 1.8164,
      "theoretical_loss": 4.534534578941903,
      "tokens_seen": 153223168
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009631895687061184,
      "loss": 2.0316,
      "theoretical_loss": 4.534283489457875,
      "tokens_seen": 153288704
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009631695085255767,
      "loss": 1.9613,
      "theoretical_loss": 4.534032537343078,
      "tokens_seen": 153354240
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009631494483450352,
      "loss": 1.8766,
      "theoretical_loss": 4.5337817224637025,
      "tokens_seen": 153419776
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009631293881644935,
      "loss": 2.105,
      "theoretical_loss": 4.533531044686123,
      "tokens_seen": 153485312
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009631093279839519,
      "loss": 1.89,
      "theoretical_loss": 4.533280503876906,
      "tokens_seen": 153550848
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009630892678034102,
      "loss": 2.0764,
      "theoretical_loss": 4.533030099902803,
      "tokens_seen": 153616384
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009630692076228686,
      "loss": 1.9391,
      "theoretical_loss": 4.532779832630748,
      "tokens_seen": 153681920
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000963049147442327,
      "loss": 2.0657,
      "theoretical_loss": 4.532529701927868,
      "tokens_seen": 153747456
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009630290872617854,
      "loss": 1.9845,
      "theoretical_loss": 4.53227970766147,
      "tokens_seen": 153812992
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009630090270812437,
      "loss": 1.9766,
      "theoretical_loss": 4.53202984969905,
      "tokens_seen": 153878528
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009629889669007021,
      "loss": 1.949,
      "theoretical_loss": 4.531780127908286,
      "tokens_seen": 153944064
    },
    {
      "epoch": 0.05,
      "objective/train/advantage_avg": 0.49506837129592896,
      "objective/train/docs_used": 63027,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.0724551677703857,
      "objective/train/original_loss": 2.0724549293518066,
      "objective/train/theoretical_loss": 4.531530542157043,
      "objective/train/tokens_used": 174469600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24586887657642365,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507569313049316,
      "objective/train/weighted_lm_loss": 2.1776838302612305,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.531530542157043,
      "tokens_seen": 154009600
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009629689067201605,
      "loss": 2.0624,
      "theoretical_loss": 4.531530542157043,
      "tokens_seen": 154009600
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009629488465396189,
      "loss": 1.997,
      "theoretical_loss": 4.53128109231337,
      "tokens_seen": 154075136
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009629287863590772,
      "loss": 2.1624,
      "theoretical_loss": 4.531031778245499,
      "tokens_seen": 154140672
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009629087261785357,
      "loss": 1.9835,
      "theoretical_loss": 4.530782599821848,
      "tokens_seen": 154206208
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009628886659979939,
      "loss": 1.8497,
      "theoretical_loss": 4.5305335569110134,
      "tokens_seen": 154271744
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009628686058174524,
      "loss": 1.9732,
      "theoretical_loss": 4.530284649381782,
      "tokens_seen": 154337280
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009628485456369107,
      "loss": 2.0882,
      "theoretical_loss": 4.530035877103115,
      "tokens_seen": 154402816
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009628284854563691,
      "loss": 2.0315,
      "theoretical_loss": 4.529787239944165,
      "tokens_seen": 154468352
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009628084252758275,
      "loss": 1.9269,
      "theoretical_loss": 4.529538737774259,
      "tokens_seen": 154533888
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009627883650952859,
      "loss": 2.0243,
      "theoretical_loss": 4.52929037046291,
      "tokens_seen": 154599424
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009627683049147442,
      "loss": 1.9704,
      "theoretical_loss": 4.529042137879809,
      "tokens_seen": 154664960
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009627482447342027,
      "loss": 1.9537,
      "theoretical_loss": 4.528794039894832,
      "tokens_seen": 154730496
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009627281845536609,
      "loss": 1.9413,
      "theoretical_loss": 4.528546076378033,
      "tokens_seen": 154796032
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009627081243731194,
      "loss": 1.8682,
      "theoretical_loss": 4.528298247199649,
      "tokens_seen": 154861568
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009626880641925777,
      "loss": 1.9414,
      "theoretical_loss": 4.528050552230092,
      "tokens_seen": 154927104
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009626680040120361,
      "loss": 1.8703,
      "theoretical_loss": 4.52780299133996,
      "tokens_seen": 154992640
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009626479438314945,
      "loss": 2.079,
      "theoretical_loss": 4.527555564400026,
      "tokens_seen": 155058176
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009626278836509528,
      "loss": 2.0089,
      "theoretical_loss": 4.527308271281243,
      "tokens_seen": 155123712
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009626078234704112,
      "loss": 1.9527,
      "theoretical_loss": 4.527061111854746,
      "tokens_seen": 155189248
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009625877632898697,
      "loss": 2.0116,
      "theoretical_loss": 4.526814085991843,
      "tokens_seen": 155254784
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009625677031093279,
      "loss": 2.0765,
      "theoretical_loss": 4.5265671935640235,
      "tokens_seen": 155320320
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009625476429287864,
      "loss": 1.9764,
      "theoretical_loss": 4.526320434442955,
      "tokens_seen": 155385856
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009625275827482446,
      "loss": 1.9078,
      "theoretical_loss": 4.526073808500481,
      "tokens_seen": 155451392
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009625075225677031,
      "loss": 1.8425,
      "theoretical_loss": 4.525827315608623,
      "tokens_seen": 155516928
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009624874623871615,
      "loss": 1.7777,
      "theoretical_loss": 4.5255809556395805,
      "tokens_seen": 155582464
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009624674022066198,
      "loss": 2.0199,
      "theoretical_loss": 4.525334728465727,
      "tokens_seen": 155648000
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009624473420260783,
      "loss": 2.0546,
      "theoretical_loss": 4.525088633959613,
      "tokens_seen": 155713536
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009624272818455367,
      "loss": 1.8745,
      "theoretical_loss": 4.524842671993966,
      "tokens_seen": 155779072
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000962407221664995,
      "loss": 1.9245,
      "theoretical_loss": 4.524596842441689,
      "tokens_seen": 155844608
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009623871614844534,
      "loss": 2.1329,
      "theoretical_loss": 4.524351145175858,
      "tokens_seen": 155910144
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009623671013039118,
      "loss": 1.8084,
      "theoretical_loss": 4.524105580069728,
      "tokens_seen": 155975680
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009623470411233701,
      "loss": 1.9775,
      "theoretical_loss": 4.523860146996725,
      "tokens_seen": 156041216
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009623269809428286,
      "loss": 1.8121,
      "theoretical_loss": 4.5236148458304495,
      "tokens_seen": 156106752
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009623069207622868,
      "loss": 1.9081,
      "theoretical_loss": 4.523369676444679,
      "tokens_seen": 156172288
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009622868605817453,
      "loss": 1.9997,
      "theoretical_loss": 4.523124638713361,
      "tokens_seen": 156237824
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009622668004012036,
      "loss": 2.0708,
      "theoretical_loss": 4.52287973251062,
      "tokens_seen": 156303360
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000962246740220662,
      "loss": 1.901,
      "theoretical_loss": 4.52263495771075,
      "tokens_seen": 156368896
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009622266800401204,
      "loss": 1.9521,
      "theoretical_loss": 4.522390314188221,
      "tokens_seen": 156434432
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009622066198595788,
      "loss": 1.9751,
      "theoretical_loss": 4.522145801817673,
      "tokens_seen": 156499968
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009621865596790371,
      "loss": 1.9347,
      "theoretical_loss": 4.5219014204739185,
      "tokens_seen": 156565504
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009621664994984956,
      "loss": 1.7942,
      "theoretical_loss": 4.521657170031944,
      "tokens_seen": 156631040
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009621464393179539,
      "loss": 1.9491,
      "theoretical_loss": 4.521413050366905,
      "tokens_seen": 156696576
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009621263791374123,
      "loss": 1.8449,
      "theoretical_loss": 4.521169061354129,
      "tokens_seen": 156762112
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009621063189568706,
      "loss": 1.8015,
      "theoretical_loss": 4.520925202869115,
      "tokens_seen": 156827648
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000962086258776329,
      "loss": 1.9523,
      "theoretical_loss": 4.520681474787532,
      "tokens_seen": 156893184
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009620661985957874,
      "loss": 1.9431,
      "theoretical_loss": 4.520437876985218,
      "tokens_seen": 156958720
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009620461384152458,
      "loss": 1.8395,
      "theoretical_loss": 4.520194409338185,
      "tokens_seen": 157024256
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009620260782347041,
      "loss": 2.0112,
      "theoretical_loss": 4.519951071722611,
      "tokens_seen": 157089792
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009620060180541625,
      "loss": 1.8646,
      "theoretical_loss": 4.519707864014844,
      "tokens_seen": 157155328
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009619859578736209,
      "loss": 1.8881,
      "theoretical_loss": 4.519464786091401,
      "tokens_seen": 157220864
    },
    {
      "epoch": 0.05,
      "objective/train/advantage_avg": 0.4935777187347412,
      "objective/train/docs_used": 64097,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 2.037578582763672,
      "objective/train/original_loss": 2.037578582763672,
      "objective/train/theoretical_loss": 4.519221837828971,
      "objective/train/tokens_used": 177746400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24397633969783783,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505980253219604,
      "objective/train/weighted_lm_loss": 2.1407289505004883,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 4.519221837828971,
      "tokens_seen": 157286400
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009619658976930793,
      "loss": 2.1463,
      "theoretical_loss": 4.519221837828971,
      "tokens_seen": 157286400
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009619458375125376,
      "loss": 1.8261,
      "theoretical_loss": 4.518979019104407,
      "tokens_seen": 157351936
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009619257773319961,
      "loss": 2.0235,
      "theoretical_loss": 4.518736329794732,
      "tokens_seen": 157417472
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009619057171514543,
      "loss": 1.9399,
      "theoretical_loss": 4.5184937697771375,
      "tokens_seen": 157483008
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009618856569709128,
      "loss": 2.0169,
      "theoretical_loss": 4.51825133892898,
      "tokens_seen": 157548544
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009618655967903711,
      "loss": 1.9519,
      "theoretical_loss": 4.51800903712779,
      "tokens_seen": 157614080
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009618455366098295,
      "loss": 1.9318,
      "theoretical_loss": 4.517766864251255,
      "tokens_seen": 157679616
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009618254764292879,
      "loss": 1.878,
      "theoretical_loss": 4.517524820177239,
      "tokens_seen": 157745152
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009618054162487463,
      "loss": 1.8226,
      "theoretical_loss": 4.517282904783764,
      "tokens_seen": 157810688
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009617853560682046,
      "loss": 1.9889,
      "theoretical_loss": 4.5170411179490255,
      "tokens_seen": 157876224
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009617652958876631,
      "loss": 1.9731,
      "theoretical_loss": 4.516799459551378,
      "tokens_seen": 157941760
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009617452357071213,
      "loss": 1.8943,
      "theoretical_loss": 4.516557929469347,
      "tokens_seen": 158007296
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009617251755265798,
      "loss": 1.9482,
      "theoretical_loss": 4.516316527581621,
      "tokens_seen": 158072832
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009617051153460381,
      "loss": 1.9452,
      "theoretical_loss": 4.5160752537670525,
      "tokens_seen": 158138368
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009616850551654965,
      "loss": 1.9644,
      "theoretical_loss": 4.515834107904659,
      "tokens_seen": 158203904
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009616649949849549,
      "loss": 2.0442,
      "theoretical_loss": 4.515593089873624,
      "tokens_seen": 158269440
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009616449348044132,
      "loss": 1.9659,
      "theoretical_loss": 4.515352199553295,
      "tokens_seen": 158334976
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009616248746238716,
      "loss": 2.0096,
      "theoretical_loss": 4.51511143682318,
      "tokens_seen": 158400512
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009616048144433301,
      "loss": 1.9158,
      "theoretical_loss": 4.514870801562955,
      "tokens_seen": 158466048
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009615847542627883,
      "loss": 1.904,
      "theoretical_loss": 4.514630293652454,
      "tokens_seen": 158531584
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009615646940822468,
      "loss": 1.9249,
      "theoretical_loss": 4.514389912971679,
      "tokens_seen": 158597120
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000961544633901705,
      "loss": 1.8967,
      "theoretical_loss": 4.5141496594007915,
      "tokens_seen": 158662656
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009615245737211635,
      "loss": 1.9135,
      "theoretical_loss": 4.513909532820117,
      "tokens_seen": 158728192
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009615045135406219,
      "loss": 1.9641,
      "theoretical_loss": 4.513669533110141,
      "tokens_seen": 158793728
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009614844533600802,
      "loss": 1.9775,
      "theoretical_loss": 4.513429660151513,
      "tokens_seen": 158859264
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009614643931795386,
      "loss": 1.9063,
      "theoretical_loss": 4.513189913825043,
      "tokens_seen": 158924800
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009614443329989971,
      "loss": 1.9086,
      "theoretical_loss": 4.5129502940117,
      "tokens_seen": 158990336
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009614242728184553,
      "loss": 1.803,
      "theoretical_loss": 4.512710800592618,
      "tokens_seen": 159055872
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009614042126379138,
      "loss": 1.919,
      "theoretical_loss": 4.51247143344909,
      "tokens_seen": 159121408
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000961384152457372,
      "loss": 1.8886,
      "theoretical_loss": 4.512232192462566,
      "tokens_seen": 159186944
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009613640922768305,
      "loss": 1.9788,
      "theoretical_loss": 4.511993077514662,
      "tokens_seen": 159252480
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009613440320962889,
      "loss": 1.9061,
      "theoretical_loss": 4.51175408848715,
      "tokens_seen": 159318016
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009613239719157472,
      "loss": 1.9861,
      "theoretical_loss": 4.511515225261961,
      "tokens_seen": 159383552
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009613039117352056,
      "loss": 1.8786,
      "theoretical_loss": 4.511276487721187,
      "tokens_seen": 159449088
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000961283851554664,
      "loss": 1.9009,
      "theoretical_loss": 4.511037875747078,
      "tokens_seen": 159514624
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009612637913741223,
      "loss": 1.9976,
      "theoretical_loss": 4.5107993892220435,
      "tokens_seen": 159580160
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009612437311935808,
      "loss": 1.9536,
      "theoretical_loss": 4.5105610280286506,
      "tokens_seen": 159645696
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009612236710130391,
      "loss": 1.9695,
      "theoretical_loss": 4.5103227920496245,
      "tokens_seen": 159711232
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009612036108324975,
      "loss": 1.9963,
      "theoretical_loss": 4.510084681167848,
      "tokens_seen": 159776768
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009611835506519559,
      "loss": 1.8545,
      "theoretical_loss": 4.509846695266361,
      "tokens_seen": 159842304
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009611634904714143,
      "loss": 1.9444,
      "theoretical_loss": 4.509608834228365,
      "tokens_seen": 159907840
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009611434302908726,
      "loss": 1.897,
      "theoretical_loss": 4.50937109793721,
      "tokens_seen": 159973376
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000961123370110331,
      "loss": 1.7706,
      "theoretical_loss": 4.5091334862764105,
      "tokens_seen": 160038912
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009611033099297893,
      "loss": 1.9551,
      "theoretical_loss": 4.508895999129634,
      "tokens_seen": 160104448
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009610832497492478,
      "loss": 1.9135,
      "theoretical_loss": 4.508658636380705,
      "tokens_seen": 160169984
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009610631895687061,
      "loss": 1.8921,
      "theoretical_loss": 4.508421397913603,
      "tokens_seen": 160235520
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009610431293881645,
      "loss": 2.0087,
      "theoretical_loss": 4.508184283612462,
      "tokens_seen": 160301056
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009610230692076228,
      "loss": 1.9208,
      "theoretical_loss": 4.5079472933615765,
      "tokens_seen": 160366592
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009610030090270813,
      "loss": 2.1351,
      "theoretical_loss": 4.507710427045389,
      "tokens_seen": 160432128
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009609829488465396,
      "loss": 2.0943,
      "theoretical_loss": 4.507473684548502,
      "tokens_seen": 160497664
    },
    {
      "epoch": 0.05,
      "objective/train/advantage_avg": 0.49687397480010986,
      "objective/train/docs_used": 65316,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8788886070251465,
      "objective/train/original_loss": 1.8788886070251465,
      "objective/train/theoretical_loss": 4.507237065755672,
      "objective/train/tokens_used": 181023200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24701941013336182,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0509432554244995,
      "objective/train/weighted_lm_loss": 1.97454035282135,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.507237065755672,
      "tokens_seen": 160563200
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000960962888665998,
      "loss": 1.985,
      "theoretical_loss": 4.507237065755672,
      "tokens_seen": 160563200
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009609428284854565,
      "loss": 1.8503,
      "theoretical_loss": 4.507000570551806,
      "tokens_seen": 160628736
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009609227683049147,
      "loss": 1.8316,
      "theoretical_loss": 4.50676419882197,
      "tokens_seen": 160694272
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009609027081243732,
      "loss": 2.0776,
      "theoretical_loss": 4.506527950451379,
      "tokens_seen": 160759808
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009608826479438315,
      "loss": 1.8979,
      "theoretical_loss": 4.506291825325405,
      "tokens_seen": 160825344
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009608625877632899,
      "loss": 1.9678,
      "theoretical_loss": 4.5060558233295716,
      "tokens_seen": 160890880
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009608425275827483,
      "loss": 1.8766,
      "theoretical_loss": 4.505819944349556,
      "tokens_seen": 160956416
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009608224674022067,
      "loss": 1.95,
      "theoretical_loss": 4.505584188271186,
      "tokens_seen": 161021952
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000960802407221665,
      "loss": 1.9329,
      "theoretical_loss": 4.505348554980444,
      "tokens_seen": 161087488
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009607823470411235,
      "loss": 1.8921,
      "theoretical_loss": 4.5051130443634655,
      "tokens_seen": 161153024
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009607622868605817,
      "loss": 1.7781,
      "theoretical_loss": 4.504877656306535,
      "tokens_seen": 161218560
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009607422266800402,
      "loss": 2.0076,
      "theoretical_loss": 4.504642390696089,
      "tokens_seen": 161284096
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009607221664994985,
      "loss": 1.9829,
      "theoretical_loss": 4.504407247418717,
      "tokens_seen": 161349632
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009607021063189569,
      "loss": 1.9316,
      "theoretical_loss": 4.50417222636116,
      "tokens_seen": 161415168
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009606820461384153,
      "loss": 1.973,
      "theoretical_loss": 4.503937327410306,
      "tokens_seen": 161480704
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009606619859578736,
      "loss": 1.9379,
      "theoretical_loss": 4.5037025504531965,
      "tokens_seen": 161546240
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000960641925777332,
      "loss": 1.9915,
      "theoretical_loss": 4.503467895377025,
      "tokens_seen": 161611776
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009606218655967905,
      "loss": 1.9256,
      "theoretical_loss": 4.50323336206913,
      "tokens_seen": 161677312
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009606018054162487,
      "loss": 1.984,
      "theoretical_loss": 4.502998950417004,
      "tokens_seen": 161742848
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009605817452357072,
      "loss": 1.8998,
      "theoretical_loss": 4.502764660308286,
      "tokens_seen": 161808384
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009605616850551654,
      "loss": 2.0802,
      "theoretical_loss": 4.502530491630768,
      "tokens_seen": 161873920
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009605416248746239,
      "loss": 2.0452,
      "theoretical_loss": 4.502296444272387,
      "tokens_seen": 161939456
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009605215646940823,
      "loss": 1.8631,
      "theoretical_loss": 4.502062518121232,
      "tokens_seen": 162004992
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009605015045135406,
      "loss": 1.9151,
      "theoretical_loss": 4.501828713065537,
      "tokens_seen": 162070528
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000960481444332999,
      "loss": 1.8775,
      "theoretical_loss": 4.501595028993686,
      "tokens_seen": 162136064
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009604613841524575,
      "loss": 1.9531,
      "theoretical_loss": 4.501361465794213,
      "tokens_seen": 162201600
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009604413239719157,
      "loss": 1.9826,
      "theoretical_loss": 4.501128023355796,
      "tokens_seen": 162267136
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009604212637913742,
      "loss": 1.9779,
      "theoretical_loss": 4.500894701567263,
      "tokens_seen": 162332672
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009604012036108325,
      "loss": 1.8989,
      "theoretical_loss": 4.500661500317587,
      "tokens_seen": 162398208
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009603811434302909,
      "loss": 1.8547,
      "theoretical_loss": 4.500428419495892,
      "tokens_seen": 162463744
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009603610832497493,
      "loss": 1.9796,
      "theoretical_loss": 4.500195458991443,
      "tokens_seen": 162529280
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009603410230692076,
      "loss": 2.0481,
      "theoretical_loss": 4.4999626186936545,
      "tokens_seen": 162594816
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000960320962888666,
      "loss": 2.0033,
      "theoretical_loss": 4.499729898492089,
      "tokens_seen": 162660352
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009603009027081244,
      "loss": 2.0799,
      "theoretical_loss": 4.4994972982764505,
      "tokens_seen": 162725888
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009602808425275827,
      "loss": 1.8236,
      "theoretical_loss": 4.499264817936593,
      "tokens_seen": 162791424
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009602607823470412,
      "loss": 2.1797,
      "theoretical_loss": 4.499032457362514,
      "tokens_seen": 162856960
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009602407221664995,
      "loss": 1.9804,
      "theoretical_loss": 4.498800216444353,
      "tokens_seen": 162922496
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009602206619859579,
      "loss": 1.9246,
      "theoretical_loss": 4.498568095072401,
      "tokens_seen": 162988032
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009602006018054163,
      "loss": 1.9251,
      "theoretical_loss": 4.498336093137089,
      "tokens_seen": 163053568
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009601805416248747,
      "loss": 1.9896,
      "theoretical_loss": 4.498104210528993,
      "tokens_seen": 163119104
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000960160481444333,
      "loss": 1.9048,
      "theoretical_loss": 4.497872447138834,
      "tokens_seen": 163184640
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009601404212637914,
      "loss": 1.9159,
      "theoretical_loss": 4.497640802857477,
      "tokens_seen": 163250176
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009601203610832497,
      "loss": 2.0265,
      "theoretical_loss": 4.49740927757593,
      "tokens_seen": 163315712
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009601003009027082,
      "loss": 1.7832,
      "theoretical_loss": 4.497177871185343,
      "tokens_seen": 163381248
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009600802407221665,
      "loss": 2.0497,
      "theoretical_loss": 4.4969465835770155,
      "tokens_seen": 163446784
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009600601805416249,
      "loss": 1.8603,
      "theoretical_loss": 4.496715414642381,
      "tokens_seen": 163512320
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009600401203610832,
      "loss": 1.9919,
      "theoretical_loss": 4.496484364273021,
      "tokens_seen": 163577856
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009600200601805417,
      "loss": 2.0578,
      "theoretical_loss": 4.49625343236066,
      "tokens_seen": 163643392
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00096,
      "loss": 1.97,
      "theoretical_loss": 4.496022618797161,
      "tokens_seen": 163708928
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009599799398194584,
      "loss": 1.8953,
      "theoretical_loss": 4.495791923474533,
      "tokens_seen": 163774464
    },
    {
      "epoch": 0.05,
      "objective/train/advantage_avg": 0.4943779706954956,
      "objective/train/docs_used": 66569,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7600229978561401,
      "objective/train/original_loss": 1.7600231170654297,
      "objective/train/theoretical_loss": 4.495561346284925,
      "objective/train/tokens_used": 184300000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2446385622024536,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050681471824646,
      "objective/train/weighted_lm_loss": 1.849135160446167,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.495561346284925,
      "tokens_seen": 163840000
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009599598796389167,
      "loss": 1.7395,
      "theoretical_loss": 4.495561346284925,
      "tokens_seen": 163840000
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009599398194583751,
      "loss": 1.8734,
      "theoretical_loss": 4.495330887120624,
      "tokens_seen": 163905536
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009599197592778335,
      "loss": 2.0076,
      "theoretical_loss": 4.495100545874065,
      "tokens_seen": 163971072
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009598996990972919,
      "loss": 1.955,
      "theoretical_loss": 4.4948703224378175,
      "tokens_seen": 164036608
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009598796389167502,
      "loss": 1.9929,
      "theoretical_loss": 4.494640216704598,
      "tokens_seen": 164102144
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009598595787362087,
      "loss": 2.0399,
      "theoretical_loss": 4.494410228567258,
      "tokens_seen": 164167680
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000959839518555667,
      "loss": 1.9737,
      "theoretical_loss": 4.494180357918792,
      "tokens_seen": 164233216
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009598194583751254,
      "loss": 1.9149,
      "theoretical_loss": 4.493950604652333,
      "tokens_seen": 164298752
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009597993981945838,
      "loss": 1.9785,
      "theoretical_loss": 4.493720968661158,
      "tokens_seen": 164364288
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009597793380140421,
      "loss": 1.8819,
      "theoretical_loss": 4.493491449838676,
      "tokens_seen": 164429824
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009597592778335005,
      "loss": 1.9185,
      "theoretical_loss": 4.493262048078442,
      "tokens_seen": 164495360
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000959739217652959,
      "loss": 1.928,
      "theoretical_loss": 4.493032763274148,
      "tokens_seen": 164560896
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009597191574724172,
      "loss": 1.9115,
      "theoretical_loss": 4.492803595319623,
      "tokens_seen": 164626432
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009596990972918757,
      "loss": 1.8792,
      "theoretical_loss": 4.492574544108838,
      "tokens_seen": 164691968
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009596790371113339,
      "loss": 1.7851,
      "theoretical_loss": 4.492345609535899,
      "tokens_seen": 164757504
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009596589769307924,
      "loss": 2.0436,
      "theoretical_loss": 4.492116791495052,
      "tokens_seen": 164823040
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009596389167502508,
      "loss": 1.8955,
      "theoretical_loss": 4.49188808988068,
      "tokens_seen": 164888576
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009596188565697091,
      "loss": 1.7672,
      "theoretical_loss": 4.491659504587304,
      "tokens_seen": 164954112
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009595987963891675,
      "loss": 1.8993,
      "theoretical_loss": 4.491431035509584,
      "tokens_seen": 165019648
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009595787362086258,
      "loss": 1.8166,
      "theoretical_loss": 4.4912026825423155,
      "tokens_seen": 165085184
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009595586760280842,
      "loss": 1.8962,
      "theoretical_loss": 4.490974445580429,
      "tokens_seen": 165150720
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009595386158475427,
      "loss": 1.8735,
      "theoretical_loss": 4.490746324518995,
      "tokens_seen": 165216256
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009595185556670009,
      "loss": 1.8615,
      "theoretical_loss": 4.490518319253221,
      "tokens_seen": 165281792
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009594984954864594,
      "loss": 1.8063,
      "theoretical_loss": 4.490290429678447,
      "tokens_seen": 165347328
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009594784353059178,
      "loss": 1.9261,
      "theoretical_loss": 4.490062655690153,
      "tokens_seen": 165412864
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009594583751253761,
      "loss": 1.8684,
      "theoretical_loss": 4.4898349971839515,
      "tokens_seen": 165478400
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009594383149448346,
      "loss": 1.7628,
      "theoretical_loss": 4.489607454055593,
      "tokens_seen": 165543936
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009594182547642929,
      "loss": 1.9646,
      "theoretical_loss": 4.489380026200961,
      "tokens_seen": 165609472
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009593981945837513,
      "loss": 1.9232,
      "theoretical_loss": 4.489152713516077,
      "tokens_seen": 165675008
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009593781344032097,
      "loss": 1.7531,
      "theoretical_loss": 4.488925515897095,
      "tokens_seen": 165740544
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000959358074222668,
      "loss": 1.6834,
      "theoretical_loss": 4.488698433240304,
      "tokens_seen": 165806080
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009593380140421264,
      "loss": 1.949,
      "theoretical_loss": 4.48847146544213,
      "tokens_seen": 165871616
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009593179538615848,
      "loss": 1.8811,
      "theoretical_loss": 4.488244612399129,
      "tokens_seen": 165937152
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009592978936810431,
      "loss": 1.8608,
      "theoretical_loss": 4.488017874007993,
      "tokens_seen": 166002688
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009592778335005016,
      "loss": 1.873,
      "theoretical_loss": 4.487791250165549,
      "tokens_seen": 166068224
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009592577733199599,
      "loss": 1.8062,
      "theoretical_loss": 4.487564740768756,
      "tokens_seen": 166133760
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009592377131394183,
      "loss": 1.8608,
      "theoretical_loss": 4.487338345714707,
      "tokens_seen": 166199296
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009592176529588767,
      "loss": 1.9687,
      "theoretical_loss": 4.487112064900627,
      "tokens_seen": 166264832
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009591975927783351,
      "loss": 1.9348,
      "theoretical_loss": 4.486885898223873,
      "tokens_seen": 166330368
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009591775325977934,
      "loss": 1.8068,
      "theoretical_loss": 4.48665984558194,
      "tokens_seen": 166395904
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009591574724172518,
      "loss": 1.8497,
      "theoretical_loss": 4.486433906872448,
      "tokens_seen": 166461440
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009591374122367101,
      "loss": 1.971,
      "theoretical_loss": 4.486208081993155,
      "tokens_seen": 166526976
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009591173520561686,
      "loss": 1.9106,
      "theoretical_loss": 4.485982370841946,
      "tokens_seen": 166592512
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009590972918756269,
      "loss": 1.878,
      "theoretical_loss": 4.485756773316844,
      "tokens_seen": 166658048
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009590772316950853,
      "loss": 1.9501,
      "theoretical_loss": 4.485531289315997,
      "tokens_seen": 166723584
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009590571715145436,
      "loss": 1.8681,
      "theoretical_loss": 4.485305918737687,
      "tokens_seen": 166789120
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009590371113340021,
      "loss": 1.8118,
      "theoretical_loss": 4.48508066148033,
      "tokens_seen": 166854656
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009590170511534604,
      "loss": 1.8352,
      "theoretical_loss": 4.484855517442467,
      "tokens_seen": 166920192
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009589969909729188,
      "loss": 1.7397,
      "theoretical_loss": 4.484630486522775,
      "tokens_seen": 166985728
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009589769307923771,
      "loss": 1.9343,
      "theoretical_loss": 4.4844055686200575,
      "tokens_seen": 167051264
    },
    {
      "epoch": 0.05,
      "objective/train/advantage_avg": 0.4909823536872864,
      "objective/train/docs_used": 67770,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.918248176574707,
      "objective/train/original_loss": 1.9182484149932861,
      "objective/train/theoretical_loss": 4.48418076363325,
      "objective/train/tokens_used": 187576800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2417370080947876,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503270626068115,
      "objective/train/weighted_lm_loss": 2.014829158782959,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.48418076363325,
      "tokens_seen": 167116800
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009589568706118355,
      "loss": 1.9665,
      "theoretical_loss": 4.48418076363325,
      "tokens_seen": 167116800
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009589368104312939,
      "loss": 1.9255,
      "theoretical_loss": 4.483956071461419,
      "tokens_seen": 167182336
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009589167502507523,
      "loss": 2.0147,
      "theoretical_loss": 4.483731492003757,
      "tokens_seen": 167247872
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009588966900702106,
      "loss": 1.825,
      "theoretical_loss": 4.483507025159591,
      "tokens_seen": 167313408
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009588766298896691,
      "loss": 1.8508,
      "theoretical_loss": 4.483282670828374,
      "tokens_seen": 167378944
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009588565697091274,
      "loss": 1.9944,
      "theoretical_loss": 4.483058428909688,
      "tokens_seen": 167444480
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009588365095285858,
      "loss": 1.8854,
      "theoretical_loss": 4.482834299303246,
      "tokens_seen": 167510016
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009588164493480442,
      "loss": 1.9292,
      "theoretical_loss": 4.482610281908888,
      "tokens_seen": 167575552
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009587963891675025,
      "loss": 1.9696,
      "theoretical_loss": 4.482386376626582,
      "tokens_seen": 167641088
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009587763289869609,
      "loss": 1.8661,
      "theoretical_loss": 4.482162583356427,
      "tokens_seen": 167706624
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009587562688064194,
      "loss": 1.7834,
      "theoretical_loss": 4.481938901998647,
      "tokens_seen": 167772160
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009587362086258776,
      "loss": 2.0171,
      "theoretical_loss": 4.481715332453595,
      "tokens_seen": 167837696
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009587161484453361,
      "loss": 1.8928,
      "theoretical_loss": 4.481491874621751,
      "tokens_seen": 167903232
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009586960882647943,
      "loss": 1.6732,
      "theoretical_loss": 4.481268528403723,
      "tokens_seen": 167968768
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009586760280842528,
      "loss": 1.8741,
      "theoretical_loss": 4.481045293700248,
      "tokens_seen": 168034304
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009586559679037112,
      "loss": 1.8851,
      "theoretical_loss": 4.480822170412186,
      "tokens_seen": 168099840
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009586359077231695,
      "loss": 1.8634,
      "theoretical_loss": 4.480599158440526,
      "tokens_seen": 168165376
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009586158475426279,
      "loss": 1.7684,
      "theoretical_loss": 4.480376257686384,
      "tokens_seen": 168230912
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009585957873620862,
      "loss": 2.012,
      "theoretical_loss": 4.480153468051001,
      "tokens_seen": 168296448
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009585757271815446,
      "loss": 1.9126,
      "theoretical_loss": 4.479930789435746,
      "tokens_seen": 168361984
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009585556670010031,
      "loss": 1.9654,
      "theoretical_loss": 4.479708221742111,
      "tokens_seen": 168427520
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009585356068204613,
      "loss": 1.8602,
      "theoretical_loss": 4.479485764871717,
      "tokens_seen": 168493056
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009585155466399198,
      "loss": 1.9912,
      "theoretical_loss": 4.4792634187263065,
      "tokens_seen": 168558592
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009584954864593782,
      "loss": 2.0017,
      "theoretical_loss": 4.479041183207752,
      "tokens_seen": 168624128
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009584754262788365,
      "loss": 1.8021,
      "theoretical_loss": 4.478819058218047,
      "tokens_seen": 168689664
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009584553660982949,
      "loss": 1.9251,
      "theoretical_loss": 4.478597043659311,
      "tokens_seen": 168755200
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009584353059177533,
      "loss": 1.9435,
      "theoretical_loss": 4.4783751394337905,
      "tokens_seen": 168820736
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009584152457372116,
      "loss": 1.7251,
      "theoretical_loss": 4.478153345443854,
      "tokens_seen": 168886272
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009583951855566701,
      "loss": 1.964,
      "theoretical_loss": 4.477931661591995,
      "tokens_seen": 168951808
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009583751253761283,
      "loss": 1.7283,
      "theoretical_loss": 4.47771008778083,
      "tokens_seen": 169017344
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009583550651955868,
      "loss": 1.7976,
      "theoretical_loss": 4.4774886239131,
      "tokens_seen": 169082880
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009583350050150451,
      "loss": 2.0903,
      "theoretical_loss": 4.477267269891671,
      "tokens_seen": 169148416
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009583149448345035,
      "loss": 1.8024,
      "theoretical_loss": 4.477046025619531,
      "tokens_seen": 169213952
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009582948846539619,
      "loss": 1.9521,
      "theoretical_loss": 4.47682489099979,
      "tokens_seen": 169279488
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009582748244734203,
      "loss": 1.9436,
      "theoretical_loss": 4.476603865935683,
      "tokens_seen": 169345024
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009582547642928786,
      "loss": 1.9069,
      "theoretical_loss": 4.476382950330569,
      "tokens_seen": 169410560
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009582347041123371,
      "loss": 1.9511,
      "theoretical_loss": 4.476162144087925,
      "tokens_seen": 169476096
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009582146439317953,
      "loss": 1.9537,
      "theoretical_loss": 4.475941447111355,
      "tokens_seen": 169541632
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009581945837512538,
      "loss": 2.0403,
      "theoretical_loss": 4.475720859304583,
      "tokens_seen": 169607168
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009581745235707121,
      "loss": 1.9219,
      "theoretical_loss": 4.4755003805714555,
      "tokens_seen": 169672704
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009581544633901705,
      "loss": 1.8615,
      "theoretical_loss": 4.47528001081594,
      "tokens_seen": 169738240
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009581344032096289,
      "loss": 1.9764,
      "theoretical_loss": 4.475059749942126,
      "tokens_seen": 169803776
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009581143430290873,
      "loss": 1.6865,
      "theoretical_loss": 4.474839597854226,
      "tokens_seen": 169869312
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009580942828485456,
      "loss": 2.0214,
      "theoretical_loss": 4.474619554456571,
      "tokens_seen": 169934848
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000958074222668004,
      "loss": 1.8969,
      "theoretical_loss": 4.4743996196536155,
      "tokens_seen": 170000384
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009580541624874624,
      "loss": 1.9962,
      "theoretical_loss": 4.474179793349933,
      "tokens_seen": 170065920
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009580341023069208,
      "loss": 1.7917,
      "theoretical_loss": 4.473960075450218,
      "tokens_seen": 170131456
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009580140421263791,
      "loss": 1.9129,
      "theoretical_loss": 4.473740465859285,
      "tokens_seen": 170196992
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009579939819458375,
      "loss": 1.993,
      "theoretical_loss": 4.47352096448207,
      "tokens_seen": 170262528
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009579739217652958,
      "loss": 2.0726,
      "theoretical_loss": 4.473301571223629,
      "tokens_seen": 170328064
    },
    {
      "epoch": 0.05,
      "objective/train/advantage_avg": 0.4946981370449066,
      "objective/train/docs_used": 69022,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.9940156936645508,
      "objective/train/original_loss": 1.9940155744552612,
      "objective/train/theoretical_loss": 4.473082285989134,
      "objective/train/tokens_used": 190853600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24500080943107605,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507153272628784,
      "objective/train/weighted_lm_loss": 2.094956636428833,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.473082285989134,
      "tokens_seen": 170393600
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009579538615847543,
      "loss": 1.9899,
      "theoretical_loss": 4.473082285989134,
      "tokens_seen": 170393600
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009579338014042127,
      "loss": 2.0756,
      "theoretical_loss": 4.472863108683884,
      "tokens_seen": 170459136
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000957913741223671,
      "loss": 1.9973,
      "theoretical_loss": 4.47264403921329,
      "tokens_seen": 170524672
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009578936810431295,
      "loss": 1.8127,
      "theoretical_loss": 4.472425077482886,
      "tokens_seen": 170590208
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009578736208625878,
      "loss": 1.8438,
      "theoretical_loss": 4.472206223398325,
      "tokens_seen": 170655744
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009578535606820462,
      "loss": 1.9837,
      "theoretical_loss": 4.471987476865376,
      "tokens_seen": 170721280
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009578335005015046,
      "loss": 1.8991,
      "theoretical_loss": 4.471768837789931,
      "tokens_seen": 170786816
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009578134403209629,
      "loss": 1.8756,
      "theoretical_loss": 4.4715503060779955,
      "tokens_seen": 170852352
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009577933801404213,
      "loss": 1.8956,
      "theoretical_loss": 4.471331881635698,
      "tokens_seen": 170917888
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009577733199598798,
      "loss": 1.9264,
      "theoretical_loss": 4.471113564369282,
      "tokens_seen": 170983424
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000957753259779338,
      "loss": 1.7323,
      "theoretical_loss": 4.47089535418511,
      "tokens_seen": 171048960
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009577331995987965,
      "loss": 2.1019,
      "theoretical_loss": 4.470677250989661,
      "tokens_seen": 171114496
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009577131394182547,
      "loss": 1.8691,
      "theoretical_loss": 4.470459254689533,
      "tokens_seen": 171180032
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009576930792377132,
      "loss": 1.8631,
      "theoretical_loss": 4.470241365191441,
      "tokens_seen": 171245568
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009576730190571716,
      "loss": 1.9141,
      "theoretical_loss": 4.4700235824022165,
      "tokens_seen": 171311104
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009576529588766299,
      "loss": 1.9757,
      "theoretical_loss": 4.469805906228807,
      "tokens_seen": 171376640
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009576328986960883,
      "loss": 1.8581,
      "theoretical_loss": 4.469588336578277,
      "tokens_seen": 171442176
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009576128385155466,
      "loss": 1.905,
      "theoretical_loss": 4.469370873357811,
      "tokens_seen": 171507712
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000957592778335005,
      "loss": 1.7967,
      "theoretical_loss": 4.469153516474703,
      "tokens_seen": 171573248
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009575727181544635,
      "loss": 1.893,
      "theoretical_loss": 4.468936265836371,
      "tokens_seen": 171638784
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009575526579739217,
      "loss": 1.8793,
      "theoretical_loss": 4.468719121350343,
      "tokens_seen": 171704320
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009575325977933802,
      "loss": 1.816,
      "theoretical_loss": 4.468502082924266,
      "tokens_seen": 171769856
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009575125376128386,
      "loss": 1.905,
      "theoretical_loss": 4.468285150465899,
      "tokens_seen": 171835392
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009574924774322969,
      "loss": 1.888,
      "theoretical_loss": 4.468068323883122,
      "tokens_seen": 171900928
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009574724172517553,
      "loss": 1.8005,
      "theoretical_loss": 4.467851603083923,
      "tokens_seen": 171966464
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009574523570712137,
      "loss": 1.8414,
      "theoretical_loss": 4.467634987976411,
      "tokens_seen": 172032000
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000957432296890672,
      "loss": 2.0145,
      "theoretical_loss": 4.467418478468808,
      "tokens_seen": 172097536
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009574122367101305,
      "loss": 1.9675,
      "theoretical_loss": 4.467202074469449,
      "tokens_seen": 172163072
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009573921765295887,
      "loss": 1.8405,
      "theoretical_loss": 4.466985775886784,
      "tokens_seen": 172228608
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009573721163490472,
      "loss": 1.9227,
      "theoretical_loss": 4.4667695826293805,
      "tokens_seen": 172294144
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009573520561685055,
      "loss": 1.8589,
      "theoretical_loss": 4.4665534946059156,
      "tokens_seen": 172359680
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009573319959879639,
      "loss": 1.9755,
      "theoretical_loss": 4.4663375117251825,
      "tokens_seen": 172425216
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009573119358074223,
      "loss": 1.8297,
      "theoretical_loss": 4.466121633896087,
      "tokens_seen": 172490752
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009572918756268807,
      "loss": 2.0146,
      "theoretical_loss": 4.4659058610276485,
      "tokens_seen": 172556288
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000957271815446339,
      "loss": 1.837,
      "theoretical_loss": 4.465690193029003,
      "tokens_seen": 172621824
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009572517552657975,
      "loss": 2.0488,
      "theoretical_loss": 4.465474629809394,
      "tokens_seen": 172687360
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009572316950852557,
      "loss": 1.8981,
      "theoretical_loss": 4.465259171278182,
      "tokens_seen": 172752896
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009572116349047142,
      "loss": 1.9211,
      "theoretical_loss": 4.4650438173448395,
      "tokens_seen": 172818432
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009571915747241725,
      "loss": 1.886,
      "theoretical_loss": 4.464828567918951,
      "tokens_seen": 172883968
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009571715145436309,
      "loss": 1.8676,
      "theoretical_loss": 4.464613422910213,
      "tokens_seen": 172949504
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009571514543630893,
      "loss": 1.8263,
      "theoretical_loss": 4.464398382228435,
      "tokens_seen": 173015040
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009571313941825477,
      "loss": 1.9294,
      "theoretical_loss": 4.464183445783537,
      "tokens_seen": 173080576
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000957111334002006,
      "loss": 1.9157,
      "theoretical_loss": 4.463968613485555,
      "tokens_seen": 173146112
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009570912738214644,
      "loss": 2.0705,
      "theoretical_loss": 4.463753885244631,
      "tokens_seen": 173211648
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009570712136409228,
      "loss": 1.8366,
      "theoretical_loss": 4.463539260971023,
      "tokens_seen": 173277184
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009570511534603812,
      "loss": 2.0209,
      "theoretical_loss": 4.463324740575098,
      "tokens_seen": 173342720
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009570310932798395,
      "loss": 1.8134,
      "theoretical_loss": 4.463110323967335,
      "tokens_seen": 173408256
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000957011033099298,
      "loss": 2.0315,
      "theoretical_loss": 4.462896011058323,
      "tokens_seen": 173473792
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009569909729187562,
      "loss": 1.936,
      "theoretical_loss": 4.462681801758762,
      "tokens_seen": 173539328
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009569709127382147,
      "loss": 1.8807,
      "theoretical_loss": 4.4624676959794645,
      "tokens_seen": 173604864
    },
    {
      "epoch": 0.05,
      "objective/train/advantage_avg": 0.4944310486316681,
      "objective/train/docs_used": 70257,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.9424092769622803,
      "objective/train/original_loss": 1.9424092769622803,
      "objective/train/theoretical_loss": 4.462253693631352,
      "objective/train/tokens_used": 194130400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24486778676509857,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506879091262817,
      "objective/train/weighted_lm_loss": 2.0407896041870117,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 4.462253693631352,
      "tokens_seen": 173670400
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000956950852557673,
      "loss": 1.9977,
      "theoretical_loss": 4.462253693631352,
      "tokens_seen": 173670400
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009569307923771314,
      "loss": 1.9687,
      "theoretical_loss": 4.462039794625454,
      "tokens_seen": 173735936
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009569107321965898,
      "loss": 1.9434,
      "theoretical_loss": 4.461825998872914,
      "tokens_seen": 173801472
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009568906720160482,
      "loss": 1.8063,
      "theoretical_loss": 4.461612306284984,
      "tokens_seen": 173867008
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009568706118355065,
      "loss": 1.8382,
      "theoretical_loss": 4.461398716773022,
      "tokens_seen": 173932544
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000956850551654965,
      "loss": 1.9081,
      "theoretical_loss": 4.461185230248503,
      "tokens_seen": 173998080
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009568304914744232,
      "loss": 1.7065,
      "theoretical_loss": 4.460971846623005,
      "tokens_seen": 174063616
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009568104312938817,
      "loss": 1.854,
      "theoretical_loss": 4.4607585658082165,
      "tokens_seen": 174129152
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00095679037111334,
      "loss": 1.8371,
      "theoretical_loss": 4.460545387715937,
      "tokens_seen": 174194688
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009567703109327984,
      "loss": 1.9075,
      "theoretical_loss": 4.460332312258075,
      "tokens_seen": 174260224
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009567502507522568,
      "loss": 1.9443,
      "theoretical_loss": 4.460119339346643,
      "tokens_seen": 174325760
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009567301905717151,
      "loss": 1.9899,
      "theoretical_loss": 4.4599064688937675,
      "tokens_seen": 174391296
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009567101303911735,
      "loss": 1.7698,
      "theoretical_loss": 4.45969370081168,
      "tokens_seen": 174456832
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000956690070210632,
      "loss": 1.7344,
      "theoretical_loss": 4.459481035012722,
      "tokens_seen": 174522368
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009566700100300902,
      "loss": 1.8361,
      "theoretical_loss": 4.45926847140934,
      "tokens_seen": 174587904
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009566499498495487,
      "loss": 1.8969,
      "theoretical_loss": 4.459056009914093,
      "tokens_seen": 174653440
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009566298896690069,
      "loss": 1.7649,
      "theoretical_loss": 4.458843650439643,
      "tokens_seen": 174718976
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009566098294884654,
      "loss": 1.8389,
      "theoretical_loss": 4.458631392898763,
      "tokens_seen": 174784512
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009565897693079238,
      "loss": 1.9442,
      "theoretical_loss": 4.45841923720433,
      "tokens_seen": 174850048
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009565697091273821,
      "loss": 2.0796,
      "theoretical_loss": 4.45820718326933,
      "tokens_seen": 174915584
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009565496489468405,
      "loss": 2.0647,
      "theoretical_loss": 4.457995231006857,
      "tokens_seen": 174981120
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000956529588766299,
      "loss": 1.8885,
      "theoretical_loss": 4.457783380330108,
      "tokens_seen": 175046656
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009565095285857572,
      "loss": 1.8958,
      "theoretical_loss": 4.4575716311523905,
      "tokens_seen": 175112192
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009564894684052157,
      "loss": 1.8034,
      "theoretical_loss": 4.457359983387116,
      "tokens_seen": 175177728
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000956469408224674,
      "loss": 1.9068,
      "theoretical_loss": 4.457148436947804,
      "tokens_seen": 175243264
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009564493480441324,
      "loss": 1.9427,
      "theoretical_loss": 4.456936991748078,
      "tokens_seen": 175308800
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009564292878635909,
      "loss": 1.9902,
      "theoretical_loss": 4.456725647701669,
      "tokens_seen": 175374336
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009564092276830491,
      "loss": 1.8762,
      "theoretical_loss": 4.456514404722413,
      "tokens_seen": 175439872
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009563891675025076,
      "loss": 1.9261,
      "theoretical_loss": 4.456303262724252,
      "tokens_seen": 175505408
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009563691073219659,
      "loss": 1.8886,
      "theoretical_loss": 4.456092221621233,
      "tokens_seen": 175570944
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009563490471414243,
      "loss": 1.8757,
      "theoretical_loss": 4.455881281327508,
      "tokens_seen": 175636480
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009563289869608827,
      "loss": 1.9739,
      "theoretical_loss": 4.455670441757334,
      "tokens_seen": 175702016
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009563089267803411,
      "loss": 1.8367,
      "theoretical_loss": 4.455459702825074,
      "tokens_seen": 175767552
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009562888665997994,
      "loss": 1.9812,
      "theoretical_loss": 4.455249064445195,
      "tokens_seen": 175833088
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009562688064192579,
      "loss": 1.9976,
      "theoretical_loss": 4.4550385265322685,
      "tokens_seen": 175898624
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009562487462387161,
      "loss": 1.9515,
      "theoretical_loss": 4.454828089000969,
      "tokens_seen": 175964160
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009562286860581746,
      "loss": 1.8709,
      "theoretical_loss": 4.454617751766078,
      "tokens_seen": 176029696
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009562086258776329,
      "loss": 1.8376,
      "theoretical_loss": 4.454407514742479,
      "tokens_seen": 176095232
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009561885656970913,
      "loss": 1.9017,
      "theoretical_loss": 4.45419737784516,
      "tokens_seen": 176160768
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009561685055165497,
      "loss": 1.9527,
      "theoretical_loss": 4.4539873409892135,
      "tokens_seen": 176226304
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009561484453360081,
      "loss": 2.0055,
      "theoretical_loss": 4.453777404089832,
      "tokens_seen": 176291840
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009561283851554664,
      "loss": 1.8616,
      "theoretical_loss": 4.453567567062318,
      "tokens_seen": 176357376
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009561083249749248,
      "loss": 1.9151,
      "theoretical_loss": 4.45335782982207,
      "tokens_seen": 176422912
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009560882647943832,
      "loss": 1.865,
      "theoretical_loss": 4.453148192284595,
      "tokens_seen": 176488448
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009560682046138416,
      "loss": 2.0749,
      "theoretical_loss": 4.4529386543655,
      "tokens_seen": 176553984
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009560481444332999,
      "loss": 2.0762,
      "theoretical_loss": 4.452729215980495,
      "tokens_seen": 176619520
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009560280842527584,
      "loss": 1.8006,
      "theoretical_loss": 4.452519877045393,
      "tokens_seen": 176685056
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009560080240722166,
      "loss": 2.0119,
      "theoretical_loss": 4.452310637476112,
      "tokens_seen": 176750592
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009559879638916751,
      "loss": 1.9248,
      "theoretical_loss": 4.452101497188666,
      "tokens_seen": 176816128
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009559679037111334,
      "loss": 2.002,
      "theoretical_loss": 4.451892456099176,
      "tokens_seen": 176881664
    },
    {
      "epoch": 0.05,
      "objective/train/advantage_avg": 0.49354079365730286,
      "objective/train/docs_used": 71338,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6465367078781128,
      "objective/train/original_loss": 1.6465368270874023,
      "objective/train/theoretical_loss": 4.451683514123864,
      "objective/train/tokens_used": 197407200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2438439577817917,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505937337875366,
      "objective/train/weighted_lm_loss": 1.7298942804336548,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.451683514123864,
      "tokens_seen": 176947200
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009559478435305918,
      "loss": 1.8323,
      "theoretical_loss": 4.451683514123864,
      "tokens_seen": 176947200
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009559277833500502,
      "loss": 1.8433,
      "theoretical_loss": 4.451474671179055,
      "tokens_seen": 177012736
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009559077231695086,
      "loss": 1.9551,
      "theoretical_loss": 4.451265927181171,
      "tokens_seen": 177078272
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009558876629889669,
      "loss": 2.0386,
      "theoretical_loss": 4.451057282046741,
      "tokens_seen": 177143808
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009558676028084254,
      "loss": 1.8875,
      "theoretical_loss": 4.450848735692391,
      "tokens_seen": 177209344
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009558475426278836,
      "loss": 2.0172,
      "theoretical_loss": 4.45064028803485,
      "tokens_seen": 177274880
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009558274824473421,
      "loss": 1.8569,
      "theoretical_loss": 4.450431938990949,
      "tokens_seen": 177340416
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009558074222668004,
      "loss": 1.8493,
      "theoretical_loss": 4.450223688477617,
      "tokens_seen": 177405952
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009557873620862588,
      "loss": 1.9685,
      "theoretical_loss": 4.450015536411886,
      "tokens_seen": 177471488
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009557673019057172,
      "loss": 1.906,
      "theoretical_loss": 4.449807482710888,
      "tokens_seen": 177537024
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009557472417251755,
      "loss": 1.8314,
      "theoretical_loss": 4.449599527291854,
      "tokens_seen": 177602560
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009557271815446339,
      "loss": 1.8091,
      "theoretical_loss": 4.449391670072117,
      "tokens_seen": 177668096
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009557071213640924,
      "loss": 1.9665,
      "theoretical_loss": 4.449183910969108,
      "tokens_seen": 177733632
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009556870611835506,
      "loss": 1.9209,
      "theoretical_loss": 4.44897624990036,
      "tokens_seen": 177799168
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009556670010030091,
      "loss": 1.7531,
      "theoretical_loss": 4.448768686783504,
      "tokens_seen": 177864704
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009556469408224673,
      "loss": 1.8657,
      "theoretical_loss": 4.448561221536272,
      "tokens_seen": 177930240
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009556268806419258,
      "loss": 1.8918,
      "theoretical_loss": 4.448353854076494,
      "tokens_seen": 177995776
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009556068204613842,
      "loss": 1.9234,
      "theoretical_loss": 4.4481465843220995,
      "tokens_seen": 178061312
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009555867602808425,
      "loss": 1.9232,
      "theoretical_loss": 4.447939412191117,
      "tokens_seen": 178126848
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009555667001003009,
      "loss": 1.9091,
      "theoretical_loss": 4.447732337601675,
      "tokens_seen": 178192384
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009555466399197594,
      "loss": 1.8798,
      "theoretical_loss": 4.4475253604719995,
      "tokens_seen": 178257920
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009555265797392176,
      "loss": 1.7858,
      "theoretical_loss": 4.447318480720416,
      "tokens_seen": 178323456
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009555065195586761,
      "loss": 2.0034,
      "theoretical_loss": 4.447111698265346,
      "tokens_seen": 178388992
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009554864593781343,
      "loss": 1.7928,
      "theoretical_loss": 4.446905013025313,
      "tokens_seen": 178454528
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009554663991975928,
      "loss": 1.7196,
      "theoretical_loss": 4.446698424918937,
      "tokens_seen": 178520064
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009554463390170512,
      "loss": 1.9755,
      "theoretical_loss": 4.446491933864936,
      "tokens_seen": 178585600
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009554262788365095,
      "loss": 1.849,
      "theoretical_loss": 4.4462855397821235,
      "tokens_seen": 178651136
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009554062186559679,
      "loss": 1.9696,
      "theoretical_loss": 4.4460792425894144,
      "tokens_seen": 178716672
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009553861584754263,
      "loss": 1.8554,
      "theoretical_loss": 4.44587304220582,
      "tokens_seen": 178782208
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009553660982948846,
      "loss": 1.7727,
      "theoretical_loss": 4.445666938550447,
      "tokens_seen": 178847744
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009553460381143431,
      "loss": 1.8321,
      "theoretical_loss": 4.445460931542502,
      "tokens_seen": 178913280
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009553259779338014,
      "loss": 1.6822,
      "theoretical_loss": 4.4452550211012865,
      "tokens_seen": 178978816
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009553059177532598,
      "loss": 1.9822,
      "theoretical_loss": 4.4450492071462,
      "tokens_seen": 179044352
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009552858575727181,
      "loss": 1.8321,
      "theoretical_loss": 4.444843489596739,
      "tokens_seen": 179109888
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009552657973921765,
      "loss": 2.0049,
      "theoretical_loss": 4.444637868372495,
      "tokens_seen": 179175424
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009552457372116349,
      "loss": 1.9795,
      "theoretical_loss": 4.444432343393158,
      "tokens_seen": 179240960
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009552256770310933,
      "loss": 2.046,
      "theoretical_loss": 4.444226914578513,
      "tokens_seen": 179306496
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009552056168505516,
      "loss": 2.0291,
      "theoretical_loss": 4.444021581848442,
      "tokens_seen": 179372032
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009551855566700101,
      "loss": 2.0813,
      "theoretical_loss": 4.443816345122922,
      "tokens_seen": 179437568
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009551654964894684,
      "loss": 1.7821,
      "theoretical_loss": 4.443611204322026,
      "tokens_seen": 179503104
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009551454363089268,
      "loss": 1.916,
      "theoretical_loss": 4.4434061593659235,
      "tokens_seen": 179568640
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009551253761283851,
      "loss": 1.9274,
      "theoretical_loss": 4.4432012101748795,
      "tokens_seen": 179634176
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009551053159478436,
      "loss": 1.8894,
      "theoretical_loss": 4.442996356669253,
      "tokens_seen": 179699712
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009550852557673019,
      "loss": 1.9832,
      "theoretical_loss": 4.4427915987695,
      "tokens_seen": 179765248
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009550651955867603,
      "loss": 1.8851,
      "theoretical_loss": 4.442586936396171,
      "tokens_seen": 179830784
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009550451354062186,
      "loss": 1.8177,
      "theoretical_loss": 4.442382369469909,
      "tokens_seen": 179896320
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000955025075225677,
      "loss": 1.9614,
      "theoretical_loss": 4.442177897911458,
      "tokens_seen": 179961856
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009550050150451354,
      "loss": 1.9223,
      "theoretical_loss": 4.441973521641649,
      "tokens_seen": 180027392
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009549849548645938,
      "loss": 1.9658,
      "theoretical_loss": 4.441769240581412,
      "tokens_seen": 180092928
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009549648946840521,
      "loss": 2.0016,
      "theoretical_loss": 4.441565054651772,
      "tokens_seen": 180158464
    },
    {
      "epoch": 0.05,
      "objective/train/advantage_avg": 0.4942334294319153,
      "objective/train/docs_used": 72550,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7501457929611206,
      "objective/train/original_loss": 1.7501459121704102,
      "objective/train/theoretical_loss": 4.441360963773846,
      "objective/train/tokens_used": 200684000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2446201890707016,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506670475006104,
      "objective/train/weighted_lm_loss": 1.8388445377349854,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.441360963773846,
      "tokens_seen": 180224000
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009549448345035106,
      "loss": 1.9136,
      "theoretical_loss": 4.441360963773846,
      "tokens_seen": 180224000
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000954924774322969,
      "loss": 2.0606,
      "theoretical_loss": 4.441156967868846,
      "tokens_seen": 180289536
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009549047141424273,
      "loss": 2.0416,
      "theoretical_loss": 4.440953066858077,
      "tokens_seen": 180355072
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009548846539618858,
      "loss": 2.0186,
      "theoretical_loss": 4.440749260662939,
      "tokens_seen": 180420608
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000954864593781344,
      "loss": 1.8847,
      "theoretical_loss": 4.440545549204925,
      "tokens_seen": 180486144
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009548445336008025,
      "loss": 1.7716,
      "theoretical_loss": 4.440341932405623,
      "tokens_seen": 180551680
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009548244734202608,
      "loss": 1.8777,
      "theoretical_loss": 4.4401384101867105,
      "tokens_seen": 180617216
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009548044132397192,
      "loss": 1.8929,
      "theoretical_loss": 4.439934982469962,
      "tokens_seen": 180682752
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009547843530591776,
      "loss": 1.9978,
      "theoretical_loss": 4.439731649177244,
      "tokens_seen": 180748288
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009547642928786359,
      "loss": 1.9419,
      "theoretical_loss": 4.439528410230514,
      "tokens_seen": 180813824
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009547442326980943,
      "loss": 1.767,
      "theoretical_loss": 4.439325265551826,
      "tokens_seen": 180879360
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009547241725175528,
      "loss": 1.9141,
      "theoretical_loss": 4.439122215063323,
      "tokens_seen": 180944896
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000954704112337011,
      "loss": 1.9685,
      "theoretical_loss": 4.438919258687242,
      "tokens_seen": 181010432
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009546840521564695,
      "loss": 1.826,
      "theoretical_loss": 4.438716396345913,
      "tokens_seen": 181075968
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009546639919759277,
      "loss": 1.9419,
      "theoretical_loss": 4.438513627961757,
      "tokens_seen": 181141504
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009546439317953862,
      "loss": 1.8724,
      "theoretical_loss": 4.438310953457286,
      "tokens_seen": 181207040
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009546238716148446,
      "loss": 1.9956,
      "theoretical_loss": 4.438108372755108,
      "tokens_seen": 181272576
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009546038114343029,
      "loss": 1.9181,
      "theoretical_loss": 4.43790588577792,
      "tokens_seen": 181338112
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009545837512537613,
      "loss": 1.9048,
      "theoretical_loss": 4.437703492448509,
      "tokens_seen": 181403648
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0009545636910732198,
      "loss": 2.0959,
      "theoretical_loss": 4.437501192689755,
      "tokens_seen": 181469184
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000954543630892678,
      "loss": 1.9178,
      "theoretical_loss": 4.437298986424631,
      "tokens_seen": 181534720
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009545235707121365,
      "loss": 1.9351,
      "theoretical_loss": 4.437096873576199,
      "tokens_seen": 181600256
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009545035105315947,
      "loss": 1.8552,
      "theoretical_loss": 4.436894854067614,
      "tokens_seen": 181665792
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009544834503510532,
      "loss": 1.8738,
      "theoretical_loss": 4.4366929278221185,
      "tokens_seen": 181731328
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009544633901705116,
      "loss": 2.0089,
      "theoretical_loss": 4.436491094763049,
      "tokens_seen": 181796864
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009544433299899699,
      "loss": 1.8173,
      "theoretical_loss": 4.436289354813832,
      "tokens_seen": 181862400
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009544232698094283,
      "loss": 1.9785,
      "theoretical_loss": 4.436087707897984,
      "tokens_seen": 181927936
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009544032096288867,
      "loss": 1.8956,
      "theoretical_loss": 4.435886153939111,
      "tokens_seen": 181993472
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000954383149448345,
      "loss": 1.901,
      "theoretical_loss": 4.435684692860912,
      "tokens_seen": 182059008
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009543630892678035,
      "loss": 1.9325,
      "theoretical_loss": 4.435483324587172,
      "tokens_seen": 182124544
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009543430290872618,
      "loss": 1.9463,
      "theoretical_loss": 4.435282049041769,
      "tokens_seen": 182190080
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009543229689067202,
      "loss": 1.8567,
      "theoretical_loss": 4.43508086614867,
      "tokens_seen": 182255616
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009543029087261785,
      "loss": 2.0057,
      "theoretical_loss": 4.4348797758319325,
      "tokens_seen": 182321152
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000954282848545637,
      "loss": 1.8366,
      "theoretical_loss": 4.434678778015701,
      "tokens_seen": 182386688
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009542627883650953,
      "loss": 1.8694,
      "theoretical_loss": 4.434477872624212,
      "tokens_seen": 182452224
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009542427281845537,
      "loss": 2.0037,
      "theoretical_loss": 4.434277059581789,
      "tokens_seen": 182517760
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000954222668004012,
      "loss": 1.925,
      "theoretical_loss": 4.434076338812848,
      "tokens_seen": 182583296
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009542026078234705,
      "loss": 1.9509,
      "theoretical_loss": 4.43387571024189,
      "tokens_seen": 182648832
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009541825476429288,
      "loss": 1.9147,
      "theoretical_loss": 4.433675173793507,
      "tokens_seen": 182714368
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009541624874623872,
      "loss": 2.0302,
      "theoretical_loss": 4.433474729392379,
      "tokens_seen": 182779904
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009541424272818455,
      "loss": 2.053,
      "theoretical_loss": 4.433274376963276,
      "tokens_seen": 182845440
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000954122367101304,
      "loss": 1.8329,
      "theoretical_loss": 4.4330741164310545,
      "tokens_seen": 182910976
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009541023069207623,
      "loss": 1.7745,
      "theoretical_loss": 4.43287394772066,
      "tokens_seen": 182976512
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009540822467402207,
      "loss": 1.9235,
      "theoretical_loss": 4.432673870757127,
      "tokens_seen": 183042048
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000954062186559679,
      "loss": 1.956,
      "theoretical_loss": 4.432473885465576,
      "tokens_seen": 183107584
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009540421263791374,
      "loss": 2.0946,
      "theoretical_loss": 4.432273991771218,
      "tokens_seen": 183173120
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009540220661985958,
      "loss": 1.9536,
      "theoretical_loss": 4.43207418959935,
      "tokens_seen": 183238656
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009540020060180542,
      "loss": 1.8331,
      "theoretical_loss": 4.431874478875356,
      "tokens_seen": 183304192
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009539819458375125,
      "loss": 1.8043,
      "theoretical_loss": 4.431674859524709,
      "tokens_seen": 183369728
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000953961885656971,
      "loss": 1.8608,
      "theoretical_loss": 4.43147533147297,
      "tokens_seen": 183435264
    },
    {
      "epoch": 0.06,
      "objective/train/advantage_avg": 0.4887767434120178,
      "objective/train/docs_used": 73723,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8926588296890259,
      "objective/train/original_loss": 1.8926584720611572,
      "objective/train/theoretical_loss": 4.431275894645784,
      "objective/train/tokens_used": 203960800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24062563478946686,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501008033752441,
      "objective/train/weighted_lm_loss": 1.9868710041046143,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.431275894645784,
      "tokens_seen": 183500800
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009539418254764293,
      "loss": 1.91,
      "theoretical_loss": 4.431275894645784,
      "tokens_seen": 183500800
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009539217652958877,
      "loss": 1.9714,
      "theoretical_loss": 4.431076548968887,
      "tokens_seen": 183566336
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000953901705115346,
      "loss": 1.9316,
      "theoretical_loss": 4.430877294368099,
      "tokens_seen": 183631872
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009538816449348044,
      "loss": 1.8594,
      "theoretical_loss": 4.430678130769329,
      "tokens_seen": 183697408
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009538615847542628,
      "loss": 1.8918,
      "theoretical_loss": 4.43047905809857,
      "tokens_seen": 183762944
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009538415245737212,
      "loss": 1.9531,
      "theoretical_loss": 4.430280076281904,
      "tokens_seen": 183828480
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009538214643931795,
      "loss": 1.7753,
      "theoretical_loss": 4.430081185245498,
      "tokens_seen": 183894016
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000953801404212638,
      "loss": 1.9356,
      "theoretical_loss": 4.429882384915607,
      "tokens_seen": 183959552
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009537813440320962,
      "loss": 1.9047,
      "theoretical_loss": 4.42968367521857,
      "tokens_seen": 184025088
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009537612838515547,
      "loss": 1.9452,
      "theoretical_loss": 4.4294850560808126,
      "tokens_seen": 184090624
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009537412236710131,
      "loss": 1.8756,
      "theoretical_loss": 4.429286527428848,
      "tokens_seen": 184156160
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009537211634904714,
      "loss": 1.9095,
      "theoretical_loss": 4.429088089189273,
      "tokens_seen": 184221696
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009537011033099298,
      "loss": 1.9531,
      "theoretical_loss": 4.428889741288771,
      "tokens_seen": 184287232
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009536810431293881,
      "loss": 2.0226,
      "theoretical_loss": 4.428691483654109,
      "tokens_seen": 184352768
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009536609829488465,
      "loss": 1.9104,
      "theoretical_loss": 4.428493316212144,
      "tokens_seen": 184418304
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000953640922768305,
      "loss": 1.9173,
      "theoretical_loss": 4.428295238889815,
      "tokens_seen": 184483840
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009536208625877632,
      "loss": 1.9288,
      "theoretical_loss": 4.428097251614145,
      "tokens_seen": 184549376
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009536008024072217,
      "loss": 1.8834,
      "theoretical_loss": 4.4278993543122445,
      "tokens_seen": 184614912
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009535807422266801,
      "loss": 2.0234,
      "theoretical_loss": 4.427701546911307,
      "tokens_seen": 184680448
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009535606820461384,
      "loss": 1.846,
      "theoretical_loss": 4.427503829338613,
      "tokens_seen": 184745984
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009535406218655968,
      "loss": 1.8591,
      "theoretical_loss": 4.427306201521524,
      "tokens_seen": 184811520
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009535205616850551,
      "loss": 1.8564,
      "theoretical_loss": 4.427108663387489,
      "tokens_seen": 184877056
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009535005015045135,
      "loss": 1.85,
      "theoretical_loss": 4.426911214864043,
      "tokens_seen": 184942592
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000953480441323972,
      "loss": 1.8551,
      "theoretical_loss": 4.426713855878798,
      "tokens_seen": 185008128
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009534603811434302,
      "loss": 2.0234,
      "theoretical_loss": 4.426516586359458,
      "tokens_seen": 185073664
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009534403209628887,
      "loss": 2.0515,
      "theoretical_loss": 4.426319406233807,
      "tokens_seen": 185139200
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009534202607823471,
      "loss": 1.9792,
      "theoretical_loss": 4.426122315429713,
      "tokens_seen": 185204736
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009534002006018054,
      "loss": 1.8541,
      "theoretical_loss": 4.425925313875129,
      "tokens_seen": 185270272
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009533801404212639,
      "loss": 1.8656,
      "theoretical_loss": 4.425728401498089,
      "tokens_seen": 185335808
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009533600802407222,
      "loss": 2.041,
      "theoretical_loss": 4.425531578226715,
      "tokens_seen": 185401344
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009533400200601806,
      "loss": 1.9088,
      "theoretical_loss": 4.425334843989207,
      "tokens_seen": 185466880
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009533199598796389,
      "loss": 2.0045,
      "theoretical_loss": 4.425138198713853,
      "tokens_seen": 185532416
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009532998996990973,
      "loss": 1.8641,
      "theoretical_loss": 4.424941642329019,
      "tokens_seen": 185597952
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009532798395185557,
      "loss": 2.0065,
      "theoretical_loss": 4.42474517476316,
      "tokens_seen": 185663488
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009532597793380141,
      "loss": 1.8994,
      "theoretical_loss": 4.424548795944808,
      "tokens_seen": 185729024
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009532397191574724,
      "loss": 1.8996,
      "theoretical_loss": 4.424352505802582,
      "tokens_seen": 185794560
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009532196589769309,
      "loss": 1.898,
      "theoretical_loss": 4.42415630426518,
      "tokens_seen": 185860096
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009531995987963892,
      "loss": 1.9314,
      "theoretical_loss": 4.423960191261387,
      "tokens_seen": 185925632
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009531795386158476,
      "loss": 1.8369,
      "theoretical_loss": 4.423764166720066,
      "tokens_seen": 185991168
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009531594784353059,
      "loss": 1.8829,
      "theoretical_loss": 4.4235682305701625,
      "tokens_seen": 186056704
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009531394182547644,
      "loss": 1.8661,
      "theoretical_loss": 4.423372382740707,
      "tokens_seen": 186122240
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009531193580742227,
      "loss": 1.8,
      "theoretical_loss": 4.423176623160811,
      "tokens_seen": 186187776
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009530992978936811,
      "loss": 1.8912,
      "theoretical_loss": 4.422980951759665,
      "tokens_seen": 186253312
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009530792377131394,
      "loss": 1.8467,
      "theoretical_loss": 4.4227853684665455,
      "tokens_seen": 186318848
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009530591775325978,
      "loss": 1.8339,
      "theoretical_loss": 4.422589873210806,
      "tokens_seen": 186384384
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009530391173520562,
      "loss": 1.8041,
      "theoretical_loss": 4.422394465921887,
      "tokens_seen": 186449920
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009530190571715146,
      "loss": 1.8145,
      "theoretical_loss": 4.422199146529305,
      "tokens_seen": 186515456
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009529989969909729,
      "loss": 1.8298,
      "theoretical_loss": 4.42200391496266,
      "tokens_seen": 186580992
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009529789368104314,
      "loss": 1.9237,
      "theoretical_loss": 4.4218087711516345,
      "tokens_seen": 186646528
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009529588766298897,
      "loss": 1.9574,
      "theoretical_loss": 4.421613715025988,
      "tokens_seen": 186712064
    },
    {
      "epoch": 0.06,
      "objective/train/advantage_avg": 0.49304935336112976,
      "objective/train/docs_used": 74904,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7934205532073975,
      "objective/train/original_loss": 1.7934205532073975,
      "objective/train/theoretical_loss": 4.421418746515567,
      "objective/train/tokens_used": 207237600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24341203272342682,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505423545837402,
      "objective/train/weighted_lm_loss": 1.8839329481124878,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.421418746515567,
      "tokens_seen": 186777600
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009529388164493481,
      "loss": 1.8573,
      "theoretical_loss": 4.421418746515567,
      "tokens_seen": 186777600
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009529187562688064,
      "loss": 2.0024,
      "theoretical_loss": 4.421223865550291,
      "tokens_seen": 186843136
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009528986960882648,
      "loss": 1.9979,
      "theoretical_loss": 4.421029072060167,
      "tokens_seen": 186908672
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009528786359077232,
      "loss": 2.0378,
      "theoretical_loss": 4.420834365975278,
      "tokens_seen": 186974208
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009528585757271816,
      "loss": 1.8248,
      "theoretical_loss": 4.42063974722579,
      "tokens_seen": 187039744
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009528385155466399,
      "loss": 1.8471,
      "theoretical_loss": 4.420445215741949,
      "tokens_seen": 187105280
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009528184553660984,
      "loss": 1.8749,
      "theoretical_loss": 4.420250771454078,
      "tokens_seen": 187170816
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009527983951855566,
      "loss": 1.9967,
      "theoretical_loss": 4.420056414292584,
      "tokens_seen": 187236352
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009527783350050151,
      "loss": 1.9333,
      "theoretical_loss": 4.419862144187952,
      "tokens_seen": 187301888
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009527582748244735,
      "loss": 1.93,
      "theoretical_loss": 4.419667961070747,
      "tokens_seen": 187367424
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009527382146439318,
      "loss": 2.0093,
      "theoretical_loss": 4.419473864871613,
      "tokens_seen": 187432960
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009527181544633902,
      "loss": 1.9498,
      "theoretical_loss": 4.419279855521276,
      "tokens_seen": 187498496
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009526980942828485,
      "loss": 2.0281,
      "theoretical_loss": 4.419085932950537,
      "tokens_seen": 187564032
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009526780341023069,
      "loss": 1.9989,
      "theoretical_loss": 4.41889209709028,
      "tokens_seen": 187629568
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009526579739217654,
      "loss": 1.8987,
      "theoretical_loss": 4.4186983478714685,
      "tokens_seen": 187695104
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009526379137412236,
      "loss": 1.9608,
      "theoretical_loss": 4.418504685225142,
      "tokens_seen": 187760640
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009526178535606821,
      "loss": 1.9515,
      "theoretical_loss": 4.4183111090824205,
      "tokens_seen": 187826176
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009525977933801405,
      "loss": 2.0248,
      "theoretical_loss": 4.418117619374502,
      "tokens_seen": 187891712
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009525777331995988,
      "loss": 1.9202,
      "theoretical_loss": 4.417924216032667,
      "tokens_seen": 187957248
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009525576730190572,
      "loss": 1.9305,
      "theoretical_loss": 4.417730898988268,
      "tokens_seen": 188022784
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009525376128385155,
      "loss": 1.9592,
      "theoretical_loss": 4.417537668172743,
      "tokens_seen": 188088320
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009525175526579739,
      "loss": 1.8921,
      "theoretical_loss": 4.417344523517602,
      "tokens_seen": 188153856
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009524974924774324,
      "loss": 1.9449,
      "theoretical_loss": 4.417151464954437,
      "tokens_seen": 188219392
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009524774322968906,
      "loss": 1.9775,
      "theoretical_loss": 4.416958492414918,
      "tokens_seen": 188284928
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009524573721163491,
      "loss": 1.9599,
      "theoretical_loss": 4.416765605830792,
      "tokens_seen": 188350464
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009524373119358074,
      "loss": 1.8162,
      "theoretical_loss": 4.4165728051338835,
      "tokens_seen": 188416000
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009524172517552658,
      "loss": 1.9151,
      "theoretical_loss": 4.416380090256095,
      "tokens_seen": 188481536
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009523971915747242,
      "loss": 1.8405,
      "theoretical_loss": 4.416187461129409,
      "tokens_seen": 188547072
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009523771313941826,
      "loss": 2.0353,
      "theoretical_loss": 4.415994917685881,
      "tokens_seen": 188612608
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009523570712136409,
      "loss": 2.0348,
      "theoretical_loss": 4.415802459857648,
      "tokens_seen": 188678144
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009523370110330993,
      "loss": 2.0916,
      "theoretical_loss": 4.415610087576923,
      "tokens_seen": 188743680
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009523169508525576,
      "loss": 1.8956,
      "theoretical_loss": 4.4154178007759946,
      "tokens_seen": 188809216
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009522968906720161,
      "loss": 2.0073,
      "theoretical_loss": 4.415225599387231,
      "tokens_seen": 188874752
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009522768304914744,
      "loss": 1.916,
      "theoretical_loss": 4.415033483343075,
      "tokens_seen": 188940288
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009522567703109328,
      "loss": 1.9299,
      "theoretical_loss": 4.414841452576049,
      "tokens_seen": 189005824
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009522367101303912,
      "loss": 1.9453,
      "theoretical_loss": 4.4146495070187495,
      "tokens_seen": 189071360
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009522166499498496,
      "loss": 1.961,
      "theoretical_loss": 4.41445764660385,
      "tokens_seen": 189136896
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009521965897693079,
      "loss": 1.7822,
      "theoretical_loss": 4.414265871264102,
      "tokens_seen": 189202432
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009521765295887663,
      "loss": 1.8333,
      "theoretical_loss": 4.414074180932333,
      "tokens_seen": 189267968
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009521564694082246,
      "loss": 1.9526,
      "theoretical_loss": 4.413882575541446,
      "tokens_seen": 189333504
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009521364092276831,
      "loss": 1.8715,
      "theoretical_loss": 4.41369105502442,
      "tokens_seen": 189399040
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009521163490471414,
      "loss": 1.8555,
      "theoretical_loss": 4.41349961931431,
      "tokens_seen": 189464576
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009520962888665998,
      "loss": 1.8327,
      "theoretical_loss": 4.413308268344249,
      "tokens_seen": 189530112
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009520762286860581,
      "loss": 1.989,
      "theoretical_loss": 4.413117002047443,
      "tokens_seen": 189595648
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009520561685055166,
      "loss": 1.9247,
      "theoretical_loss": 4.4129258203571755,
      "tokens_seen": 189661184
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009520361083249749,
      "loss": 1.8879,
      "theoretical_loss": 4.412734723206805,
      "tokens_seen": 189726720
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009520160481444333,
      "loss": 1.9158,
      "theoretical_loss": 4.412543710529766,
      "tokens_seen": 189792256
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009519959879638917,
      "loss": 1.9396,
      "theoretical_loss": 4.412352782259568,
      "tokens_seen": 189857792
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009519759277833501,
      "loss": 1.7889,
      "theoretical_loss": 4.4121619383297945,
      "tokens_seen": 189923328
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009519558676028084,
      "loss": 2.0035,
      "theoretical_loss": 4.411971178674106,
      "tokens_seen": 189988864
    },
    {
      "epoch": 0.06,
      "objective/train/advantage_avg": 0.4934732913970947,
      "objective/train/docs_used": 76077,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8127355575561523,
      "objective/train/original_loss": 1.8127355575561523,
      "objective/train/theoretical_loss": 4.411780503226238,
      "objective/train/tokens_used": 210514400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24385586380958557,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505869388580322,
      "objective/train/weighted_lm_loss": 1.9044004678726196,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.411780503226238,
      "tokens_seen": 190054400
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009519358074222668,
      "loss": 1.9662,
      "theoretical_loss": 4.411780503226238,
      "tokens_seen": 190054400
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009519157472417252,
      "loss": 1.9113,
      "theoretical_loss": 4.411589911919999,
      "tokens_seen": 190119936
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009518956870611836,
      "loss": 1.9649,
      "theoretical_loss": 4.411399404689275,
      "tokens_seen": 190185472
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000951875626880642,
      "loss": 2.0698,
      "theoretical_loss": 4.411208981468025,
      "tokens_seen": 190251008
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009518555667001003,
      "loss": 1.8534,
      "theoretical_loss": 4.4110186421902835,
      "tokens_seen": 190316544
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009518355065195588,
      "loss": 1.7358,
      "theoretical_loss": 4.410828386790158,
      "tokens_seen": 190382080
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000951815446339017,
      "loss": 1.7901,
      "theoretical_loss": 4.410638215201831,
      "tokens_seen": 190447616
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009517953861584755,
      "loss": 1.9016,
      "theoretical_loss": 4.410448127359561,
      "tokens_seen": 190513152
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009517753259779339,
      "loss": 2.0252,
      "theoretical_loss": 4.4102581231976785,
      "tokens_seen": 190578688
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009517552657973922,
      "loss": 1.9728,
      "theoretical_loss": 4.410068202650588,
      "tokens_seen": 190644224
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009517352056168506,
      "loss": 1.9752,
      "theoretical_loss": 4.40987836565277,
      "tokens_seen": 190709760
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009517151454363089,
      "loss": 1.8996,
      "theoretical_loss": 4.409688612138778,
      "tokens_seen": 190775296
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009516950852557673,
      "loss": 1.9769,
      "theoretical_loss": 4.409498942043237,
      "tokens_seen": 190840832
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009516750250752258,
      "loss": 2.0145,
      "theoretical_loss": 4.4093093553008496,
      "tokens_seen": 190906368
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000951654964894684,
      "loss": 1.9347,
      "theoretical_loss": 4.409119851846388,
      "tokens_seen": 190971904
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009516349047141425,
      "loss": 1.8087,
      "theoretical_loss": 4.4089304316147,
      "tokens_seen": 191037440
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009516148445336009,
      "loss": 1.9947,
      "theoretical_loss": 4.408741094540707,
      "tokens_seen": 191102976
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009515947843530592,
      "loss": 1.8363,
      "theoretical_loss": 4.408551840559403,
      "tokens_seen": 191168512
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009515747241725176,
      "loss": 1.8745,
      "theoretical_loss": 4.408362669605854,
      "tokens_seen": 191234048
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000951554663991976,
      "loss": 1.9859,
      "theoretical_loss": 4.408173581615199,
      "tokens_seen": 191299584
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009515346038114343,
      "loss": 1.8372,
      "theoretical_loss": 4.407984576522653,
      "tokens_seen": 191365120
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009515145436308928,
      "loss": 2.0925,
      "theoretical_loss": 4.4077956542635,
      "tokens_seen": 191430656
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000951494483450351,
      "loss": 1.9144,
      "theoretical_loss": 4.407606814773099,
      "tokens_seen": 191496192
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009514744232698095,
      "loss": 1.7939,
      "theoretical_loss": 4.407418057986881,
      "tokens_seen": 191561728
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009514543630892678,
      "loss": 1.9196,
      "theoretical_loss": 4.407229383840347,
      "tokens_seen": 191627264
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009514343029087262,
      "loss": 1.9596,
      "theoretical_loss": 4.407040792269075,
      "tokens_seen": 191692800
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009514142427281846,
      "loss": 1.9563,
      "theoretical_loss": 4.406852283208711,
      "tokens_seen": 191758336
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000951394182547643,
      "loss": 1.8223,
      "theoretical_loss": 4.406663856594977,
      "tokens_seen": 191823872
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009513741223671013,
      "loss": 1.8939,
      "theoretical_loss": 4.406475512363663,
      "tokens_seen": 191889408
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009513540621865597,
      "loss": 1.8934,
      "theoretical_loss": 4.4062872504506325,
      "tokens_seen": 191954944
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000951334002006018,
      "loss": 1.7905,
      "theoretical_loss": 4.406099070791822,
      "tokens_seen": 192020480
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009513139418254765,
      "loss": 1.768,
      "theoretical_loss": 4.405910973323239,
      "tokens_seen": 192086016
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009512938816449348,
      "loss": 1.8877,
      "theoretical_loss": 4.405722957980962,
      "tokens_seen": 192151552
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009512738214643932,
      "loss": 1.9167,
      "theoretical_loss": 4.405535024701141,
      "tokens_seen": 192217088
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009512537612838516,
      "loss": 1.9675,
      "theoretical_loss": 4.405347173419998,
      "tokens_seen": 192282624
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00095123370110331,
      "loss": 1.9097,
      "theoretical_loss": 4.405159404073826,
      "tokens_seen": 192348160
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009512136409227683,
      "loss": 1.8425,
      "theoretical_loss": 4.40497171659899,
      "tokens_seen": 192413696
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009511935807422267,
      "loss": 2.008,
      "theoretical_loss": 4.404784110931924,
      "tokens_seen": 192479232
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000951173520561685,
      "loss": 1.7656,
      "theoretical_loss": 4.404596587009136,
      "tokens_seen": 192544768
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009511534603811435,
      "loss": 1.7931,
      "theoretical_loss": 4.404409144767201,
      "tokens_seen": 192610304
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009511334002006018,
      "loss": 1.8794,
      "theoretical_loss": 4.404221784142768,
      "tokens_seen": 192675840
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009511133400200602,
      "loss": 1.8634,
      "theoretical_loss": 4.404034505072556,
      "tokens_seen": 192741376
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009510932798395185,
      "loss": 1.8689,
      "theoretical_loss": 4.403847307493353,
      "tokens_seen": 192806912
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000951073219658977,
      "loss": 1.9959,
      "theoretical_loss": 4.40366019134202,
      "tokens_seen": 192872448
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009510531594784353,
      "loss": 1.9952,
      "theoretical_loss": 4.403473156555487,
      "tokens_seen": 192937984
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009510330992978937,
      "loss": 1.9028,
      "theoretical_loss": 4.403286203070751,
      "tokens_seen": 193003520
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009510130391173521,
      "loss": 1.7787,
      "theoretical_loss": 4.403099330824887,
      "tokens_seen": 193069056
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009509929789368105,
      "loss": 1.9463,
      "theoretical_loss": 4.4029125397550315,
      "tokens_seen": 193134592
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009509729187562688,
      "loss": 1.834,
      "theoretical_loss": 4.402725829798397,
      "tokens_seen": 193200128
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009509528585757272,
      "loss": 1.9024,
      "theoretical_loss": 4.402539200892264,
      "tokens_seen": 193265664
    },
    {
      "epoch": 0.06,
      "objective/train/advantage_avg": 0.49331486225128174,
      "objective/train/docs_used": 77266,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8808577060699463,
      "objective/train/original_loss": 1.8808577060699463,
      "objective/train/theoretical_loss": 4.40235265297398,
      "objective/train/tokens_used": 213791200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24369393289089203,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505704879760742,
      "objective/train/weighted_lm_loss": 1.9757757186889648,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.40235265297398,
      "tokens_seen": 193331200
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009509327983951855,
      "loss": 1.8993,
      "theoretical_loss": 4.40235265297398,
      "tokens_seen": 193331200
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000950912738214644,
      "loss": 1.8289,
      "theoretical_loss": 4.402166185980967,
      "tokens_seen": 193396736
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009508926780341023,
      "loss": 1.8115,
      "theoretical_loss": 4.4019797998507135,
      "tokens_seen": 193462272
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009508726178535607,
      "loss": 1.8387,
      "theoretical_loss": 4.401793494520778,
      "tokens_seen": 193527808
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009508525576730191,
      "loss": 1.7944,
      "theoretical_loss": 4.401607269928788,
      "tokens_seen": 193593344
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009508324974924774,
      "loss": 1.7242,
      "theoretical_loss": 4.4014211260124405,
      "tokens_seen": 193658880
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009508124373119358,
      "loss": 1.8879,
      "theoretical_loss": 4.401235062709502,
      "tokens_seen": 193724416
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009507923771313943,
      "loss": 1.8485,
      "theoretical_loss": 4.401049079957809,
      "tokens_seen": 193789952
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009507723169508525,
      "loss": 1.8177,
      "theoretical_loss": 4.400863177695264,
      "tokens_seen": 193855488
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000950752256770311,
      "loss": 1.8554,
      "theoretical_loss": 4.400677355859841,
      "tokens_seen": 193921024
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009507321965897692,
      "loss": 2.0593,
      "theoretical_loss": 4.400491614389582,
      "tokens_seen": 193986560
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009507121364092277,
      "loss": 1.8376,
      "theoretical_loss": 4.400305953222596,
      "tokens_seen": 194052096
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009506920762286861,
      "loss": 1.9119,
      "theoretical_loss": 4.4001203722970645,
      "tokens_seen": 194117632
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009506720160481444,
      "loss": 1.8258,
      "theoretical_loss": 4.399934871551234,
      "tokens_seen": 194183168
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009506519558676028,
      "loss": 1.6996,
      "theoretical_loss": 4.3997494509234185,
      "tokens_seen": 194248704
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009506318956870613,
      "loss": 1.8192,
      "theoretical_loss": 4.399564110352005,
      "tokens_seen": 194314240
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009506118355065195,
      "loss": 1.9074,
      "theoretical_loss": 4.399378849775443,
      "tokens_seen": 194379776
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000950591775325978,
      "loss": 1.8622,
      "theoretical_loss": 4.399193669132255,
      "tokens_seen": 194445312
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009505717151454362,
      "loss": 1.9398,
      "theoretical_loss": 4.399008568361027,
      "tokens_seen": 194510848
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009505516549648947,
      "loss": 1.8967,
      "theoretical_loss": 4.398823547400417,
      "tokens_seen": 194576384
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009505315947843531,
      "loss": 1.9654,
      "theoretical_loss": 4.398638606189149,
      "tokens_seen": 194641920
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009505115346038114,
      "loss": 1.8578,
      "theoretical_loss": 4.398453744666012,
      "tokens_seen": 194707456
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009504914744232698,
      "loss": 1.9715,
      "theoretical_loss": 4.398268962769867,
      "tokens_seen": 194772992
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009504714142427282,
      "loss": 2.0093,
      "theoretical_loss": 4.39808426043964,
      "tokens_seen": 194838528
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009504513540621865,
      "loss": 1.8911,
      "theoretical_loss": 4.397899637614326,
      "tokens_seen": 194904064
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000950431293881645,
      "loss": 1.9758,
      "theoretical_loss": 4.397715094232984,
      "tokens_seen": 194969600
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009504112337011034,
      "loss": 1.9613,
      "theoretical_loss": 4.397530630234744,
      "tokens_seen": 195035136
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009503911735205617,
      "loss": 1.9392,
      "theoretical_loss": 4.3973462455588,
      "tokens_seen": 195100672
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009503711133400201,
      "loss": 1.8308,
      "theoretical_loss": 4.397161940144415,
      "tokens_seen": 195166208
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009503510531594784,
      "loss": 1.7817,
      "theoretical_loss": 4.396977713930919,
      "tokens_seen": 195231744
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009503309929789369,
      "loss": 1.9156,
      "theoretical_loss": 4.396793566857708,
      "tokens_seen": 195297280
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009503109327983952,
      "loss": 1.8466,
      "theoretical_loss": 4.396609498864243,
      "tokens_seen": 195362816
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009502908726178536,
      "loss": 1.8075,
      "theoretical_loss": 4.396425509890054,
      "tokens_seen": 195428352
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000950270812437312,
      "loss": 1.8059,
      "theoretical_loss": 4.396241599874738,
      "tokens_seen": 195493888
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009502507522567704,
      "loss": 1.8906,
      "theoretical_loss": 4.396057768757957,
      "tokens_seen": 195559424
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009502306920762287,
      "loss": 1.8618,
      "theoretical_loss": 4.395874016479438,
      "tokens_seen": 195624960
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009502106318956871,
      "loss": 1.8689,
      "theoretical_loss": 4.395690342978977,
      "tokens_seen": 195690496
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009501905717151454,
      "loss": 1.7489,
      "theoretical_loss": 4.395506748196435,
      "tokens_seen": 195756032
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009501705115346039,
      "loss": 2.0602,
      "theoretical_loss": 4.395323232071737,
      "tokens_seen": 195821568
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009501504513540622,
      "loss": 1.8469,
      "theoretical_loss": 4.395139794544878,
      "tokens_seen": 195887104
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009501303911735206,
      "loss": 1.9692,
      "theoretical_loss": 4.394956435555916,
      "tokens_seen": 195952640
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009501103309929789,
      "loss": 1.9052,
      "theoretical_loss": 4.394773155044976,
      "tokens_seen": 196018176
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009500902708124374,
      "loss": 1.8058,
      "theoretical_loss": 4.394589952952247,
      "tokens_seen": 196083712
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009500702106318957,
      "loss": 1.787,
      "theoretical_loss": 4.394406829217983,
      "tokens_seen": 196149248
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009500501504513541,
      "loss": 1.8961,
      "theoretical_loss": 4.394223783782509,
      "tokens_seen": 196214784
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009500300902708125,
      "loss": 1.7633,
      "theoretical_loss": 4.394040816586209,
      "tokens_seen": 196280320
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009500100300902709,
      "loss": 1.7421,
      "theoretical_loss": 4.393857927569534,
      "tokens_seen": 196345856
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009499899699097292,
      "loss": 1.9565,
      "theoretical_loss": 4.393675116673003,
      "tokens_seen": 196411392
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009499699097291876,
      "loss": 1.8156,
      "theoretical_loss": 4.393492383837196,
      "tokens_seen": 196476928
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009499498495486459,
      "loss": 1.8552,
      "theoretical_loss": 4.393309729002761,
      "tokens_seen": 196542464
    },
    {
      "epoch": 0.06,
      "objective/train/advantage_avg": 0.4933859407901764,
      "objective/train/docs_used": 78313,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5727931261062622,
      "objective/train/original_loss": 1.5727932453155518,
      "objective/train/theoretical_loss": 4.393127152110409,
      "objective/train/tokens_used": 217068000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24412523210048676,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05057954788208,
      "objective/train/weighted_lm_loss": 1.652202844619751,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.393127152110409,
      "tokens_seen": 196608000
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009499297893681044,
      "loss": 1.7561,
      "theoretical_loss": 4.393127152110409,
      "tokens_seen": 196608000
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009499097291875627,
      "loss": 1.855,
      "theoretical_loss": 4.392944653100918,
      "tokens_seen": 196673536
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009498896690070211,
      "loss": 1.9183,
      "theoretical_loss": 4.392762231915128,
      "tokens_seen": 196739072
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009498696088264795,
      "loss": 1.8333,
      "theoretical_loss": 4.392579888493946,
      "tokens_seen": 196804608
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009498495486459378,
      "loss": 1.8654,
      "theoretical_loss": 4.392397622778343,
      "tokens_seen": 196870144
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009498294884653962,
      "loss": 1.8546,
      "theoretical_loss": 4.392215434709351,
      "tokens_seen": 196935680
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009498094282848547,
      "loss": 1.9817,
      "theoretical_loss": 4.392033324228073,
      "tokens_seen": 197001216
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009497893681043129,
      "loss": 1.8226,
      "theoretical_loss": 4.391851291275671,
      "tokens_seen": 197066752
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009497693079237714,
      "loss": 1.8504,
      "theoretical_loss": 4.391669335793372,
      "tokens_seen": 197132288
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009497492477432296,
      "loss": 1.7561,
      "theoretical_loss": 4.391487457722469,
      "tokens_seen": 197197824
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009497291875626881,
      "loss": 1.671,
      "theoretical_loss": 4.391305657004317,
      "tokens_seen": 197263360
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009497091273821465,
      "loss": 1.9367,
      "theoretical_loss": 4.391123933580336,
      "tokens_seen": 197328896
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009496890672016048,
      "loss": 1.9203,
      "theoretical_loss": 4.39094228739201,
      "tokens_seen": 197394432
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009496690070210632,
      "loss": 1.8818,
      "theoretical_loss": 4.390760718380884,
      "tokens_seen": 197459968
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009496489468405217,
      "loss": 1.8564,
      "theoretical_loss": 4.390579226488571,
      "tokens_seen": 197525504
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009496288866599799,
      "loss": 1.9317,
      "theoretical_loss": 4.3903978116567455,
      "tokens_seen": 197591040
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009496088264794384,
      "loss": 1.9909,
      "theoretical_loss": 4.390216473827143,
      "tokens_seen": 197656576
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009495887662988966,
      "loss": 1.7772,
      "theoretical_loss": 4.390035212941568,
      "tokens_seen": 197722112
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009495687061183551,
      "loss": 1.7826,
      "theoretical_loss": 4.3898540289418815,
      "tokens_seen": 197787648
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009495486459378135,
      "loss": 1.8518,
      "theoretical_loss": 4.389672921770013,
      "tokens_seen": 197853184
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009495285857572718,
      "loss": 1.7901,
      "theoretical_loss": 4.389491891367953,
      "tokens_seen": 197918720
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009495085255767302,
      "loss": 1.8022,
      "theoretical_loss": 4.389310937677754,
      "tokens_seen": 197984256
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009494884653961886,
      "loss": 1.9155,
      "theoretical_loss": 4.389130060641534,
      "tokens_seen": 198049792
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009494684052156469,
      "loss": 1.8942,
      "theoretical_loss": 4.388949260201471,
      "tokens_seen": 198115328
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009494483450351054,
      "loss": 1.8778,
      "theoretical_loss": 4.388768536299808,
      "tokens_seen": 198180864
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009494282848545636,
      "loss": 1.9305,
      "theoretical_loss": 4.388587888878849,
      "tokens_seen": 198246400
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009494082246740221,
      "loss": 1.9257,
      "theoretical_loss": 4.388407317880962,
      "tokens_seen": 198311936
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009493881644934804,
      "loss": 1.9663,
      "theoretical_loss": 4.388226823248576,
      "tokens_seen": 198377472
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009493681043129388,
      "loss": 1.9069,
      "theoretical_loss": 4.388046404924184,
      "tokens_seen": 198443008
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009493480441323972,
      "loss": 1.8286,
      "theoretical_loss": 4.387866062850339,
      "tokens_seen": 198508544
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009493279839518556,
      "loss": 1.915,
      "theoretical_loss": 4.3876857969696585,
      "tokens_seen": 198574080
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009493079237713139,
      "loss": 1.8648,
      "theoretical_loss": 4.387505607224821,
      "tokens_seen": 198639616
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009492878635907724,
      "loss": 1.8246,
      "theoretical_loss": 4.387325493558566,
      "tokens_seen": 198705152
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009492678034102307,
      "loss": 1.8195,
      "theoretical_loss": 4.387145455913698,
      "tokens_seen": 198770688
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009492477432296891,
      "loss": 1.7988,
      "theoretical_loss": 4.38696549423308,
      "tokens_seen": 198836224
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009492276830491474,
      "loss": 1.9393,
      "theoretical_loss": 4.386785608459639,
      "tokens_seen": 198901760
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009492076228686058,
      "loss": 1.9559,
      "theoretical_loss": 4.386605798536362,
      "tokens_seen": 198967296
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009491875626880642,
      "loss": 1.945,
      "theoretical_loss": 4.3864260644063,
      "tokens_seen": 199032832
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009491675025075226,
      "loss": 1.8763,
      "theoretical_loss": 4.386246406012562,
      "tokens_seen": 199098368
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009491474423269809,
      "loss": 1.9405,
      "theoretical_loss": 4.3860668232983215,
      "tokens_seen": 199163904
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009491273821464393,
      "loss": 1.962,
      "theoretical_loss": 4.385887316206812,
      "tokens_seen": 199229440
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009491073219658977,
      "loss": 1.8279,
      "theoretical_loss": 4.385707884681327,
      "tokens_seen": 199294976
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009490872617853561,
      "loss": 1.7523,
      "theoretical_loss": 4.385528528665225,
      "tokens_seen": 199360512
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009490672016048144,
      "loss": 1.8032,
      "theoretical_loss": 4.385349248101922,
      "tokens_seen": 199426048
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009490471414242729,
      "loss": 1.8202,
      "theoretical_loss": 4.385170042934896,
      "tokens_seen": 199491584
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009490270812437311,
      "loss": 1.8708,
      "theoretical_loss": 4.384990913107686,
      "tokens_seen": 199557120
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009490070210631896,
      "loss": 1.8222,
      "theoretical_loss": 4.384811858563891,
      "tokens_seen": 199622656
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009489869608826479,
      "loss": 1.7764,
      "theoretical_loss": 4.384632879247173,
      "tokens_seen": 199688192
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009489669007021063,
      "loss": 1.865,
      "theoretical_loss": 4.384453975101251,
      "tokens_seen": 199753728
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009489468405215648,
      "loss": 1.8772,
      "theoretical_loss": 4.384275146069909,
      "tokens_seen": 199819264
    },
    {
      "epoch": 0.06,
      "objective/train/advantage_avg": 0.4942682981491089,
      "objective/train/docs_used": 79488,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6611281633377075,
      "objective/train/original_loss": 1.661128282546997,
      "objective/train/theoretical_loss": 4.384096392096988,
      "objective/train/tokens_used": 220344800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24450337886810303,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05066978931427,
      "objective/train/weighted_lm_loss": 1.7454839944839478,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.384096392096988,
      "tokens_seen": 199884800
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009489267803410231,
      "loss": 1.8157,
      "theoretical_loss": 4.384096392096988,
      "tokens_seen": 199884800
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009489067201604815,
      "loss": 1.9058,
      "theoretical_loss": 4.38391771312639,
      "tokens_seen": 199950336
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009488866599799399,
      "loss": 1.9625,
      "theoretical_loss": 4.38373910910208,
      "tokens_seen": 200015872
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009488665997993982,
      "loss": 1.8666,
      "theoretical_loss": 4.383560579968078,
      "tokens_seen": 200081408
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009488465396188566,
      "loss": 1.7205,
      "theoretical_loss": 4.3833821256684695,
      "tokens_seen": 200146944
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009488264794383151,
      "loss": 1.8827,
      "theoretical_loss": 4.383203746147397,
      "tokens_seen": 200212480
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009488064192577733,
      "loss": 1.9402,
      "theoretical_loss": 4.383025441349063,
      "tokens_seen": 200278016
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009487863590772318,
      "loss": 1.8922,
      "theoretical_loss": 4.382847211217731,
      "tokens_seen": 200343552
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00094876629889669,
      "loss": 1.9755,
      "theoretical_loss": 4.3826690556977255,
      "tokens_seen": 200409088
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009487462387161485,
      "loss": 1.9237,
      "theoretical_loss": 4.382490974733425,
      "tokens_seen": 200474624
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009487261785356069,
      "loss": 2.0019,
      "theoretical_loss": 4.382312968269276,
      "tokens_seen": 200540160
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009487061183550652,
      "loss": 1.8139,
      "theoretical_loss": 4.382135036249777,
      "tokens_seen": 200605696
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009486860581745236,
      "loss": 1.8938,
      "theoretical_loss": 4.3819571786194915,
      "tokens_seen": 200671232
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009486659979939821,
      "loss": 1.8056,
      "theoretical_loss": 4.381779395323038,
      "tokens_seen": 200736768
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009486459378134403,
      "loss": 1.8419,
      "theoretical_loss": 4.381601686305098,
      "tokens_seen": 200802304
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009486258776328988,
      "loss": 1.7723,
      "theoretical_loss": 4.381424051510409,
      "tokens_seen": 200867840
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000948605817452357,
      "loss": 1.8334,
      "theoretical_loss": 4.38124649088377,
      "tokens_seen": 200933376
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009485857572718155,
      "loss": 1.8382,
      "theoretical_loss": 4.381069004370039,
      "tokens_seen": 200998912
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009485656970912739,
      "loss": 1.8665,
      "theoretical_loss": 4.38089159191413,
      "tokens_seen": 201064448
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009485456369107322,
      "loss": 1.7653,
      "theoretical_loss": 4.38071425346102,
      "tokens_seen": 201129984
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009485255767301906,
      "loss": 1.8926,
      "theoretical_loss": 4.380536988955742,
      "tokens_seen": 201195520
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000948505516549649,
      "loss": 1.7422,
      "theoretical_loss": 4.380359798343388,
      "tokens_seen": 201261056
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009484854563691073,
      "loss": 1.707,
      "theoretical_loss": 4.380182681569111,
      "tokens_seen": 201326592
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009484653961885658,
      "loss": 1.8416,
      "theoretical_loss": 4.380005638578119,
      "tokens_seen": 201392128
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000948445336008024,
      "loss": 1.8569,
      "theoretical_loss": 4.37982866931568,
      "tokens_seen": 201457664
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009484252758274825,
      "loss": 1.82,
      "theoretical_loss": 4.379651773727122,
      "tokens_seen": 201523200
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009484052156469408,
      "loss": 1.978,
      "theoretical_loss": 4.379474951757829,
      "tokens_seen": 201588736
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009483851554663992,
      "loss": 1.7683,
      "theoretical_loss": 4.379298203353244,
      "tokens_seen": 201654272
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009483650952858576,
      "loss": 1.9227,
      "theoretical_loss": 4.379121528458869,
      "tokens_seen": 201719808
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000948345035105316,
      "loss": 1.8948,
      "theoretical_loss": 4.378944927020262,
      "tokens_seen": 201785344
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009483249749247743,
      "loss": 1.939,
      "theoretical_loss": 4.378768398983042,
      "tokens_seen": 201850880
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009483049147442328,
      "loss": 1.8403,
      "theoretical_loss": 4.378591944292882,
      "tokens_seen": 201916416
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000948284854563691,
      "loss": 1.8222,
      "theoretical_loss": 4.378415562895518,
      "tokens_seen": 201981952
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009482647943831495,
      "loss": 1.7433,
      "theoretical_loss": 4.378239254736738,
      "tokens_seen": 202047488
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009482447342026078,
      "loss": 1.9094,
      "theoretical_loss": 4.378063019762392,
      "tokens_seen": 202113024
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009482246740220662,
      "loss": 1.7928,
      "theoretical_loss": 4.377886857918385,
      "tokens_seen": 202178560
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009482046138415246,
      "loss": 1.8734,
      "theoretical_loss": 4.377710769150681,
      "tokens_seen": 202244096
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000948184553660983,
      "loss": 1.9099,
      "theoretical_loss": 4.3775347534053015,
      "tokens_seen": 202309632
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009481644934804413,
      "loss": 1.7542,
      "theoretical_loss": 4.377358810628324,
      "tokens_seen": 202375168
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009481444332998997,
      "loss": 1.8399,
      "theoretical_loss": 4.377182940765883,
      "tokens_seen": 202440704
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009481243731193581,
      "loss": 1.7524,
      "theoretical_loss": 4.377007143764175,
      "tokens_seen": 202506240
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009481043129388165,
      "loss": 1.8736,
      "theoretical_loss": 4.376831419569445,
      "tokens_seen": 202571776
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009480842527582748,
      "loss": 1.9028,
      "theoretical_loss": 4.3766557681280025,
      "tokens_seen": 202637312
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009480641925777333,
      "loss": 1.9198,
      "theoretical_loss": 4.3764801893862115,
      "tokens_seen": 202702848
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009480441323971915,
      "loss": 1.8034,
      "theoretical_loss": 4.376304683290491,
      "tokens_seen": 202768384
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00094802407221665,
      "loss": 1.8115,
      "theoretical_loss": 4.37612924978732,
      "tokens_seen": 202833920
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009480040120361083,
      "loss": 1.8673,
      "theoretical_loss": 4.375953888823233,
      "tokens_seen": 202899456
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009479839518555667,
      "loss": 1.7649,
      "theoretical_loss": 4.375778600344819,
      "tokens_seen": 202964992
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009479638916750251,
      "loss": 1.7969,
      "theoretical_loss": 4.375603384298726,
      "tokens_seen": 203030528
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009479438314944835,
      "loss": 1.8562,
      "theoretical_loss": 4.375428240631659,
      "tokens_seen": 203096064
    },
    {
      "epoch": 0.06,
      "objective/train/advantage_avg": 0.49045994877815247,
      "objective/train/docs_used": 80605,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8460792303085327,
      "objective/train/original_loss": 1.8460791110992432,
      "objective/train/theoretical_loss": 4.375253169290376,
      "objective/train/tokens_used": 223621600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24119678139686584,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050271987915039,
      "objective/train/weighted_lm_loss": 1.9390350580215454,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.375253169290376,
      "tokens_seen": 203161600
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009479237713139418,
      "loss": 1.9466,
      "theoretical_loss": 4.375253169290376,
      "tokens_seen": 203161600
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009479037111334003,
      "loss": 1.9524,
      "theoretical_loss": 4.3750781702216965,
      "tokens_seen": 203227136
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009478836509528585,
      "loss": 1.7591,
      "theoretical_loss": 4.37490324337249,
      "tokens_seen": 203292672
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000947863590772317,
      "loss": 1.8594,
      "theoretical_loss": 4.374728388689688,
      "tokens_seen": 203358208
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009478435305917753,
      "loss": 1.8902,
      "theoretical_loss": 4.374553606120274,
      "tokens_seen": 203423744
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009478234704112337,
      "loss": 1.8355,
      "theoretical_loss": 4.374378895611289,
      "tokens_seen": 203489280
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009478034102306921,
      "loss": 1.9065,
      "theoretical_loss": 4.374204257109832,
      "tokens_seen": 203554816
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009477833500501504,
      "loss": 1.7489,
      "theoretical_loss": 4.374029690563053,
      "tokens_seen": 203620352
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009477632898696088,
      "loss": 1.8441,
      "theoretical_loss": 4.373855195918162,
      "tokens_seen": 203685888
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009477432296890673,
      "loss": 1.9567,
      "theoretical_loss": 4.373680773122424,
      "tokens_seen": 203751424
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009477231695085255,
      "loss": 1.8593,
      "theoretical_loss": 4.3735064221231585,
      "tokens_seen": 203816960
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000947703109327984,
      "loss": 1.9252,
      "theoretical_loss": 4.373332142867739,
      "tokens_seen": 203882496
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009476830491474422,
      "loss": 1.8711,
      "theoretical_loss": 4.3731579353036,
      "tokens_seen": 203948032
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009476629889669007,
      "loss": 1.8338,
      "theoretical_loss": 4.372983799378225,
      "tokens_seen": 204013568
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009476429287863591,
      "loss": 1.8558,
      "theoretical_loss": 4.372809735039157,
      "tokens_seen": 204079104
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009476228686058174,
      "loss": 1.8181,
      "theoretical_loss": 4.372635742233991,
      "tokens_seen": 204144640
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009476028084252758,
      "loss": 1.7501,
      "theoretical_loss": 4.372461820910382,
      "tokens_seen": 204210176
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009475827482447343,
      "loss": 1.8748,
      "theoretical_loss": 4.372287971016036,
      "tokens_seen": 204275712
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009475626880641925,
      "loss": 1.7815,
      "theoretical_loss": 4.372114192498714,
      "tokens_seen": 204341248
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000947542627883651,
      "loss": 1.9619,
      "theoretical_loss": 4.371940485306234,
      "tokens_seen": 204406784
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009475225677031093,
      "loss": 1.8881,
      "theoretical_loss": 4.371766849386468,
      "tokens_seen": 204472320
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009475025075225677,
      "loss": 1.9036,
      "theoretical_loss": 4.371593284687343,
      "tokens_seen": 204537856
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009474824473420261,
      "loss": 1.7205,
      "theoretical_loss": 4.371419791156841,
      "tokens_seen": 204603392
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009474623871614844,
      "loss": 1.8684,
      "theoretical_loss": 4.371246368742996,
      "tokens_seen": 204668928
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009474423269809429,
      "loss": 2.0226,
      "theoretical_loss": 4.3710730173939005,
      "tokens_seen": 204734464
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009474222668004012,
      "loss": 1.7672,
      "theoretical_loss": 4.3708997370577,
      "tokens_seen": 204800000
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009474022066198596,
      "loss": 1.9649,
      "theoretical_loss": 4.370726527682593,
      "tokens_seen": 204865536
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000947382146439318,
      "loss": 1.9361,
      "theoretical_loss": 4.370553389216834,
      "tokens_seen": 204931072
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009473620862587764,
      "loss": 1.8473,
      "theoretical_loss": 4.370380321608731,
      "tokens_seen": 204996608
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009473420260782347,
      "loss": 1.978,
      "theoretical_loss": 4.370207324806645,
      "tokens_seen": 205062144
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009473219658976932,
      "loss": 1.8565,
      "theoretical_loss": 4.370034398758996,
      "tokens_seen": 205127680
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009473019057171515,
      "loss": 1.7406,
      "theoretical_loss": 4.369861543414251,
      "tokens_seen": 205193216
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009472818455366099,
      "loss": 2.0285,
      "theoretical_loss": 4.369688758720937,
      "tokens_seen": 205258752
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009472617853560682,
      "loss": 1.851,
      "theoretical_loss": 4.36951604462763,
      "tokens_seen": 205324288
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009472417251755266,
      "loss": 1.8151,
      "theoretical_loss": 4.369343401082966,
      "tokens_seen": 205389824
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000947221664994985,
      "loss": 1.7439,
      "theoretical_loss": 4.369170828035628,
      "tokens_seen": 205455360
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009472016048144434,
      "loss": 1.8962,
      "theoretical_loss": 4.368998325434355,
      "tokens_seen": 205520896
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009471815446339017,
      "loss": 1.8172,
      "theoretical_loss": 4.368825893227943,
      "tokens_seen": 205586432
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009471614844533601,
      "loss": 1.7986,
      "theoretical_loss": 4.368653531365237,
      "tokens_seen": 205651968
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009471414242728185,
      "loss": 1.8557,
      "theoretical_loss": 4.3684812397951385,
      "tokens_seen": 205717504
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009471213640922769,
      "loss": 1.7482,
      "theoretical_loss": 4.3683090184666,
      "tokens_seen": 205783040
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009471013039117352,
      "loss": 1.7962,
      "theoretical_loss": 4.368136867328628,
      "tokens_seen": 205848576
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009470812437311937,
      "loss": 1.7665,
      "theoretical_loss": 4.367964786330285,
      "tokens_seen": 205914112
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009470611835506519,
      "loss": 1.8658,
      "theoretical_loss": 4.367792775420682,
      "tokens_seen": 205979648
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009470411233701104,
      "loss": 1.7218,
      "theoretical_loss": 4.367620834548987,
      "tokens_seen": 206045184
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009470210631895687,
      "loss": 1.8305,
      "theoretical_loss": 4.367448963664419,
      "tokens_seen": 206110720
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009470010030090271,
      "loss": 1.8604,
      "theoretical_loss": 4.367277162716249,
      "tokens_seen": 206176256
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009469809428284855,
      "loss": 1.8282,
      "theoretical_loss": 4.367105431653806,
      "tokens_seen": 206241792
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009469608826479439,
      "loss": 1.8322,
      "theoretical_loss": 4.3669337704264635,
      "tokens_seen": 206307328
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009469408224674022,
      "loss": 1.891,
      "theoretical_loss": 4.366762178983656,
      "tokens_seen": 206372864
    },
    {
      "epoch": 0.06,
      "objective/train/advantage_avg": 0.4905458092689514,
      "objective/train/docs_used": 81724,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7321252822875977,
      "objective/train/original_loss": 1.7321252822875977,
      "objective/train/theoretical_loss": 4.366590657274865,
      "objective/train/tokens_used": 226898400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24100235104560852,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502796173095703,
      "objective/train/weighted_lm_loss": 1.8194676637649536,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 4.366590657274865,
      "tokens_seen": 206438400
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009469207622868607,
      "loss": 1.82,
      "theoretical_loss": 4.366590657274865,
      "tokens_seen": 206438400
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009469007021063189,
      "loss": 1.8988,
      "theoretical_loss": 4.366419205249628,
      "tokens_seen": 206503936
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009468806419257774,
      "loss": 1.9824,
      "theoretical_loss": 4.366247822857533,
      "tokens_seen": 206569472
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009468605817452357,
      "loss": 1.8839,
      "theoretical_loss": 4.36607651004822,
      "tokens_seen": 206635008
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009468405215646941,
      "loss": 1.8377,
      "theoretical_loss": 4.365905266771383,
      "tokens_seen": 206700544
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009468204613841525,
      "loss": 1.9625,
      "theoretical_loss": 4.36573409297677,
      "tokens_seen": 206766080
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009468004012036108,
      "loss": 1.7877,
      "theoretical_loss": 4.365562988614176,
      "tokens_seen": 206831616
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009467803410230692,
      "loss": 1.9814,
      "theoretical_loss": 4.365391953633453,
      "tokens_seen": 206897152
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009467602808425277,
      "loss": 1.8504,
      "theoretical_loss": 4.365220987984502,
      "tokens_seen": 206962688
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009467402206619859,
      "loss": 1.5948,
      "theoretical_loss": 4.365050091617277,
      "tokens_seen": 207028224
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009467201604814444,
      "loss": 1.8719,
      "theoretical_loss": 4.364879264481787,
      "tokens_seen": 207093760
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009467001003009026,
      "loss": 1.9276,
      "theoretical_loss": 4.364708506528087,
      "tokens_seen": 207159296
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009466800401203611,
      "loss": 1.7526,
      "theoretical_loss": 4.364537817706289,
      "tokens_seen": 207224832
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009466599799398195,
      "loss": 1.8202,
      "theoretical_loss": 4.364367197966552,
      "tokens_seen": 207290368
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009466399197592778,
      "loss": 1.7798,
      "theoretical_loss": 4.364196647259092,
      "tokens_seen": 207355904
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009466198595787362,
      "loss": 1.8522,
      "theoretical_loss": 4.364026165534174,
      "tokens_seen": 207421440
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009465997993981947,
      "loss": 1.7636,
      "theoretical_loss": 4.363855752742112,
      "tokens_seen": 207486976
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009465797392176529,
      "loss": 1.9009,
      "theoretical_loss": 4.363685408833276,
      "tokens_seen": 207552512
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009465596790371114,
      "loss": 1.9654,
      "theoretical_loss": 4.363515133758084,
      "tokens_seen": 207618048
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009465396188565697,
      "loss": 1.9998,
      "theoretical_loss": 4.363344927467009,
      "tokens_seen": 207683584
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009465195586760281,
      "loss": 1.9099,
      "theoretical_loss": 4.36317478991057,
      "tokens_seen": 207749120
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009464994984954865,
      "loss": 1.8827,
      "theoretical_loss": 4.363004721039341,
      "tokens_seen": 207814656
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009464794383149448,
      "loss": 1.8781,
      "theoretical_loss": 4.3628347208039475,
      "tokens_seen": 207880192
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009464593781344032,
      "loss": 1.7533,
      "theoretical_loss": 4.362664789155063,
      "tokens_seen": 207945728
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009464393179538616,
      "loss": 1.661,
      "theoretical_loss": 4.362494926043414,
      "tokens_seen": 208011264
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009464192577733199,
      "loss": 1.7946,
      "theoretical_loss": 4.362325131419779,
      "tokens_seen": 208076800
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009463991975927784,
      "loss": 1.9939,
      "theoretical_loss": 4.362155405234985,
      "tokens_seen": 208142336
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009463791374122367,
      "loss": 1.8894,
      "theoretical_loss": 4.361985747439911,
      "tokens_seen": 208207872
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009463590772316951,
      "loss": 1.9331,
      "theoretical_loss": 4.361816157985485,
      "tokens_seen": 208273408
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009463390170511535,
      "loss": 1.8647,
      "theoretical_loss": 4.361646636822689,
      "tokens_seen": 208338944
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009463189568706119,
      "loss": 2.0205,
      "theoretical_loss": 4.361477183902554,
      "tokens_seen": 208404480
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009462988966900702,
      "loss": 1.8844,
      "theoretical_loss": 4.361307799176159,
      "tokens_seen": 208470016
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009462788365095286,
      "loss": 1.8218,
      "theoretical_loss": 4.361138482594637,
      "tokens_seen": 208535552
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009462587763289869,
      "loss": 1.8755,
      "theoretical_loss": 4.360969234109169,
      "tokens_seen": 208601088
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009462387161484454,
      "loss": 1.8704,
      "theoretical_loss": 4.360800053670989,
      "tokens_seen": 208666624
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009462186559679037,
      "loss": 1.7776,
      "theoretical_loss": 4.360630941231378,
      "tokens_seen": 208732160
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009461985957873621,
      "loss": 1.7952,
      "theoretical_loss": 4.360461896741668,
      "tokens_seen": 208797696
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009461785356068204,
      "loss": 1.8615,
      "theoretical_loss": 4.360292920153244,
      "tokens_seen": 208863232
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009461584754262789,
      "loss": 1.8697,
      "theoretical_loss": 4.360124011417536,
      "tokens_seen": 208928768
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009461384152457372,
      "loss": 1.9524,
      "theoretical_loss": 4.359955170486029,
      "tokens_seen": 208994304
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009461183550651956,
      "loss": 1.7694,
      "theoretical_loss": 4.359786397310254,
      "tokens_seen": 209059840
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000946098294884654,
      "loss": 1.7871,
      "theoretical_loss": 4.359617691841795,
      "tokens_seen": 209125376
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009460782347041123,
      "loss": 1.9203,
      "theoretical_loss": 4.359449054032282,
      "tokens_seen": 209190912
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009460581745235707,
      "loss": 1.8939,
      "theoretical_loss": 4.3592804838334,
      "tokens_seen": 209256448
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009460381143430291,
      "loss": 1.7299,
      "theoretical_loss": 4.359111981196877,
      "tokens_seen": 209321984
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009460180541624874,
      "loss": 1.9227,
      "theoretical_loss": 4.358943546074497,
      "tokens_seen": 209387520
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009459979939819459,
      "loss": 1.8355,
      "theoretical_loss": 4.358775178418089,
      "tokens_seen": 209453056
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009459779338014042,
      "loss": 1.9658,
      "theoretical_loss": 4.358606878179533,
      "tokens_seen": 209518592
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009459578736208626,
      "loss": 1.8921,
      "theoretical_loss": 4.358438645310758,
      "tokens_seen": 209584128
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009459378134403211,
      "loss": 1.7757,
      "theoretical_loss": 4.3582704797637435,
      "tokens_seen": 209649664
    },
    {
      "epoch": 0.06,
      "objective/train/advantage_avg": 0.49177494645118713,
      "objective/train/docs_used": 82917,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6986033916473389,
      "objective/train/original_loss": 1.6986033916473389,
      "objective/train/theoretical_loss": 4.358102381490517,
      "objective/train/tokens_used": 230175200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24229730665683746,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050409197807312,
      "objective/train/weighted_lm_loss": 1.7842323780059814,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.358102381490517,
      "tokens_seen": 209715200
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009459177532597793,
      "loss": 1.8477,
      "theoretical_loss": 4.358102381490517,
      "tokens_seen": 209715200
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009458976930792378,
      "loss": 1.8678,
      "theoretical_loss": 4.357934350443156,
      "tokens_seen": 209780736
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009458776328986961,
      "loss": 1.8636,
      "theoretical_loss": 4.357766386573785,
      "tokens_seen": 209846272
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009458575727181545,
      "loss": 1.8975,
      "theoretical_loss": 4.357598489834581,
      "tokens_seen": 209911808
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009458375125376129,
      "loss": 1.8138,
      "theoretical_loss": 4.3574306601777675,
      "tokens_seen": 209977344
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009458174523570712,
      "loss": 1.8093,
      "theoretical_loss": 4.357262897555617,
      "tokens_seen": 210042880
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009457973921765296,
      "loss": 1.8011,
      "theoretical_loss": 4.357095201920452,
      "tokens_seen": 210108416
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009457773319959881,
      "loss": 1.7639,
      "theoretical_loss": 4.356927573224643,
      "tokens_seen": 210173952
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009457572718154463,
      "loss": 1.7829,
      "theoretical_loss": 4.356760011420608,
      "tokens_seen": 210239488
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009457372116349048,
      "loss": 1.9724,
      "theoretical_loss": 4.356592516460817,
      "tokens_seen": 210305024
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000945717151454363,
      "loss": 1.992,
      "theoretical_loss": 4.356425088297786,
      "tokens_seen": 210370560
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009456970912738215,
      "loss": 1.8751,
      "theoretical_loss": 4.3562577268840785,
      "tokens_seen": 210436096
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009456770310932799,
      "loss": 1.9101,
      "theoretical_loss": 4.3560904321723095,
      "tokens_seen": 210501632
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009456569709127382,
      "loss": 1.9617,
      "theoretical_loss": 4.355923204115141,
      "tokens_seen": 210567168
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009456369107321966,
      "loss": 1.9317,
      "theoretical_loss": 4.3557560426652815,
      "tokens_seen": 210632704
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009456168505516551,
      "loss": 1.7983,
      "theoretical_loss": 4.355588947775491,
      "tokens_seen": 210698240
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009455967903711133,
      "loss": 1.8471,
      "theoretical_loss": 4.355421919398576,
      "tokens_seen": 210763776
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009455767301905718,
      "loss": 1.8127,
      "theoretical_loss": 4.35525495748739,
      "tokens_seen": 210829312
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00094555667001003,
      "loss": 1.8958,
      "theoretical_loss": 4.355088061994837,
      "tokens_seen": 210894848
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009455366098294885,
      "loss": 1.9029,
      "theoretical_loss": 4.354921232873868,
      "tokens_seen": 210960384
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009455165496489469,
      "loss": 1.876,
      "theoretical_loss": 4.354754470077481,
      "tokens_seen": 211025920
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009454964894684052,
      "loss": 1.9396,
      "theoretical_loss": 4.354587773558722,
      "tokens_seen": 211091456
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009454764292878636,
      "loss": 1.8428,
      "theoretical_loss": 4.354421143270687,
      "tokens_seen": 211156992
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000945456369107322,
      "loss": 1.7118,
      "theoretical_loss": 4.354254579166517,
      "tokens_seen": 211222528
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009454363089267803,
      "loss": 1.8104,
      "theoretical_loss": 4.354088081199402,
      "tokens_seen": 211288064
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009454162487462388,
      "loss": 1.8452,
      "theoretical_loss": 4.353921649322579,
      "tokens_seen": 211353600
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009453961885656971,
      "loss": 1.8087,
      "theoretical_loss": 4.353755283489334,
      "tokens_seen": 211419136
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009453761283851555,
      "loss": 1.8845,
      "theoretical_loss": 4.353588983652999,
      "tokens_seen": 211484672
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009453560682046139,
      "loss": 1.8368,
      "theoretical_loss": 4.3534227497669535,
      "tokens_seen": 211550208
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009453360080240723,
      "loss": 1.9266,
      "theoretical_loss": 4.353256581784625,
      "tokens_seen": 211615744
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009453159478435306,
      "loss": 1.7949,
      "theoretical_loss": 4.353090479659488,
      "tokens_seen": 211681280
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000945295887662989,
      "loss": 1.8264,
      "theoretical_loss": 4.3529244433450645,
      "tokens_seen": 211746816
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009452758274824473,
      "loss": 1.7892,
      "theoretical_loss": 4.352758472794923,
      "tokens_seen": 211812352
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009452557673019058,
      "loss": 1.8234,
      "theoretical_loss": 4.35259256796268,
      "tokens_seen": 211877888
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009452357071213641,
      "loss": 1.8501,
      "theoretical_loss": 4.352426728801999,
      "tokens_seen": 211943424
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009452156469408225,
      "loss": 1.8889,
      "theoretical_loss": 4.352260955266589,
      "tokens_seen": 212008960
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009451955867602808,
      "loss": 1.9609,
      "theoretical_loss": 4.352095247310208,
      "tokens_seen": 212074496
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009451755265797393,
      "loss": 1.7887,
      "theoretical_loss": 4.351929604886658,
      "tokens_seen": 212140032
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009451554663991976,
      "loss": 1.8833,
      "theoretical_loss": 4.351764027949793,
      "tokens_seen": 212205568
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000945135406218656,
      "loss": 1.9028,
      "theoretical_loss": 4.3515985164535085,
      "tokens_seen": 212271104
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009451153460381143,
      "loss": 1.8628,
      "theoretical_loss": 4.351433070351748,
      "tokens_seen": 212336640
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009450952858575727,
      "loss": 1.9107,
      "theoretical_loss": 4.351267689598503,
      "tokens_seen": 212402176
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009450752256770311,
      "loss": 1.9297,
      "theoretical_loss": 4.3511023741478105,
      "tokens_seen": 212467712
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009450551654964895,
      "loss": 1.8527,
      "theoretical_loss": 4.350937123953755,
      "tokens_seen": 212533248
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009450351053159478,
      "loss": 1.8359,
      "theoretical_loss": 4.350771938970466,
      "tokens_seen": 212598784
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009450150451354063,
      "loss": 1.9545,
      "theoretical_loss": 4.35060681915212,
      "tokens_seen": 212664320
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009449949849548646,
      "loss": 1.6537,
      "theoretical_loss": 4.35044176445294,
      "tokens_seen": 212729856
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000944974924774323,
      "loss": 1.9184,
      "theoretical_loss": 4.350276774827195,
      "tokens_seen": 212795392
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009449548645937814,
      "loss": 1.7896,
      "theoretical_loss": 4.350111850229202,
      "tokens_seen": 212860928
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009449348044132397,
      "loss": 1.8893,
      "theoretical_loss": 4.3499469906133195,
      "tokens_seen": 212926464
    },
    {
      "epoch": 0.06,
      "objective/train/advantage_avg": 0.49271607398986816,
      "objective/train/docs_used": 84140,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6249154806137085,
      "objective/train/original_loss": 1.624915599822998,
      "objective/train/theoretical_loss": 4.349782195933957,
      "objective/train/tokens_used": 233452000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24310225248336792,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505075454711914,
      "objective/train/weighted_lm_loss": 1.7071648836135864,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.349782195933957,
      "tokens_seen": 212992000
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009449147442326981,
      "loss": 1.723,
      "theoretical_loss": 4.349782195933957,
      "tokens_seen": 212992000
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009448946840521565,
      "loss": 1.7778,
      "theoretical_loss": 4.349617466145568,
      "tokens_seen": 213057536
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009448746238716148,
      "loss": 1.7859,
      "theoretical_loss": 4.34945280120265,
      "tokens_seen": 213123072
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009448545636910733,
      "loss": 1.8328,
      "theoretical_loss": 4.349288201059751,
      "tokens_seen": 213188608
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009448345035105315,
      "loss": 1.7755,
      "theoretical_loss": 4.34912366567146,
      "tokens_seen": 213254144
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00094481444332999,
      "loss": 1.9151,
      "theoretical_loss": 4.348959194992415,
      "tokens_seen": 213319680
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009447943831494484,
      "loss": 1.8385,
      "theoretical_loss": 4.348794788977298,
      "tokens_seen": 213385216
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009447743229689067,
      "loss": 1.8569,
      "theoretical_loss": 4.348630447580839,
      "tokens_seen": 213450752
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009447542627883651,
      "loss": 1.8746,
      "theoretical_loss": 4.34846617075781,
      "tokens_seen": 213516288
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009447342026078234,
      "loss": 1.7508,
      "theoretical_loss": 4.34830195846303,
      "tokens_seen": 213581824
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009447141424272818,
      "loss": 1.9108,
      "theoretical_loss": 4.348137810651366,
      "tokens_seen": 213647360
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009446940822467403,
      "loss": 1.8,
      "theoretical_loss": 4.347973727277726,
      "tokens_seen": 213712896
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009446740220661985,
      "loss": 1.8573,
      "theoretical_loss": 4.347809708297067,
      "tokens_seen": 213778432
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000944653961885657,
      "loss": 1.7998,
      "theoretical_loss": 4.347645753664389,
      "tokens_seen": 213843968
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009446339017051154,
      "loss": 1.7913,
      "theoretical_loss": 4.347481863334738,
      "tokens_seen": 213909504
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009446138415245737,
      "loss": 1.849,
      "theoretical_loss": 4.347318037263206,
      "tokens_seen": 213975040
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009445937813440321,
      "loss": 1.8244,
      "theoretical_loss": 4.34715427540493,
      "tokens_seen": 214040576
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009445737211634905,
      "loss": 1.8991,
      "theoretical_loss": 4.34699057771509,
      "tokens_seen": 214106112
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009445536609829488,
      "loss": 1.8225,
      "theoretical_loss": 4.346826944148912,
      "tokens_seen": 214171648
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009445336008024073,
      "loss": 1.8634,
      "theoretical_loss": 4.346663374661669,
      "tokens_seen": 214237184
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009445135406218655,
      "loss": 1.8842,
      "theoretical_loss": 4.346499869208676,
      "tokens_seen": 214302720
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000944493480441324,
      "loss": 1.888,
      "theoretical_loss": 4.346336427745294,
      "tokens_seen": 214368256
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009444734202607823,
      "loss": 1.8777,
      "theoretical_loss": 4.3461730502269305,
      "tokens_seen": 214433792
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0009444533600802407,
      "loss": 1.9699,
      "theoretical_loss": 4.346009736609034,
      "tokens_seen": 214499328
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009444332998996992,
      "loss": 1.7312,
      "theoretical_loss": 4.345846486847099,
      "tokens_seen": 214564864
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009444132397191575,
      "loss": 1.89,
      "theoretical_loss": 4.345683300896668,
      "tokens_seen": 214630400
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009443931795386159,
      "loss": 1.7744,
      "theoretical_loss": 4.345520178713323,
      "tokens_seen": 214695936
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009443731193580743,
      "loss": 1.8952,
      "theoretical_loss": 4.345357120252692,
      "tokens_seen": 214761472
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009443530591775327,
      "loss": 1.988,
      "theoretical_loss": 4.34519412547045,
      "tokens_seen": 214827008
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000944332998996991,
      "loss": 1.908,
      "theoretical_loss": 4.345031194322313,
      "tokens_seen": 214892544
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009443129388164494,
      "loss": 1.9555,
      "theoretical_loss": 4.344868326764045,
      "tokens_seen": 214958080
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009442928786359077,
      "loss": 1.9617,
      "theoretical_loss": 4.344705522751449,
      "tokens_seen": 215023616
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009442728184553662,
      "loss": 1.8555,
      "theoretical_loss": 4.344542782240376,
      "tokens_seen": 215089152
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009442527582748245,
      "loss": 1.8447,
      "theoretical_loss": 4.344380105186721,
      "tokens_seen": 215154688
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009442326980942829,
      "loss": 1.8825,
      "theoretical_loss": 4.344217491546422,
      "tokens_seen": 215220224
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009442126379137412,
      "loss": 2.0065,
      "theoretical_loss": 4.344054941275459,
      "tokens_seen": 215285760
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009441925777331997,
      "loss": 1.8606,
      "theoretical_loss": 4.343892454329863,
      "tokens_seen": 215351296
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000944172517552658,
      "loss": 1.8549,
      "theoretical_loss": 4.343730030665699,
      "tokens_seen": 215416832
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009441524573721164,
      "loss": 1.691,
      "theoretical_loss": 4.343567670239084,
      "tokens_seen": 215482368
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009441323971915747,
      "loss": 1.7313,
      "theoretical_loss": 4.343405373006175,
      "tokens_seen": 215547904
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009441123370110331,
      "loss": 1.9053,
      "theoretical_loss": 4.343243138923173,
      "tokens_seen": 215613440
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009440922768304915,
      "loss": 2.0116,
      "theoretical_loss": 4.343080967946323,
      "tokens_seen": 215678976
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009440722166499499,
      "loss": 1.9259,
      "theoretical_loss": 4.342918860031914,
      "tokens_seen": 215744512
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009440521564694082,
      "loss": 1.7324,
      "theoretical_loss": 4.342756815136278,
      "tokens_seen": 215810048
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009440320962888667,
      "loss": 1.9033,
      "theoretical_loss": 4.3425948332157915,
      "tokens_seen": 215875584
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000944012036108325,
      "loss": 1.7331,
      "theoretical_loss": 4.3424329142268725,
      "tokens_seen": 215941120
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009439919759277834,
      "loss": 1.7929,
      "theoretical_loss": 4.342271058125983,
      "tokens_seen": 216006656
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009439719157472418,
      "loss": 1.9002,
      "theoretical_loss": 4.342109264869631,
      "tokens_seen": 216072192
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009439518555667001,
      "loss": 1.9297,
      "theoretical_loss": 4.341947534414364,
      "tokens_seen": 216137728
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009439317953861585,
      "loss": 1.83,
      "theoretical_loss": 4.341785866716774,
      "tokens_seen": 216203264
    },
    {
      "epoch": 0.07,
      "objective/train/advantage_avg": 0.4911940395832062,
      "objective/train/docs_used": 85191,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.79802668094635,
      "objective/train/original_loss": 1.79802668094635,
      "objective/train/theoretical_loss": 4.341624261733497,
      "objective/train/tokens_used": 236728800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24179859459400177,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503485202789307,
      "objective/train/weighted_lm_loss": 1.8884522914886475,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.341624261733497,
      "tokens_seen": 216268800
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000943911735205617,
      "loss": 1.8136,
      "theoretical_loss": 4.341624261733497,
      "tokens_seen": 216268800
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009438916750250752,
      "loss": 1.8533,
      "theoretical_loss": 4.341462719421211,
      "tokens_seen": 216334336
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009438716148445337,
      "loss": 1.8464,
      "theoretical_loss": 4.341301239736639,
      "tokens_seen": 216399872
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009438515546639919,
      "loss": 1.9912,
      "theoretical_loss": 4.341139822636544,
      "tokens_seen": 216465408
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009438314944834504,
      "loss": 1.8442,
      "theoretical_loss": 4.340978468077735,
      "tokens_seen": 216530944
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009438114343029088,
      "loss": 1.9045,
      "theoretical_loss": 4.34081717601706,
      "tokens_seen": 216596480
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009437913741223671,
      "loss": 1.7926,
      "theoretical_loss": 4.340655946411413,
      "tokens_seen": 216662016
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009437713139418255,
      "loss": 1.896,
      "theoretical_loss": 4.340494779217732,
      "tokens_seen": 216727552
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009437512537612838,
      "loss": 1.8232,
      "theoretical_loss": 4.340333674392992,
      "tokens_seen": 216793088
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009437311935807422,
      "loss": 1.9165,
      "theoretical_loss": 4.340172631894218,
      "tokens_seen": 216858624
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009437111334002007,
      "loss": 1.9249,
      "theoretical_loss": 4.340011651678472,
      "tokens_seen": 216924160
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009436910732196589,
      "loss": 1.757,
      "theoretical_loss": 4.33985073370286,
      "tokens_seen": 216989696
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009436710130391174,
      "loss": 1.8661,
      "theoretical_loss": 4.339689877924531,
      "tokens_seen": 217055232
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009436509528585758,
      "loss": 1.8468,
      "theoretical_loss": 4.339529084300676,
      "tokens_seen": 217120768
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009436308926780341,
      "loss": 1.8,
      "theoretical_loss": 4.339368352788531,
      "tokens_seen": 217186304
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009436108324974925,
      "loss": 1.8137,
      "theoretical_loss": 4.339207683345369,
      "tokens_seen": 217251840
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009435907723169509,
      "loss": 1.8983,
      "theoretical_loss": 4.3390470759285105,
      "tokens_seen": 217317376
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009435707121364092,
      "loss": 1.8911,
      "theoretical_loss": 4.338886530495314,
      "tokens_seen": 217382912
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009435506519558677,
      "loss": 1.8343,
      "theoretical_loss": 4.338726047003183,
      "tokens_seen": 217448448
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009435305917753259,
      "loss": 1.797,
      "theoretical_loss": 4.338565625409563,
      "tokens_seen": 217513984
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009435105315947844,
      "loss": 1.8644,
      "theoretical_loss": 4.338405265671941,
      "tokens_seen": 217579520
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009434904714142427,
      "loss": 1.6793,
      "theoretical_loss": 4.338244967747844,
      "tokens_seen": 217645056
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009434704112337011,
      "loss": 1.834,
      "theoretical_loss": 4.338084731594844,
      "tokens_seen": 217710592
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009434503510531595,
      "loss": 1.7865,
      "theoretical_loss": 4.3379245571705525,
      "tokens_seen": 217776128
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009434302908726179,
      "loss": 1.7227,
      "theoretical_loss": 4.337764444432625,
      "tokens_seen": 217841664
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009434102306920762,
      "loss": 1.8882,
      "theoretical_loss": 4.337604393338758,
      "tokens_seen": 217907200
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009433901705115347,
      "loss": 1.8603,
      "theoretical_loss": 4.337444403846687,
      "tokens_seen": 217972736
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000943370110330993,
      "loss": 1.9331,
      "theoretical_loss": 4.3372844759141955,
      "tokens_seen": 218038272
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009433500501504514,
      "loss": 1.7618,
      "theoretical_loss": 4.337124609499101,
      "tokens_seen": 218103808
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009433299899699097,
      "loss": 1.8607,
      "theoretical_loss": 4.336964804559269,
      "tokens_seen": 218169344
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009433099297893681,
      "loss": 1.8718,
      "theoretical_loss": 4.336805061052601,
      "tokens_seen": 218234880
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009432898696088265,
      "loss": 1.6282,
      "theoretical_loss": 4.336645378937046,
      "tokens_seen": 218300416
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009432698094282849,
      "loss": 1.7722,
      "theoretical_loss": 4.336485758170589,
      "tokens_seen": 218365952
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009432497492477432,
      "loss": 1.8395,
      "theoretical_loss": 4.336326198711259,
      "tokens_seen": 218431488
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009432296890672016,
      "loss": 1.8712,
      "theoretical_loss": 4.336166700517125,
      "tokens_seen": 218497024
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00094320962888666,
      "loss": 1.8655,
      "theoretical_loss": 4.3360072635463,
      "tokens_seen": 218562560
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009431895687061184,
      "loss": 1.989,
      "theoretical_loss": 4.335847887756934,
      "tokens_seen": 218628096
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009431695085255767,
      "loss": 1.8191,
      "theoretical_loss": 4.335688573107221,
      "tokens_seen": 218693632
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009431494483450351,
      "loss": 1.8258,
      "theoretical_loss": 4.3355293195553966,
      "tokens_seen": 218759168
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009431293881644934,
      "loss": 1.8348,
      "theoretical_loss": 4.335370127059735,
      "tokens_seen": 218824704
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009431093279839519,
      "loss": 1.7862,
      "theoretical_loss": 4.335210995578553,
      "tokens_seen": 218890240
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009430892678034102,
      "loss": 1.8832,
      "theoretical_loss": 4.335051925070209,
      "tokens_seen": 218955776
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009430692076228686,
      "loss": 1.8708,
      "theoretical_loss": 4.334892915493099,
      "tokens_seen": 219021312
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000943049147442327,
      "loss": 1.73,
      "theoretical_loss": 4.334733966805665,
      "tokens_seen": 219086848
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009430290872617854,
      "loss": 1.9105,
      "theoretical_loss": 4.334575078966383,
      "tokens_seen": 219152384
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009430090270812437,
      "loss": 1.8589,
      "theoretical_loss": 4.334416251933778,
      "tokens_seen": 219217920
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009429889669007022,
      "loss": 1.8095,
      "theoretical_loss": 4.334257485666408,
      "tokens_seen": 219283456
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009429689067201604,
      "loss": 1.9426,
      "theoretical_loss": 4.334098780122876,
      "tokens_seen": 219348992
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009429488465396189,
      "loss": 1.9143,
      "theoretical_loss": 4.333940135261823,
      "tokens_seen": 219414528
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009429287863590774,
      "loss": 1.8288,
      "theoretical_loss": 4.333781551041935,
      "tokens_seen": 219480064
    },
    {
      "epoch": 0.07,
      "objective/train/advantage_avg": 0.4948246479034424,
      "objective/train/docs_used": 86440,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6815640926361084,
      "objective/train/original_loss": 1.6815643310546875,
      "objective/train/theoretical_loss": 4.3336230274219325,
      "objective/train/tokens_used": 240005600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24505706131458282,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507283210754395,
      "objective/train/weighted_lm_loss": 1.7667181491851807,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.3336230274219325,
      "tokens_seen": 219545600
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009429087261785356,
      "loss": 1.7774,
      "theoretical_loss": 4.3336230274219325,
      "tokens_seen": 219545600
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009428886659979941,
      "loss": 1.7581,
      "theoretical_loss": 4.333464564360581,
      "tokens_seen": 219611136
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009428686058174523,
      "loss": 1.9298,
      "theoretical_loss": 4.333306161816684,
      "tokens_seen": 219676672
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009428485456369108,
      "loss": 1.9582,
      "theoretical_loss": 4.333147819749085,
      "tokens_seen": 219742208
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009428284854563692,
      "loss": 1.9185,
      "theoretical_loss": 4.332989538116671,
      "tokens_seen": 219807744
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009428084252758275,
      "loss": 1.8663,
      "theoretical_loss": 4.332831316878364,
      "tokens_seen": 219873280
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009427883650952859,
      "loss": 1.8499,
      "theoretical_loss": 4.332673155993131,
      "tokens_seen": 219938816
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009427683049147442,
      "loss": 1.894,
      "theoretical_loss": 4.332515055419976,
      "tokens_seen": 220004352
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009427482447342026,
      "loss": 1.7753,
      "theoretical_loss": 4.332357015117946,
      "tokens_seen": 220069888
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009427281845536611,
      "loss": 1.8143,
      "theoretical_loss": 4.332199035046124,
      "tokens_seen": 220135424
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009427081243731193,
      "loss": 1.7558,
      "theoretical_loss": 4.332041115163636,
      "tokens_seen": 220200960
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009426880641925778,
      "loss": 1.8822,
      "theoretical_loss": 4.3318832554296485,
      "tokens_seen": 220266496
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009426680040120362,
      "loss": 1.8592,
      "theoretical_loss": 4.3317254558033635,
      "tokens_seen": 220332032
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009426479438314945,
      "loss": 1.9003,
      "theoretical_loss": 4.331567716244027,
      "tokens_seen": 220397568
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009426278836509529,
      "loss": 1.9003,
      "theoretical_loss": 4.331410036710925,
      "tokens_seen": 220463104
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009426078234704113,
      "loss": 1.792,
      "theoretical_loss": 4.331252417163379,
      "tokens_seen": 220528640
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009425877632898696,
      "loss": 1.8564,
      "theoretical_loss": 4.331094857560754,
      "tokens_seen": 220594176
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009425677031093281,
      "loss": 1.8892,
      "theoretical_loss": 4.330937357862453,
      "tokens_seen": 220659712
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009425476429287863,
      "loss": 1.8259,
      "theoretical_loss": 4.330779918027919,
      "tokens_seen": 220725248
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009425275827482448,
      "loss": 1.9108,
      "theoretical_loss": 4.330622538016635,
      "tokens_seen": 220790784
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009425075225677031,
      "loss": 1.8322,
      "theoretical_loss": 4.330465217788122,
      "tokens_seen": 220856320
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009424874623871615,
      "loss": 1.933,
      "theoretical_loss": 4.33030795730194,
      "tokens_seen": 220921856
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009424674022066199,
      "loss": 1.808,
      "theoretical_loss": 4.330150756517692,
      "tokens_seen": 220987392
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009424473420260783,
      "loss": 1.716,
      "theoretical_loss": 4.329993615395017,
      "tokens_seen": 221052928
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009424272818455366,
      "loss": 1.8106,
      "theoretical_loss": 4.329836533893593,
      "tokens_seen": 221118464
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009424072216649951,
      "loss": 1.9543,
      "theoretical_loss": 4.329679511973138,
      "tokens_seen": 221184000
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009423871614844533,
      "loss": 1.8496,
      "theoretical_loss": 4.3295225495934115,
      "tokens_seen": 221249536
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009423671013039118,
      "loss": 1.6599,
      "theoretical_loss": 4.3293656467142085,
      "tokens_seen": 221315072
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009423470411233701,
      "loss": 1.8111,
      "theoretical_loss": 4.329208803295365,
      "tokens_seen": 221380608
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009423269809428285,
      "loss": 1.734,
      "theoretical_loss": 4.329052019296754,
      "tokens_seen": 221446144
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009423069207622869,
      "loss": 1.795,
      "theoretical_loss": 4.328895294678292,
      "tokens_seen": 221511680
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009422868605817453,
      "loss": 1.9343,
      "theoretical_loss": 4.328738629399928,
      "tokens_seen": 221577216
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009422668004012036,
      "loss": 1.7639,
      "theoretical_loss": 4.328582023421655,
      "tokens_seen": 221642752
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000942246740220662,
      "loss": 1.8256,
      "theoretical_loss": 4.328425476703503,
      "tokens_seen": 221708288
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009422266800401204,
      "loss": 1.8873,
      "theoretical_loss": 4.32826898920554,
      "tokens_seen": 221773824
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009422066198595788,
      "loss": 1.8476,
      "theoretical_loss": 4.328112560887875,
      "tokens_seen": 221839360
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009421865596790371,
      "loss": 1.9737,
      "theoretical_loss": 4.327956191710651,
      "tokens_seen": 221904896
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009421664994984955,
      "loss": 1.884,
      "theoretical_loss": 4.327799881634056,
      "tokens_seen": 221970432
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009421464393179538,
      "loss": 1.8819,
      "theoretical_loss": 4.3276436306183115,
      "tokens_seen": 222035968
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009421263791374123,
      "loss": 1.7373,
      "theoretical_loss": 4.327487438623679,
      "tokens_seen": 222101504
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009421063189568706,
      "loss": 1.8465,
      "theoretical_loss": 4.32733130561046,
      "tokens_seen": 222167040
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000942086258776329,
      "loss": 1.7857,
      "theoretical_loss": 4.327175231538991,
      "tokens_seen": 222232576
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009420661985957874,
      "loss": 1.7946,
      "theoretical_loss": 4.327019216369651,
      "tokens_seen": 222298112
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009420461384152458,
      "loss": 1.8871,
      "theoretical_loss": 4.326863260062854,
      "tokens_seen": 222363648
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009420260782347041,
      "loss": 1.7786,
      "theoretical_loss": 4.326707362579054,
      "tokens_seen": 222429184
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009420060180541626,
      "loss": 1.7791,
      "theoretical_loss": 4.326551523878742,
      "tokens_seen": 222494720
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009419859578736208,
      "loss": 1.7992,
      "theoretical_loss": 4.32639574392245,
      "tokens_seen": 222560256
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009419658976930793,
      "loss": 1.735,
      "theoretical_loss": 4.326240022670743,
      "tokens_seen": 222625792
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009419458375125376,
      "loss": 1.8356,
      "theoretical_loss": 4.326084360084229,
      "tokens_seen": 222691328
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000941925777331996,
      "loss": 1.72,
      "theoretical_loss": 4.325928756123551,
      "tokens_seen": 222756864
    },
    {
      "epoch": 0.07,
      "objective/train/advantage_avg": 0.4933840334415436,
      "objective/train/docs_used": 87601,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.578743815422058,
      "objective/train/original_loss": 1.5787436962127686,
      "objective/train/theoretical_loss": 4.325773210749392,
      "objective/train/tokens_used": 243282400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24385935068130493,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505781173706055,
      "objective/train/weighted_lm_loss": 1.658752679824829,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.325773210749392,
      "tokens_seen": 222822400
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009419057171514544,
      "loss": 1.7028,
      "theoretical_loss": 4.325773210749392,
      "tokens_seen": 222822400
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009418856569709127,
      "loss": 1.9186,
      "theoretical_loss": 4.325617723922472,
      "tokens_seen": 222887936
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009418655967903711,
      "loss": 1.7035,
      "theoretical_loss": 4.325462295603547,
      "tokens_seen": 222953472
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009418455366098296,
      "loss": 1.8367,
      "theoretical_loss": 4.325306925753415,
      "tokens_seen": 223019008
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009418254764292878,
      "loss": 1.8699,
      "theoretical_loss": 4.325151614332908,
      "tokens_seen": 223084544
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009418054162487463,
      "loss": 1.8548,
      "theoretical_loss": 4.324996361302899,
      "tokens_seen": 223150080
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009417853560682045,
      "loss": 1.8911,
      "theoretical_loss": 4.324841166624294,
      "tokens_seen": 223215616
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000941765295887663,
      "loss": 1.8347,
      "theoretical_loss": 4.324686030258042,
      "tokens_seen": 223281152
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009417452357071214,
      "loss": 1.8127,
      "theoretical_loss": 4.3245309521651265,
      "tokens_seen": 223346688
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009417251755265797,
      "loss": 1.8398,
      "theoretical_loss": 4.324375932306568,
      "tokens_seen": 223412224
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009417051153460381,
      "loss": 1.8697,
      "theoretical_loss": 4.324220970643426,
      "tokens_seen": 223477760
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009416850551654966,
      "loss": 1.8429,
      "theoretical_loss": 4.324066067136798,
      "tokens_seen": 223543296
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009416649949849548,
      "loss": 1.807,
      "theoretical_loss": 4.323911221747817,
      "tokens_seen": 223608832
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009416449348044133,
      "loss": 1.9073,
      "theoretical_loss": 4.323756434437655,
      "tokens_seen": 223674368
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009416248746238715,
      "loss": 1.7969,
      "theoretical_loss": 4.323601705167521,
      "tokens_seen": 223739904
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00094160481444333,
      "loss": 1.8434,
      "theoretical_loss": 4.323447033898661,
      "tokens_seen": 223805440
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009415847542627884,
      "loss": 1.9763,
      "theoretical_loss": 4.323292420592356,
      "tokens_seen": 223870976
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009415646940822467,
      "loss": 1.7727,
      "theoretical_loss": 4.323137865209928,
      "tokens_seen": 223936512
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009415446339017051,
      "loss": 1.8935,
      "theoretical_loss": 4.322983367712735,
      "tokens_seen": 224002048
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009415245737211635,
      "loss": 1.7683,
      "theoretical_loss": 4.322828928062171,
      "tokens_seen": 224067584
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009415045135406218,
      "loss": 1.918,
      "theoretical_loss": 4.322674546219666,
      "tokens_seen": 224133120
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009414844533600803,
      "loss": 1.8897,
      "theoretical_loss": 4.322520222146689,
      "tokens_seen": 224198656
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009414643931795386,
      "loss": 1.811,
      "theoretical_loss": 4.322365955804749,
      "tokens_seen": 224264192
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000941444332998997,
      "loss": 1.8297,
      "theoretical_loss": 4.322211747155383,
      "tokens_seen": 224329728
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009414242728184555,
      "loss": 1.8643,
      "theoretical_loss": 4.322057596160174,
      "tokens_seen": 224395264
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009414042126379137,
      "loss": 1.6737,
      "theoretical_loss": 4.321903502780736,
      "tokens_seen": 224460800
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009413841524573722,
      "loss": 1.8232,
      "theoretical_loss": 4.321749466978722,
      "tokens_seen": 224526336
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009413640922768305,
      "loss": 1.7283,
      "theoretical_loss": 4.321595488715822,
      "tokens_seen": 224591872
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009413440320962889,
      "loss": 1.8719,
      "theoretical_loss": 4.321441567953762,
      "tokens_seen": 224657408
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009413239719157473,
      "loss": 1.8208,
      "theoretical_loss": 4.3212877046543055,
      "tokens_seen": 224722944
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009413039117352057,
      "loss": 1.7285,
      "theoretical_loss": 4.32113389877925,
      "tokens_seen": 224788480
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000941283851554664,
      "loss": 1.7993,
      "theoretical_loss": 4.320980150290432,
      "tokens_seen": 224854016
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009412637913741224,
      "loss": 1.9261,
      "theoretical_loss": 4.320826459149725,
      "tokens_seen": 224919552
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009412437311935808,
      "loss": 1.911,
      "theoretical_loss": 4.320672825319035,
      "tokens_seen": 224985088
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009412236710130392,
      "loss": 1.9533,
      "theoretical_loss": 4.320519248760309,
      "tokens_seen": 225050624
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009412036108324975,
      "loss": 1.9988,
      "theoretical_loss": 4.3203657294355295,
      "tokens_seen": 225116160
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000941183550651956,
      "loss": 1.8758,
      "theoretical_loss": 4.3202122673067125,
      "tokens_seen": 225181696
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009411634904714142,
      "loss": 1.8191,
      "theoretical_loss": 4.3200588623359115,
      "tokens_seen": 225247232
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009411434302908727,
      "loss": 1.8512,
      "theoretical_loss": 4.319905514485218,
      "tokens_seen": 225312768
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000941123370110331,
      "loss": 1.7275,
      "theoretical_loss": 4.319752223716758,
      "tokens_seen": 225378304
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009411033099297894,
      "loss": 1.7934,
      "theoretical_loss": 4.319598989992695,
      "tokens_seen": 225443840
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009410832497492478,
      "loss": 1.8376,
      "theoretical_loss": 4.319445813275225,
      "tokens_seen": 225509376
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009410631895687062,
      "loss": 2.0349,
      "theoretical_loss": 4.3192926935265845,
      "tokens_seen": 225574912
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009410431293881645,
      "loss": 1.9335,
      "theoretical_loss": 4.319139630709043,
      "tokens_seen": 225640448
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000941023069207623,
      "loss": 1.7679,
      "theoretical_loss": 4.318986624784908,
      "tokens_seen": 225705984
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009410030090270812,
      "loss": 1.8803,
      "theoretical_loss": 4.3188336757165215,
      "tokens_seen": 225771520
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009409829488465397,
      "loss": 1.9487,
      "theoretical_loss": 4.318680783466261,
      "tokens_seen": 225837056
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000940962888665998,
      "loss": 1.7355,
      "theoretical_loss": 4.318527947996542,
      "tokens_seen": 225902592
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009409428284854564,
      "loss": 1.8677,
      "theoretical_loss": 4.318375169269813,
      "tokens_seen": 225968128
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009409227683049148,
      "loss": 1.8113,
      "theoretical_loss": 4.3182224472485595,
      "tokens_seen": 226033664
    },
    {
      "epoch": 0.07,
      "objective/train/advantage_avg": 0.49421149492263794,
      "objective/train/docs_used": 88830,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.722211480140686,
      "objective/train/original_loss": 1.7222118377685547,
      "objective/train/theoretical_loss": 4.3180697818953035,
      "objective/train/tokens_used": 246559200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24472033977508545,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050665259361267,
      "objective/train/weighted_lm_loss": 1.809336543083191,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.3180697818953035,
      "tokens_seen": 226099200
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009409027081243731,
      "loss": 1.7996,
      "theoretical_loss": 4.3180697818953035,
      "tokens_seen": 226099200
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009408826479438315,
      "loss": 1.9476,
      "theoretical_loss": 4.317917173172602,
      "tokens_seen": 226164736
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00094086258776329,
      "loss": 1.7379,
      "theoretical_loss": 4.317764621043046,
      "tokens_seen": 226230272
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009408425275827482,
      "loss": 1.8394,
      "theoretical_loss": 4.317612125469264,
      "tokens_seen": 226295808
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009408224674022067,
      "loss": 1.7664,
      "theoretical_loss": 4.31745968641392,
      "tokens_seen": 226361344
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009408024072216649,
      "loss": 1.8748,
      "theoretical_loss": 4.317307303839712,
      "tokens_seen": 226426880
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009407823470411234,
      "loss": 1.7081,
      "theoretical_loss": 4.317154977709375,
      "tokens_seen": 226492416
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009407622868605818,
      "loss": 1.6976,
      "theoretical_loss": 4.3170027079856785,
      "tokens_seen": 226557952
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009407422266800401,
      "loss": 1.795,
      "theoretical_loss": 4.316850494631427,
      "tokens_seen": 226623488
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009407221664994985,
      "loss": 1.881,
      "theoretical_loss": 4.316698337609461,
      "tokens_seen": 226689024
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000940702106318957,
      "loss": 1.8121,
      "theoretical_loss": 4.3165462368826555,
      "tokens_seen": 226754560
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009406820461384152,
      "loss": 1.729,
      "theoretical_loss": 4.316394192413922,
      "tokens_seen": 226820096
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009406619859578737,
      "loss": 1.8137,
      "theoretical_loss": 4.316242204166205,
      "tokens_seen": 226885632
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000940641925777332,
      "loss": 1.8044,
      "theoretical_loss": 4.316090272102486,
      "tokens_seen": 226951168
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009406218655967904,
      "loss": 1.8453,
      "theoretical_loss": 4.315938396185782,
      "tokens_seen": 227016704
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009406018054162488,
      "loss": 1.7808,
      "theoretical_loss": 4.315786576379143,
      "tokens_seen": 227082240
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009405817452357071,
      "loss": 1.7636,
      "theoretical_loss": 4.315634812645654,
      "tokens_seen": 227147776
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009405616850551655,
      "loss": 1.7784,
      "theoretical_loss": 4.315483104948438,
      "tokens_seen": 227213312
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009405416248746239,
      "loss": 1.885,
      "theoretical_loss": 4.315331453250648,
      "tokens_seen": 227278848
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009405215646940822,
      "loss": 1.7154,
      "theoretical_loss": 4.315179857515477,
      "tokens_seen": 227344384
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009405015045135407,
      "loss": 1.7473,
      "theoretical_loss": 4.315028317706149,
      "tokens_seen": 227409920
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000940481444332999,
      "loss": 1.8366,
      "theoretical_loss": 4.314876833785926,
      "tokens_seen": 227475456
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009404613841524574,
      "loss": 1.8375,
      "theoretical_loss": 4.314725405718099,
      "tokens_seen": 227540992
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009404413239719157,
      "loss": 1.7129,
      "theoretical_loss": 4.314574033466001,
      "tokens_seen": 227606528
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009404212637913741,
      "loss": 1.789,
      "theoretical_loss": 4.314422716992994,
      "tokens_seen": 227672064
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009404012036108325,
      "loss": 1.8073,
      "theoretical_loss": 4.314271456262479,
      "tokens_seen": 227737600
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009403811434302909,
      "loss": 1.8551,
      "theoretical_loss": 4.314120251237887,
      "tokens_seen": 227803136
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009403610832497492,
      "loss": 1.8417,
      "theoretical_loss": 4.313969101882687,
      "tokens_seen": 227868672
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009403410230692077,
      "loss": 1.7539,
      "theoretical_loss": 4.31381800816038,
      "tokens_seen": 227934208
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000940320962888666,
      "loss": 1.8444,
      "theoretical_loss": 4.313666970034505,
      "tokens_seen": 227999744
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009403009027081244,
      "loss": 1.9733,
      "theoretical_loss": 4.31351598746863,
      "tokens_seen": 228065280
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009402808425275827,
      "loss": 1.7642,
      "theoretical_loss": 4.313365060426363,
      "tokens_seen": 228130816
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009402607823470412,
      "loss": 1.7757,
      "theoretical_loss": 4.313214188871341,
      "tokens_seen": 228196352
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009402407221664995,
      "loss": 1.8742,
      "theoretical_loss": 4.31306337276724,
      "tokens_seen": 228261888
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009402206619859579,
      "loss": 1.8931,
      "theoretical_loss": 4.312912612077767,
      "tokens_seen": 228327424
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009402006018054162,
      "loss": 1.848,
      "theoretical_loss": 4.312761906766664,
      "tokens_seen": 228392960
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009401805416248746,
      "loss": 1.8055,
      "theoretical_loss": 4.312611256797708,
      "tokens_seen": 228458496
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000940160481444333,
      "loss": 1.8979,
      "theoretical_loss": 4.3124606621347095,
      "tokens_seen": 228524032
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009401404212637914,
      "loss": 1.8403,
      "theoretical_loss": 4.312310122741512,
      "tokens_seen": 228589568
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009401203610832497,
      "loss": 1.8505,
      "theoretical_loss": 4.312159638581995,
      "tokens_seen": 228655104
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009401003009027082,
      "loss": 1.8639,
      "theoretical_loss": 4.31200920962007,
      "tokens_seen": 228720640
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009400802407221665,
      "loss": 1.8759,
      "theoretical_loss": 4.311858835819684,
      "tokens_seen": 228786176
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009400601805416249,
      "loss": 1.8338,
      "theoretical_loss": 4.311708517144817,
      "tokens_seen": 228851712
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009400401203610832,
      "loss": 1.838,
      "theoretical_loss": 4.3115582535594825,
      "tokens_seen": 228917248
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009400200601805416,
      "loss": 1.796,
      "theoretical_loss": 4.311408045027728,
      "tokens_seen": 228982784
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00094,
      "loss": 1.9516,
      "theoretical_loss": 4.3112578915136375,
      "tokens_seen": 229048320
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009399799398194584,
      "loss": 1.8333,
      "theoretical_loss": 4.311107792981323,
      "tokens_seen": 229113856
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009399598796389167,
      "loss": 1.8905,
      "theoretical_loss": 4.310957749394935,
      "tokens_seen": 229179392
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009399398194583752,
      "loss": 1.8143,
      "theoretical_loss": 4.310807760718656,
      "tokens_seen": 229244928
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009399197592778335,
      "loss": 1.8554,
      "theoretical_loss": 4.310657826916701,
      "tokens_seen": 229310464
    },
    {
      "epoch": 0.07,
      "objective/train/advantage_avg": 0.4928604066371918,
      "objective/train/docs_used": 90061,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8705188035964966,
      "objective/train/original_loss": 1.8705189228057861,
      "objective/train/theoretical_loss": 4.310507947953321,
      "objective/train/tokens_used": 249836000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24335232377052307,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050523281097412,
      "objective/train/weighted_lm_loss": 1.9649579524993896,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.310507947953321,
      "tokens_seen": 229376000
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009398996990972919,
      "loss": 1.8946,
      "theoretical_loss": 4.310507947953321,
      "tokens_seen": 229376000
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009398796389167504,
      "loss": 1.7751,
      "theoretical_loss": 4.310358123792798,
      "tokens_seen": 229441536
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009398595787362086,
      "loss": 1.9335,
      "theoretical_loss": 4.31020835439945,
      "tokens_seen": 229507072
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009398395185556671,
      "loss": 1.9429,
      "theoretical_loss": 4.310058639737626,
      "tokens_seen": 229572608
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009398194583751253,
      "loss": 1.8902,
      "theoretical_loss": 4.309908979771709,
      "tokens_seen": 229638144
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009397993981945838,
      "loss": 1.7913,
      "theoretical_loss": 4.3097593744661165,
      "tokens_seen": 229703680
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009397793380140422,
      "loss": 1.9064,
      "theoretical_loss": 4.309609823785298,
      "tokens_seen": 229769216
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009397592778335005,
      "loss": 1.8685,
      "theoretical_loss": 4.309460327693737,
      "tokens_seen": 229834752
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009397392176529589,
      "loss": 1.9508,
      "theoretical_loss": 4.3093108861559495,
      "tokens_seen": 229900288
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009397191574724174,
      "loss": 1.839,
      "theoretical_loss": 4.3091614991364855,
      "tokens_seen": 229965824
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009396990972918756,
      "loss": 1.8797,
      "theoretical_loss": 4.309012166599929,
      "tokens_seen": 230031360
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009396790371113341,
      "loss": 1.8448,
      "theoretical_loss": 4.308862888510893,
      "tokens_seen": 230096896
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009396589769307923,
      "loss": 1.7767,
      "theoretical_loss": 4.308713664834029,
      "tokens_seen": 230162432
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009396389167502508,
      "loss": 1.6558,
      "theoretical_loss": 4.308564495534017,
      "tokens_seen": 230227968
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009396188565697092,
      "loss": 1.8051,
      "theoretical_loss": 4.308415380575573,
      "tokens_seen": 230293504
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009395987963891675,
      "loss": 1.71,
      "theoretical_loss": 4.308266319923445,
      "tokens_seen": 230359040
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009395787362086259,
      "loss": 1.669,
      "theoretical_loss": 4.308117313542413,
      "tokens_seen": 230424576
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009395586760280843,
      "loss": 1.7442,
      "theoretical_loss": 4.3079683613972914,
      "tokens_seen": 230490112
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009395386158475426,
      "loss": 2.0136,
      "theoretical_loss": 4.307819463452925,
      "tokens_seen": 230555648
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009395185556670011,
      "loss": 1.7937,
      "theoretical_loss": 4.3076706196741945,
      "tokens_seen": 230621184
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009394984954864594,
      "loss": 1.8568,
      "theoretical_loss": 4.30752183002601,
      "tokens_seen": 230686720
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009394784353059178,
      "loss": 1.7888,
      "theoretical_loss": 4.307373094473318,
      "tokens_seen": 230752256
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009394583751253761,
      "loss": 1.758,
      "theoretical_loss": 4.307224412981095,
      "tokens_seen": 230817792
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009394383149448345,
      "loss": 1.7212,
      "theoretical_loss": 4.3070757855143516,
      "tokens_seen": 230883328
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009394182547642929,
      "loss": 1.8631,
      "theoretical_loss": 4.3069272120381275,
      "tokens_seen": 230948864
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009393981945837513,
      "loss": 1.7568,
      "theoretical_loss": 4.3067786925175,
      "tokens_seen": 231014400
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009393781344032096,
      "loss": 1.8251,
      "theoretical_loss": 4.3066302269175765,
      "tokens_seen": 231079936
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009393580742226681,
      "loss": 1.9332,
      "theoretical_loss": 4.306481815203496,
      "tokens_seen": 231145472
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009393380140421264,
      "loss": 1.8374,
      "theoretical_loss": 4.30633345734043,
      "tokens_seen": 231211008
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009393179538615848,
      "loss": 1.8336,
      "theoretical_loss": 4.306185153293584,
      "tokens_seen": 231276544
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009392978936810431,
      "loss": 1.6981,
      "theoretical_loss": 4.306036903028196,
      "tokens_seen": 231342080
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009392778335005016,
      "loss": 1.7224,
      "theoretical_loss": 4.305888706509534,
      "tokens_seen": 231407616
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009392577733199599,
      "loss": 1.9788,
      "theoretical_loss": 4.3057405637029,
      "tokens_seen": 231473152
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009392377131394183,
      "loss": 1.7976,
      "theoretical_loss": 4.305592474573627,
      "tokens_seen": 231538688
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009392176529588766,
      "loss": 1.9443,
      "theoretical_loss": 4.305444439087082,
      "tokens_seen": 231604224
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000939197592778335,
      "loss": 1.7595,
      "theoretical_loss": 4.305296457208662,
      "tokens_seen": 231669760
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009391775325977934,
      "loss": 1.8406,
      "theoretical_loss": 4.305148528903798,
      "tokens_seen": 231735296
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009391574724172518,
      "loss": 1.9642,
      "theoretical_loss": 4.3050006541379515,
      "tokens_seen": 231800832
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009391374122367101,
      "loss": 1.7075,
      "theoretical_loss": 4.304852832876618,
      "tokens_seen": 231866368
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009391173520561686,
      "loss": 1.7245,
      "theoretical_loss": 4.304705065085322,
      "tokens_seen": 231931904
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009390972918756269,
      "loss": 1.972,
      "theoretical_loss": 4.304557350729623,
      "tokens_seen": 231997440
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009390772316950853,
      "loss": 1.9232,
      "theoretical_loss": 4.304409689775111,
      "tokens_seen": 232062976
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009390571715145436,
      "loss": 1.8148,
      "theoretical_loss": 4.304262082187407,
      "tokens_seen": 232128512
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000939037111334002,
      "loss": 1.6574,
      "theoretical_loss": 4.304114527932166,
      "tokens_seen": 232194048
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009390170511534604,
      "loss": 1.8246,
      "theoretical_loss": 4.303967026975072,
      "tokens_seen": 232259584
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009389969909729188,
      "loss": 1.7932,
      "theoretical_loss": 4.303819579281844,
      "tokens_seen": 232325120
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009389769307923771,
      "loss": 1.8462,
      "theoretical_loss": 4.30367218481823,
      "tokens_seen": 232390656
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009389568706118356,
      "loss": 1.8204,
      "theoretical_loss": 4.30352484355001,
      "tokens_seen": 232456192
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009389368104312938,
      "loss": 1.9215,
      "theoretical_loss": 4.303377555442998,
      "tokens_seen": 232521728
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009389167502507523,
      "loss": 1.7366,
      "theoretical_loss": 4.3032303204630376,
      "tokens_seen": 232587264
    },
    {
      "epoch": 0.07,
      "objective/train/advantage_avg": 0.49400627613067627,
      "objective/train/docs_used": 91258,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6414120197296143,
      "objective/train/original_loss": 1.6414121389389038,
      "objective/train/theoretical_loss": 4.303083138576003,
      "objective/train/tokens_used": 253112800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24473948776721954,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506447553634644,
      "objective/train/weighted_lm_loss": 1.7243413925170898,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 4.303083138576003,
      "tokens_seen": 232652800
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009388966900702107,
      "loss": 1.8242,
      "theoretical_loss": 4.303083138576003,
      "tokens_seen": 232652800
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000938876629889669,
      "loss": 1.7083,
      "theoretical_loss": 4.302936009747803,
      "tokens_seen": 232718336
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009388565697091274,
      "loss": 1.7864,
      "theoretical_loss": 4.302788933944375,
      "tokens_seen": 232783872
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009388365095285857,
      "loss": 1.8209,
      "theoretical_loss": 4.302641911131689,
      "tokens_seen": 232849408
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009388164493480441,
      "loss": 1.8624,
      "theoretical_loss": 4.3024949412757465,
      "tokens_seen": 232914944
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009387963891675026,
      "loss": 1.7925,
      "theoretical_loss": 4.302348024342581,
      "tokens_seen": 232980480
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009387763289869608,
      "loss": 1.8408,
      "theoretical_loss": 4.302201160298255,
      "tokens_seen": 233046016
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009387562688064193,
      "loss": 1.8639,
      "theoretical_loss": 4.302054349108865,
      "tokens_seen": 233111552
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009387362086258777,
      "loss": 1.7454,
      "theoretical_loss": 4.301907590740538,
      "tokens_seen": 233177088
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000938716148445336,
      "loss": 1.7967,
      "theoretical_loss": 4.301760885159431,
      "tokens_seen": 233242624
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009386960882647944,
      "loss": 1.8907,
      "theoretical_loss": 4.301614232331733,
      "tokens_seen": 233308160
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009386760280842527,
      "loss": 1.817,
      "theoretical_loss": 4.301467632223664,
      "tokens_seen": 233373696
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009386559679037111,
      "loss": 1.9417,
      "theoretical_loss": 4.301321084801478,
      "tokens_seen": 233439232
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009386359077231696,
      "loss": 1.744,
      "theoretical_loss": 4.301174590031453,
      "tokens_seen": 233504768
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009386158475426278,
      "loss": 1.7738,
      "theoretical_loss": 4.301028147879904,
      "tokens_seen": 233570304
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009385957873620863,
      "loss": 1.6949,
      "theoretical_loss": 4.300881758313178,
      "tokens_seen": 233635840
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009385757271815446,
      "loss": 1.7507,
      "theoretical_loss": 4.300735421297647,
      "tokens_seen": 233701376
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000938555667001003,
      "loss": 1.7811,
      "theoretical_loss": 4.300589136799719,
      "tokens_seen": 233766912
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009385356068204614,
      "loss": 1.8286,
      "theoretical_loss": 4.300442904785831,
      "tokens_seen": 233832448
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009385155466399198,
      "loss": 1.7496,
      "theoretical_loss": 4.300296725222449,
      "tokens_seen": 233897984
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009384954864593781,
      "loss": 1.8731,
      "theoretical_loss": 4.300150598076074,
      "tokens_seen": 233963520
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009384754262788365,
      "loss": 1.8883,
      "theoretical_loss": 4.300004523313237,
      "tokens_seen": 234029056
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009384553660982948,
      "loss": 1.8071,
      "theoretical_loss": 4.299858500900495,
      "tokens_seen": 234094592
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009384353059177533,
      "loss": 1.9614,
      "theoretical_loss": 4.299712530804441,
      "tokens_seen": 234160128
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009384152457372117,
      "loss": 1.6487,
      "theoretical_loss": 4.299566612991695,
      "tokens_seen": 234225664
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00093839518555667,
      "loss": 1.8937,
      "theoretical_loss": 4.299420747428911,
      "tokens_seen": 234291200
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009383751253761285,
      "loss": 1.867,
      "theoretical_loss": 4.2992749340827725,
      "tokens_seen": 234356736
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009383550651955868,
      "loss": 1.828,
      "theoretical_loss": 4.299129172919991,
      "tokens_seen": 234422272
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009383350050150452,
      "loss": 1.8159,
      "theoretical_loss": 4.2989834639073115,
      "tokens_seen": 234487808
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009383149448345035,
      "loss": 1.8431,
      "theoretical_loss": 4.298837807011508,
      "tokens_seen": 234553344
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000938294884653962,
      "loss": 1.955,
      "theoretical_loss": 4.298692202199386,
      "tokens_seen": 234618880
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009382748244734203,
      "loss": 1.9081,
      "theoretical_loss": 4.29854664943778,
      "tokens_seen": 234684416
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009382547642928787,
      "loss": 1.7294,
      "theoretical_loss": 4.298401148693557,
      "tokens_seen": 234749952
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000938234704112337,
      "loss": 1.8228,
      "theoretical_loss": 4.298255699933612,
      "tokens_seen": 234815488
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009382146439317954,
      "loss": 1.8377,
      "theoretical_loss": 4.298110303124871,
      "tokens_seen": 234881024
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009381945837512538,
      "loss": 1.9015,
      "theoretical_loss": 4.297964958234292,
      "tokens_seen": 234946560
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009381745235707122,
      "loss": 1.8647,
      "theoretical_loss": 4.297819665228861,
      "tokens_seen": 235012096
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009381544633901705,
      "loss": 1.8087,
      "theoretical_loss": 4.297674424075594,
      "tokens_seen": 235077632
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000938134403209629,
      "loss": 1.812,
      "theoretical_loss": 4.29752923474154,
      "tokens_seen": 235143168
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009381143430290873,
      "loss": 1.823,
      "theoretical_loss": 4.297384097193775,
      "tokens_seen": 235208704
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009380942828485457,
      "loss": 1.8137,
      "theoretical_loss": 4.297239011399407,
      "tokens_seen": 235274240
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000938074222668004,
      "loss": 1.8754,
      "theoretical_loss": 4.297093977325573,
      "tokens_seen": 235339776
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009380541624874624,
      "loss": 1.7469,
      "theoretical_loss": 4.29694899493944,
      "tokens_seen": 235405312
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009380341023069208,
      "loss": 1.7643,
      "theoretical_loss": 4.296804064208207,
      "tokens_seen": 235470848
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009380140421263792,
      "loss": 1.8098,
      "theoretical_loss": 4.296659185099099,
      "tokens_seen": 235536384
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009379939819458375,
      "loss": 1.8432,
      "theoretical_loss": 4.296514357579375,
      "tokens_seen": 235601920
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000937973921765296,
      "loss": 1.9429,
      "theoretical_loss": 4.2963695816163225,
      "tokens_seen": 235667456
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009379538615847542,
      "loss": 1.7737,
      "theoretical_loss": 4.296224857177257,
      "tokens_seen": 235732992
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009379338014042127,
      "loss": 1.8518,
      "theoretical_loss": 4.296080184229526,
      "tokens_seen": 235798528
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009379137412236711,
      "loss": 1.7253,
      "theoretical_loss": 4.295935562740505,
      "tokens_seen": 235864064
    },
    {
      "epoch": 0.07,
      "objective/train/advantage_avg": 0.4898257851600647,
      "objective/train/docs_used": 92471,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8431349992752075,
      "objective/train/original_loss": 1.843135118484497,
      "objective/train/theoretical_loss": 4.295790992677603,
      "objective/train/tokens_used": 256389600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24101541936397552,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502077341079712,
      "objective/train/weighted_lm_loss": 1.9356428384780884,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.295790992677603,
      "tokens_seen": 235929600
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009378936810431294,
      "loss": 1.856,
      "theoretical_loss": 4.295790992677603,
      "tokens_seen": 235929600
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009378736208625878,
      "loss": 1.8341,
      "theoretical_loss": 4.295646474008252,
      "tokens_seen": 235995136
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009378535606820461,
      "loss": 1.9499,
      "theoretical_loss": 4.29550200669992,
      "tokens_seen": 236060672
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009378335005015045,
      "loss": 1.7604,
      "theoretical_loss": 4.295357590720101,
      "tokens_seen": 236126208
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000937813440320963,
      "loss": 1.7573,
      "theoretical_loss": 4.2952132260363225,
      "tokens_seen": 236191744
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009377933801404212,
      "loss": 1.8687,
      "theoretical_loss": 4.295068912616135,
      "tokens_seen": 236257280
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009377733199598797,
      "loss": 1.7367,
      "theoretical_loss": 4.294924650427125,
      "tokens_seen": 236322816
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009377532597793381,
      "loss": 1.7755,
      "theoretical_loss": 4.294780439436904,
      "tokens_seen": 236388352
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009377331995987964,
      "loss": 1.7929,
      "theoretical_loss": 4.294636279613117,
      "tokens_seen": 236453888
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009377131394182548,
      "loss": 1.8914,
      "theoretical_loss": 4.2944921709234345,
      "tokens_seen": 236519424
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009376930792377131,
      "loss": 1.897,
      "theoretical_loss": 4.294348113335559,
      "tokens_seen": 236584960
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009376730190571715,
      "loss": 2.0073,
      "theoretical_loss": 4.29420410681722,
      "tokens_seen": 236650496
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00093765295887663,
      "loss": 1.7469,
      "theoretical_loss": 4.294060151336178,
      "tokens_seen": 236716032
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009376328986960882,
      "loss": 1.8601,
      "theoretical_loss": 4.293916246860224,
      "tokens_seen": 236781568
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009376128385155467,
      "loss": 1.8224,
      "theoretical_loss": 4.293772393357176,
      "tokens_seen": 236847104
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000937592778335005,
      "loss": 1.9164,
      "theoretical_loss": 4.2936285907948815,
      "tokens_seen": 236912640
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009375727181544634,
      "loss": 1.8051,
      "theoretical_loss": 4.293484839141217,
      "tokens_seen": 236978176
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009375526579739218,
      "loss": 1.7483,
      "theoretical_loss": 4.293341138364091,
      "tokens_seen": 237043712
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009375325977933802,
      "loss": 1.7834,
      "theoretical_loss": 4.293197488431435,
      "tokens_seen": 237109248
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009375125376128385,
      "loss": 1.8088,
      "theoretical_loss": 4.293053889311217,
      "tokens_seen": 237174784
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009374924774322969,
      "loss": 1.8689,
      "theoretical_loss": 4.29291034097143,
      "tokens_seen": 237240320
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009374724172517552,
      "loss": 1.7728,
      "theoretical_loss": 4.292766843380094,
      "tokens_seen": 237305856
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009374523570712137,
      "loss": 1.7375,
      "theoretical_loss": 4.292623396505262,
      "tokens_seen": 237371392
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000937432296890672,
      "loss": 1.7408,
      "theoretical_loss": 4.292480000315014,
      "tokens_seen": 237436928
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009374122367101304,
      "loss": 1.7242,
      "theoretical_loss": 4.2923366547774595,
      "tokens_seen": 237502464
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009373921765295888,
      "loss": 1.7053,
      "theoretical_loss": 4.292193359860737,
      "tokens_seen": 237568000
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009373721163490472,
      "loss": 1.7752,
      "theoretical_loss": 4.292050115533012,
      "tokens_seen": 237633536
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009373520561685055,
      "loss": 1.7755,
      "theoretical_loss": 4.291906921762481,
      "tokens_seen": 237699072
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009373319959879639,
      "loss": 1.9598,
      "theoretical_loss": 4.2917637785173675,
      "tokens_seen": 237764608
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009373119358074222,
      "loss": 1.7152,
      "theoretical_loss": 4.291620685765926,
      "tokens_seen": 237830144
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009372918756268807,
      "loss": 1.884,
      "theoretical_loss": 4.291477643476438,
      "tokens_seen": 237895680
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000937271815446339,
      "loss": 1.8008,
      "theoretical_loss": 4.291334651617213,
      "tokens_seen": 237961216
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009372517552657974,
      "loss": 1.9097,
      "theoretical_loss": 4.291191710156591,
      "tokens_seen": 238026752
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009372316950852557,
      "loss": 1.8682,
      "theoretical_loss": 4.291048819062939,
      "tokens_seen": 238092288
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009372116349047142,
      "loss": 1.7768,
      "theoretical_loss": 4.290905978304655,
      "tokens_seen": 238157824
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009371915747241725,
      "loss": 1.7492,
      "theoretical_loss": 4.290763187850161,
      "tokens_seen": 238223360
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009371715145436309,
      "loss": 1.8498,
      "theoretical_loss": 4.290620447667912,
      "tokens_seen": 238288896
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009371514543630893,
      "loss": 1.7209,
      "theoretical_loss": 4.29047775772639,
      "tokens_seen": 238354432
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009371313941825476,
      "loss": 1.7331,
      "theoretical_loss": 4.290335117994104,
      "tokens_seen": 238419968
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000937111334002006,
      "loss": 1.738,
      "theoretical_loss": 4.290192528439594,
      "tokens_seen": 238485504
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009370912738214644,
      "loss": 1.9051,
      "theoretical_loss": 4.290049989031424,
      "tokens_seen": 238551040
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009370712136409227,
      "loss": 1.9187,
      "theoretical_loss": 4.289907499738192,
      "tokens_seen": 238616576
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009370511534603812,
      "loss": 1.8264,
      "theoretical_loss": 4.289765060528521,
      "tokens_seen": 238682112
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009370310932798395,
      "loss": 1.9482,
      "theoretical_loss": 4.2896226713710615,
      "tokens_seen": 238747648
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009370110330992979,
      "loss": 1.8173,
      "theoretical_loss": 4.289480332234493,
      "tokens_seen": 238813184
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009369909729187563,
      "loss": 1.7753,
      "theoretical_loss": 4.289338043087527,
      "tokens_seen": 238878720
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009369709127382146,
      "loss": 1.7691,
      "theoretical_loss": 4.289195803898897,
      "tokens_seen": 238944256
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000936950852557673,
      "loss": 1.9161,
      "theoretical_loss": 4.289053614637368,
      "tokens_seen": 239009792
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009369307923771315,
      "loss": 1.8554,
      "theoretical_loss": 4.288911475271731,
      "tokens_seen": 239075328
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009369107321965898,
      "loss": 1.8071,
      "theoretical_loss": 4.288769385770809,
      "tokens_seen": 239140864
    },
    {
      "epoch": 0.07,
      "objective/train/advantage_avg": 0.49386152625083923,
      "objective/train/docs_used": 93570,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8087902069091797,
      "objective/train/original_loss": 1.8087904453277588,
      "objective/train/theoretical_loss": 4.28862734610345,
      "objective/train/tokens_used": 259666400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24422965943813324,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506278276443481,
      "objective/train/weighted_lm_loss": 1.9002867937088013,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.28862734610345,
      "tokens_seen": 239206400
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009368906720160482,
      "loss": 1.8793,
      "theoretical_loss": 4.28862734610345,
      "tokens_seen": 239206400
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009368706118355065,
      "loss": 1.8012,
      "theoretical_loss": 4.288485356238529,
      "tokens_seen": 239271936
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009368505516549649,
      "loss": 1.7934,
      "theoretical_loss": 4.288343416144952,
      "tokens_seen": 239337472
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009368304914744234,
      "loss": 1.9502,
      "theoretical_loss": 4.288201525791651,
      "tokens_seen": 239403008
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009368104312938816,
      "loss": 1.8676,
      "theoretical_loss": 4.288059685147586,
      "tokens_seen": 239468544
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009367903711133401,
      "loss": 1.8661,
      "theoretical_loss": 4.287917894181746,
      "tokens_seen": 239534080
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009367703109327985,
      "loss": 1.7173,
      "theoretical_loss": 4.287776152863146,
      "tokens_seen": 239599616
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009367502507522568,
      "loss": 1.8081,
      "theoretical_loss": 4.28763446116083,
      "tokens_seen": 239665152
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009367301905717152,
      "loss": 1.8705,
      "theoretical_loss": 4.28749281904387,
      "tokens_seen": 239730688
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009367101303911735,
      "loss": 1.8234,
      "theoretical_loss": 4.287351226481366,
      "tokens_seen": 239796224
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009366900702106319,
      "loss": 1.8895,
      "theoretical_loss": 4.287209683442444,
      "tokens_seen": 239861760
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009366700100300904,
      "loss": 1.8936,
      "theoretical_loss": 4.2870681898962575,
      "tokens_seen": 239927296
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009366499498495486,
      "loss": 1.8068,
      "theoretical_loss": 4.286926745811991,
      "tokens_seen": 239992832
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009366298896690071,
      "loss": 1.7208,
      "theoretical_loss": 4.286785351158852,
      "tokens_seen": 240058368
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009366098294884654,
      "loss": 1.8455,
      "theoretical_loss": 4.286644005906081,
      "tokens_seen": 240123904
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009365897693079238,
      "loss": 1.8656,
      "theoretical_loss": 4.2865027100229405,
      "tokens_seen": 240189440
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009365697091273822,
      "loss": 1.9201,
      "theoretical_loss": 4.286361463478723,
      "tokens_seen": 240254976
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009365496489468406,
      "loss": 1.8351,
      "theoretical_loss": 4.28622026624275,
      "tokens_seen": 240320512
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009365295887662989,
      "loss": 1.7223,
      "theoretical_loss": 4.286079118284368,
      "tokens_seen": 240386048
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009365095285857573,
      "loss": 1.7905,
      "theoretical_loss": 4.285938019572951,
      "tokens_seen": 240451584
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009364894684052156,
      "loss": 1.922,
      "theoretical_loss": 4.285796970077903,
      "tokens_seen": 240517120
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009364694082246741,
      "loss": 1.8986,
      "theoretical_loss": 4.285655969768651,
      "tokens_seen": 240582656
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009364493480441324,
      "loss": 1.9152,
      "theoretical_loss": 4.285515018614655,
      "tokens_seen": 240648192
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009364292878635908,
      "loss": 1.7653,
      "theoretical_loss": 4.2853741165853965,
      "tokens_seen": 240713728
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009364092276830492,
      "loss": 1.6739,
      "theoretical_loss": 4.285233263650387,
      "tokens_seen": 240779264
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009363891675025076,
      "loss": 1.7319,
      "theoretical_loss": 4.285092459779166,
      "tokens_seen": 240844800
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009363691073219659,
      "loss": 1.7374,
      "theoretical_loss": 4.2849517049412995,
      "tokens_seen": 240910336
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009363490471414243,
      "loss": 1.7947,
      "theoretical_loss": 4.28481099910638,
      "tokens_seen": 240975872
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009363289869608826,
      "loss": 1.7014,
      "theoretical_loss": 4.284670342244027,
      "tokens_seen": 241041408
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009363089267803411,
      "loss": 1.8506,
      "theoretical_loss": 4.284529734323887,
      "tokens_seen": 241106944
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009362888665997994,
      "loss": 1.7271,
      "theoretical_loss": 4.284389175315636,
      "tokens_seen": 241172480
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009362688064192578,
      "loss": 1.7538,
      "theoretical_loss": 4.284248665188974,
      "tokens_seen": 241238016
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009362487462387161,
      "loss": 1.7851,
      "theoretical_loss": 4.284108203913629,
      "tokens_seen": 241303552
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009362286860581746,
      "loss": 1.8025,
      "theoretical_loss": 4.283967791459356,
      "tokens_seen": 241369088
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009362086258776329,
      "loss": 1.8707,
      "theoretical_loss": 4.283827427795939,
      "tokens_seen": 241434624
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009361885656970913,
      "loss": 1.8208,
      "theoretical_loss": 4.283687112893184,
      "tokens_seen": 241500160
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009361685055165497,
      "loss": 1.7022,
      "theoretical_loss": 4.283546846720929,
      "tokens_seen": 241565696
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000936148445336008,
      "loss": 1.798,
      "theoretical_loss": 4.283406629249036,
      "tokens_seen": 241631232
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009361283851554664,
      "loss": 1.7899,
      "theoretical_loss": 4.283266460447394,
      "tokens_seen": 241696768
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009361083249749248,
      "loss": 1.7569,
      "theoretical_loss": 4.28312634028592,
      "tokens_seen": 241762304
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009360882647943831,
      "loss": 1.7866,
      "theoretical_loss": 4.282986268734556,
      "tokens_seen": 241827840
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009360682046138416,
      "loss": 1.6588,
      "theoretical_loss": 4.282846245763273,
      "tokens_seen": 241893376
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009360481444332999,
      "loss": 1.7699,
      "theoretical_loss": 4.282706271342066,
      "tokens_seen": 241958912
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009360280842527583,
      "loss": 1.8742,
      "theoretical_loss": 4.2825663454409595,
      "tokens_seen": 242024448
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009360080240722167,
      "loss": 1.8847,
      "theoretical_loss": 4.2824264680300015,
      "tokens_seen": 242089984
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000935987963891675,
      "loss": 1.9179,
      "theoretical_loss": 4.2822866390792695,
      "tokens_seen": 242155520
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009359679037111334,
      "loss": 1.8009,
      "theoretical_loss": 4.282146858558866,
      "tokens_seen": 242221056
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009359478435305919,
      "loss": 1.7736,
      "theoretical_loss": 4.28200712643892,
      "tokens_seen": 242286592
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009359277833500501,
      "loss": 1.627,
      "theoretical_loss": 4.281867442689587,
      "tokens_seen": 242352128
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009359077231695086,
      "loss": 1.87,
      "theoretical_loss": 4.281727807281051,
      "tokens_seen": 242417664
    },
    {
      "epoch": 0.07,
      "objective/train/advantage_avg": 0.49168694019317627,
      "objective/train/docs_used": 94878,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7456352710723877,
      "objective/train/original_loss": 1.7456352710723877,
      "objective/train/theoretical_loss": 4.281588220183519,
      "objective/train/tokens_used": 262943200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24250559508800507,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504014492034912,
      "objective/train/weighted_lm_loss": 1.8333992958068848,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.281588220183519,
      "tokens_seen": 242483200
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009358876629889668,
      "loss": 1.7932,
      "theoretical_loss": 4.281588220183519,
      "tokens_seen": 242483200
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009358676028084253,
      "loss": 1.8316,
      "theoretical_loss": 4.281448681367226,
      "tokens_seen": 242548736
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009358475426278837,
      "loss": 1.7968,
      "theoretical_loss": 4.2813091908024345,
      "tokens_seen": 242614272
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000935827482447342,
      "loss": 1.7241,
      "theoretical_loss": 4.281169748459432,
      "tokens_seen": 242679808
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009358074222668004,
      "loss": 1.8417,
      "theoretical_loss": 4.281030354308533,
      "tokens_seen": 242745344
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009357873620862589,
      "loss": 1.811,
      "theoretical_loss": 4.2808910083200775,
      "tokens_seen": 242810880
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009357673019057171,
      "loss": 1.7931,
      "theoretical_loss": 4.280751710464432,
      "tokens_seen": 242876416
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009357472417251756,
      "loss": 1.7608,
      "theoretical_loss": 4.280612460711989,
      "tokens_seen": 242941952
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009357271815446338,
      "loss": 1.8087,
      "theoretical_loss": 4.280473259033169,
      "tokens_seen": 243007488
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009357071213640923,
      "loss": 1.8108,
      "theoretical_loss": 4.280334105398417,
      "tokens_seen": 243073024
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009356870611835507,
      "loss": 1.7234,
      "theoretical_loss": 4.280194999778203,
      "tokens_seen": 243138560
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000935667001003009,
      "loss": 1.7381,
      "theoretical_loss": 4.280055942143027,
      "tokens_seen": 243204096
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009356469408224674,
      "loss": 1.7984,
      "theoretical_loss": 4.27991693246341,
      "tokens_seen": 243269632
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009356268806419258,
      "loss": 1.8265,
      "theoretical_loss": 4.279777970709902,
      "tokens_seen": 243335168
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009356068204613841,
      "loss": 1.8135,
      "theoretical_loss": 4.27963905685308,
      "tokens_seen": 243400704
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009355867602808426,
      "loss": 1.7835,
      "theoretical_loss": 4.279500190863544,
      "tokens_seen": 243466240
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009355667001003008,
      "loss": 1.7395,
      "theoretical_loss": 4.279361372711923,
      "tokens_seen": 243531776
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009355466399197593,
      "loss": 1.899,
      "theoretical_loss": 4.27922260236887,
      "tokens_seen": 243597312
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009355265797392176,
      "loss": 1.8655,
      "theoretical_loss": 4.279083879805064,
      "tokens_seen": 243662848
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000935506519558676,
      "loss": 1.8957,
      "theoretical_loss": 4.278945204991212,
      "tokens_seen": 243728384
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009354864593781344,
      "loss": 1.8112,
      "theoretical_loss": 4.278806577898042,
      "tokens_seen": 243793920
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009354663991975928,
      "loss": 1.7751,
      "theoretical_loss": 4.278667998496313,
      "tokens_seen": 243859456
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009354463390170511,
      "loss": 1.8183,
      "theoretical_loss": 4.278529466756806,
      "tokens_seen": 243924992
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009354262788365096,
      "loss": 1.7717,
      "theoretical_loss": 4.278390982650333,
      "tokens_seen": 243990528
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000935406218655968,
      "loss": 1.7982,
      "theoretical_loss": 4.278252546147724,
      "tokens_seen": 244056064
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009353861584754263,
      "loss": 1.9031,
      "theoretical_loss": 4.278114157219839,
      "tokens_seen": 244121600
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009353660982948847,
      "loss": 1.7776,
      "theoretical_loss": 4.277975815837567,
      "tokens_seen": 244187136
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000935346038114343,
      "loss": 1.8307,
      "theoretical_loss": 4.277837521971816,
      "tokens_seen": 244252672
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009353259779338015,
      "loss": 1.7035,
      "theoretical_loss": 4.277699275593523,
      "tokens_seen": 244318208
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009353059177532598,
      "loss": 1.6069,
      "theoretical_loss": 4.277561076673652,
      "tokens_seen": 244383744
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009352858575727182,
      "loss": 1.7675,
      "theoretical_loss": 4.277422925183188,
      "tokens_seen": 244449280
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009352657973921765,
      "loss": 1.7285,
      "theoretical_loss": 4.277284821093147,
      "tokens_seen": 244514816
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000935245737211635,
      "loss": 1.7922,
      "theoretical_loss": 4.277146764374566,
      "tokens_seen": 244580352
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009352256770310933,
      "loss": 1.7639,
      "theoretical_loss": 4.27700875499851,
      "tokens_seen": 244645888
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009352056168505517,
      "loss": 1.8802,
      "theoretical_loss": 4.2768707929360685,
      "tokens_seen": 244711424
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00093518555667001,
      "loss": 1.7328,
      "theoretical_loss": 4.276732878158356,
      "tokens_seen": 244776960
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009351654964894684,
      "loss": 1.8951,
      "theoretical_loss": 4.276595010636514,
      "tokens_seen": 244842496
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009351454363089268,
      "loss": 1.7935,
      "theoretical_loss": 4.276457190341707,
      "tokens_seen": 244908032
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009351253761283852,
      "loss": 1.8199,
      "theoretical_loss": 4.2763194172451255,
      "tokens_seen": 244973568
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009351053159478435,
      "loss": 1.7277,
      "theoretical_loss": 4.276181691317987,
      "tokens_seen": 245039104
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000935085255767302,
      "loss": 1.8886,
      "theoretical_loss": 4.276044012531534,
      "tokens_seen": 245104640
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009350651955867603,
      "loss": 2.0016,
      "theoretical_loss": 4.275906380857031,
      "tokens_seen": 245170176
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009350451354062187,
      "loss": 1.7677,
      "theoretical_loss": 4.275768796265771,
      "tokens_seen": 245235712
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009350250752256771,
      "loss": 1.7328,
      "theoretical_loss": 4.275631258729071,
      "tokens_seen": 245301248
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009350050150451354,
      "loss": 1.83,
      "theoretical_loss": 4.275493768218274,
      "tokens_seen": 245366784
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009349849548645938,
      "loss": 1.7851,
      "theoretical_loss": 4.275356324704745,
      "tokens_seen": 245432320
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009349648946840523,
      "loss": 1.8365,
      "theoretical_loss": 4.27521892815988,
      "tokens_seen": 245497856
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009349448345035105,
      "loss": 1.6734,
      "theoretical_loss": 4.275081578555094,
      "tokens_seen": 245563392
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000934924774322969,
      "loss": 1.7928,
      "theoretical_loss": 4.274944275861828,
      "tokens_seen": 245628928
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009349047141424272,
      "loss": 1.7817,
      "theoretical_loss": 4.274807020051552,
      "tokens_seen": 245694464
    },
    {
      "epoch": 0.07,
      "objective/train/advantage_avg": 0.4900068938732147,
      "objective/train/docs_used": 96014,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7985907793045044,
      "objective/train/original_loss": 1.798590898513794,
      "objective/train/theoretical_loss": 4.274669811095759,
      "objective/train/tokens_used": 266220000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24235199391841888,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502325296401978,
      "objective/train/weighted_lm_loss": 1.888869047164917,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.274669811095759,
      "tokens_seen": 245760000
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009348846539618857,
      "loss": 1.7632,
      "theoretical_loss": 4.274669811095759,
      "tokens_seen": 245760000
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009348645937813441,
      "loss": 1.7704,
      "theoretical_loss": 4.274532648965964,
      "tokens_seen": 245825536
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009348445336008024,
      "loss": 1.7771,
      "theoretical_loss": 4.274395533633712,
      "tokens_seen": 245891072
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009348244734202608,
      "loss": 1.7508,
      "theoretical_loss": 4.274258465070567,
      "tokens_seen": 245956608
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009348044132397193,
      "loss": 1.8499,
      "theoretical_loss": 4.274121443248122,
      "tokens_seen": 246022144
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009347843530591775,
      "loss": 1.7992,
      "theoretical_loss": 4.273984468137995,
      "tokens_seen": 246087680
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000934764292878636,
      "loss": 1.5857,
      "theoretical_loss": 4.273847539711825,
      "tokens_seen": 246153216
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009347442326980942,
      "loss": 1.8579,
      "theoretical_loss": 4.273710657941282,
      "tokens_seen": 246218752
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009347241725175527,
      "loss": 1.8407,
      "theoretical_loss": 4.273573822798054,
      "tokens_seen": 246284288
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009347041123370111,
      "loss": 1.839,
      "theoretical_loss": 4.273437034253859,
      "tokens_seen": 246349824
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009346840521564694,
      "loss": 1.8247,
      "theoretical_loss": 4.273300292280435,
      "tokens_seen": 246415360
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009346639919759278,
      "loss": 1.8677,
      "theoretical_loss": 4.273163596849549,
      "tokens_seen": 246480896
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009346439317953862,
      "loss": 1.8076,
      "theoretical_loss": 4.27302694793299,
      "tokens_seen": 246546432
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009346238716148445,
      "loss": 1.748,
      "theoretical_loss": 4.272890345502573,
      "tokens_seen": 246611968
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000934603811434303,
      "loss": 1.8006,
      "theoretical_loss": 4.272753789530134,
      "tokens_seen": 246677504
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009345837512537612,
      "loss": 1.6839,
      "theoretical_loss": 4.272617279987541,
      "tokens_seen": 246743040
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009345636910732197,
      "loss": 1.6944,
      "theoretical_loss": 4.272480816846678,
      "tokens_seen": 246808576
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000934543630892678,
      "loss": 1.7108,
      "theoretical_loss": 4.27234440007946,
      "tokens_seen": 246874112
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009345235707121364,
      "loss": 1.8806,
      "theoretical_loss": 4.272208029657822,
      "tokens_seen": 246939648
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009345035105315948,
      "loss": 1.8091,
      "theoretical_loss": 4.272071705553726,
      "tokens_seen": 247005184
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009344834503510532,
      "loss": 1.7406,
      "theoretical_loss": 4.271935427739158,
      "tokens_seen": 247070720
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009344633901705115,
      "loss": 1.7875,
      "theoretical_loss": 4.2717991961861275,
      "tokens_seen": 247136256
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00093444332998997,
      "loss": 1.7592,
      "theoretical_loss": 4.271663010866669,
      "tokens_seen": 247201792
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009344232698094283,
      "loss": 1.9044,
      "theoretical_loss": 4.27152687175284,
      "tokens_seen": 247267328
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009344032096288867,
      "loss": 1.8186,
      "theoretical_loss": 4.271390778816727,
      "tokens_seen": 247332864
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000934383149448345,
      "loss": 1.8104,
      "theoretical_loss": 4.2712547320304335,
      "tokens_seen": 247398400
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0009343630892678034,
      "loss": 1.7732,
      "theoretical_loss": 4.2711187313660925,
      "tokens_seen": 247463936
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009343430290872618,
      "loss": 1.7146,
      "theoretical_loss": 4.27098277679586,
      "tokens_seen": 247529472
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009343229689067202,
      "loss": 1.8197,
      "theoretical_loss": 4.270846868291914,
      "tokens_seen": 247595008
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009343029087261785,
      "loss": 1.8096,
      "theoretical_loss": 4.270711005826461,
      "tokens_seen": 247660544
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009342828485456369,
      "loss": 1.8219,
      "theoretical_loss": 4.270575189371727,
      "tokens_seen": 247726080
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009342627883650953,
      "loss": 1.8408,
      "theoretical_loss": 4.270439418899966,
      "tokens_seen": 247791616
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009342427281845537,
      "loss": 1.772,
      "theoretical_loss": 4.270303694383453,
      "tokens_seen": 247857152
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000934222668004012,
      "loss": 1.7264,
      "theoretical_loss": 4.270168015794489,
      "tokens_seen": 247922688
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009342026078234705,
      "loss": 1.8814,
      "theoretical_loss": 4.270032383105398,
      "tokens_seen": 247988224
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009341825476429287,
      "loss": 1.7944,
      "theoretical_loss": 4.269896796288528,
      "tokens_seen": 248053760
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009341624874623872,
      "loss": 1.7638,
      "theoretical_loss": 4.269761255316252,
      "tokens_seen": 248119296
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009341424272818455,
      "loss": 1.7488,
      "theoretical_loss": 4.269625760160965,
      "tokens_seen": 248184832
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009341223671013039,
      "loss": 1.7874,
      "theoretical_loss": 4.269490310795089,
      "tokens_seen": 248250368
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009341023069207623,
      "loss": 1.899,
      "theoretical_loss": 4.269354907191065,
      "tokens_seen": 248315904
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009340822467402207,
      "loss": 1.9713,
      "theoretical_loss": 4.269219549321364,
      "tokens_seen": 248381440
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000934062186559679,
      "loss": 1.8625,
      "theoretical_loss": 4.269084237158476,
      "tokens_seen": 248446976
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009340421263791375,
      "loss": 1.8055,
      "theoretical_loss": 4.268948970674917,
      "tokens_seen": 248512512
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009340220661985957,
      "loss": 2.1709,
      "theoretical_loss": 4.268813749843226,
      "tokens_seen": 248578048
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009340020060180542,
      "loss": 1.72,
      "theoretical_loss": 4.268678574635967,
      "tokens_seen": 248643584
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009339819458375125,
      "loss": 1.7354,
      "theoretical_loss": 4.268543445025724,
      "tokens_seen": 248709120
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009339618856569709,
      "loss": 1.8053,
      "theoretical_loss": 4.268408360985109,
      "tokens_seen": 248774656
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009339418254764293,
      "loss": 1.9108,
      "theoretical_loss": 4.268273322486757,
      "tokens_seen": 248840192
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009339217652958876,
      "loss": 1.775,
      "theoretical_loss": 4.268138329503325,
      "tokens_seen": 248905728
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009339017051153461,
      "loss": 1.8734,
      "theoretical_loss": 4.2680033820074925,
      "tokens_seen": 248971264
    },
    {
      "epoch": 0.08,
      "objective/train/advantage_avg": 0.4903026223182678,
      "objective/train/docs_used": 97104,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.707381010055542,
      "objective/train/original_loss": 1.7073808908462524,
      "objective/train/theoretical_loss": 4.267868479971966,
      "objective/train/tokens_used": 269496800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24140579998493195,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502574443817139,
      "objective/train/weighted_lm_loss": 1.793129563331604,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.267868479971966,
      "tokens_seen": 249036800
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009338816449348045,
      "loss": 1.7747,
      "theoretical_loss": 4.267868479971966,
      "tokens_seen": 249036800
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009338615847542628,
      "loss": 1.7815,
      "theoretical_loss": 4.267733623369474,
      "tokens_seen": 249102336
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009338415245737212,
      "loss": 1.8381,
      "theoretical_loss": 4.267598812172769,
      "tokens_seen": 249167872
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009338214643931797,
      "loss": 1.966,
      "theoretical_loss": 4.2674640463546245,
      "tokens_seen": 249233408
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009338014042126379,
      "loss": 1.752,
      "theoretical_loss": 4.267329325887841,
      "tokens_seen": 249298944
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009337813440320964,
      "loss": 1.5597,
      "theoretical_loss": 4.26719465074524,
      "tokens_seen": 249364480
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009337612838515546,
      "loss": 1.719,
      "theoretical_loss": 4.267060020899668,
      "tokens_seen": 249430016
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009337412236710131,
      "loss": 1.804,
      "theoretical_loss": 4.2669254363239935,
      "tokens_seen": 249495552
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009337211634904715,
      "loss": 1.7289,
      "theoretical_loss": 4.266790896991109,
      "tokens_seen": 249561088
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009337011033099298,
      "loss": 1.8028,
      "theoretical_loss": 4.266656402873932,
      "tokens_seen": 249626624
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009336810431293882,
      "loss": 1.9201,
      "theoretical_loss": 4.2665219539454,
      "tokens_seen": 249692160
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009336609829488466,
      "loss": 1.9164,
      "theoretical_loss": 4.266387550178476,
      "tokens_seen": 249757696
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009336409227683049,
      "loss": 1.8046,
      "theoretical_loss": 4.266253191546146,
      "tokens_seen": 249823232
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009336208625877634,
      "loss": 1.8337,
      "theoretical_loss": 4.266118878021419,
      "tokens_seen": 249888768
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009336008024072216,
      "loss": 1.9622,
      "theoretical_loss": 4.265984609577327,
      "tokens_seen": 249954304
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009335807422266801,
      "loss": 1.8541,
      "theoretical_loss": 4.265850386186925,
      "tokens_seen": 250019840
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009335606820461384,
      "loss": 1.6784,
      "theoretical_loss": 4.265716207823292,
      "tokens_seen": 250085376
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009335406218655968,
      "loss": 1.7275,
      "theoretical_loss": 4.265582074459531,
      "tokens_seen": 250150912
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009335205616850552,
      "loss": 1.8329,
      "theoretical_loss": 4.265447986068765,
      "tokens_seen": 250216448
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009335005015045136,
      "loss": 1.8422,
      "theoretical_loss": 4.265313942624143,
      "tokens_seen": 250281984
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009334804413239719,
      "loss": 1.9325,
      "theoretical_loss": 4.2651799440988345,
      "tokens_seen": 250347520
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009334603811434304,
      "loss": 1.8524,
      "theoretical_loss": 4.265045990466035,
      "tokens_seen": 250413056
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009334403209628887,
      "loss": 1.8634,
      "theoretical_loss": 4.264912081698962,
      "tokens_seen": 250478592
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009334202607823471,
      "loss": 1.7535,
      "theoretical_loss": 4.264778217770854,
      "tokens_seen": 250544128
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009334002006018054,
      "loss": 1.7067,
      "theoretical_loss": 4.2646443986549745,
      "tokens_seen": 250609664
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009333801404212638,
      "loss": 1.767,
      "theoretical_loss": 4.264510624324609,
      "tokens_seen": 250675200
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009333600802407222,
      "loss": 1.9334,
      "theoretical_loss": 4.2643768947530685,
      "tokens_seen": 250740736
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009333400200601806,
      "loss": 1.8349,
      "theoretical_loss": 4.2642432099136816,
      "tokens_seen": 250806272
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009333199598796389,
      "loss": 1.751,
      "theoretical_loss": 4.264109569779803,
      "tokens_seen": 250871808
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009332998996990973,
      "loss": 1.9473,
      "theoretical_loss": 4.263975974324813,
      "tokens_seen": 250937344
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009332798395185557,
      "loss": 1.7429,
      "theoretical_loss": 4.263842423522109,
      "tokens_seen": 251002880
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009332597793380141,
      "loss": 1.8841,
      "theoretical_loss": 4.263708917345115,
      "tokens_seen": 251068416
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009332397191574724,
      "loss": 1.9271,
      "theoretical_loss": 4.263575455767277,
      "tokens_seen": 251133952
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009332196589769309,
      "loss": 1.9296,
      "theoretical_loss": 4.2634420387620615,
      "tokens_seen": 251199488
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009331995987963891,
      "loss": 1.8366,
      "theoretical_loss": 4.2633086663029625,
      "tokens_seen": 251265024
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009331795386158476,
      "loss": 1.8976,
      "theoretical_loss": 4.263175338363491,
      "tokens_seen": 251330560
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009331594784353059,
      "loss": 1.8301,
      "theoretical_loss": 4.263042054917186,
      "tokens_seen": 251396096
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009331394182547643,
      "loss": 2.0389,
      "theoretical_loss": 4.262908815937604,
      "tokens_seen": 251461632
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009331193580742227,
      "loss": 2.0028,
      "theoretical_loss": 4.262775621398329,
      "tokens_seen": 251527168
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009330992978936811,
      "loss": 2.0408,
      "theoretical_loss": 4.262642471272963,
      "tokens_seen": 251592704
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009330792377131394,
      "loss": 2.0631,
      "theoretical_loss": 4.262509365535134,
      "tokens_seen": 251658240
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009330591775325979,
      "loss": 1.8172,
      "theoretical_loss": 4.262376304158492,
      "tokens_seen": 251723776
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009330391173520561,
      "loss": 1.9694,
      "theoretical_loss": 4.262243287116707,
      "tokens_seen": 251789312
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009330190571715146,
      "loss": 1.8517,
      "theoretical_loss": 4.2621103143834755,
      "tokens_seen": 251854848
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000932998996990973,
      "loss": 2.0222,
      "theoretical_loss": 4.261977385932512,
      "tokens_seen": 251920384
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009329789368104313,
      "loss": 1.9252,
      "theoretical_loss": 4.261844501737556,
      "tokens_seen": 251985920
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009329588766298897,
      "loss": 1.9154,
      "theoretical_loss": 4.26171166177237,
      "tokens_seen": 252051456
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000932938816449348,
      "loss": 1.8763,
      "theoretical_loss": 4.261578866010739,
      "tokens_seen": 252116992
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009329187562688064,
      "loss": 1.9258,
      "theoretical_loss": 4.261446114426466,
      "tokens_seen": 252182528
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009328986960882649,
      "loss": 1.823,
      "theoretical_loss": 4.261313406993382,
      "tokens_seen": 252248064
    },
    {
      "epoch": 0.08,
      "objective/train/advantage_avg": 0.49495235085487366,
      "objective/train/docs_used": 98257,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8033281564712524,
      "objective/train/original_loss": 1.803328037261963,
      "objective/train/theoretical_loss": 4.261180743685337,
      "objective/train/tokens_used": 272773600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24527713656425476,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507421493530273,
      "objective/train/weighted_lm_loss": 1.8945033550262451,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 4.261180743685337,
      "tokens_seen": 252313600
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009328786359077231,
      "loss": 1.8192,
      "theoretical_loss": 4.261180743685337,
      "tokens_seen": 252313600
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009328585757271816,
      "loss": 1.9158,
      "theoretical_loss": 4.261048124476204,
      "tokens_seen": 252379136
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009328385155466398,
      "loss": 1.9291,
      "theoretical_loss": 4.260915549339879,
      "tokens_seen": 252444672
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009328184553660983,
      "loss": 1.7672,
      "theoretical_loss": 4.2607830182502795,
      "tokens_seen": 252510208
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009327983951855567,
      "loss": 1.8606,
      "theoretical_loss": 4.260650531181345,
      "tokens_seen": 252575744
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000932778335005015,
      "loss": 1.8812,
      "theoretical_loss": 4.260518088107038,
      "tokens_seen": 252641280
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009327582748244734,
      "loss": 1.9065,
      "theoretical_loss": 4.2603856890013425,
      "tokens_seen": 252706816
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009327382146439319,
      "loss": 1.8938,
      "theoretical_loss": 4.260253333838264,
      "tokens_seen": 252772352
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009327181544633901,
      "loss": 1.9165,
      "theoretical_loss": 4.260121022591832,
      "tokens_seen": 252837888
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009326980942828486,
      "loss": 1.7946,
      "theoretical_loss": 4.259988755236097,
      "tokens_seen": 252903424
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009326780341023069,
      "loss": 1.9426,
      "theoretical_loss": 4.25985653174513,
      "tokens_seen": 252968960
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009326579739217653,
      "loss": 1.7934,
      "theoretical_loss": 4.259724352093027,
      "tokens_seen": 253034496
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009326379137412237,
      "loss": 1.7676,
      "theoretical_loss": 4.259592216253903,
      "tokens_seen": 253100032
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000932617853560682,
      "loss": 1.8153,
      "theoretical_loss": 4.259460124201899,
      "tokens_seen": 253165568
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009325977933801404,
      "loss": 1.7746,
      "theoretical_loss": 4.259328075911173,
      "tokens_seen": 253231104
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009325777331995988,
      "loss": 1.7761,
      "theoretical_loss": 4.259196071355909,
      "tokens_seen": 253296640
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009325576730190571,
      "loss": 1.9124,
      "theoretical_loss": 4.259064110510311,
      "tokens_seen": 253362176
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009325376128385156,
      "loss": 1.8265,
      "theoretical_loss": 4.258932193348604,
      "tokens_seen": 253427712
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009325175526579739,
      "loss": 1.9741,
      "theoretical_loss": 4.258800319845038,
      "tokens_seen": 253493248
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009324974924774323,
      "loss": 1.8248,
      "theoretical_loss": 4.258668489973882,
      "tokens_seen": 253558784
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009324774322968907,
      "loss": 1.8307,
      "theoretical_loss": 4.258536703709426,
      "tokens_seen": 253624320
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000932457372116349,
      "loss": 1.7819,
      "theoretical_loss": 4.258404961025986,
      "tokens_seen": 253689856
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009324373119358074,
      "loss": 1.9135,
      "theoretical_loss": 4.258273261897896,
      "tokens_seen": 253755392
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009324172517552658,
      "loss": 1.8765,
      "theoretical_loss": 4.258141606299514,
      "tokens_seen": 253820928
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009323971915747242,
      "loss": 1.7368,
      "theoretical_loss": 4.258009994205218,
      "tokens_seen": 253886464
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009323771313941826,
      "loss": 1.8066,
      "theoretical_loss": 4.257878425589407,
      "tokens_seen": 253952000
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000932357071213641,
      "loss": 1.8179,
      "theoretical_loss": 4.257746900426506,
      "tokens_seen": 254017536
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009323370110330993,
      "loss": 1.7851,
      "theoretical_loss": 4.257615418690956,
      "tokens_seen": 254083072
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009323169508525577,
      "loss": 1.7856,
      "theoretical_loss": 4.2574839803572235,
      "tokens_seen": 254148608
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009322968906720161,
      "loss": 1.8776,
      "theoretical_loss": 4.257352585399795,
      "tokens_seen": 254214144
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009322768304914745,
      "loss": 1.8318,
      "theoretical_loss": 4.25722123379318,
      "tokens_seen": 254279680
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009322567703109328,
      "loss": 1.952,
      "theoretical_loss": 4.257089925511907,
      "tokens_seen": 254345216
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009322367101303913,
      "loss": 2.0118,
      "theoretical_loss": 4.256958660530529,
      "tokens_seen": 254410752
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009322166499498495,
      "loss": 1.8912,
      "theoretical_loss": 4.256827438823617,
      "tokens_seen": 254476288
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000932196589769308,
      "loss": 1.7353,
      "theoretical_loss": 4.256696260365768,
      "tokens_seen": 254541824
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009321765295887663,
      "loss": 1.8728,
      "theoretical_loss": 4.256565125131597,
      "tokens_seen": 254607360
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009321564694082247,
      "loss": 1.8088,
      "theoretical_loss": 4.25643403309574,
      "tokens_seen": 254672896
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009321364092276831,
      "loss": 1.8988,
      "theoretical_loss": 4.256302984232858,
      "tokens_seen": 254738432
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009321163490471415,
      "loss": 1.8722,
      "theoretical_loss": 4.256171978517629,
      "tokens_seen": 254803968
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009320962888665998,
      "loss": 1.6762,
      "theoretical_loss": 4.256041015924756,
      "tokens_seen": 254869504
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009320762286860583,
      "loss": 1.8221,
      "theoretical_loss": 4.255910096428962,
      "tokens_seen": 254935040
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009320561685055165,
      "loss": 1.8804,
      "theoretical_loss": 4.255779220004991,
      "tokens_seen": 255000576
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000932036108324975,
      "loss": 1.6887,
      "theoretical_loss": 4.255648386627607,
      "tokens_seen": 255066112
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009320160481444333,
      "loss": 1.8639,
      "theoretical_loss": 4.255517596271598,
      "tokens_seen": 255131648
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009319959879638917,
      "loss": 1.8197,
      "theoretical_loss": 4.255386848911772,
      "tokens_seen": 255197184
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009319759277833501,
      "loss": 1.7893,
      "theoretical_loss": 4.255256144522958,
      "tokens_seen": 255262720
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009319558676028084,
      "loss": 1.8523,
      "theoretical_loss": 4.255125483080007,
      "tokens_seen": 255328256
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009319358074222668,
      "loss": 1.8729,
      "theoretical_loss": 4.254994864557789,
      "tokens_seen": 255393792
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009319157472417253,
      "loss": 1.7798,
      "theoretical_loss": 4.254864288931198,
      "tokens_seen": 255459328
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009318956870611835,
      "loss": 1.8593,
      "theoretical_loss": 4.254733756175147,
      "tokens_seen": 255524864
    },
    {
      "epoch": 0.08,
      "objective/train/advantage_avg": 0.4922325015068054,
      "objective/train/docs_used": 99403,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7772718667984009,
      "objective/train/original_loss": 1.7772717475891113,
      "objective/train/theoretical_loss": 4.254603266264572,
      "objective/train/tokens_used": 276050400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24300870299339294,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050458550453186,
      "objective/train/weighted_lm_loss": 1.8668968677520752,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.254603266264572,
      "tokens_seen": 255590400
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000931875626880642,
      "loss": 1.7993,
      "theoretical_loss": 4.254603266264572,
      "tokens_seen": 255590400
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009318555667001002,
      "loss": 1.7996,
      "theoretical_loss": 4.254472819174428,
      "tokens_seen": 255655936
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009318355065195587,
      "loss": 1.6432,
      "theoretical_loss": 4.254342414879692,
      "tokens_seen": 255721472
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009318154463390171,
      "loss": 1.833,
      "theoretical_loss": 4.254212053355362,
      "tokens_seen": 255787008
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009317953861584754,
      "loss": 1.8269,
      "theoretical_loss": 4.254081734576458,
      "tokens_seen": 255852544
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009317753259779338,
      "loss": 1.7435,
      "theoretical_loss": 4.25395145851802,
      "tokens_seen": 255918080
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009317552657973923,
      "loss": 1.8863,
      "theoretical_loss": 4.253821225155107,
      "tokens_seen": 255983616
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009317352056168505,
      "loss": 1.7373,
      "theoretical_loss": 4.253691034462804,
      "tokens_seen": 256049152
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000931715145436309,
      "loss": 1.8167,
      "theoretical_loss": 4.253560886416212,
      "tokens_seen": 256114688
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009316950852557673,
      "loss": 1.8065,
      "theoretical_loss": 4.253430780990455,
      "tokens_seen": 256180224
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009316750250752257,
      "loss": 1.7302,
      "theoretical_loss": 4.253300718160678,
      "tokens_seen": 256245760
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009316549648946841,
      "loss": 1.8542,
      "theoretical_loss": 4.253170697902046,
      "tokens_seen": 256311296
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009316349047141424,
      "loss": 1.7814,
      "theoretical_loss": 4.253040720189746,
      "tokens_seen": 256376832
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009316148445336008,
      "loss": 1.9602,
      "theoretical_loss": 4.252910784998984,
      "tokens_seen": 256442368
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009315947843530592,
      "loss": 1.6887,
      "theoretical_loss": 4.2527808923049895,
      "tokens_seen": 256507904
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009315747241725175,
      "loss": 1.8397,
      "theoretical_loss": 4.252651042083009,
      "tokens_seen": 256573440
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000931554663991976,
      "loss": 1.7413,
      "theoretical_loss": 4.252521234308315,
      "tokens_seen": 256638976
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009315346038114343,
      "loss": 1.7299,
      "theoretical_loss": 4.252391468956194,
      "tokens_seen": 256704512
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009315145436308927,
      "loss": 1.7748,
      "theoretical_loss": 4.25226174600196,
      "tokens_seen": 256770048
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009314944834503511,
      "loss": 1.7904,
      "theoretical_loss": 4.252132065420942,
      "tokens_seen": 256835584
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009314744232698095,
      "loss": 1.8599,
      "theoretical_loss": 4.2520024271884935,
      "tokens_seen": 256901120
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009314543630892678,
      "loss": 1.6606,
      "theoretical_loss": 4.251872831279986,
      "tokens_seen": 256966656
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009314343029087262,
      "loss": 1.7963,
      "theoretical_loss": 4.251743277670815,
      "tokens_seen": 257032192
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009314142427281845,
      "loss": 1.9087,
      "theoretical_loss": 4.251613766336391,
      "tokens_seen": 257097728
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000931394182547643,
      "loss": 1.9016,
      "theoretical_loss": 4.251484297252151,
      "tokens_seen": 257163264
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009313741223671013,
      "loss": 1.9086,
      "theoretical_loss": 4.25135487039355,
      "tokens_seen": 257228800
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009313540621865597,
      "loss": 1.9209,
      "theoretical_loss": 4.251225485736063,
      "tokens_seen": 257294336
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000931334002006018,
      "loss": 1.7854,
      "theoretical_loss": 4.251096143255185,
      "tokens_seen": 257359872
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009313139418254765,
      "loss": 1.9528,
      "theoretical_loss": 4.250966842926434,
      "tokens_seen": 257425408
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009312938816449348,
      "loss": 1.808,
      "theoretical_loss": 4.250837584725346,
      "tokens_seen": 257490944
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009312738214643932,
      "loss": 1.7127,
      "theoretical_loss": 4.250708368627477,
      "tokens_seen": 257556480
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009312537612838515,
      "loss": 1.9612,
      "theoretical_loss": 4.250579194608408,
      "tokens_seen": 257622016
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009312337011033099,
      "loss": 1.8438,
      "theoretical_loss": 4.250450062643734,
      "tokens_seen": 257687552
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009312136409227683,
      "loss": 1.7651,
      "theoretical_loss": 4.2503209727090745,
      "tokens_seen": 257753088
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009311935807422267,
      "loss": 1.8181,
      "theoretical_loss": 4.250191924780069,
      "tokens_seen": 257818624
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000931173520561685,
      "loss": 1.781,
      "theoretical_loss": 4.250062918832374,
      "tokens_seen": 257884160
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009311534603811435,
      "loss": 1.6716,
      "theoretical_loss": 4.249933954841672,
      "tokens_seen": 257949696
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009311334002006018,
      "loss": 2.0471,
      "theoretical_loss": 4.249805032783661,
      "tokens_seen": 258015232
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009311133400200602,
      "loss": 1.8217,
      "theoretical_loss": 4.249676152634061,
      "tokens_seen": 258080768
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009310932798395186,
      "loss": 1.9101,
      "theoretical_loss": 4.249547314368613,
      "tokens_seen": 258146304
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009310732196589769,
      "loss": 1.8241,
      "theoretical_loss": 4.2494185179630755,
      "tokens_seen": 258211840
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009310531594784353,
      "loss": 1.8213,
      "theoretical_loss": 4.24928976339323,
      "tokens_seen": 258277376
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009310330992978937,
      "loss": 1.8407,
      "theoretical_loss": 4.249161050634876,
      "tokens_seen": 258342912
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000931013039117352,
      "loss": 1.9322,
      "theoretical_loss": 4.249032379663836,
      "tokens_seen": 258408448
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009309929789368105,
      "loss": 1.7703,
      "theoretical_loss": 4.24890375045595,
      "tokens_seen": 258473984
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009309729187562687,
      "loss": 1.8058,
      "theoretical_loss": 4.248775162987078,
      "tokens_seen": 258539520
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009309528585757272,
      "loss": 1.7423,
      "theoretical_loss": 4.248646617233103,
      "tokens_seen": 258605056
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009309327983951856,
      "loss": 1.7474,
      "theoretical_loss": 4.248518113169924,
      "tokens_seen": 258670592
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009309127382146439,
      "loss": 1.9037,
      "theoretical_loss": 4.248389650773463,
      "tokens_seen": 258736128
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009308926780341024,
      "loss": 1.7221,
      "theoretical_loss": 4.248261230019661,
      "tokens_seen": 258801664
    },
    {
      "epoch": 0.08,
      "objective/train/advantage_avg": 0.4922111928462982,
      "objective/train/docs_used": 100739,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5991153717041016,
      "objective/train/original_loss": 1.5991151332855225,
      "objective/train/theoretical_loss": 4.24813285088448,
      "objective/train/tokens_used": 279327200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24256941676139832,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504542589187622,
      "objective/train/weighted_lm_loss": 1.679618239402771,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 4.24813285088448,
      "tokens_seen": 258867200
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009308726178535606,
      "loss": 1.7685,
      "theoretical_loss": 4.24813285088448,
      "tokens_seen": 258867200
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009308525576730191,
      "loss": 1.8756,
      "theoretical_loss": 4.248004513343899,
      "tokens_seen": 258932736
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009308324974924775,
      "loss": 1.9389,
      "theoretical_loss": 4.24787621737392,
      "tokens_seen": 258998272
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009308124373119358,
      "loss": 1.7622,
      "theoretical_loss": 4.247747962950564,
      "tokens_seen": 259063808
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009307923771313942,
      "loss": 1.8245,
      "theoretical_loss": 4.247619750049871,
      "tokens_seen": 259129344
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009307723169508527,
      "loss": 1.7996,
      "theoretical_loss": 4.247491578647903,
      "tokens_seen": 259194880
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009307522567703109,
      "loss": 1.7617,
      "theoretical_loss": 4.247363448720739,
      "tokens_seen": 259260416
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009307321965897694,
      "loss": 1.7627,
      "theoretical_loss": 4.24723536024448,
      "tokens_seen": 259325952
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009307121364092277,
      "loss": 1.8327,
      "theoretical_loss": 4.247107313195246,
      "tokens_seen": 259391488
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009306920762286861,
      "loss": 1.8597,
      "theoretical_loss": 4.246979307549177,
      "tokens_seen": 259457024
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009306720160481445,
      "loss": 1.8707,
      "theoretical_loss": 4.246851343282432,
      "tokens_seen": 259522560
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009306519558676028,
      "loss": 1.9826,
      "theoretical_loss": 4.246723420371191,
      "tokens_seen": 259588096
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009306318956870612,
      "loss": 1.828,
      "theoretical_loss": 4.246595538791653,
      "tokens_seen": 259653632
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009306118355065196,
      "loss": 1.8696,
      "theoretical_loss": 4.246467698520037,
      "tokens_seen": 259719168
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009305917753259779,
      "loss": 1.823,
      "theoretical_loss": 4.246339899532582,
      "tokens_seen": 259784704
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009305717151454364,
      "loss": 1.872,
      "theoretical_loss": 4.246212141805544,
      "tokens_seen": 259850240
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009305516549648947,
      "loss": 1.8403,
      "theoretical_loss": 4.246084425315203,
      "tokens_seen": 259915776
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009305315947843531,
      "loss": 1.6529,
      "theoretical_loss": 4.245956750037855,
      "tokens_seen": 259981312
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009305115346038115,
      "loss": 1.7805,
      "theoretical_loss": 4.245829115949818,
      "tokens_seen": 260046848
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009304914744232699,
      "loss": 1.9065,
      "theoretical_loss": 4.245701523027429,
      "tokens_seen": 260112384
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009304714142427282,
      "loss": 1.8109,
      "theoretical_loss": 4.245573971247042,
      "tokens_seen": 260177920
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009304513540621866,
      "loss": 1.8887,
      "theoretical_loss": 4.245446460585035,
      "tokens_seen": 260243456
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009304312938816449,
      "loss": 1.886,
      "theoretical_loss": 4.245318991017802,
      "tokens_seen": 260308992
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009304112337011034,
      "loss": 1.7386,
      "theoretical_loss": 4.245191562521757,
      "tokens_seen": 260374528
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009303911735205617,
      "loss": 1.8463,
      "theoretical_loss": 4.2450641750733356,
      "tokens_seen": 260440064
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009303711133400201,
      "loss": 1.7975,
      "theoretical_loss": 4.24493682864899,
      "tokens_seen": 260505600
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009303510531594784,
      "loss": 1.8787,
      "theoretical_loss": 4.244809523225195,
      "tokens_seen": 260571136
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009303309929789369,
      "loss": 1.7446,
      "theoretical_loss": 4.244682258778441,
      "tokens_seen": 260636672
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009303109327983952,
      "loss": 1.8998,
      "theoretical_loss": 4.2445550352852415,
      "tokens_seen": 260702208
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009302908726178536,
      "loss": 1.919,
      "theoretical_loss": 4.244427852722127,
      "tokens_seen": 260767744
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000930270812437312,
      "loss": 1.8169,
      "theoretical_loss": 4.244300711065646,
      "tokens_seen": 260833280
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009302507522567703,
      "loss": 1.785,
      "theoretical_loss": 4.244173610292372,
      "tokens_seen": 260898816
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009302306920762287,
      "loss": 1.685,
      "theoretical_loss": 4.244046550378892,
      "tokens_seen": 260964352
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009302106318956871,
      "loss": 1.6913,
      "theoretical_loss": 4.243919531301814,
      "tokens_seen": 261029888
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009301905717151454,
      "loss": 1.7615,
      "theoretical_loss": 4.243792553037767,
      "tokens_seen": 261095424
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009301705115346039,
      "loss": 1.8247,
      "theoretical_loss": 4.243665615563396,
      "tokens_seen": 261160960
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009301504513540622,
      "loss": 1.8357,
      "theoretical_loss": 4.243538718855369,
      "tokens_seen": 261226496
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009301303911735206,
      "loss": 1.8834,
      "theoretical_loss": 4.243411862890371,
      "tokens_seen": 261292032
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000930110330992979,
      "loss": 1.7083,
      "theoretical_loss": 4.243285047645106,
      "tokens_seen": 261357568
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009300902708124373,
      "loss": 1.8269,
      "theoretical_loss": 4.243158273096297,
      "tokens_seen": 261423104
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009300702106318957,
      "loss": 1.9182,
      "theoretical_loss": 4.243031539220689,
      "tokens_seen": 261488640
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009300501504513541,
      "loss": 1.9445,
      "theoretical_loss": 4.242904845995041,
      "tokens_seen": 261554176
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009300300902708124,
      "loss": 1.8878,
      "theoretical_loss": 4.242778193396136,
      "tokens_seen": 261619712
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009300100300902709,
      "loss": 1.8315,
      "theoretical_loss": 4.242651581400774,
      "tokens_seen": 261685248
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009299899699097291,
      "loss": 1.8104,
      "theoretical_loss": 4.242525009985773,
      "tokens_seen": 261750784
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009299699097291876,
      "loss": 1.8935,
      "theoretical_loss": 4.242398479127972,
      "tokens_seen": 261816320
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000929949849548646,
      "loss": 1.7999,
      "theoretical_loss": 4.242271988804228,
      "tokens_seen": 261881856
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009299297893681043,
      "loss": 1.7533,
      "theoretical_loss": 4.242145538991417,
      "tokens_seen": 261947392
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009299097291875627,
      "loss": 1.6952,
      "theoretical_loss": 4.242019129666434,
      "tokens_seen": 262012928
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000929889669007021,
      "loss": 1.688,
      "theoretical_loss": 4.241892760806194,
      "tokens_seen": 262078464
    },
    {
      "epoch": 0.08,
      "objective/train/advantage_avg": 0.4959133267402649,
      "objective/train/docs_used": 101934,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8622653484344482,
      "objective/train/original_loss": 1.8622653484344482,
      "objective/train/theoretical_loss": 4.241766432387629,
      "objective/train/tokens_used": 282604000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24602378904819489,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050842046737671,
      "objective/train/weighted_lm_loss": 1.957066535949707,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0396549701690674,
      "theoretical_loss": 4.241766432387629,
      "tokens_seen": 262144000
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009298696088264794,
      "loss": 1.8692,
      "theoretical_loss": 4.241766432387629,
      "tokens_seen": 262144000
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009298495486459379,
      "loss": 1.8105,
      "theoretical_loss": 4.241640144387691,
      "tokens_seen": 262209536
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009298294884653961,
      "loss": 1.7178,
      "theoretical_loss": 4.241513896783352,
      "tokens_seen": 262275072
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009298094282848546,
      "loss": 1.8063,
      "theoretical_loss": 4.2413876895516,
      "tokens_seen": 262340608
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000929789368104313,
      "loss": 1.8252,
      "theoretical_loss": 4.241261522669445,
      "tokens_seen": 262406144
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009297693079237713,
      "loss": 1.8267,
      "theoretical_loss": 4.241135396113914,
      "tokens_seen": 262471680
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009297492477432297,
      "loss": 1.9232,
      "theoretical_loss": 4.2410093098620525,
      "tokens_seen": 262537216
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000929729187562688,
      "loss": 1.7269,
      "theoretical_loss": 4.240883263890925,
      "tokens_seen": 262602752
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009297091273821464,
      "loss": 1.9214,
      "theoretical_loss": 4.240757258177617,
      "tokens_seen": 262668288
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009296890672016049,
      "loss": 1.8207,
      "theoretical_loss": 4.240631292699231,
      "tokens_seen": 262733824
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009296690070210631,
      "loss": 1.7695,
      "theoretical_loss": 4.2405053674328865,
      "tokens_seen": 262799360
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009296489468405216,
      "loss": 1.789,
      "theoretical_loss": 4.240379482355724,
      "tokens_seen": 262864896
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009296288866599799,
      "loss": 1.6906,
      "theoretical_loss": 4.240253637444903,
      "tokens_seen": 262930432
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009296088264794383,
      "loss": 1.689,
      "theoretical_loss": 4.2401278326776,
      "tokens_seen": 262995968
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009295887662988967,
      "loss": 1.7909,
      "theoretical_loss": 4.2400020680310115,
      "tokens_seen": 263061504
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009295687061183551,
      "loss": 1.7541,
      "theoretical_loss": 4.239876343482352,
      "tokens_seen": 263127040
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009295486459378134,
      "loss": 1.8034,
      "theoretical_loss": 4.239750659008854,
      "tokens_seen": 263192576
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009295285857572719,
      "loss": 1.7027,
      "theoretical_loss": 4.23962501458777,
      "tokens_seen": 263258112
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009295085255767301,
      "loss": 1.923,
      "theoretical_loss": 4.2394994101963706,
      "tokens_seen": 263323648
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009294884653961886,
      "loss": 1.7502,
      "theoretical_loss": 4.239373845811944,
      "tokens_seen": 263389184
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009294684052156469,
      "loss": 1.8158,
      "theoretical_loss": 4.2392483214117975,
      "tokens_seen": 263454720
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009294483450351053,
      "loss": 1.9041,
      "theoretical_loss": 4.239122836973258,
      "tokens_seen": 263520256
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009294282848545638,
      "loss": 1.8612,
      "theoretical_loss": 4.238997392473669,
      "tokens_seen": 263585792
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009294082246740221,
      "loss": 1.8119,
      "theoretical_loss": 4.238871987890395,
      "tokens_seen": 263651328
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009293881644934805,
      "loss": 1.8197,
      "theoretical_loss": 4.238746623200815,
      "tokens_seen": 263716864
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009293681043129388,
      "loss": 1.8277,
      "theoretical_loss": 4.238621298382331,
      "tokens_seen": 263782400
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009293480441323973,
      "loss": 1.8537,
      "theoretical_loss": 4.23849601341236,
      "tokens_seen": 263847936
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009293279839518556,
      "loss": 1.7895,
      "theoretical_loss": 4.2383707682683385,
      "tokens_seen": 263913472
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000929307923771314,
      "loss": 1.8325,
      "theoretical_loss": 4.238245562927722,
      "tokens_seen": 263979008
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009292878635907723,
      "loss": 1.7732,
      "theoretical_loss": 4.238120397367984,
      "tokens_seen": 264044544
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009292678034102307,
      "loss": 1.8699,
      "theoretical_loss": 4.237995271566616,
      "tokens_seen": 264110080
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009292477432296891,
      "loss": 1.9023,
      "theoretical_loss": 4.237870185501127,
      "tokens_seen": 264175616
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009292276830491475,
      "loss": 1.7925,
      "theoretical_loss": 4.237745139149047,
      "tokens_seen": 264241152
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009292076228686058,
      "loss": 1.825,
      "theoretical_loss": 4.237620132487923,
      "tokens_seen": 264306688
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009291875626880643,
      "loss": 1.7765,
      "theoretical_loss": 4.237495165495317,
      "tokens_seen": 264372224
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009291675025075226,
      "loss": 1.7354,
      "theoretical_loss": 4.237370238148814,
      "tokens_seen": 264437760
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000929147442326981,
      "loss": 1.7925,
      "theoretical_loss": 4.237245350426015,
      "tokens_seen": 264503296
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009291273821464394,
      "loss": 1.6433,
      "theoretical_loss": 4.23712050230454,
      "tokens_seen": 264568832
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009291073219658977,
      "loss": 1.8064,
      "theoretical_loss": 4.236995693762026,
      "tokens_seen": 264634368
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009290872617853561,
      "loss": 1.6874,
      "theoretical_loss": 4.236870924776128,
      "tokens_seen": 264699904
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009290672016048145,
      "loss": 1.8004,
      "theoretical_loss": 4.236746195324523,
      "tokens_seen": 264765440
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009290471414242728,
      "loss": 1.8509,
      "theoretical_loss": 4.2366215053849,
      "tokens_seen": 264830976
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009290270812437313,
      "loss": 1.8856,
      "theoretical_loss": 4.23649685493497,
      "tokens_seen": 264896512
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009290070210631895,
      "loss": 1.8162,
      "theoretical_loss": 4.236372243952461,
      "tokens_seen": 264962048
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000928986960882648,
      "loss": 1.9025,
      "theoretical_loss": 4.2362476724151215,
      "tokens_seen": 265027584
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009289669007021064,
      "loss": 1.8805,
      "theoretical_loss": 4.236123140300713,
      "tokens_seen": 265093120
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009289468405215647,
      "loss": 1.7836,
      "theoretical_loss": 4.235998647587021,
      "tokens_seen": 265158656
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009289267803410231,
      "loss": 1.8182,
      "theoretical_loss": 4.235874194251843,
      "tokens_seen": 265224192
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009289067201604814,
      "loss": 1.737,
      "theoretical_loss": 4.235749780272998,
      "tokens_seen": 265289728
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009288866599799398,
      "loss": 1.8422,
      "theoretical_loss": 4.2356254056283245,
      "tokens_seen": 265355264
    },
    {
      "epoch": 0.08,
      "objective/train/advantage_avg": 0.49507758021354675,
      "objective/train/docs_used": 103026,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5715179443359375,
      "objective/train/original_loss": 1.5715179443359375,
      "objective/train/theoretical_loss": 4.235501070295674,
      "objective/train/tokens_used": 285880800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24532651901245117,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507550239562988,
      "objective/train/weighted_lm_loss": 1.6512619256973267,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.235501070295674,
      "tokens_seen": 265420800
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009288665997993983,
      "loss": 1.8102,
      "theoretical_loss": 4.235501070295674,
      "tokens_seen": 265420800
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009288465396188565,
      "loss": 1.8941,
      "theoretical_loss": 4.235376774252922,
      "tokens_seen": 265486336
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000928826479438315,
      "loss": 1.9101,
      "theoretical_loss": 4.235252517477956,
      "tokens_seen": 265551872
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009288064192577734,
      "loss": 1.7015,
      "theoretical_loss": 4.235128299948684,
      "tokens_seen": 265617408
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009287863590772317,
      "loss": 1.7076,
      "theoretical_loss": 4.235004121643034,
      "tokens_seen": 265682944
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009287662988966901,
      "loss": 1.8181,
      "theoretical_loss": 4.234879982538949,
      "tokens_seen": 265748480
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009287462387161485,
      "loss": 1.7664,
      "theoretical_loss": 4.23475588261439,
      "tokens_seen": 265814016
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009287261785356068,
      "loss": 1.787,
      "theoretical_loss": 4.234631821847337,
      "tokens_seen": 265879552
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009287061183550653,
      "loss": 1.7446,
      "theoretical_loss": 4.234507800215788,
      "tokens_seen": 265945088
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009286860581745235,
      "loss": 1.8445,
      "theoretical_loss": 4.234383817697755,
      "tokens_seen": 266010624
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000928665997993982,
      "loss": 1.777,
      "theoretical_loss": 4.234259874271275,
      "tokens_seen": 266076160
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009286459378134403,
      "loss": 1.8054,
      "theoretical_loss": 4.234135969914396,
      "tokens_seen": 266141696
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009286258776328987,
      "loss": 1.9136,
      "theoretical_loss": 4.234012104605187,
      "tokens_seen": 266207232
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009286058174523571,
      "loss": 1.7999,
      "theoretical_loss": 4.233888278321734,
      "tokens_seen": 266272768
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009285857572718155,
      "loss": 1.6735,
      "theoretical_loss": 4.23376449104214,
      "tokens_seen": 266338304
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009285656970912738,
      "loss": 1.8224,
      "theoretical_loss": 4.233640742744528,
      "tokens_seen": 266403840
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009285456369107323,
      "loss": 1.9214,
      "theoretical_loss": 4.233517033407035,
      "tokens_seen": 266469376
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009285255767301905,
      "loss": 1.8953,
      "theoretical_loss": 4.23339336300782,
      "tokens_seen": 266534912
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000928505516549649,
      "loss": 1.8393,
      "theoretical_loss": 4.233269731525055,
      "tokens_seen": 266600448
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009284854563691073,
      "loss": 1.8253,
      "theoretical_loss": 4.233146138936933,
      "tokens_seen": 266665984
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009284653961885657,
      "loss": 1.8904,
      "theoretical_loss": 4.233022585221662,
      "tokens_seen": 266731520
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009284453360080241,
      "loss": 1.8195,
      "theoretical_loss": 4.232899070357472,
      "tokens_seen": 266797056
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009284252758274825,
      "loss": 1.7398,
      "theoretical_loss": 4.232775594322605,
      "tokens_seen": 266862592
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009284052156469408,
      "loss": 1.8531,
      "theoretical_loss": 4.232652157095323,
      "tokens_seen": 266928128
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009283851554663992,
      "loss": 1.8012,
      "theoretical_loss": 4.232528758653906,
      "tokens_seen": 266993664
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009283650952858576,
      "loss": 1.7835,
      "theoretical_loss": 4.232405398976653,
      "tokens_seen": 267059200
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000928345035105316,
      "loss": 1.7591,
      "theoretical_loss": 4.232282078041876,
      "tokens_seen": 267124736
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009283249749247743,
      "loss": 1.7493,
      "theoretical_loss": 4.232158795827908,
      "tokens_seen": 267190272
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009283049147442327,
      "loss": 1.8331,
      "theoretical_loss": 4.232035552313097,
      "tokens_seen": 267255808
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000928284854563691,
      "loss": 1.759,
      "theoretical_loss": 4.231912347475812,
      "tokens_seen": 267321344
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009282647943831495,
      "loss": 1.9049,
      "theoretical_loss": 4.231789181294436,
      "tokens_seen": 267386880
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009282447342026078,
      "loss": 1.8436,
      "theoretical_loss": 4.231666053747371,
      "tokens_seen": 267452416
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009282246740220662,
      "loss": 1.9009,
      "theoretical_loss": 4.231542964813034,
      "tokens_seen": 267517952
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009282046138415246,
      "loss": 1.7458,
      "theoretical_loss": 4.231419914469864,
      "tokens_seen": 267583488
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000928184553660983,
      "loss": 1.9132,
      "theoretical_loss": 4.231296902696314,
      "tokens_seen": 267649024
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009281644934804413,
      "loss": 1.738,
      "theoretical_loss": 4.231173929470853,
      "tokens_seen": 267714560
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009281444332998998,
      "loss": 1.8208,
      "theoretical_loss": 4.231050994771971,
      "tokens_seen": 267780096
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000928124373119358,
      "loss": 1.7115,
      "theoretical_loss": 4.230928098578174,
      "tokens_seen": 267845632
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009281043129388165,
      "loss": 1.7666,
      "theoretical_loss": 4.230805240867982,
      "tokens_seen": 267911168
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009280842527582748,
      "loss": 1.8364,
      "theoretical_loss": 4.230682421619936,
      "tokens_seen": 267976704
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009280641925777332,
      "loss": 1.7936,
      "theoretical_loss": 4.2305596408125945,
      "tokens_seen": 268042240
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009280441323971916,
      "loss": 1.7574,
      "theoretical_loss": 4.23043689842453,
      "tokens_seen": 268107776
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009280240722166499,
      "loss": 1.7779,
      "theoretical_loss": 4.230314194434336,
      "tokens_seen": 268173312
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009280040120361083,
      "loss": 1.7898,
      "theoretical_loss": 4.23019152882062,
      "tokens_seen": 268238848
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009279839518555668,
      "loss": 1.8302,
      "theoretical_loss": 4.2300689015620065,
      "tokens_seen": 268304384
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000927963891675025,
      "loss": 1.8138,
      "theoretical_loss": 4.229946312637141,
      "tokens_seen": 268369920
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009279438314944835,
      "loss": 1.7802,
      "theoretical_loss": 4.229823762024681,
      "tokens_seen": 268435456
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009279237713139418,
      "loss": 1.8906,
      "theoretical_loss": 4.2297012497033055,
      "tokens_seen": 268500992
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009279037111334002,
      "loss": 1.8555,
      "theoretical_loss": 4.229578775651707,
      "tokens_seen": 268566528
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009278836509528587,
      "loss": 1.7775,
      "theoretical_loss": 4.229456339848598,
      "tokens_seen": 268632064
    },
    {
      "epoch": 0.08,
      "objective/train/advantage_avg": 0.49237704277038574,
      "objective/train/docs_used": 104173,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5822020769119263,
      "objective/train/original_loss": 1.5822019577026367,
      "objective/train/theoretical_loss": 4.2293339422727065,
      "objective/train/tokens_used": 289157600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24295735359191895,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504728555679321,
      "objective/train/weighted_lm_loss": 1.661965250968933,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.2293339422727065,
      "tokens_seen": 268697600
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009278635907723169,
      "loss": 1.677,
      "theoretical_loss": 4.2293339422727065,
      "tokens_seen": 268697600
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009278435305917754,
      "loss": 1.771,
      "theoretical_loss": 4.229211582902778,
      "tokens_seen": 268763136
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009278234704112338,
      "loss": 1.8272,
      "theoretical_loss": 4.229089261717574,
      "tokens_seen": 268828672
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009278034102306921,
      "loss": 1.8982,
      "theoretical_loss": 4.2289669786958735,
      "tokens_seen": 268894208
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009277833500501505,
      "loss": 1.8134,
      "theoretical_loss": 4.228844733816474,
      "tokens_seen": 268959744
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009277632898696089,
      "loss": 1.8508,
      "theoretical_loss": 4.228722527058188,
      "tokens_seen": 269025280
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009277432296890672,
      "loss": 1.7465,
      "theoretical_loss": 4.228600358399845,
      "tokens_seen": 269090816
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009277231695085257,
      "loss": 1.8031,
      "theoretical_loss": 4.228478227820292,
      "tokens_seen": 269156352
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009277031093279839,
      "loss": 1.6002,
      "theoretical_loss": 4.228356135298394,
      "tokens_seen": 269221888
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009276830491474424,
      "loss": 1.7915,
      "theoretical_loss": 4.2282340808130305,
      "tokens_seen": 269287424
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009276629889669007,
      "loss": 1.7165,
      "theoretical_loss": 4.2281120643431,
      "tokens_seen": 269352960
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009276429287863591,
      "loss": 1.6739,
      "theoretical_loss": 4.227990085867517,
      "tokens_seen": 269418496
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009276228686058175,
      "loss": 1.7462,
      "theoretical_loss": 4.227868145365211,
      "tokens_seen": 269484032
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009276028084252759,
      "loss": 1.7257,
      "theoretical_loss": 4.227746242815132,
      "tokens_seen": 269549568
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009275827482447342,
      "loss": 1.8566,
      "theoretical_loss": 4.227624378196245,
      "tokens_seen": 269615104
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009275626880641927,
      "loss": 1.8052,
      "theoretical_loss": 4.227502551487531,
      "tokens_seen": 269680640
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000927542627883651,
      "loss": 1.804,
      "theoretical_loss": 4.227380762667987,
      "tokens_seen": 269746176
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009275225677031094,
      "loss": 1.8553,
      "theoretical_loss": 4.2272590117166295,
      "tokens_seen": 269811712
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009275025075225677,
      "loss": 1.682,
      "theoretical_loss": 4.22713729861249,
      "tokens_seen": 269877248
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009274824473420261,
      "loss": 1.7293,
      "theoretical_loss": 4.2270156233346174,
      "tokens_seen": 269942784
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009274623871614845,
      "loss": 1.7395,
      "theoretical_loss": 4.226893985862076,
      "tokens_seen": 270008320
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009274423269809429,
      "loss": 1.8387,
      "theoretical_loss": 4.226772386173948,
      "tokens_seen": 270073856
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009274222668004012,
      "loss": 1.8202,
      "theoretical_loss": 4.226650824249333,
      "tokens_seen": 270139392
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009274022066198596,
      "loss": 1.8298,
      "theoretical_loss": 4.226529300067345,
      "tokens_seen": 270204928
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000927382146439318,
      "loss": 1.8401,
      "theoretical_loss": 4.226407813607116,
      "tokens_seen": 270270464
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009273620862587764,
      "loss": 1.831,
      "theoretical_loss": 4.226286364847795,
      "tokens_seen": 270336000
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009273420260782347,
      "loss": 1.7314,
      "theoretical_loss": 4.226164953768545,
      "tokens_seen": 270401536
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009273219658976931,
      "loss": 1.7272,
      "theoretical_loss": 4.226043580348549,
      "tokens_seen": 270467072
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009273019057171514,
      "loss": 1.9179,
      "theoretical_loss": 4.2259222445670055,
      "tokens_seen": 270532608
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009272818455366099,
      "loss": 1.8673,
      "theoretical_loss": 4.225800946403127,
      "tokens_seen": 270598144
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009272617853560682,
      "loss": 1.6766,
      "theoretical_loss": 4.2256796858361465,
      "tokens_seen": 270663680
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009272417251755266,
      "loss": 1.7084,
      "theoretical_loss": 4.225558462845311,
      "tokens_seen": 270729216
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000927221664994985,
      "loss": 1.6541,
      "theoretical_loss": 4.225437277409885,
      "tokens_seen": 270794752
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009272016048144434,
      "loss": 1.7796,
      "theoretical_loss": 4.225316129509147,
      "tokens_seen": 270860288
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009271815446339017,
      "loss": 1.6994,
      "theoretical_loss": 4.2251950191223955,
      "tokens_seen": 270925824
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009271614844533602,
      "loss": 1.8671,
      "theoretical_loss": 4.225073946228944,
      "tokens_seen": 270991360
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009271414242728184,
      "loss": 1.698,
      "theoretical_loss": 4.224952910808122,
      "tokens_seen": 271056896
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009271213640922769,
      "loss": 1.7853,
      "theoretical_loss": 4.224831912839275,
      "tokens_seen": 271122432
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009271013039117352,
      "loss": 1.8293,
      "theoretical_loss": 4.224710952301766,
      "tokens_seen": 271187968
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009270812437311936,
      "loss": 1.8736,
      "theoretical_loss": 4.2245900291749745,
      "tokens_seen": 271253504
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000927061183550652,
      "loss": 1.7211,
      "theoretical_loss": 4.224469143438294,
      "tokens_seen": 271319040
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009270411233701103,
      "loss": 1.9201,
      "theoretical_loss": 4.224348295071138,
      "tokens_seen": 271384576
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009270210631895687,
      "loss": 1.8416,
      "theoretical_loss": 4.224227484052934,
      "tokens_seen": 271450112
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009270010030090272,
      "loss": 1.7762,
      "theoretical_loss": 4.224106710363124,
      "tokens_seen": 271515648
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009269809428284854,
      "loss": 1.9026,
      "theoretical_loss": 4.223985973981171,
      "tokens_seen": 271581184
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009269608826479439,
      "loss": 1.9119,
      "theoretical_loss": 4.223865274886551,
      "tokens_seen": 271646720
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009269408224674021,
      "loss": 1.8214,
      "theoretical_loss": 4.223744613058755,
      "tokens_seen": 271712256
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009269207622868606,
      "loss": 1.7376,
      "theoretical_loss": 4.223623988477295,
      "tokens_seen": 271777792
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000926900702106319,
      "loss": 1.9123,
      "theoretical_loss": 4.223503401121693,
      "tokens_seen": 271843328
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009268806419257773,
      "loss": 1.741,
      "theoretical_loss": 4.223382850971494,
      "tokens_seen": 271908864
    },
    {
      "epoch": 0.08,
      "objective/train/advantage_avg": 0.4953818619251251,
      "objective/train/docs_used": 105443,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6509478092193604,
      "objective/train/original_loss": 1.6509475708007812,
      "objective/train/theoretical_loss": 4.223262338006254,
      "objective/train/tokens_used": 292434400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24556824564933777,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507866144180298,
      "objective/train/weighted_lm_loss": 1.7348129749298096,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9753099083900452,
      "theoretical_loss": 4.223262338006254,
      "tokens_seen": 271974400
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009268605817452357,
      "loss": 1.7963,
      "theoretical_loss": 4.223262338006254,
      "tokens_seen": 271974400
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009268405215646942,
      "loss": 1.6685,
      "theoretical_loss": 4.223141862205546,
      "tokens_seen": 272039936
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009268204613841524,
      "loss": 1.7575,
      "theoretical_loss": 4.223021423548962,
      "tokens_seen": 272105472
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009268004012036109,
      "loss": 1.7426,
      "theoretical_loss": 4.2229010220161065,
      "tokens_seen": 272171008
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009267803410230691,
      "loss": 1.7231,
      "theoretical_loss": 4.222780657586601,
      "tokens_seen": 272236544
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009267602808425276,
      "loss": 1.7428,
      "theoretical_loss": 4.222660330240086,
      "tokens_seen": 272302080
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000926740220661986,
      "loss": 1.9007,
      "theoretical_loss": 4.222540039956215,
      "tokens_seen": 272367616
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009267201604814443,
      "loss": 1.795,
      "theoretical_loss": 4.222419786714657,
      "tokens_seen": 272433152
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009267001003009027,
      "loss": 1.768,
      "theoretical_loss": 4.2222995704951005,
      "tokens_seen": 272498688
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009266800401203611,
      "loss": 1.7197,
      "theoretical_loss": 4.222179391277247,
      "tokens_seen": 272564224
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009266599799398194,
      "loss": 1.76,
      "theoretical_loss": 4.222059249040814,
      "tokens_seen": 272629760
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009266399197592779,
      "loss": 1.7811,
      "theoretical_loss": 4.221939143765538,
      "tokens_seen": 272695296
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009266198595787362,
      "loss": 1.842,
      "theoretical_loss": 4.221819075431168,
      "tokens_seen": 272760832
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009265997993981946,
      "loss": 1.7936,
      "theoretical_loss": 4.221699044017471,
      "tokens_seen": 272826368
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009265797392176529,
      "loss": 1.904,
      "theoretical_loss": 4.2215790495042285,
      "tokens_seen": 272891904
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009265596790371113,
      "loss": 1.8888,
      "theoretical_loss": 4.22145909187124,
      "tokens_seen": 272957440
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009265396188565697,
      "loss": 1.8071,
      "theoretical_loss": 4.221339171098318,
      "tokens_seen": 273022976
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009265195586760281,
      "loss": 1.6943,
      "theoretical_loss": 4.221219287165294,
      "tokens_seen": 273088512
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009264994984954864,
      "loss": 1.9064,
      "theoretical_loss": 4.221099440052014,
      "tokens_seen": 273154048
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009264794383149449,
      "loss": 1.8892,
      "theoretical_loss": 4.220979629738339,
      "tokens_seen": 273219584
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009264593781344032,
      "loss": 1.7186,
      "theoretical_loss": 4.220859856204147,
      "tokens_seen": 273285120
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009264393179538616,
      "loss": 1.7486,
      "theoretical_loss": 4.22074011942933,
      "tokens_seen": 273350656
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00092641925777332,
      "loss": 1.7085,
      "theoretical_loss": 4.220620419393799,
      "tokens_seen": 273416192
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009263991975927784,
      "loss": 1.7723,
      "theoretical_loss": 4.220500756077479,
      "tokens_seen": 273481728
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009263791374122368,
      "loss": 1.8846,
      "theoretical_loss": 4.22038112946031,
      "tokens_seen": 273547264
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009263590772316951,
      "loss": 1.875,
      "theoretical_loss": 4.220261539522249,
      "tokens_seen": 273612800
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009263390170511535,
      "loss": 1.7537,
      "theoretical_loss": 4.220141986243268,
      "tokens_seen": 273678336
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009263189568706118,
      "loss": 1.7963,
      "theoretical_loss": 4.220022469603356,
      "tokens_seen": 273743872
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009262988966900703,
      "loss": 1.7591,
      "theoretical_loss": 4.219902989582516,
      "tokens_seen": 273809408
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009262788365095286,
      "loss": 1.8612,
      "theoretical_loss": 4.219783546160768,
      "tokens_seen": 273874944
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000926258776328987,
      "loss": 1.8849,
      "theoretical_loss": 4.219664139318145,
      "tokens_seen": 273940480
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009262387161484454,
      "loss": 1.7991,
      "theoretical_loss": 4.219544769034701,
      "tokens_seen": 274006016
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009262186559679038,
      "loss": 1.7852,
      "theoretical_loss": 4.2194254352905,
      "tokens_seen": 274071552
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009261985957873621,
      "loss": 1.7718,
      "theoretical_loss": 4.219306138065626,
      "tokens_seen": 274137088
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009261785356068206,
      "loss": 1.8751,
      "theoretical_loss": 4.219186877340174,
      "tokens_seen": 274202624
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009261584754262788,
      "loss": 1.7946,
      "theoretical_loss": 4.219067653094261,
      "tokens_seen": 274268160
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009261384152457373,
      "loss": 1.6745,
      "theoretical_loss": 4.218948465308013,
      "tokens_seen": 274333696
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009261183550651956,
      "loss": 1.7044,
      "theoretical_loss": 4.218829313961575,
      "tokens_seen": 274399232
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000926098294884654,
      "loss": 1.8374,
      "theoretical_loss": 4.218710199035108,
      "tokens_seen": 274464768
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009260782347041124,
      "loss": 1.6713,
      "theoretical_loss": 4.218591120508787,
      "tokens_seen": 274530304
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009260581745235707,
      "loss": 1.7884,
      "theoretical_loss": 4.218472078362803,
      "tokens_seen": 274595840
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009260381143430291,
      "loss": 1.7148,
      "theoretical_loss": 4.2183530725773615,
      "tokens_seen": 274661376
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009260180541624876,
      "loss": 1.8029,
      "theoretical_loss": 4.218234103132686,
      "tokens_seen": 274726912
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009259979939819458,
      "loss": 1.9025,
      "theoretical_loss": 4.218115170009014,
      "tokens_seen": 274792448
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009259779338014043,
      "loss": 1.7295,
      "theoretical_loss": 4.217996273186598,
      "tokens_seen": 274857984
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009259578736208625,
      "loss": 1.7892,
      "theoretical_loss": 4.217877412645706,
      "tokens_seen": 274923520
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000925937813440321,
      "loss": 1.8208,
      "theoretical_loss": 4.217758588366623,
      "tokens_seen": 274989056
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009259177532597794,
      "loss": 1.8495,
      "theoretical_loss": 4.217639800329647,
      "tokens_seen": 275054592
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009258976930792377,
      "loss": 1.6701,
      "theoretical_loss": 4.217521048515094,
      "tokens_seen": 275120128
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009258776328986961,
      "loss": 1.7104,
      "theoretical_loss": 4.217402332903292,
      "tokens_seen": 275185664
    },
    {
      "epoch": 0.08,
      "objective/train/advantage_avg": 0.4921969771385193,
      "objective/train/docs_used": 106703,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6392239332199097,
      "objective/train/original_loss": 1.6392236948013306,
      "objective/train/theoretical_loss": 4.217283653474588,
      "objective/train/tokens_used": 295711200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.243023082613945,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050455093383789,
      "objective/train/weighted_lm_loss": 1.7219140529632568,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.217283653474588,
      "tokens_seen": 275251200
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009258575727181546,
      "loss": 1.7057,
      "theoretical_loss": 4.217283653474588,
      "tokens_seen": 275251200
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009258375125376128,
      "loss": 1.752,
      "theoretical_loss": 4.217165010209343,
      "tokens_seen": 275316736
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009258174523570713,
      "loss": 1.8469,
      "theoretical_loss": 4.217046403087933,
      "tokens_seen": 275382272
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009257973921765295,
      "loss": 1.7028,
      "theoretical_loss": 4.2169278320907475,
      "tokens_seen": 275447808
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000925777331995988,
      "loss": 1.86,
      "theoretical_loss": 4.216809297198195,
      "tokens_seen": 275513344
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009257572718154464,
      "loss": 1.8148,
      "theoretical_loss": 4.216690798390697,
      "tokens_seen": 275578880
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009257372116349047,
      "loss": 1.7633,
      "theoretical_loss": 4.216572335648692,
      "tokens_seen": 275644416
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009257171514543631,
      "loss": 1.8091,
      "theoretical_loss": 4.21645390895263,
      "tokens_seen": 275709952
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009256970912738215,
      "loss": 1.8062,
      "theoretical_loss": 4.21633551828298,
      "tokens_seen": 275775488
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009256770310932798,
      "loss": 1.6688,
      "theoretical_loss": 4.216217163620225,
      "tokens_seen": 275841024
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009256569709127383,
      "loss": 1.8757,
      "theoretical_loss": 4.216098844944863,
      "tokens_seen": 275906560
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009256369107321966,
      "loss": 1.8344,
      "theoretical_loss": 4.215980562237409,
      "tokens_seen": 275972096
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000925616850551655,
      "loss": 1.904,
      "theoretical_loss": 4.215862315478388,
      "tokens_seen": 276037632
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009255967903711133,
      "loss": 1.8534,
      "theoretical_loss": 4.215744104648347,
      "tokens_seen": 276103168
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009255767301905717,
      "loss": 1.7778,
      "theoretical_loss": 4.2156259297278424,
      "tokens_seen": 276168704
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009255566700100301,
      "loss": 1.9598,
      "theoretical_loss": 4.215507790697449,
      "tokens_seen": 276234240
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009255366098294885,
      "loss": 1.7881,
      "theoretical_loss": 4.2153896875377574,
      "tokens_seen": 276299776
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009255165496489468,
      "loss": 1.8789,
      "theoretical_loss": 4.21527162022937,
      "tokens_seen": 276365312
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009254964894684053,
      "loss": 1.8463,
      "theoretical_loss": 4.215153588752907,
      "tokens_seen": 276430848
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009254764292878636,
      "loss": 1.9038,
      "theoretical_loss": 4.215035593089002,
      "tokens_seen": 276496384
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000925456369107322,
      "loss": 1.6358,
      "theoretical_loss": 4.214917633218304,
      "tokens_seen": 276561920
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009254363089267803,
      "loss": 1.8301,
      "theoretical_loss": 4.214799709121479,
      "tokens_seen": 276627456
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009254162487462388,
      "loss": 1.8829,
      "theoretical_loss": 4.214681820779205,
      "tokens_seen": 276692992
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009253961885656971,
      "loss": 1.7861,
      "theoretical_loss": 4.214563968172178,
      "tokens_seen": 276758528
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009253761283851555,
      "loss": 1.7876,
      "theoretical_loss": 4.214446151281106,
      "tokens_seen": 276824064
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009253560682046138,
      "loss": 1.8599,
      "theoretical_loss": 4.214328370086713,
      "tokens_seen": 276889600
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009253360080240722,
      "loss": 1.6949,
      "theoretical_loss": 4.21421062456974,
      "tokens_seen": 276955136
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009253159478435306,
      "loss": 1.7641,
      "theoretical_loss": 4.214092914710941,
      "tokens_seen": 277020672
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000925295887662989,
      "loss": 1.7158,
      "theoretical_loss": 4.213975240491084,
      "tokens_seen": 277086208
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009252758274824473,
      "loss": 1.8548,
      "theoretical_loss": 4.213857601890955,
      "tokens_seen": 277151744
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009252557673019058,
      "loss": 1.8456,
      "theoretical_loss": 4.213739998891352,
      "tokens_seen": 277217280
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000925235707121364,
      "loss": 1.7495,
      "theoretical_loss": 4.213622431473088,
      "tokens_seen": 277282816
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009252156469408225,
      "loss": 1.7643,
      "theoretical_loss": 4.213504899616995,
      "tokens_seen": 277348352
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009251955867602808,
      "loss": 1.7701,
      "theoretical_loss": 4.213387403303912,
      "tokens_seen": 277413888
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009251755265797392,
      "loss": 1.7438,
      "theoretical_loss": 4.213269942514701,
      "tokens_seen": 277479424
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009251554663991976,
      "loss": 1.8916,
      "theoretical_loss": 4.2131525172302355,
      "tokens_seen": 277544960
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000925135406218656,
      "loss": 1.8483,
      "theoretical_loss": 4.213035127431402,
      "tokens_seen": 277610496
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009251153460381143,
      "loss": 1.7789,
      "theoretical_loss": 4.212917773099105,
      "tokens_seen": 277676032
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009250952858575728,
      "loss": 1.7993,
      "theoretical_loss": 4.2128004542142605,
      "tokens_seen": 277741568
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000925075225677031,
      "loss": 1.8079,
      "theoretical_loss": 4.212683170757802,
      "tokens_seen": 277807104
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009250551654964895,
      "loss": 1.8363,
      "theoretical_loss": 4.212565922710677,
      "tokens_seen": 277872640
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009250351053159479,
      "loss": 1.8011,
      "theoretical_loss": 4.212448710053847,
      "tokens_seen": 277938176
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009250150451354062,
      "loss": 1.8875,
      "theoretical_loss": 4.212331532768289,
      "tokens_seen": 278003712
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009249949849548646,
      "loss": 1.7518,
      "theoretical_loss": 4.212214390834994,
      "tokens_seen": 278069248
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009249749247743229,
      "loss": 1.7445,
      "theoretical_loss": 4.21209728423497,
      "tokens_seen": 278134784
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009249548645937813,
      "loss": 1.9204,
      "theoretical_loss": 4.211980212949234,
      "tokens_seen": 278200320
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009249348044132398,
      "loss": 1.7544,
      "theoretical_loss": 4.2118631769588255,
      "tokens_seen": 278265856
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009249147442326981,
      "loss": 1.8568,
      "theoretical_loss": 4.211746176244792,
      "tokens_seen": 278331392
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009248946840521565,
      "loss": 1.8516,
      "theoretical_loss": 4.2116292107882,
      "tokens_seen": 278396928
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000924874623871615,
      "loss": 1.7939,
      "theoretical_loss": 4.211512280570127,
      "tokens_seen": 278462464
    },
    {
      "epoch": 0.08,
      "objective/train/advantage_avg": 0.4929666519165039,
      "objective/train/docs_used": 107929,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.715563178062439,
      "objective/train/original_loss": 1.7155632972717285,
      "objective/train/theoretical_loss": 4.211395385571668,
      "objective/train/tokens_used": 298988000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2441163808107376,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505377054214478,
      "objective/train/weighted_lm_loss": 1.8023045063018799,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.211395385571668,
      "tokens_seen": 278528000
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009248545636910732,
      "loss": 1.8125,
      "theoretical_loss": 4.211395385571668,
      "tokens_seen": 278528000
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009248345035105317,
      "loss": 1.8219,
      "theoretical_loss": 4.211278525773933,
      "tokens_seen": 278593536
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00092481444332999,
      "loss": 1.838,
      "theoretical_loss": 4.211161701158042,
      "tokens_seen": 278659072
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009247943831494484,
      "loss": 1.8463,
      "theoretical_loss": 4.211044911705135,
      "tokens_seen": 278724608
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009247743229689068,
      "loss": 1.8635,
      "theoretical_loss": 4.2109281573963635,
      "tokens_seen": 278790144
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009247542627883651,
      "loss": 1.8983,
      "theoretical_loss": 4.210811438212894,
      "tokens_seen": 278855680
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009247342026078235,
      "loss": 1.739,
      "theoretical_loss": 4.2106947541359085,
      "tokens_seen": 278921216
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009247141424272819,
      "loss": 1.9169,
      "theoretical_loss": 4.2105781051466025,
      "tokens_seen": 278986752
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009246940822467402,
      "loss": 1.7906,
      "theoretical_loss": 4.2104614912261855,
      "tokens_seen": 279052288
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009246740220661987,
      "loss": 1.6945,
      "theoretical_loss": 4.210344912355883,
      "tokens_seen": 279117824
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000924653961885657,
      "loss": 1.7868,
      "theoretical_loss": 4.210228368516935,
      "tokens_seen": 279183360
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009246339017051154,
      "loss": 1.8477,
      "theoretical_loss": 4.210111859690594,
      "tokens_seen": 279248896
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009246138415245737,
      "loss": 1.9463,
      "theoretical_loss": 4.209995385858129,
      "tokens_seen": 279314432
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009245937813440321,
      "loss": 1.9702,
      "theoretical_loss": 4.20987894700082,
      "tokens_seen": 279379968
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009245737211634905,
      "loss": 1.8656,
      "theoretical_loss": 4.209762543099966,
      "tokens_seen": 279445504
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009245536609829489,
      "loss": 1.821,
      "theoretical_loss": 4.209646174136879,
      "tokens_seen": 279511040
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009245336008024072,
      "loss": 1.8142,
      "theoretical_loss": 4.209529840092884,
      "tokens_seen": 279576576
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009245135406218657,
      "loss": 1.911,
      "theoretical_loss": 4.209413540949319,
      "tokens_seen": 279642112
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000924493480441324,
      "loss": 1.7045,
      "theoretical_loss": 4.209297276687541,
      "tokens_seen": 279707648
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009244734202607824,
      "loss": 1.8239,
      "theoretical_loss": 4.209181047288917,
      "tokens_seen": 279773184
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009244533600802407,
      "loss": 1.7827,
      "theoretical_loss": 4.209064852734831,
      "tokens_seen": 279838720
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009244332998996992,
      "loss": 1.962,
      "theoretical_loss": 4.20894869300668,
      "tokens_seen": 279904256
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009244132397191575,
      "loss": 1.8998,
      "theoretical_loss": 4.2088325680858745,
      "tokens_seen": 279969792
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009243931795386159,
      "loss": 1.7249,
      "theoretical_loss": 4.208716477953843,
      "tokens_seen": 280035328
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009243731193580742,
      "loss": 1.8494,
      "theoretical_loss": 4.2086004225920215,
      "tokens_seen": 280100864
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009243530591775326,
      "loss": 1.8086,
      "theoretical_loss": 4.208484401981868,
      "tokens_seen": 280166400
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000924332998996991,
      "loss": 1.7255,
      "theoretical_loss": 4.208368416104849,
      "tokens_seen": 280231936
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009243129388164494,
      "loss": 1.7742,
      "theoretical_loss": 4.208252464942449,
      "tokens_seen": 280297472
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009242928786359077,
      "loss": 1.8617,
      "theoretical_loss": 4.208136548476163,
      "tokens_seen": 280363008
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009242728184553662,
      "loss": 1.8351,
      "theoretical_loss": 4.208020666687503,
      "tokens_seen": 280428544
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0009242527582748244,
      "loss": 1.9316,
      "theoretical_loss": 4.207904819557995,
      "tokens_seen": 280494080
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009242326980942829,
      "loss": 1.7367,
      "theoretical_loss": 4.207789007069177,
      "tokens_seen": 280559616
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009242126379137412,
      "loss": 1.8257,
      "theoretical_loss": 4.207673229202604,
      "tokens_seen": 280625152
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009241925777331996,
      "loss": 1.7784,
      "theoretical_loss": 4.2075574859398435,
      "tokens_seen": 280690688
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000924172517552658,
      "loss": 1.8542,
      "theoretical_loss": 4.207441777262477,
      "tokens_seen": 280756224
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009241524573721164,
      "loss": 1.703,
      "theoretical_loss": 4.2073261031521,
      "tokens_seen": 280821760
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009241323971915747,
      "loss": 1.7389,
      "theoretical_loss": 4.207210463590323,
      "tokens_seen": 280887296
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009241123370110332,
      "loss": 1.9065,
      "theoretical_loss": 4.207094858558771,
      "tokens_seen": 280952832
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009240922768304914,
      "loss": 1.774,
      "theoretical_loss": 4.206979288039081,
      "tokens_seen": 281018368
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009240722166499499,
      "loss": 1.7895,
      "theoretical_loss": 4.2068637520129055,
      "tokens_seen": 281083904
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009240521564694083,
      "loss": 1.7552,
      "theoretical_loss": 4.206748250461911,
      "tokens_seen": 281149440
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009240320962888666,
      "loss": 1.771,
      "theoretical_loss": 4.206632783367777,
      "tokens_seen": 281214976
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000924012036108325,
      "loss": 1.7158,
      "theoretical_loss": 4.206517350712199,
      "tokens_seen": 281280512
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009239919759277833,
      "loss": 1.9191,
      "theoretical_loss": 4.2064019524768845,
      "tokens_seen": 281346048
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009239719157472417,
      "loss": 1.8433,
      "theoretical_loss": 4.206286588643556,
      "tokens_seen": 281411584
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009239518555667002,
      "loss": 1.7216,
      "theoretical_loss": 4.206171259193948,
      "tokens_seen": 281477120
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009239317953861584,
      "loss": 1.7115,
      "theoretical_loss": 4.206055964109813,
      "tokens_seen": 281542656
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009239117352056169,
      "loss": 1.705,
      "theoretical_loss": 4.205940703372915,
      "tokens_seen": 281608192
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009238916750250753,
      "loss": 1.7713,
      "theoretical_loss": 4.20582547696503,
      "tokens_seen": 281673728
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009238716148445336,
      "loss": 1.8652,
      "theoretical_loss": 4.205710284867951,
      "tokens_seen": 281739264
    },
    {
      "epoch": 0.09,
      "objective/train/advantage_avg": 0.49427810311317444,
      "objective/train/docs_used": 109222,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6125668287277222,
      "objective/train/original_loss": 1.6125669479370117,
      "objective/train/theoretical_loss": 4.205595127063485,
      "objective/train/tokens_used": 302264800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24455593526363373,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050671100616455,
      "objective/train/weighted_lm_loss": 1.694186806678772,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.205595127063485,
      "tokens_seen": 281804800
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000923851554663992,
      "loss": 1.7848,
      "theoretical_loss": 4.205595127063485,
      "tokens_seen": 281804800
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009238314944834503,
      "loss": 1.7997,
      "theoretical_loss": 4.20548000353345,
      "tokens_seen": 281870336
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009238114343029087,
      "loss": 1.7206,
      "theoretical_loss": 4.205364914259679,
      "tokens_seen": 281935872
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009237913741223672,
      "loss": 1.7657,
      "theoretical_loss": 4.205249859224021,
      "tokens_seen": 282001408
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009237713139418254,
      "loss": 1.7049,
      "theoretical_loss": 4.205134838408337,
      "tokens_seen": 282066944
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009237512537612839,
      "loss": 1.7584,
      "theoretical_loss": 4.2050198517945,
      "tokens_seen": 282132480
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009237311935807422,
      "loss": 1.7654,
      "theoretical_loss": 4.204904899364402,
      "tokens_seen": 282198016
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009237111334002006,
      "loss": 1.7786,
      "theoretical_loss": 4.2047899810999425,
      "tokens_seen": 282263552
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000923691073219659,
      "loss": 1.8459,
      "theoretical_loss": 4.20467509698304,
      "tokens_seen": 282329088
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009236710130391174,
      "loss": 1.7859,
      "theoretical_loss": 4.204560246995625,
      "tokens_seen": 282394624
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009236509528585757,
      "loss": 1.7339,
      "theoretical_loss": 4.204445431119639,
      "tokens_seen": 282460160
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009236308926780341,
      "loss": 1.7904,
      "theoretical_loss": 4.204330649337042,
      "tokens_seen": 282525696
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009236108324974924,
      "loss": 1.7929,
      "theoretical_loss": 4.204215901629803,
      "tokens_seen": 282591232
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009235907723169509,
      "loss": 1.8564,
      "theoretical_loss": 4.2041011879799095,
      "tokens_seen": 282656768
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009235707121364092,
      "loss": 1.7214,
      "theoretical_loss": 4.2039865083693595,
      "tokens_seen": 282722304
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009235506519558676,
      "loss": 1.8249,
      "theoretical_loss": 4.203871862780165,
      "tokens_seen": 282787840
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000923530591775326,
      "loss": 1.7751,
      "theoretical_loss": 4.203757251194353,
      "tokens_seen": 282853376
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009235105315947844,
      "loss": 1.7531,
      "theoretical_loss": 4.203642673593962,
      "tokens_seen": 282918912
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009234904714142427,
      "loss": 1.7935,
      "theoretical_loss": 4.203528129961047,
      "tokens_seen": 282984448
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009234704112337011,
      "loss": 1.7967,
      "theoretical_loss": 4.203413620277673,
      "tokens_seen": 283049984
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009234503510531594,
      "loss": 1.7649,
      "theoretical_loss": 4.203299144525923,
      "tokens_seen": 283115520
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009234302908726179,
      "loss": 1.8882,
      "theoretical_loss": 4.20318470268789,
      "tokens_seen": 283181056
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009234102306920763,
      "loss": 1.7575,
      "theoretical_loss": 4.203070294745682,
      "tokens_seen": 283246592
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009233901705115346,
      "loss": 1.7621,
      "theoretical_loss": 4.202955920681421,
      "tokens_seen": 283312128
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000923370110330993,
      "loss": 1.7461,
      "theoretical_loss": 4.202841580477241,
      "tokens_seen": 283377664
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009233500501504514,
      "loss": 1.9422,
      "theoretical_loss": 4.202727274115292,
      "tokens_seen": 283443200
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009233299899699098,
      "loss": 1.798,
      "theoretical_loss": 4.202613001577735,
      "tokens_seen": 283508736
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009233099297893681,
      "loss": 1.8495,
      "theoretical_loss": 4.202498762846745,
      "tokens_seen": 283574272
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009232898696088266,
      "loss": 1.7852,
      "theoretical_loss": 4.202384557904513,
      "tokens_seen": 283639808
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009232698094282848,
      "loss": 1.7845,
      "theoretical_loss": 4.202270386733241,
      "tokens_seen": 283705344
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009232497492477433,
      "loss": 1.7888,
      "theoretical_loss": 4.202156249315145,
      "tokens_seen": 283770880
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009232296890672016,
      "loss": 1.8123,
      "theoretical_loss": 4.202042145632453,
      "tokens_seen": 283836416
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00092320962888666,
      "loss": 1.7403,
      "theoretical_loss": 4.201928075667411,
      "tokens_seen": 283901952
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009231895687061184,
      "loss": 1.735,
      "theoretical_loss": 4.201814039402273,
      "tokens_seen": 283967488
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009231695085255768,
      "loss": 1.7603,
      "theoretical_loss": 4.201700036819311,
      "tokens_seen": 284033024
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009231494483450351,
      "loss": 1.9366,
      "theoretical_loss": 4.201586067900806,
      "tokens_seen": 284098560
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009231293881644936,
      "loss": 1.6707,
      "theoretical_loss": 4.201472132629057,
      "tokens_seen": 284164096
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009231093279839518,
      "loss": 1.9417,
      "theoretical_loss": 4.201358230986372,
      "tokens_seen": 284229632
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009230892678034103,
      "loss": 1.82,
      "theoretical_loss": 4.201244362955077,
      "tokens_seen": 284295168
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009230692076228687,
      "loss": 1.8958,
      "theoretical_loss": 4.201130528517506,
      "tokens_seen": 284360704
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000923049147442327,
      "loss": 1.8833,
      "theoretical_loss": 4.201016727656012,
      "tokens_seen": 284426240
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009230290872617854,
      "loss": 1.784,
      "theoretical_loss": 4.200902960352956,
      "tokens_seen": 284491776
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009230090270812437,
      "loss": 1.8496,
      "theoretical_loss": 4.200789226590717,
      "tokens_seen": 284557312
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009229889669007021,
      "loss": 1.7737,
      "theoretical_loss": 4.200675526351684,
      "tokens_seen": 284622848
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009229689067201606,
      "loss": 1.893,
      "theoretical_loss": 4.2005618596182615,
      "tokens_seen": 284688384
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009229488465396188,
      "loss": 1.8352,
      "theoretical_loss": 4.200448226372865,
      "tokens_seen": 284753920
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009229287863590773,
      "loss": 1.8389,
      "theoretical_loss": 4.200334626597925,
      "tokens_seen": 284819456
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009229087261785357,
      "loss": 1.8186,
      "theoretical_loss": 4.200221060275885,
      "tokens_seen": 284884992
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000922888665997994,
      "loss": 1.8161,
      "theoretical_loss": 4.200107527389202,
      "tokens_seen": 284950528
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009228686058174524,
      "loss": 1.7495,
      "theoretical_loss": 4.1999940279203445,
      "tokens_seen": 285016064
    },
    {
      "epoch": 0.09,
      "objective/train/advantage_avg": 0.4956050515174866,
      "objective/train/docs_used": 110428,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.646780252456665,
      "objective/train/original_loss": 1.646780252456665,
      "objective/train/theoretical_loss": 4.1998805618517965,
      "objective/train/tokens_used": 305541600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24578233063220978,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508100986480713,
      "objective/train/weighted_lm_loss": 1.730436086654663,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.1998805618517965,
      "tokens_seen": 285081600
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009228485456369107,
      "loss": 1.7924,
      "theoretical_loss": 4.1998805618517965,
      "tokens_seen": 285081600
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009228284854563691,
      "loss": 1.7654,
      "theoretical_loss": 4.199767129166054,
      "tokens_seen": 285147136
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009228084252758276,
      "loss": 1.871,
      "theoretical_loss": 4.199653729845626,
      "tokens_seen": 285212672
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009227883650952858,
      "loss": 1.8636,
      "theoretical_loss": 4.199540363873036,
      "tokens_seen": 285278208
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009227683049147443,
      "loss": 1.7923,
      "theoretical_loss": 4.19942703123082,
      "tokens_seen": 285343744
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009227482447342026,
      "loss": 1.7764,
      "theoretical_loss": 4.199313731901524,
      "tokens_seen": 285409280
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000922728184553661,
      "loss": 1.791,
      "theoretical_loss": 4.199200465867714,
      "tokens_seen": 285474816
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009227081243731194,
      "loss": 1.8081,
      "theoretical_loss": 4.199087233111962,
      "tokens_seen": 285540352
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009226880641925778,
      "loss": 1.6973,
      "theoretical_loss": 4.198974033616859,
      "tokens_seen": 285605888
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009226680040120361,
      "loss": 1.8117,
      "theoretical_loss": 4.1988608673650045,
      "tokens_seen": 285671424
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009226479438314945,
      "loss": 1.8261,
      "theoretical_loss": 4.198747734339013,
      "tokens_seen": 285736960
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009226278836509528,
      "loss": 1.8351,
      "theoretical_loss": 4.198634634521513,
      "tokens_seen": 285802496
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009226078234704113,
      "loss": 1.7645,
      "theoretical_loss": 4.198521567895144,
      "tokens_seen": 285868032
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009225877632898696,
      "loss": 1.7425,
      "theoretical_loss": 4.198408534442561,
      "tokens_seen": 285933568
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000922567703109328,
      "loss": 1.7141,
      "theoretical_loss": 4.198295534146429,
      "tokens_seen": 285999104
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009225476429287864,
      "loss": 1.7746,
      "theoretical_loss": 4.19818256698943,
      "tokens_seen": 286064640
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009225275827482448,
      "loss": 1.7737,
      "theoretical_loss": 4.198069632954255,
      "tokens_seen": 286130176
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009225075225677031,
      "loss": 1.7182,
      "theoretical_loss": 4.19795673202361,
      "tokens_seen": 286195712
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009224874623871615,
      "loss": 1.8036,
      "theoretical_loss": 4.197843864180214,
      "tokens_seen": 286261248
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009224674022066198,
      "loss": 1.8232,
      "theoretical_loss": 4.1977310294068,
      "tokens_seen": 286326784
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009224473420260783,
      "loss": 1.7982,
      "theoretical_loss": 4.19761822768611,
      "tokens_seen": 286392320
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009224272818455366,
      "loss": 1.735,
      "theoretical_loss": 4.1975054590009035,
      "tokens_seen": 286457856
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000922407221664995,
      "loss": 1.7124,
      "theoretical_loss": 4.197392723333951,
      "tokens_seen": 286523392
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009223871614844533,
      "loss": 1.8675,
      "theoretical_loss": 4.1972800206680345,
      "tokens_seen": 286588928
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009223671013039118,
      "loss": 1.7668,
      "theoretical_loss": 4.197167350985952,
      "tokens_seen": 286654464
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009223470411233701,
      "loss": 1.812,
      "theoretical_loss": 4.197054714270513,
      "tokens_seen": 286720000
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009223269809428285,
      "loss": 1.8564,
      "theoretical_loss": 4.196942110504538,
      "tokens_seen": 286785536
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009223069207622869,
      "loss": 1.7916,
      "theoretical_loss": 4.196829539670864,
      "tokens_seen": 286851072
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009222868605817452,
      "loss": 1.7991,
      "theoretical_loss": 4.1967170017523365,
      "tokens_seen": 286916608
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009222668004012036,
      "loss": 1.7301,
      "theoretical_loss": 4.196604496731818,
      "tokens_seen": 286982144
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000922246740220662,
      "loss": 1.8278,
      "theoretical_loss": 4.196492024592183,
      "tokens_seen": 287047680
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009222266800401203,
      "loss": 1.891,
      "theoretical_loss": 4.196379585316315,
      "tokens_seen": 287113216
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009222066198595788,
      "loss": 1.7694,
      "theoretical_loss": 4.1962671788871155,
      "tokens_seen": 287178752
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009221865596790371,
      "loss": 1.7081,
      "theoretical_loss": 4.196154805287497,
      "tokens_seen": 287244288
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009221664994984955,
      "loss": 1.7177,
      "theoretical_loss": 4.196042464500382,
      "tokens_seen": 287309824
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009221464393179539,
      "loss": 1.7366,
      "theoretical_loss": 4.195930156508711,
      "tokens_seen": 287375360
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009221263791374122,
      "loss": 1.7228,
      "theoretical_loss": 4.195817881295431,
      "tokens_seen": 287440896
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009221063189568706,
      "loss": 1.8829,
      "theoretical_loss": 4.195705638843508,
      "tokens_seen": 287506432
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000922086258776329,
      "loss": 1.7452,
      "theoretical_loss": 4.195593429135916,
      "tokens_seen": 287571968
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009220661985957873,
      "loss": 2.0069,
      "theoretical_loss": 4.195481252155645,
      "tokens_seen": 287637504
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009220461384152458,
      "loss": 1.9142,
      "theoretical_loss": 4.1953691078856945,
      "tokens_seen": 287703040
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000922026078234704,
      "loss": 1.8221,
      "theoretical_loss": 4.19525699630908,
      "tokens_seen": 287768576
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009220060180541625,
      "loss": 1.823,
      "theoretical_loss": 4.195144917408828,
      "tokens_seen": 287834112
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009219859578736209,
      "loss": 1.7773,
      "theoretical_loss": 4.195032871167977,
      "tokens_seen": 287899648
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009219658976930792,
      "loss": 1.706,
      "theoretical_loss": 4.1949208575695796,
      "tokens_seen": 287965184
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009219458375125376,
      "loss": 1.7919,
      "theoretical_loss": 4.1948088765967,
      "tokens_seen": 288030720
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009219257773319961,
      "loss": 1.7851,
      "theoretical_loss": 4.194696928232417,
      "tokens_seen": 288096256
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009219057171514544,
      "loss": 1.7673,
      "theoretical_loss": 4.194585012459818,
      "tokens_seen": 288161792
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009218856569709128,
      "loss": 1.7555,
      "theoretical_loss": 4.194473129262008,
      "tokens_seen": 288227328
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009218655967903711,
      "loss": 1.7409,
      "theoretical_loss": 4.194361278622099,
      "tokens_seen": 288292864
    },
    {
      "epoch": 0.09,
      "objective/train/advantage_avg": 0.4943329095840454,
      "objective/train/docs_used": 111591,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6546730995178223,
      "objective/train/original_loss": 1.6546730995178223,
      "objective/train/theoretical_loss": 4.194249460523222,
      "objective/train/tokens_used": 308818400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2445804327726364,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050676703453064,
      "objective/train/weighted_lm_loss": 1.7384644746780396,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.194249460523222,
      "tokens_seen": 288358400
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009218455366098295,
      "loss": 1.713,
      "theoretical_loss": 4.194249460523222,
      "tokens_seen": 288358400
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000921825476429288,
      "loss": 1.8013,
      "theoretical_loss": 4.194137674948516,
      "tokens_seen": 288423936
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009218054162487462,
      "loss": 1.8124,
      "theoretical_loss": 4.194025921881134,
      "tokens_seen": 288489472
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009217853560682047,
      "loss": 1.7621,
      "theoretical_loss": 4.19391420130424,
      "tokens_seen": 288555008
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000921765295887663,
      "loss": 1.8619,
      "theoretical_loss": 4.193802513201015,
      "tokens_seen": 288620544
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009217452357071214,
      "loss": 1.8636,
      "theoretical_loss": 4.193690857554646,
      "tokens_seen": 288686080
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009217251755265798,
      "loss": 1.9187,
      "theoretical_loss": 4.193579234348338,
      "tokens_seen": 288751616
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009217051153460382,
      "loss": 1.8693,
      "theoretical_loss": 4.193467643565306,
      "tokens_seen": 288817152
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009216850551654965,
      "loss": 1.8022,
      "theoretical_loss": 4.193356085188778,
      "tokens_seen": 288882688
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009216649949849549,
      "loss": 1.7514,
      "theoretical_loss": 4.193244559201995,
      "tokens_seen": 288948224
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009216449348044132,
      "loss": 1.8518,
      "theoretical_loss": 4.193133065588209,
      "tokens_seen": 289013760
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009216248746238717,
      "loss": 2.0078,
      "theoretical_loss": 4.193021604330686,
      "tokens_seen": 289079296
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00092160481444333,
      "loss": 1.8522,
      "theoretical_loss": 4.1929101754127025,
      "tokens_seen": 289144832
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009215847542627884,
      "loss": 1.7833,
      "theoretical_loss": 4.192798778817551,
      "tokens_seen": 289210368
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009215646940822468,
      "loss": 1.8018,
      "theoretical_loss": 4.192687414528532,
      "tokens_seen": 289275904
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009215446339017052,
      "loss": 1.8112,
      "theoretical_loss": 4.192576082528962,
      "tokens_seen": 289341440
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009215245737211635,
      "loss": 1.8178,
      "theoretical_loss": 4.192464782802167,
      "tokens_seen": 289406976
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009215045135406219,
      "loss": 1.8044,
      "theoretical_loss": 4.192353515331489,
      "tokens_seen": 289472512
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009214844533600802,
      "loss": 1.7338,
      "theoretical_loss": 4.192242280100279,
      "tokens_seen": 289538048
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009214643931795387,
      "loss": 1.8753,
      "theoretical_loss": 4.192131077091901,
      "tokens_seen": 289603584
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000921444332998997,
      "loss": 1.7773,
      "theoretical_loss": 4.192019906289733,
      "tokens_seen": 289669120
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009214242728184554,
      "loss": 1.9292,
      "theoretical_loss": 4.191908767677163,
      "tokens_seen": 289734656
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009214042126379137,
      "loss": 1.7483,
      "theoretical_loss": 4.191797661237594,
      "tokens_seen": 289800192
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009213841524573722,
      "loss": 1.8835,
      "theoretical_loss": 4.191686586954439,
      "tokens_seen": 289865728
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009213640922768305,
      "loss": 1.9106,
      "theoretical_loss": 4.1915755448111245,
      "tokens_seen": 289931264
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009213440320962889,
      "loss": 1.8306,
      "theoretical_loss": 4.191464534791089,
      "tokens_seen": 289996800
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009213239719157473,
      "loss": 1.8738,
      "theoretical_loss": 4.191353556877782,
      "tokens_seen": 290062336
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009213039117352056,
      "loss": 1.7921,
      "theoretical_loss": 4.191242611054668,
      "tokens_seen": 290127872
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000921283851554664,
      "loss": 1.8376,
      "theoretical_loss": 4.191131697305222,
      "tokens_seen": 290193408
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009212637913741224,
      "loss": 1.8509,
      "theoretical_loss": 4.191020815612932,
      "tokens_seen": 290258944
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009212437311935807,
      "loss": 1.8397,
      "theoretical_loss": 4.190909965961296,
      "tokens_seen": 290324480
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009212236710130392,
      "loss": 1.7438,
      "theoretical_loss": 4.190799148333827,
      "tokens_seen": 290390016
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009212036108324975,
      "loss": 1.9381,
      "theoretical_loss": 4.1906883627140505,
      "tokens_seen": 290455552
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009211835506519559,
      "loss": 1.7074,
      "theoretical_loss": 4.1905776090855,
      "tokens_seen": 290521088
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009211634904714143,
      "loss": 1.7709,
      "theoretical_loss": 4.190466887431725,
      "tokens_seen": 290586624
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009211434302908726,
      "loss": 1.8717,
      "theoretical_loss": 4.190356197736288,
      "tokens_seen": 290652160
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000921123370110331,
      "loss": 1.7057,
      "theoretical_loss": 4.19024553998276,
      "tokens_seen": 290717696
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009211033099297895,
      "loss": 1.8125,
      "theoretical_loss": 4.190134914154727,
      "tokens_seen": 290783232
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009210832497492477,
      "loss": 1.889,
      "theoretical_loss": 4.190024320235784,
      "tokens_seen": 290848768
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009210631895687062,
      "loss": 1.8423,
      "theoretical_loss": 4.189913758209543,
      "tokens_seen": 290914304
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009210431293881644,
      "loss": 1.7914,
      "theoretical_loss": 4.189803228059623,
      "tokens_seen": 290979840
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009210230692076229,
      "loss": 1.9284,
      "theoretical_loss": 4.189692729769659,
      "tokens_seen": 291045376
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009210030090270813,
      "loss": 1.9621,
      "theoretical_loss": 4.189582263323296,
      "tokens_seen": 291110912
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009209829488465396,
      "loss": 1.7024,
      "theoretical_loss": 4.1894718287041925,
      "tokens_seen": 291176448
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000920962888665998,
      "loss": 1.7627,
      "theoretical_loss": 4.189361425896016,
      "tokens_seen": 291241984
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009209428284854565,
      "loss": 1.7543,
      "theoretical_loss": 4.18925105488245,
      "tokens_seen": 291307520
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009209227683049147,
      "loss": 1.8712,
      "theoretical_loss": 4.189140715647189,
      "tokens_seen": 291373056
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009209027081243732,
      "loss": 1.7619,
      "theoretical_loss": 4.189030408173936,
      "tokens_seen": 291438592
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009208826479438314,
      "loss": 1.7172,
      "theoretical_loss": 4.188920132446411,
      "tokens_seen": 291504128
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009208625877632899,
      "loss": 1.8162,
      "theoretical_loss": 4.1888098884483425,
      "tokens_seen": 291569664
    },
    {
      "epoch": 0.09,
      "objective/train/advantage_avg": 0.49275243282318115,
      "objective/train/docs_used": 112844,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.677362322807312,
      "objective/train/original_loss": 1.6773622035980225,
      "objective/train/theoretical_loss": 4.188699676163473,
      "objective/train/tokens_used": 312095200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24318408966064453,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505115985870361,
      "objective/train/weighted_lm_loss": 1.7620242834091187,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.188699676163473,
      "tokens_seen": 291635200
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009208425275827483,
      "loss": 1.865,
      "theoretical_loss": 4.188699676163473,
      "tokens_seen": 291635200
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009208224674022066,
      "loss": 1.8001,
      "theoretical_loss": 4.188589495575557,
      "tokens_seen": 291700736
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000920802407221665,
      "loss": 1.7084,
      "theoretical_loss": 4.188479346668359,
      "tokens_seen": 291766272
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009207823470411234,
      "loss": 1.6633,
      "theoretical_loss": 4.188369229425657,
      "tokens_seen": 291831808
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009207622868605817,
      "loss": 1.9012,
      "theoretical_loss": 4.188259143831241,
      "tokens_seen": 291897344
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009207422266800402,
      "loss": 1.8598,
      "theoretical_loss": 4.188149089868912,
      "tokens_seen": 291962880
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009207221664994984,
      "loss": 1.7914,
      "theoretical_loss": 4.188039067522484,
      "tokens_seen": 292028416
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009207021063189569,
      "loss": 1.818,
      "theoretical_loss": 4.1879290767757835,
      "tokens_seen": 292093952
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009206820461384152,
      "loss": 1.8285,
      "theoretical_loss": 4.187819117612645,
      "tokens_seen": 292159488
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009206619859578736,
      "loss": 1.7891,
      "theoretical_loss": 4.18770919001692,
      "tokens_seen": 292225024
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000920641925777332,
      "loss": 1.8172,
      "theoretical_loss": 4.18759929397247,
      "tokens_seen": 292290560
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009206218655967904,
      "loss": 1.8366,
      "theoretical_loss": 4.187489429463166,
      "tokens_seen": 292356096
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009206018054162487,
      "loss": 1.87,
      "theoretical_loss": 4.187379596472894,
      "tokens_seen": 292421632
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009205817452357072,
      "loss": 1.8121,
      "theoretical_loss": 4.18726979498555,
      "tokens_seen": 292487168
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009205616850551655,
      "loss": 1.8013,
      "theoretical_loss": 4.187160024985044,
      "tokens_seen": 292552704
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009205416248746239,
      "loss": 1.8087,
      "theoretical_loss": 4.187050286455294,
      "tokens_seen": 292618240
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009205215646940822,
      "loss": 1.9665,
      "theoretical_loss": 4.1869405793802335,
      "tokens_seen": 292683776
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009205015045135406,
      "loss": 1.8839,
      "theoretical_loss": 4.1868309037438065,
      "tokens_seen": 292749312
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000920481444332999,
      "loss": 1.7737,
      "theoretical_loss": 4.1867212595299685,
      "tokens_seen": 292814848
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009204613841524574,
      "loss": 1.7755,
      "theoretical_loss": 4.186611646722687,
      "tokens_seen": 292880384
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009204413239719157,
      "loss": 1.9252,
      "theoretical_loss": 4.186502065305942,
      "tokens_seen": 292945920
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009204212637913741,
      "loss": 1.7296,
      "theoretical_loss": 4.186392515263723,
      "tokens_seen": 293011456
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009204012036108326,
      "loss": 1.8694,
      "theoretical_loss": 4.186282996580034,
      "tokens_seen": 293076992
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009203811434302909,
      "loss": 1.9082,
      "theoretical_loss": 4.1861735092388885,
      "tokens_seen": 293142528
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009203610832497493,
      "loss": 1.9323,
      "theoretical_loss": 4.186064053224314,
      "tokens_seen": 293208064
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009203410230692077,
      "loss": 1.7652,
      "theoretical_loss": 4.185954628520347,
      "tokens_seen": 293273600
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000920320962888666,
      "loss": 1.8631,
      "theoretical_loss": 4.185845235111037,
      "tokens_seen": 293339136
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009203009027081244,
      "loss": 1.7549,
      "theoretical_loss": 4.185735872980447,
      "tokens_seen": 293404672
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009202808425275828,
      "loss": 1.927,
      "theoretical_loss": 4.185626542112649,
      "tokens_seen": 293470208
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009202607823470411,
      "loss": 1.7839,
      "theoretical_loss": 4.185517242491728,
      "tokens_seen": 293535744
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009202407221664996,
      "loss": 1.8171,
      "theoretical_loss": 4.185407974101779,
      "tokens_seen": 293601280
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009202206619859579,
      "loss": 1.8146,
      "theoretical_loss": 4.185298736926912,
      "tokens_seen": 293666816
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009202006018054163,
      "loss": 1.8778,
      "theoretical_loss": 4.185189530951245,
      "tokens_seen": 293732352
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009201805416248747,
      "loss": 1.6603,
      "theoretical_loss": 4.18508035615891,
      "tokens_seen": 293797888
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000920160481444333,
      "loss": 1.813,
      "theoretical_loss": 4.184971212534048,
      "tokens_seen": 293863424
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009201404212637914,
      "loss": 1.8153,
      "theoretical_loss": 4.184862100060816,
      "tokens_seen": 293928960
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009201203610832499,
      "loss": 1.8382,
      "theoretical_loss": 4.184753018723379,
      "tokens_seen": 293994496
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009201003009027081,
      "loss": 1.7472,
      "theoretical_loss": 4.184643968505914,
      "tokens_seen": 294060032
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009200802407221666,
      "loss": 1.8905,
      "theoretical_loss": 4.184534949392611,
      "tokens_seen": 294125568
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009200601805416248,
      "loss": 1.7706,
      "theoretical_loss": 4.18442596136767,
      "tokens_seen": 294191104
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009200401203610833,
      "loss": 1.8976,
      "theoretical_loss": 4.184317004415303,
      "tokens_seen": 294256640
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009200200601805417,
      "loss": 1.8011,
      "theoretical_loss": 4.184208078519735,
      "tokens_seen": 294322176
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00092,
      "loss": 1.7645,
      "theoretical_loss": 4.184099183665199,
      "tokens_seen": 294387712
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009199799398194584,
      "loss": 1.9067,
      "theoretical_loss": 4.183990319835944,
      "tokens_seen": 294453248
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009199598796389169,
      "loss": 1.7968,
      "theoretical_loss": 4.183881487016227,
      "tokens_seen": 294518784
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009199398194583751,
      "loss": 1.856,
      "theoretical_loss": 4.183772685190318,
      "tokens_seen": 294584320
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009199197592778336,
      "loss": 1.925,
      "theoretical_loss": 4.1836639143425,
      "tokens_seen": 294649856
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009198996990972918,
      "loss": 1.787,
      "theoretical_loss": 4.183555174457063,
      "tokens_seen": 294715392
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009198796389167503,
      "loss": 1.7992,
      "theoretical_loss": 4.183446465518312,
      "tokens_seen": 294780928
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009198595787362087,
      "loss": 1.7961,
      "theoretical_loss": 4.183337787510563,
      "tokens_seen": 294846464
    },
    {
      "epoch": 0.09,
      "objective/train/advantage_avg": 0.4940253496170044,
      "objective/train/docs_used": 114046,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7510039806365967,
      "objective/train/original_loss": 1.7510039806365967,
      "objective/train/theoretical_loss": 4.1832291404181445,
      "objective/train/tokens_used": 315372000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24425998330116272,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506443977355957,
      "objective/train/weighted_lm_loss": 1.83975088596344,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 4.1832291404181445,
      "tokens_seen": 294912000
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000919839518555667,
      "loss": 1.8329,
      "theoretical_loss": 4.1832291404181445,
      "tokens_seen": 294912000
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009198194583751254,
      "loss": 1.7224,
      "theoretical_loss": 4.183120524225393,
      "tokens_seen": 294977536
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009197993981945838,
      "loss": 1.9731,
      "theoretical_loss": 4.183011938916657,
      "tokens_seen": 295043072
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009197793380140421,
      "loss": 1.7446,
      "theoretical_loss": 4.182903384476301,
      "tokens_seen": 295108608
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009197592778335006,
      "loss": 1.7929,
      "theoretical_loss": 4.182794860888696,
      "tokens_seen": 295174144
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009197392176529588,
      "loss": 1.8788,
      "theoretical_loss": 4.182686368138226,
      "tokens_seen": 295239680
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009197191574724173,
      "loss": 1.8534,
      "theoretical_loss": 4.182577906209287,
      "tokens_seen": 295305216
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009196990972918756,
      "loss": 1.7757,
      "theoretical_loss": 4.182469475086286,
      "tokens_seen": 295370752
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000919679037111334,
      "loss": 1.8442,
      "theoretical_loss": 4.18236107475364,
      "tokens_seen": 295436288
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009196589769307924,
      "loss": 1.9694,
      "theoretical_loss": 4.182252705195778,
      "tokens_seen": 295501824
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009196389167502508,
      "loss": 1.6922,
      "theoretical_loss": 4.182144366397143,
      "tokens_seen": 295567360
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009196188565697091,
      "loss": 1.8983,
      "theoretical_loss": 4.182036058342186,
      "tokens_seen": 295632896
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009195987963891676,
      "loss": 1.8164,
      "theoretical_loss": 4.18192778101537,
      "tokens_seen": 295698432
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009195787362086259,
      "loss": 1.8424,
      "theoretical_loss": 4.1818195344011695,
      "tokens_seen": 295763968
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009195586760280843,
      "loss": 1.7505,
      "theoretical_loss": 4.181711318484071,
      "tokens_seen": 295829504
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009195386158475426,
      "loss": 1.769,
      "theoretical_loss": 4.181603133248572,
      "tokens_seen": 295895040
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000919518555667001,
      "loss": 1.7624,
      "theoretical_loss": 4.181494978679181,
      "tokens_seen": 295960576
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009194984954864594,
      "loss": 1.7339,
      "theoretical_loss": 4.181386854760417,
      "tokens_seen": 296026112
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009194784353059178,
      "loss": 1.8652,
      "theoretical_loss": 4.181278761476811,
      "tokens_seen": 296091648
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009194583751253761,
      "loss": 1.7627,
      "theoretical_loss": 4.181170698812906,
      "tokens_seen": 296157184
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009194383149448345,
      "loss": 1.8077,
      "theoretical_loss": 4.181062666753256,
      "tokens_seen": 296222720
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009194182547642929,
      "loss": 1.8082,
      "theoretical_loss": 4.180954665282423,
      "tokens_seen": 296288256
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009193981945837513,
      "loss": 1.7497,
      "theoretical_loss": 4.180846694384986,
      "tokens_seen": 296353792
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009193781344032096,
      "loss": 1.8343,
      "theoretical_loss": 4.18073875404553,
      "tokens_seen": 296419328
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000919358074222668,
      "loss": 1.7863,
      "theoretical_loss": 4.180630844248653,
      "tokens_seen": 296484864
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009193380140421263,
      "loss": 1.8319,
      "theoretical_loss": 4.180522964978966,
      "tokens_seen": 296550400
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009193179538615848,
      "loss": 1.7604,
      "theoretical_loss": 4.180415116221088,
      "tokens_seen": 296615936
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009192978936810431,
      "loss": 1.9332,
      "theoretical_loss": 4.180307297959651,
      "tokens_seen": 296681472
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009192778335005015,
      "loss": 1.8325,
      "theoretical_loss": 4.180199510179299,
      "tokens_seen": 296747008
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009192577733199599,
      "loss": 1.6463,
      "theoretical_loss": 4.180091752864684,
      "tokens_seen": 296812544
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009192377131394183,
      "loss": 1.8372,
      "theoretical_loss": 4.179984026000473,
      "tokens_seen": 296878080
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009192176529588766,
      "loss": 1.7951,
      "theoretical_loss": 4.179876329571341,
      "tokens_seen": 296943616
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009191975927783351,
      "loss": 1.867,
      "theoretical_loss": 4.179768663561975,
      "tokens_seen": 297009152
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009191775325977933,
      "loss": 1.7988,
      "theoretical_loss": 4.179661027957074,
      "tokens_seen": 297074688
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009191574724172518,
      "loss": 1.8474,
      "theoretical_loss": 4.179553422741347,
      "tokens_seen": 297140224
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009191374122367101,
      "loss": 1.725,
      "theoretical_loss": 4.179445847899515,
      "tokens_seen": 297205760
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009191173520561685,
      "loss": 1.603,
      "theoretical_loss": 4.1793383034163085,
      "tokens_seen": 297271296
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009190972918756269,
      "loss": 1.6909,
      "theoretical_loss": 4.17923078927647,
      "tokens_seen": 297336832
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009190772316950852,
      "loss": 1.7892,
      "theoretical_loss": 4.1791233054647545,
      "tokens_seen": 297402368
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009190571715145436,
      "loss": 1.8126,
      "theoretical_loss": 4.179015851965926,
      "tokens_seen": 297467904
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009190371113340021,
      "loss": 1.7874,
      "theoretical_loss": 4.178908428764759,
      "tokens_seen": 297533440
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009190170511534603,
      "loss": 1.7865,
      "theoretical_loss": 4.178801035846041,
      "tokens_seen": 297598976
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009189969909729188,
      "loss": 1.81,
      "theoretical_loss": 4.178693673194569,
      "tokens_seen": 297664512
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000918976930792377,
      "loss": 1.7445,
      "theoretical_loss": 4.178586340795152,
      "tokens_seen": 297730048
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009189568706118355,
      "loss": 1.8662,
      "theoretical_loss": 4.17847903863261,
      "tokens_seen": 297795584
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009189368104312939,
      "loss": 1.7746,
      "theoretical_loss": 4.178371766691773,
      "tokens_seen": 297861120
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009189167502507522,
      "loss": 1.7787,
      "theoretical_loss": 4.1782645249574815,
      "tokens_seen": 297926656
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009188966900702107,
      "loss": 1.8718,
      "theoretical_loss": 4.178157313414589,
      "tokens_seen": 297992192
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009188766298896691,
      "loss": 1.8258,
      "theoretical_loss": 4.178050132047958,
      "tokens_seen": 298057728
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009188565697091274,
      "loss": 1.7314,
      "theoretical_loss": 4.177942980842464,
      "tokens_seen": 298123264
    },
    {
      "epoch": 0.09,
      "objective/train/advantage_avg": 0.4918665587902069,
      "objective/train/docs_used": 115247,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.66420578956604,
      "objective/train/original_loss": 1.66420578956604,
      "objective/train/theoretical_loss": 4.1778358597829905,
      "objective/train/tokens_used": 318648800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2422984093427658,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504183769226074,
      "objective/train/weighted_lm_loss": 1.7481639385223389,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.1778358597829905,
      "tokens_seen": 298188800
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009188365095285858,
      "loss": 1.825,
      "theoretical_loss": 4.1778358597829905,
      "tokens_seen": 298188800
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009188164493480442,
      "loss": 1.7152,
      "theoretical_loss": 4.1777287688544344,
      "tokens_seen": 298254336
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009187963891675025,
      "loss": 1.8607,
      "theoretical_loss": 4.177621708041703,
      "tokens_seen": 298319872
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000918776328986961,
      "loss": 1.8997,
      "theoretical_loss": 4.177514677329713,
      "tokens_seen": 298385408
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009187562688064192,
      "loss": 1.8691,
      "theoretical_loss": 4.177407676703393,
      "tokens_seen": 298450944
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009187362086258777,
      "loss": 1.7494,
      "theoretical_loss": 4.177300706147683,
      "tokens_seen": 298516480
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000918716148445336,
      "loss": 1.8863,
      "theoretical_loss": 4.177193765647534,
      "tokens_seen": 298582016
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009186960882647944,
      "loss": 1.7996,
      "theoretical_loss": 4.1770868551879055,
      "tokens_seen": 298647552
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009186760280842528,
      "loss": 1.8353,
      "theoretical_loss": 4.17697997475377,
      "tokens_seen": 298713088
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009186559679037112,
      "loss": 1.8755,
      "theoretical_loss": 4.176873124330111,
      "tokens_seen": 298778624
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009186359077231695,
      "loss": 2.0181,
      "theoretical_loss": 4.176766303901922,
      "tokens_seen": 298844160
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000918615847542628,
      "loss": 2.0517,
      "theoretical_loss": 4.176659513454207,
      "tokens_seen": 298909696
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009185957873620863,
      "loss": 1.8836,
      "theoretical_loss": 4.176552752971981,
      "tokens_seen": 298975232
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009185757271815447,
      "loss": 1.875,
      "theoretical_loss": 4.1764460224402695,
      "tokens_seen": 299040768
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000918555667001003,
      "loss": 1.8465,
      "theoretical_loss": 4.17633932184411,
      "tokens_seen": 299106304
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009185356068204614,
      "loss": 1.9421,
      "theoretical_loss": 4.176232651168551,
      "tokens_seen": 299171840
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009185155466399198,
      "loss": 1.8236,
      "theoretical_loss": 4.176126010398647,
      "tokens_seen": 299237376
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009184954864593782,
      "loss": 1.9052,
      "theoretical_loss": 4.176019399519471,
      "tokens_seen": 299302912
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009184754262788365,
      "loss": 1.8264,
      "theoretical_loss": 4.1759128185161005,
      "tokens_seen": 299368448
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009184553660982949,
      "loss": 1.7301,
      "theoretical_loss": 4.175806267373627,
      "tokens_seen": 299433984
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009184353059177533,
      "loss": 1.8618,
      "theoretical_loss": 4.17569974607715,
      "tokens_seen": 299499520
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009184152457372117,
      "loss": 1.7564,
      "theoretical_loss": 4.1755932546117815,
      "tokens_seen": 299565056
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00091839518555667,
      "loss": 1.7976,
      "theoretical_loss": 4.175486792962646,
      "tokens_seen": 299630592
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009183751253761285,
      "loss": 1.8856,
      "theoretical_loss": 4.175380361114874,
      "tokens_seen": 299696128
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009183550651955867,
      "loss": 1.7783,
      "theoretical_loss": 4.1752739590536105,
      "tokens_seen": 299761664
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009183350050150452,
      "loss": 1.7971,
      "theoretical_loss": 4.17516758676401,
      "tokens_seen": 299827200
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009183149448345035,
      "loss": 1.924,
      "theoretical_loss": 4.175061244231237,
      "tokens_seen": 299892736
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009182948846539619,
      "loss": 1.9111,
      "theoretical_loss": 4.174954931440468,
      "tokens_seen": 299958272
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009182748244734203,
      "loss": 1.947,
      "theoretical_loss": 4.174848648376889,
      "tokens_seen": 300023808
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009182547642928787,
      "loss": 1.8443,
      "theoretical_loss": 4.174742395025696,
      "tokens_seen": 300089344
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000918234704112337,
      "loss": 1.8847,
      "theoretical_loss": 4.174636171372097,
      "tokens_seen": 300154880
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009182146439317955,
      "loss": 1.8018,
      "theoretical_loss": 4.174529977401312,
      "tokens_seen": 300220416
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009181945837512537,
      "loss": 1.7974,
      "theoretical_loss": 4.174423813098566,
      "tokens_seen": 300285952
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009181745235707122,
      "loss": 1.8399,
      "theoretical_loss": 4.174317678449101,
      "tokens_seen": 300351488
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009181544633901705,
      "loss": 1.7105,
      "theoretical_loss": 4.174211573438166,
      "tokens_seen": 300417024
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009181344032096289,
      "loss": 1.9141,
      "theoretical_loss": 4.174105498051022,
      "tokens_seen": 300482560
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009181143430290873,
      "loss": 1.9205,
      "theoretical_loss": 4.173999452272939,
      "tokens_seen": 300548096
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009180942828485456,
      "loss": 1.9612,
      "theoretical_loss": 4.173893436089199,
      "tokens_seen": 300613632
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000918074222668004,
      "loss": 1.92,
      "theoretical_loss": 4.173787449485094,
      "tokens_seen": 300679168
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009180541624874625,
      "loss": 2.085,
      "theoretical_loss": 4.173681492445926,
      "tokens_seen": 300744704
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009180341023069207,
      "loss": 1.9086,
      "theoretical_loss": 4.173575564957009,
      "tokens_seen": 300810240
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009180140421263792,
      "loss": 1.8754,
      "theoretical_loss": 4.1734696670036655,
      "tokens_seen": 300875776
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009179939819458374,
      "loss": 2.1293,
      "theoretical_loss": 4.17336379857123,
      "tokens_seen": 300941312
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009179739217652959,
      "loss": 1.9747,
      "theoretical_loss": 4.1732579596450465,
      "tokens_seen": 301006848
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009179538615847543,
      "loss": 1.8285,
      "theoretical_loss": 4.17315215021047,
      "tokens_seen": 301072384
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009179338014042126,
      "loss": 2.0572,
      "theoretical_loss": 4.173046370252866,
      "tokens_seen": 301137920
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000917913741223671,
      "loss": 1.832,
      "theoretical_loss": 4.172940619757611,
      "tokens_seen": 301203456
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009178936810431295,
      "loss": 1.8726,
      "theoretical_loss": 4.172834898710089,
      "tokens_seen": 301268992
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009178736208625877,
      "loss": 1.8577,
      "theoretical_loss": 4.1727292070957,
      "tokens_seen": 301334528
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009178535606820462,
      "loss": 1.934,
      "theoretical_loss": 4.172623544899849,
      "tokens_seen": 301400064
    },
    {
      "epoch": 0.09,
      "objective/train/advantage_avg": 0.4897353947162628,
      "objective/train/docs_used": 116418,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8441132307052612,
      "objective/train/original_loss": 1.8441133499145508,
      "objective/train/theoretical_loss": 4.172517912107954,
      "objective/train/tokens_used": 321925600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24035406112670898,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050195336341858,
      "objective/train/weighted_lm_loss": 1.9364567995071411,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.172517912107954,
      "tokens_seen": 301465600
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009178335005015045,
      "loss": 2.0693,
      "theoretical_loss": 4.172517912107954,
      "tokens_seen": 301465600
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009178134403209629,
      "loss": 1.8463,
      "theoretical_loss": 4.172412308705443,
      "tokens_seen": 301531136
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009177933801404213,
      "loss": 1.8559,
      "theoretical_loss": 4.172306734677753,
      "tokens_seen": 301596672
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009177733199598796,
      "loss": 1.9022,
      "theoretical_loss": 4.172201190010334,
      "tokens_seen": 301662208
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000917753259779338,
      "loss": 1.814,
      "theoretical_loss": 4.172095674688645,
      "tokens_seen": 301727744
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009177331995987964,
      "loss": 2.0133,
      "theoretical_loss": 4.171990188698155,
      "tokens_seen": 301793280
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009177131394182547,
      "loss": 1.8285,
      "theoretical_loss": 4.171884732024344,
      "tokens_seen": 301858816
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009176930792377132,
      "loss": 1.8609,
      "theoretical_loss": 4.171779304652702,
      "tokens_seen": 301924352
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009176730190571715,
      "loss": 1.8516,
      "theoretical_loss": 4.171673906568729,
      "tokens_seen": 301989888
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009176529588766299,
      "loss": 1.8591,
      "theoretical_loss": 4.171568537757936,
      "tokens_seen": 302055424
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009176328986960883,
      "loss": 2.047,
      "theoretical_loss": 4.171463198205844,
      "tokens_seen": 302120960
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009176128385155467,
      "loss": 1.8739,
      "theoretical_loss": 4.171357887897985,
      "tokens_seen": 302186496
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000917592778335005,
      "loss": 1.8573,
      "theoretical_loss": 4.171252606819899,
      "tokens_seen": 302252032
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009175727181544634,
      "loss": 1.9282,
      "theoretical_loss": 4.17114735495714,
      "tokens_seen": 302317568
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009175526579739217,
      "loss": 1.8217,
      "theoretical_loss": 4.171042132295268,
      "tokens_seen": 302383104
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009175325977933802,
      "loss": 1.8833,
      "theoretical_loss": 4.170936938819857,
      "tokens_seen": 302448640
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009175125376128385,
      "loss": 1.779,
      "theoretical_loss": 4.170831774516489,
      "tokens_seen": 302514176
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009174924774322969,
      "loss": 1.8598,
      "theoretical_loss": 4.170726639370757,
      "tokens_seen": 302579712
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009174724172517552,
      "loss": 1.855,
      "theoretical_loss": 4.170621533368265,
      "tokens_seen": 302645248
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009174523570712137,
      "loss": 1.8664,
      "theoretical_loss": 4.170516456494624,
      "tokens_seen": 302710784
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000917432296890672,
      "loss": 1.853,
      "theoretical_loss": 4.170411408735461,
      "tokens_seen": 302776320
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009174122367101304,
      "loss": 1.7922,
      "theoretical_loss": 4.170306390076408,
      "tokens_seen": 302841856
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009173921765295889,
      "loss": 2.0018,
      "theoretical_loss": 4.1702014005031085,
      "tokens_seen": 302907392
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009173721163490471,
      "loss": 1.7791,
      "theoretical_loss": 4.170096440001217,
      "tokens_seen": 302972928
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009173520561685056,
      "loss": 1.9441,
      "theoretical_loss": 4.169991508556398,
      "tokens_seen": 303038464
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009173319959879639,
      "loss": 1.8292,
      "theoretical_loss": 4.169886606154328,
      "tokens_seen": 303104000
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009173119358074223,
      "loss": 1.8355,
      "theoretical_loss": 4.169781732780688,
      "tokens_seen": 303169536
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009172918756268807,
      "loss": 1.861,
      "theoretical_loss": 4.169676888421175,
      "tokens_seen": 303235072
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009172718154463391,
      "loss": 1.7574,
      "theoretical_loss": 4.169572073061493,
      "tokens_seen": 303300608
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009172517552657974,
      "loss": 1.9617,
      "theoretical_loss": 4.169467286687359,
      "tokens_seen": 303366144
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009172316950852559,
      "loss": 1.8277,
      "theoretical_loss": 4.169362529284497,
      "tokens_seen": 303431680
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009172116349047141,
      "loss": 1.8347,
      "theoretical_loss": 4.169257800838642,
      "tokens_seen": 303497216
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009171915747241726,
      "loss": 2.0021,
      "theoretical_loss": 4.16915310133554,
      "tokens_seen": 303562752
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000917171514543631,
      "loss": 1.9621,
      "theoretical_loss": 4.169048430760946,
      "tokens_seen": 303628288
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009171514543630893,
      "loss": 1.8753,
      "theoretical_loss": 4.168943789100626,
      "tokens_seen": 303693824
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009171313941825477,
      "loss": 1.8549,
      "theoretical_loss": 4.168839176340356,
      "tokens_seen": 303759360
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000917111334002006,
      "loss": 1.7648,
      "theoretical_loss": 4.1687345924659205,
      "tokens_seen": 303824896
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009170912738214644,
      "loss": 1.7129,
      "theoretical_loss": 4.168630037463117,
      "tokens_seen": 303890432
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009170712136409229,
      "loss": 1.8143,
      "theoretical_loss": 4.168525511317751,
      "tokens_seen": 303955968
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009170511534603811,
      "loss": 1.8852,
      "theoretical_loss": 4.168421014015637,
      "tokens_seen": 304021504
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009170310932798396,
      "loss": 1.8363,
      "theoretical_loss": 4.168316545542602,
      "tokens_seen": 304087040
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009170110330992978,
      "loss": 1.8495,
      "theoretical_loss": 4.168212105884482,
      "tokens_seen": 304152576
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009169909729187563,
      "loss": 1.8456,
      "theoretical_loss": 4.168107695027123,
      "tokens_seen": 304218112
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009169709127382147,
      "loss": 1.7831,
      "theoretical_loss": 4.1680033129563805,
      "tokens_seen": 304283648
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000916950852557673,
      "loss": 1.75,
      "theoretical_loss": 4.167898959658121,
      "tokens_seen": 304349184
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009169307923771314,
      "loss": 1.7698,
      "theoretical_loss": 4.167794635118219,
      "tokens_seen": 304414720
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009169107321965899,
      "loss": 1.8279,
      "theoretical_loss": 4.167690339322562,
      "tokens_seen": 304480256
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009168906720160481,
      "loss": 1.6521,
      "theoretical_loss": 4.167586072257047,
      "tokens_seen": 304545792
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009168706118355066,
      "loss": 1.7137,
      "theoretical_loss": 4.167481833907576,
      "tokens_seen": 304611328
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009168505516549649,
      "loss": 1.931,
      "theoretical_loss": 4.167377624260068,
      "tokens_seen": 304676864
    },
    {
      "epoch": 0.09,
      "objective/train/advantage_avg": 0.4910045862197876,
      "objective/train/docs_used": 117554,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6303279399871826,
      "objective/train/original_loss": 1.6303277015686035,
      "objective/train/theoretical_loss": 4.167273443300447,
      "objective/train/tokens_used": 325202400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.241532102227211,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050328254699707,
      "objective/train/weighted_lm_loss": 1.7122821807861328,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.167273443300447,
      "tokens_seen": 304742400
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009168304914744233,
      "loss": 1.7162,
      "theoretical_loss": 4.167273443300447,
      "tokens_seen": 304742400
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009168104312938817,
      "loss": 1.9333,
      "theoretical_loss": 4.167169291014649,
      "tokens_seen": 304807936
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00091679037111334,
      "loss": 2.0274,
      "theoretical_loss": 4.16706516738862,
      "tokens_seen": 304873472
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009167703109327984,
      "loss": 1.7796,
      "theoretical_loss": 4.1669610724083155,
      "tokens_seen": 304939008
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009167502507522568,
      "loss": 1.8187,
      "theoretical_loss": 4.1668570060597006,
      "tokens_seen": 305004544
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009167301905717151,
      "loss": 1.8278,
      "theoretical_loss": 4.16675296832875,
      "tokens_seen": 305070080
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009167101303911736,
      "loss": 1.7686,
      "theoretical_loss": 4.166648959201449,
      "tokens_seen": 305135616
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009166900702106319,
      "loss": 1.9126,
      "theoretical_loss": 4.1665449786637945,
      "tokens_seen": 305201152
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009166700100300903,
      "loss": 1.8431,
      "theoretical_loss": 4.166441026701788,
      "tokens_seen": 305266688
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009166499498495487,
      "loss": 1.8467,
      "theoretical_loss": 4.166337103301447,
      "tokens_seen": 305332224
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000916629889669007,
      "loss": 1.8731,
      "theoretical_loss": 4.166233208448794,
      "tokens_seen": 305397760
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009166098294884654,
      "loss": 1.8152,
      "theoretical_loss": 4.166129342129865,
      "tokens_seen": 305463296
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009165897693079238,
      "loss": 1.8298,
      "theoretical_loss": 4.166025504330703,
      "tokens_seen": 305528832
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009165697091273821,
      "loss": 1.8831,
      "theoretical_loss": 4.165921695037363,
      "tokens_seen": 305594368
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009165496489468406,
      "loss": 1.8718,
      "theoretical_loss": 4.165817914235908,
      "tokens_seen": 305659904
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009165295887662989,
      "loss": 1.8804,
      "theoretical_loss": 4.165714161912412,
      "tokens_seen": 305725440
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009165095285857573,
      "loss": 1.7289,
      "theoretical_loss": 4.165610438052958,
      "tokens_seen": 305790976
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009164894684052156,
      "loss": 1.7683,
      "theoretical_loss": 4.1655067426436405,
      "tokens_seen": 305856512
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009164694082246741,
      "loss": 1.8967,
      "theoretical_loss": 4.165403075670562,
      "tokens_seen": 305922048
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009164493480441324,
      "loss": 1.8074,
      "theoretical_loss": 4.165299437119834,
      "tokens_seen": 305987584
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009164292878635908,
      "loss": 1.8561,
      "theoretical_loss": 4.1651958269775795,
      "tokens_seen": 306053120
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009164092276830491,
      "loss": 1.7487,
      "theoretical_loss": 4.165092245229932,
      "tokens_seen": 306118656
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009163891675025075,
      "loss": 1.7637,
      "theoretical_loss": 4.164988691863032,
      "tokens_seen": 306184192
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009163691073219659,
      "loss": 1.9437,
      "theoretical_loss": 4.164885166863032,
      "tokens_seen": 306249728
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009163490471414243,
      "loss": 1.8181,
      "theoretical_loss": 4.164781670216092,
      "tokens_seen": 306315264
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009163289869608826,
      "loss": 1.7637,
      "theoretical_loss": 4.1646782019083854,
      "tokens_seen": 306380800
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009163089267803411,
      "loss": 1.6582,
      "theoretical_loss": 4.164574761926092,
      "tokens_seen": 306446336
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009162888665997994,
      "loss": 1.76,
      "theoretical_loss": 4.164471350255402,
      "tokens_seen": 306511872
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009162688064192578,
      "loss": 2.0577,
      "theoretical_loss": 4.164367966882517,
      "tokens_seen": 306577408
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009162487462387162,
      "loss": 1.574,
      "theoretical_loss": 4.164264611793644,
      "tokens_seen": 306642944
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009162286860581745,
      "loss": 1.8299,
      "theoretical_loss": 4.164161284975005,
      "tokens_seen": 306708480
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009162086258776329,
      "loss": 1.7307,
      "theoretical_loss": 4.164057986412828,
      "tokens_seen": 306774016
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009161885656970913,
      "loss": 1.8212,
      "theoretical_loss": 4.163954716093354,
      "tokens_seen": 306839552
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009161685055165496,
      "loss": 1.8538,
      "theoretical_loss": 4.163851474002829,
      "tokens_seen": 306905088
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009161484453360081,
      "loss": 1.7853,
      "theoretical_loss": 4.1637482601275115,
      "tokens_seen": 306970624
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009161283851554663,
      "loss": 1.8917,
      "theoretical_loss": 4.16364507445367,
      "tokens_seen": 307036160
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009161083249749248,
      "loss": 1.831,
      "theoretical_loss": 4.163541916967581,
      "tokens_seen": 307101696
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009160882647943832,
      "loss": 1.9265,
      "theoretical_loss": 4.163438787655533,
      "tokens_seen": 307167232
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009160682046138415,
      "loss": 1.8527,
      "theoretical_loss": 4.163335686503822,
      "tokens_seen": 307232768
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009160481444332999,
      "loss": 1.868,
      "theoretical_loss": 4.163232613498753,
      "tokens_seen": 307298304
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009160280842527582,
      "loss": 1.7818,
      "theoretical_loss": 4.163129568626642,
      "tokens_seen": 307363840
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009160080240722166,
      "loss": 1.7847,
      "theoretical_loss": 4.1630265518738145,
      "tokens_seen": 307429376
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009159879638916751,
      "loss": 1.8922,
      "theoretical_loss": 4.162923563226607,
      "tokens_seen": 307494912
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009159679037111333,
      "loss": 1.7713,
      "theoretical_loss": 4.162820602671362,
      "tokens_seen": 307560448
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009159478435305918,
      "loss": 1.8753,
      "theoretical_loss": 4.162717670194433,
      "tokens_seen": 307625984
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009159277833500502,
      "loss": 1.673,
      "theoretical_loss": 4.162614765782186,
      "tokens_seen": 307691520
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009159077231695085,
      "loss": 1.805,
      "theoretical_loss": 4.1625118894209905,
      "tokens_seen": 307757056
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000915887662988967,
      "loss": 1.7618,
      "theoretical_loss": 4.1624090410972325,
      "tokens_seen": 307822592
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009158676028084253,
      "loss": 1.7028,
      "theoretical_loss": 4.162306220797301,
      "tokens_seen": 307888128
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009158475426278837,
      "loss": 1.8686,
      "theoretical_loss": 4.162203428507601,
      "tokens_seen": 307953664
    },
    {
      "epoch": 0.09,
      "objective/train/advantage_avg": 0.49156174063682556,
      "objective/train/docs_used": 118729,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7690374851226807,
      "objective/train/original_loss": 1.7690374851226807,
      "objective/train/theoretical_loss": 4.16210066421454,
      "objective/train/tokens_used": 328479200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2423403114080429,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050388216972351,
      "objective/train/weighted_lm_loss": 1.858426570892334,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.16210066421454,
      "tokens_seen": 308019200
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009158274824473421,
      "loss": 1.7856,
      "theoretical_loss": 4.16210066421454,
      "tokens_seen": 308019200
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009158074222668004,
      "loss": 1.8842,
      "theoretical_loss": 4.16199792790454,
      "tokens_seen": 308084736
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009157873620862588,
      "loss": 1.8463,
      "theoretical_loss": 4.161895219564032,
      "tokens_seen": 308150272
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009157673019057172,
      "loss": 1.7313,
      "theoretical_loss": 4.161792539179454,
      "tokens_seen": 308215808
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009157472417251755,
      "loss": 1.7317,
      "theoretical_loss": 4.161689886737255,
      "tokens_seen": 308281344
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000915727181544634,
      "loss": 1.9005,
      "theoretical_loss": 4.161587262223895,
      "tokens_seen": 308346880
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009157071213640923,
      "loss": 1.7685,
      "theoretical_loss": 4.161484665625839,
      "tokens_seen": 308412416
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009156870611835507,
      "loss": 1.8842,
      "theoretical_loss": 4.161382096929565,
      "tokens_seen": 308477952
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009156670010030091,
      "loss": 1.8179,
      "theoretical_loss": 4.161279556121562,
      "tokens_seen": 308543488
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009156469408224675,
      "loss": 1.8155,
      "theoretical_loss": 4.1611770431883235,
      "tokens_seen": 308609024
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009156268806419258,
      "loss": 1.8621,
      "theoretical_loss": 4.161074558116356,
      "tokens_seen": 308674560
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009156068204613842,
      "loss": 1.8453,
      "theoretical_loss": 4.160972100892174,
      "tokens_seen": 308740096
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009155867602808425,
      "loss": 1.737,
      "theoretical_loss": 4.160869671502302,
      "tokens_seen": 308805632
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000915566700100301,
      "loss": 1.767,
      "theoretical_loss": 4.160767269933274,
      "tokens_seen": 308871168
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009155466399197593,
      "loss": 1.814,
      "theoretical_loss": 4.160664896171632,
      "tokens_seen": 308936704
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009155265797392177,
      "loss": 1.8207,
      "theoretical_loss": 4.160562550203928,
      "tokens_seen": 309002240
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000915506519558676,
      "loss": 1.9213,
      "theoretical_loss": 4.160460232016725,
      "tokens_seen": 309067776
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009154864593781345,
      "loss": 1.8559,
      "theoretical_loss": 4.160357941596594,
      "tokens_seen": 309133312
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009154663991975928,
      "loss": 1.7862,
      "theoretical_loss": 4.160255678930115,
      "tokens_seen": 309198848
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009154463390170512,
      "loss": 1.8416,
      "theoretical_loss": 4.1601534440038765,
      "tokens_seen": 309264384
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009154262788365095,
      "loss": 1.8704,
      "theoretical_loss": 4.16005123680448,
      "tokens_seen": 309329920
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009154062186559679,
      "loss": 1.7641,
      "theoretical_loss": 4.159949057318532,
      "tokens_seen": 309395456
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009153861584754263,
      "loss": 1.7451,
      "theoretical_loss": 4.1598469055326515,
      "tokens_seen": 309460992
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009153660982948847,
      "loss": 1.7089,
      "theoretical_loss": 4.159744781433464,
      "tokens_seen": 309526528
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000915346038114343,
      "loss": 1.8765,
      "theoretical_loss": 4.159642685007606,
      "tokens_seen": 309592064
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009153259779338015,
      "loss": 1.8021,
      "theoretical_loss": 4.159540616241723,
      "tokens_seen": 309657600
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009153059177532598,
      "loss": 1.7355,
      "theoretical_loss": 4.159438575122471,
      "tokens_seen": 309723136
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009152858575727182,
      "loss": 1.7621,
      "theoretical_loss": 4.159336561636513,
      "tokens_seen": 309788672
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009152657973921766,
      "loss": 1.7003,
      "theoretical_loss": 4.1592345757705225,
      "tokens_seen": 309854208
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009152457372116349,
      "loss": 1.6826,
      "theoretical_loss": 4.159132617511181,
      "tokens_seen": 309919744
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009152256770310933,
      "loss": 1.7535,
      "theoretical_loss": 4.159030686845182,
      "tokens_seen": 309985280
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009152056168505517,
      "loss": 1.776,
      "theoretical_loss": 4.158928783759226,
      "tokens_seen": 310050816
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00091518555667001,
      "loss": 1.7716,
      "theoretical_loss": 4.158826908240022,
      "tokens_seen": 310116352
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009151654964894685,
      "loss": 1.7732,
      "theoretical_loss": 4.158725060274291,
      "tokens_seen": 310181888
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009151454363089267,
      "loss": 1.7397,
      "theoretical_loss": 4.158623239848762,
      "tokens_seen": 310247424
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009151253761283852,
      "loss": 1.7299,
      "theoretical_loss": 4.15852144695017,
      "tokens_seen": 310312960
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009151053159478436,
      "loss": 1.8316,
      "theoretical_loss": 4.158419681565265,
      "tokens_seen": 310378496
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009150852557673019,
      "loss": 1.7404,
      "theoretical_loss": 4.158317943680801,
      "tokens_seen": 310444032
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009150651955867603,
      "loss": 1.6477,
      "theoretical_loss": 4.158216233283545,
      "tokens_seen": 310509568
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009150451354062186,
      "loss": 1.8772,
      "theoretical_loss": 4.158114550360271,
      "tokens_seen": 310575104
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000915025075225677,
      "loss": 1.925,
      "theoretical_loss": 4.1580128948977615,
      "tokens_seen": 310640640
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009150050150451355,
      "loss": 1.7451,
      "theoretical_loss": 4.157911266882811,
      "tokens_seen": 310706176
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009149849548645937,
      "loss": 1.7826,
      "theoretical_loss": 4.157809666302221,
      "tokens_seen": 310771712
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009149648946840522,
      "loss": 1.7177,
      "theoretical_loss": 4.157708093142801,
      "tokens_seen": 310837248
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009149448345035106,
      "loss": 1.9009,
      "theoretical_loss": 4.157606547391374,
      "tokens_seen": 310902784
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009149247743229689,
      "loss": 1.7678,
      "theoretical_loss": 4.157505029034767,
      "tokens_seen": 310968320
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009149047141424273,
      "loss": 1.9484,
      "theoretical_loss": 4.157403538059819,
      "tokens_seen": 311033856
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009148846539618857,
      "loss": 1.8541,
      "theoretical_loss": 4.157302074453379,
      "tokens_seen": 311099392
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000914864593781344,
      "loss": 1.8846,
      "theoretical_loss": 4.157200638202301,
      "tokens_seen": 311164928
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009148445336008025,
      "loss": 1.5425,
      "theoretical_loss": 4.157099229293453,
      "tokens_seen": 311230464
    },
    {
      "epoch": 0.09,
      "objective/train/advantage_avg": 0.4924384355545044,
      "objective/train/docs_used": 120076,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.819303274154663,
      "objective/train/original_loss": 1.8193035125732422,
      "objective/train/theoretical_loss": 4.156997847713709,
      "objective/train/tokens_used": 331756000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24328577518463135,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050480604171753,
      "objective/train/weighted_lm_loss": 1.911105751991272,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.156997847713709,
      "tokens_seen": 311296000
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009148244734202607,
      "loss": 1.7748,
      "theoretical_loss": 4.156997847713709,
      "tokens_seen": 311296000
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009148044132397192,
      "loss": 1.7985,
      "theoretical_loss": 4.156896493449951,
      "tokens_seen": 311361536
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009147843530591775,
      "loss": 1.7454,
      "theoretical_loss": 4.156795166489074,
      "tokens_seen": 311427072
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009147642928786359,
      "loss": 1.8097,
      "theoretical_loss": 4.15669386681798,
      "tokens_seen": 311492608
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009147442326980943,
      "loss": 1.8527,
      "theoretical_loss": 4.156592594423579,
      "tokens_seen": 311558144
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009147241725175527,
      "loss": 1.7922,
      "theoretical_loss": 4.156491349292789,
      "tokens_seen": 311623680
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000914704112337011,
      "loss": 1.7492,
      "theoretical_loss": 4.156390131412543,
      "tokens_seen": 311689216
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009146840521564695,
      "loss": 1.6228,
      "theoretical_loss": 4.156288940769778,
      "tokens_seen": 311754752
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009146639919759277,
      "loss": 1.8304,
      "theoretical_loss": 4.156187777351438,
      "tokens_seen": 311820288
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009146439317953862,
      "loss": 1.7774,
      "theoretical_loss": 4.156086641144483,
      "tokens_seen": 311885824
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009146238716148445,
      "loss": 1.8537,
      "theoretical_loss": 4.155985532135875,
      "tokens_seen": 311951360
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009146038114343029,
      "loss": 1.7901,
      "theoretical_loss": 4.155884450312589,
      "tokens_seen": 312016896
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009145837512537613,
      "loss": 1.8438,
      "theoretical_loss": 4.155783395661609,
      "tokens_seen": 312082432
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009145636910732197,
      "loss": 1.8641,
      "theoretical_loss": 4.155682368169925,
      "tokens_seen": 312147968
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000914543630892678,
      "loss": 1.8986,
      "theoretical_loss": 4.1555813678245395,
      "tokens_seen": 312213504
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009145235707121364,
      "loss": 1.9761,
      "theoretical_loss": 4.155480394612462,
      "tokens_seen": 312279040
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009145035105315948,
      "loss": 1.644,
      "theoretical_loss": 4.15537944852071,
      "tokens_seen": 312344576
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009144834503510532,
      "loss": 1.8129,
      "theoretical_loss": 4.155278529536313,
      "tokens_seen": 312410112
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009144633901705115,
      "loss": 1.6904,
      "theoretical_loss": 4.155177637646306,
      "tokens_seen": 312475648
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00091444332998997,
      "loss": 1.8086,
      "theoretical_loss": 4.1550767728377345,
      "tokens_seen": 312541184
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009144232698094282,
      "loss": 1.619,
      "theoretical_loss": 4.154975935097655,
      "tokens_seen": 312606720
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009144032096288867,
      "loss": 1.7375,
      "theoretical_loss": 4.154875124413128,
      "tokens_seen": 312672256
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009143831494483451,
      "loss": 1.779,
      "theoretical_loss": 4.154774340771228,
      "tokens_seen": 312737792
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009143630892678034,
      "loss": 1.6979,
      "theoretical_loss": 4.154673584159036,
      "tokens_seen": 312803328
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009143430290872619,
      "loss": 1.8726,
      "theoretical_loss": 4.15457285456364,
      "tokens_seen": 312868864
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009143229689067202,
      "loss": 1.7921,
      "theoretical_loss": 4.154472151972142,
      "tokens_seen": 312934400
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009143029087261786,
      "loss": 1.8944,
      "theoretical_loss": 4.154371476371646,
      "tokens_seen": 312999936
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000914282848545637,
      "loss": 1.9317,
      "theoretical_loss": 4.154270827749271,
      "tokens_seen": 313065472
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009142627883650953,
      "loss": 1.8379,
      "theoretical_loss": 4.154170206092143,
      "tokens_seen": 313131008
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009142427281845537,
      "loss": 1.6154,
      "theoretical_loss": 4.154069611387394,
      "tokens_seen": 313196544
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009142226680040121,
      "loss": 1.7201,
      "theoretical_loss": 4.153969043622169,
      "tokens_seen": 313262080
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009142026078234704,
      "loss": 1.9128,
      "theoretical_loss": 4.153868502783618,
      "tokens_seen": 313327616
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009141825476429289,
      "loss": 1.868,
      "theoretical_loss": 4.153767988858904,
      "tokens_seen": 313393152
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0009141624874623871,
      "loss": 1.7828,
      "theoretical_loss": 4.153667501835195,
      "tokens_seen": 313458688
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009141424272818456,
      "loss": 1.7955,
      "theoretical_loss": 4.15356704169967,
      "tokens_seen": 313524224
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000914122367101304,
      "loss": 1.8222,
      "theoretical_loss": 4.1534666084395155,
      "tokens_seen": 313589760
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009141023069207623,
      "loss": 1.7689,
      "theoretical_loss": 4.153366202041927,
      "tokens_seen": 313655296
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009140822467402207,
      "loss": 1.7823,
      "theoretical_loss": 4.153265822494111,
      "tokens_seen": 313720832
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000914062186559679,
      "loss": 2.0244,
      "theoretical_loss": 4.153165469783279,
      "tokens_seen": 313786368
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009140421263791374,
      "loss": 1.8247,
      "theoretical_loss": 4.153065143896655,
      "tokens_seen": 313851904
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009140220661985959,
      "loss": 1.9073,
      "theoretical_loss": 4.152964844821469,
      "tokens_seen": 313917440
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009140020060180541,
      "loss": 1.8277,
      "theoretical_loss": 4.15286457254496,
      "tokens_seen": 313982976
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009139819458375126,
      "loss": 1.8345,
      "theoretical_loss": 4.152764327054376,
      "tokens_seen": 314048512
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000913961885656971,
      "loss": 1.7599,
      "theoretical_loss": 4.152664108336976,
      "tokens_seen": 314114048
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009139418254764293,
      "loss": 1.8354,
      "theoretical_loss": 4.152563916380026,
      "tokens_seen": 314179584
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009139217652958877,
      "loss": 1.818,
      "theoretical_loss": 4.152463751170799,
      "tokens_seen": 314245120
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000913901705115346,
      "loss": 1.778,
      "theoretical_loss": 4.152363612696579,
      "tokens_seen": 314310656
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009138816449348044,
      "loss": 1.8639,
      "theoretical_loss": 4.152263500944658,
      "tokens_seen": 314376192
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009138615847542629,
      "loss": 1.7843,
      "theoretical_loss": 4.152163415902336,
      "tokens_seen": 314441728
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009138415245737211,
      "loss": 1.6987,
      "theoretical_loss": 4.152063357556923,
      "tokens_seen": 314507264
    },
    {
      "epoch": 0.1,
      "objective/train/advantage_avg": 0.49289390444755554,
      "objective/train/docs_used": 121169,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7357332706451416,
      "objective/train/original_loss": 1.735733151435852,
      "objective/train/theoretical_loss": 4.151963325895737,
      "objective/train/tokens_used": 335032800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24352145195007324,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505274534225464,
      "objective/train/weighted_lm_loss": 1.8233038187026978,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.151963325895737,
      "tokens_seen": 314572800
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009138214643931796,
      "loss": 1.7403,
      "theoretical_loss": 4.151963325895737,
      "tokens_seen": 314572800
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009138014042126379,
      "loss": 1.7728,
      "theoretical_loss": 4.151863320906105,
      "tokens_seen": 314638336
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009137813440320963,
      "loss": 1.7417,
      "theoretical_loss": 4.151763342575362,
      "tokens_seen": 314703872
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009137612838515547,
      "loss": 1.7573,
      "theoretical_loss": 4.151663390890851,
      "tokens_seen": 314769408
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009137412236710131,
      "loss": 1.8931,
      "theoretical_loss": 4.151563465839927,
      "tokens_seen": 314834944
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009137211634904714,
      "loss": 1.7439,
      "theoretical_loss": 4.151463567409949,
      "tokens_seen": 314900480
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009137011033099299,
      "loss": 1.8137,
      "theoretical_loss": 4.151363695588287,
      "tokens_seen": 314966016
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009136810431293881,
      "loss": 1.7893,
      "theoretical_loss": 4.151263850362321,
      "tokens_seen": 315031552
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009136609829488466,
      "loss": 1.844,
      "theoretical_loss": 4.151164031719437,
      "tokens_seen": 315097088
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009136409227683049,
      "loss": 1.8051,
      "theoretical_loss": 4.151064239647032,
      "tokens_seen": 315162624
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009136208625877633,
      "loss": 1.7564,
      "theoretical_loss": 4.1509644741325085,
      "tokens_seen": 315228160
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009136008024072217,
      "loss": 1.8875,
      "theoretical_loss": 4.150864735163281,
      "tokens_seen": 315293696
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009135807422266801,
      "loss": 1.8228,
      "theoretical_loss": 4.15076502272677,
      "tokens_seen": 315359232
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009135606820461384,
      "loss": 1.7781,
      "theoretical_loss": 4.150665336810405,
      "tokens_seen": 315424768
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009135406218655968,
      "loss": 1.7661,
      "theoretical_loss": 4.150565677401627,
      "tokens_seen": 315490304
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009135205616850552,
      "loss": 1.7718,
      "theoretical_loss": 4.15046604448788,
      "tokens_seen": 315555840
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009135005015045136,
      "loss": 1.7745,
      "theoretical_loss": 4.150366438056622,
      "tokens_seen": 315621376
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009134804413239719,
      "loss": 1.7982,
      "theoretical_loss": 4.150266858095317,
      "tokens_seen": 315686912
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009134603811434303,
      "loss": 1.674,
      "theoretical_loss": 4.150167304591437,
      "tokens_seen": 315752448
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009134403209628886,
      "loss": 1.7676,
      "theoretical_loss": 4.150067777532464,
      "tokens_seen": 315817984
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009134202607823471,
      "loss": 1.7478,
      "theoretical_loss": 4.149968276905888,
      "tokens_seen": 315883520
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009134002006018054,
      "loss": 1.797,
      "theoretical_loss": 4.149868802699206,
      "tokens_seen": 315949056
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009133801404212638,
      "loss": 2.1356,
      "theoretical_loss": 4.149769354899927,
      "tokens_seen": 316014592
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009133600802407222,
      "loss": 1.8429,
      "theoretical_loss": 4.149669933495565,
      "tokens_seen": 316080128
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009133400200601806,
      "loss": 1.7527,
      "theoretical_loss": 4.149570538473644,
      "tokens_seen": 316145664
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009133199598796389,
      "loss": 1.731,
      "theoretical_loss": 4.149471169821696,
      "tokens_seen": 316211200
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009132998996990974,
      "loss": 1.7687,
      "theoretical_loss": 4.149371827527264,
      "tokens_seen": 316276736
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009132798395185556,
      "loss": 1.6522,
      "theoretical_loss": 4.149272511577895,
      "tokens_seen": 316342272
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009132597793380141,
      "loss": 1.7597,
      "theoretical_loss": 4.149173221961146,
      "tokens_seen": 316407808
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009132397191574724,
      "loss": 1.6395,
      "theoretical_loss": 4.1490739586645855,
      "tokens_seen": 316473344
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009132196589769308,
      "loss": 1.87,
      "theoretical_loss": 4.148974721675787,
      "tokens_seen": 316538880
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009131995987963892,
      "loss": 1.7709,
      "theoretical_loss": 4.148875510982333,
      "tokens_seen": 316604416
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009131795386158475,
      "loss": 1.8424,
      "theoretical_loss": 4.1487763265718165,
      "tokens_seen": 316669952
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009131594784353059,
      "loss": 1.7587,
      "theoretical_loss": 4.148677168431836,
      "tokens_seen": 316735488
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009131394182547644,
      "loss": 1.7732,
      "theoretical_loss": 4.148578036550001,
      "tokens_seen": 316801024
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009131193580742226,
      "loss": 1.8576,
      "theoretical_loss": 4.148478930913927,
      "tokens_seen": 316866560
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009130992978936811,
      "loss": 1.7591,
      "theoretical_loss": 4.148379851511241,
      "tokens_seen": 316932096
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009130792377131393,
      "loss": 1.8007,
      "theoretical_loss": 4.148280798329575,
      "tokens_seen": 316997632
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009130591775325978,
      "loss": 1.8282,
      "theoretical_loss": 4.148181771356571,
      "tokens_seen": 317063168
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009130391173520562,
      "loss": 1.7429,
      "theoretical_loss": 4.14808277057988,
      "tokens_seen": 317128704
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009130190571715145,
      "loss": 1.7629,
      "theoretical_loss": 4.147983795987161,
      "tokens_seen": 317194240
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009129989969909729,
      "loss": 1.7354,
      "theoretical_loss": 4.147884847566079,
      "tokens_seen": 317259776
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009129789368104314,
      "loss": 1.7621,
      "theoretical_loss": 4.147785925304311,
      "tokens_seen": 317325312
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009129588766298896,
      "loss": 1.7322,
      "theoretical_loss": 4.147687029189541,
      "tokens_seen": 317390848
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009129388164493481,
      "loss": 1.901,
      "theoretical_loss": 4.14758815920946,
      "tokens_seen": 317456384
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009129187562688063,
      "loss": 1.7886,
      "theoretical_loss": 4.147489315351769,
      "tokens_seen": 317521920
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009128986960882648,
      "loss": 1.6663,
      "theoretical_loss": 4.147390497604177,
      "tokens_seen": 317587456
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009128786359077233,
      "loss": 1.7448,
      "theoretical_loss": 4.147291705954401,
      "tokens_seen": 317652992
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009128585757271815,
      "loss": 1.717,
      "theoretical_loss": 4.147192940390165,
      "tokens_seen": 317718528
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00091283851554664,
      "loss": 1.947,
      "theoretical_loss": 4.147094200899205,
      "tokens_seen": 317784064
    },
    {
      "epoch": 0.1,
      "objective/train/advantage_avg": 0.49450305104255676,
      "objective/train/docs_used": 122440,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.54131281375885,
      "objective/train/original_loss": 1.5413126945495605,
      "objective/train/theoretical_loss": 4.146995487469262,
      "objective/train/tokens_used": 338309600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24474811553955078,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506945848464966,
      "objective/train/weighted_lm_loss": 1.619470477104187,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 4.146995487469262,
      "tokens_seen": 317849600
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009128184553660983,
      "loss": 1.7732,
      "theoretical_loss": 4.146995487469262,
      "tokens_seen": 317849600
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009127983951855567,
      "loss": 1.8489,
      "theoretical_loss": 4.146896800088085,
      "tokens_seen": 317915136
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009127783350050151,
      "loss": 1.7643,
      "theoretical_loss": 4.146798138743433,
      "tokens_seen": 317980672
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009127582748244735,
      "loss": 1.707,
      "theoretical_loss": 4.146699503423074,
      "tokens_seen": 318046208
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009127382146439318,
      "loss": 1.6486,
      "theoretical_loss": 4.1466008941147825,
      "tokens_seen": 318111744
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009127181544633903,
      "loss": 1.8341,
      "theoretical_loss": 4.146502310806342,
      "tokens_seen": 318177280
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009126980942828485,
      "loss": 1.6462,
      "theoretical_loss": 4.146403753485544,
      "tokens_seen": 318242816
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000912678034102307,
      "loss": 1.7509,
      "theoretical_loss": 4.146305222140187,
      "tokens_seen": 318308352
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009126579739217653,
      "loss": 1.7517,
      "theoretical_loss": 4.146206716758081,
      "tokens_seen": 318373888
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009126379137412237,
      "loss": 1.7286,
      "theoretical_loss": 4.1461082373270415,
      "tokens_seen": 318439424
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009126178535606821,
      "loss": 1.8237,
      "theoretical_loss": 4.146009783834892,
      "tokens_seen": 318504960
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009125977933801405,
      "loss": 1.8733,
      "theoretical_loss": 4.145911356269468,
      "tokens_seen": 318570496
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009125777331995988,
      "loss": 1.8255,
      "theoretical_loss": 4.1458129546186075,
      "tokens_seen": 318636032
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009125576730190572,
      "loss": 1.7995,
      "theoretical_loss": 4.1457145788701615,
      "tokens_seen": 318701568
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009125376128385156,
      "loss": 1.7193,
      "theoretical_loss": 4.145616229011987,
      "tokens_seen": 318767104
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000912517552657974,
      "loss": 1.828,
      "theoretical_loss": 4.145517905031948,
      "tokens_seen": 318832640
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009124974924774323,
      "loss": 1.7877,
      "theoretical_loss": 4.145419606917921,
      "tokens_seen": 318898176
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009124774322968907,
      "loss": 1.9088,
      "theoretical_loss": 4.145321334657785,
      "tokens_seen": 318963712
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000912457372116349,
      "loss": 1.7268,
      "theoretical_loss": 4.145223088239432,
      "tokens_seen": 319029248
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009124373119358075,
      "loss": 1.801,
      "theoretical_loss": 4.145124867650759,
      "tokens_seen": 319094784
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009124172517552658,
      "loss": 1.7628,
      "theoretical_loss": 4.145026672879674,
      "tokens_seen": 319160320
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009123971915747242,
      "loss": 1.7966,
      "theoretical_loss": 4.14492850391409,
      "tokens_seen": 319225856
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009123771313941826,
      "loss": 1.6315,
      "theoretical_loss": 4.14483036074193,
      "tokens_seen": 319291392
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000912357071213641,
      "loss": 1.7785,
      "theoretical_loss": 4.144732243351125,
      "tokens_seen": 319356928
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009123370110330993,
      "loss": 1.8006,
      "theoretical_loss": 4.144634151729615,
      "tokens_seen": 319422464
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009123169508525578,
      "loss": 1.7751,
      "theoretical_loss": 4.144536085865344,
      "tokens_seen": 319488000
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000912296890672016,
      "loss": 1.9455,
      "theoretical_loss": 4.14443804574627,
      "tokens_seen": 319553536
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009122768304914745,
      "loss": 1.7975,
      "theoretical_loss": 4.144340031360354,
      "tokens_seen": 319619072
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009122567703109328,
      "loss": 1.8911,
      "theoretical_loss": 4.14424204269557,
      "tokens_seen": 319684608
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009122367101303912,
      "loss": 1.6413,
      "theoretical_loss": 4.1441440797398945,
      "tokens_seen": 319750144
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009122166499498496,
      "loss": 1.6841,
      "theoretical_loss": 4.144046142481317,
      "tokens_seen": 319815680
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009121965897693079,
      "loss": 1.9622,
      "theoretical_loss": 4.143948230907832,
      "tokens_seen": 319881216
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009121765295887663,
      "loss": 1.7783,
      "theoretical_loss": 4.1438503450074435,
      "tokens_seen": 319946752
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009121564694082248,
      "loss": 1.7984,
      "theoretical_loss": 4.1437524847681635,
      "tokens_seen": 320012288
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000912136409227683,
      "loss": 1.7974,
      "theoretical_loss": 4.143654650178012,
      "tokens_seen": 320077824
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009121163490471415,
      "loss": 1.8931,
      "theoretical_loss": 4.143556841225015,
      "tokens_seen": 320143360
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009120962888665997,
      "loss": 1.8957,
      "theoretical_loss": 4.14345905789721,
      "tokens_seen": 320208896
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009120762286860582,
      "loss": 1.9032,
      "theoretical_loss": 4.14336130018264,
      "tokens_seen": 320274432
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009120561685055166,
      "loss": 1.6911,
      "theoretical_loss": 4.143263568069358,
      "tokens_seen": 320339968
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009120361083249749,
      "loss": 1.7567,
      "theoretical_loss": 4.143165861545421,
      "tokens_seen": 320405504
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009120160481444333,
      "loss": 1.6795,
      "theoretical_loss": 4.143068180598902,
      "tokens_seen": 320471040
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009119959879638918,
      "loss": 1.7473,
      "theoretical_loss": 4.142970525217872,
      "tokens_seen": 320536576
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00091197592778335,
      "loss": 1.8411,
      "theoretical_loss": 4.142872895390417,
      "tokens_seen": 320602112
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009119558676028085,
      "loss": 1.8678,
      "theoretical_loss": 4.142775291104629,
      "tokens_seen": 320667648
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009119358074222667,
      "loss": 1.7909,
      "theoretical_loss": 4.142677712348608,
      "tokens_seen": 320733184
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009119157472417252,
      "loss": 1.748,
      "theoretical_loss": 4.14258015911046,
      "tokens_seen": 320798720
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009118956870611836,
      "loss": 1.7048,
      "theoretical_loss": 4.142482631378303,
      "tokens_seen": 320864256
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009118756268806419,
      "loss": 1.7556,
      "theoretical_loss": 4.142385129140259,
      "tokens_seen": 320929792
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009118555667001003,
      "loss": 1.6847,
      "theoretical_loss": 4.142287652384462,
      "tokens_seen": 320995328
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009118355065195587,
      "loss": 1.6888,
      "theoretical_loss": 4.142190201099049,
      "tokens_seen": 321060864
    },
    {
      "epoch": 0.1,
      "objective/train/advantage_avg": 0.48887574672698975,
      "objective/train/docs_used": 123582,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6981501579284668,
      "objective/train/original_loss": 1.6981501579284668,
      "objective/train/theoretical_loss": 4.142092775272169,
      "objective/train/tokens_used": 341586400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23958638310432434,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050105333328247,
      "objective/train/weighted_lm_loss": 1.7833449840545654,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.142092775272169,
      "tokens_seen": 321126400
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000911815446339017,
      "loss": 1.7979,
      "theoretical_loss": 4.142092775272169,
      "tokens_seen": 321126400
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009117953861584755,
      "loss": 1.783,
      "theoretical_loss": 4.141995374891978,
      "tokens_seen": 321191936
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009117753259779338,
      "loss": 1.5579,
      "theoretical_loss": 4.1418979999466385,
      "tokens_seen": 321257472
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009117552657973922,
      "loss": 1.656,
      "theoretical_loss": 4.141800650424322,
      "tokens_seen": 321323008
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009117352056168505,
      "loss": 1.7395,
      "theoretical_loss": 4.141703326313209,
      "tokens_seen": 321388544
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000911715145436309,
      "loss": 1.8497,
      "theoretical_loss": 4.141606027601485,
      "tokens_seen": 321454080
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009116950852557673,
      "loss": 1.7712,
      "theoretical_loss": 4.141508754277346,
      "tokens_seen": 321519616
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009116750250752257,
      "loss": 1.6483,
      "theoretical_loss": 4.141411506328995,
      "tokens_seen": 321585152
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000911654964894684,
      "loss": 1.8046,
      "theoretical_loss": 4.141314283744643,
      "tokens_seen": 321650688
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009116349047141425,
      "loss": 1.8253,
      "theoretical_loss": 4.141217086512508,
      "tokens_seen": 321716224
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009116148445336008,
      "loss": 1.8028,
      "theoretical_loss": 4.141119914620817,
      "tokens_seen": 321781760
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009115947843530592,
      "loss": 1.7437,
      "theoretical_loss": 4.141022768057805,
      "tokens_seen": 321847296
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009115747241725175,
      "loss": 1.6649,
      "theoretical_loss": 4.140925646811714,
      "tokens_seen": 321912832
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000911554663991976,
      "loss": 1.7697,
      "theoretical_loss": 4.140828550870793,
      "tokens_seen": 321978368
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009115346038114343,
      "loss": 1.9109,
      "theoretical_loss": 4.140731480223303,
      "tokens_seen": 322043904
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009115145436308927,
      "loss": 1.7703,
      "theoretical_loss": 4.140634434857506,
      "tokens_seen": 322109440
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000911494483450351,
      "loss": 1.7073,
      "theoretical_loss": 4.1405374147616785,
      "tokens_seen": 322174976
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009114744232698094,
      "loss": 1.7863,
      "theoretical_loss": 4.140440419924101,
      "tokens_seen": 322240512
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009114543630892678,
      "loss": 1.8862,
      "theoretical_loss": 4.140343450333064,
      "tokens_seen": 322306048
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009114343029087262,
      "loss": 1.5811,
      "theoretical_loss": 4.1402465059768625,
      "tokens_seen": 322371584
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009114142427281845,
      "loss": 1.7835,
      "theoretical_loss": 4.140149586843803,
      "tokens_seen": 322437120
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000911394182547643,
      "loss": 1.7408,
      "theoretical_loss": 4.140052692922199,
      "tokens_seen": 322502656
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009113741223671014,
      "loss": 1.8122,
      "theoretical_loss": 4.139955824200369,
      "tokens_seen": 322568192
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009113540621865597,
      "loss": 1.8127,
      "theoretical_loss": 4.1398589806666415,
      "tokens_seen": 322633728
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009113340020060182,
      "loss": 1.7441,
      "theoretical_loss": 4.139762162309355,
      "tokens_seen": 322699264
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009113139418254764,
      "loss": 1.8992,
      "theoretical_loss": 4.13966536911685,
      "tokens_seen": 322764800
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009112938816449349,
      "loss": 1.7507,
      "theoretical_loss": 4.139568601077481,
      "tokens_seen": 322830336
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009112738214643932,
      "loss": 1.7393,
      "theoretical_loss": 4.139471858179606,
      "tokens_seen": 322895872
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009112537612838516,
      "loss": 1.6676,
      "theoretical_loss": 4.139375140411592,
      "tokens_seen": 322961408
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00091123370110331,
      "loss": 1.7302,
      "theoretical_loss": 4.139278447761814,
      "tokens_seen": 323026944
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009112136409227683,
      "loss": 1.6805,
      "theoretical_loss": 4.139181780218656,
      "tokens_seen": 323092480
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009111935807422267,
      "loss": 1.6964,
      "theoretical_loss": 4.139085137770507,
      "tokens_seen": 323158016
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009111735205616852,
      "loss": 1.7933,
      "theoretical_loss": 4.138988520405764,
      "tokens_seen": 323223552
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009111534603811434,
      "loss": 1.6671,
      "theoretical_loss": 4.138891928112835,
      "tokens_seen": 323289088
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009111334002006019,
      "loss": 1.8928,
      "theoretical_loss": 4.138795360880132,
      "tokens_seen": 323354624
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009111133400200601,
      "loss": 1.6507,
      "theoretical_loss": 4.138698818696076,
      "tokens_seen": 323420160
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009110932798395186,
      "loss": 1.7278,
      "theoretical_loss": 4.138602301549097,
      "tokens_seen": 323485696
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000911073219658977,
      "loss": 1.7415,
      "theoretical_loss": 4.138505809427632,
      "tokens_seen": 323551232
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009110531594784353,
      "loss": 1.8598,
      "theoretical_loss": 4.138409342320124,
      "tokens_seen": 323616768
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009110330992978937,
      "loss": 1.5982,
      "theoretical_loss": 4.138312900215024,
      "tokens_seen": 323682304
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009110130391173522,
      "loss": 1.7677,
      "theoretical_loss": 4.138216483100795,
      "tokens_seen": 323747840
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009109929789368104,
      "loss": 1.6368,
      "theoretical_loss": 4.1381200909659,
      "tokens_seen": 323813376
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009109729187562689,
      "loss": 1.7068,
      "theoretical_loss": 4.138023723798817,
      "tokens_seen": 323878912
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009109528585757271,
      "loss": 1.7321,
      "theoretical_loss": 4.137927381588026,
      "tokens_seen": 323944448
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009109327983951856,
      "loss": 1.7413,
      "theoretical_loss": 4.137831064322021,
      "tokens_seen": 324009984
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000910912738214644,
      "loss": 1.8037,
      "theoretical_loss": 4.1377347719892965,
      "tokens_seen": 324075520
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009108926780341023,
      "loss": 1.6356,
      "theoretical_loss": 4.137638504578359,
      "tokens_seen": 324141056
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009108726178535607,
      "loss": 1.7395,
      "theoretical_loss": 4.137542262077721,
      "tokens_seen": 324206592
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009108525576730191,
      "loss": 1.6488,
      "theoretical_loss": 4.1374460444759045,
      "tokens_seen": 324272128
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009108324974924774,
      "loss": 1.7526,
      "theoretical_loss": 4.137349851761437,
      "tokens_seen": 324337664
    },
    {
      "epoch": 0.1,
      "objective/train/advantage_avg": 0.49557074904441833,
      "objective/train/docs_used": 124782,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.604373812675476,
      "objective/train/original_loss": 1.6043736934661865,
      "objective/train/theoretical_loss": 4.137253683922854,
      "objective/train/tokens_used": 344863200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2460704892873764,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508081912994385,
      "objective/train/weighted_lm_loss": 1.6858175992965698,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 4.137253683922854,
      "tokens_seen": 324403200
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009108124373119359,
      "loss": 1.7217,
      "theoretical_loss": 4.137253683922854,
      "tokens_seen": 324403200
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009107923771313942,
      "loss": 1.7176,
      "theoretical_loss": 4.137157540948699,
      "tokens_seen": 324468736
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009107723169508526,
      "loss": 1.7777,
      "theoretical_loss": 4.137061422827525,
      "tokens_seen": 324534272
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009107522567703109,
      "loss": 1.817,
      "theoretical_loss": 4.136965329547888,
      "tokens_seen": 324599808
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009107321965897693,
      "loss": 1.6928,
      "theoretical_loss": 4.136869261098356,
      "tokens_seen": 324665344
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009107121364092277,
      "loss": 1.7311,
      "theoretical_loss": 4.136773217467502,
      "tokens_seen": 324730880
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009106920762286861,
      "loss": 1.863,
      "theoretical_loss": 4.136677198643908,
      "tokens_seen": 324796416
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009106720160481444,
      "loss": 1.7369,
      "theoretical_loss": 4.136581204616162,
      "tokens_seen": 324861952
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009106519558676029,
      "loss": 1.7076,
      "theoretical_loss": 4.136485235372861,
      "tokens_seen": 324927488
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009106318956870612,
      "loss": 1.7839,
      "theoretical_loss": 4.13638929090261,
      "tokens_seen": 324993024
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009106118355065196,
      "loss": 1.6871,
      "theoretical_loss": 4.13629337119402,
      "tokens_seen": 325058560
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009105917753259779,
      "loss": 1.6558,
      "theoretical_loss": 4.136197476235708,
      "tokens_seen": 325124096
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009105717151454364,
      "loss": 1.8245,
      "theoretical_loss": 4.136101606016304,
      "tokens_seen": 325189632
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009105516549648947,
      "loss": 1.7017,
      "theoretical_loss": 4.136005760524439,
      "tokens_seen": 325255168
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009105315947843531,
      "loss": 1.7789,
      "theoretical_loss": 4.135909939748757,
      "tokens_seen": 325320704
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009105115346038114,
      "loss": 1.7688,
      "theoretical_loss": 4.135814143677906,
      "tokens_seen": 325386240
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009104914744232698,
      "loss": 1.7062,
      "theoretical_loss": 4.135718372300543,
      "tokens_seen": 325451776
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009104714142427282,
      "loss": 1.7736,
      "theoretical_loss": 4.135622625605333,
      "tokens_seen": 325517312
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009104513540621866,
      "loss": 1.698,
      "theoretical_loss": 4.135526903580946,
      "tokens_seen": 325582848
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009104312938816449,
      "loss": 1.8322,
      "theoretical_loss": 4.135431206216061,
      "tokens_seen": 325648384
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009104112337011034,
      "loss": 1.7194,
      "theoretical_loss": 4.135335533499366,
      "tokens_seen": 325713920
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009103911735205616,
      "loss": 1.7395,
      "theoretical_loss": 4.135239885419554,
      "tokens_seen": 325779456
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009103711133400201,
      "loss": 1.7968,
      "theoretical_loss": 4.135144261965327,
      "tokens_seen": 325844992
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009103510531594784,
      "loss": 1.666,
      "theoretical_loss": 4.135048663125394,
      "tokens_seen": 325910528
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009103309929789368,
      "loss": 1.8125,
      "theoretical_loss": 4.134953088888471,
      "tokens_seen": 325976064
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009103109327983952,
      "loss": 1.8514,
      "theoretical_loss": 4.134857539243283,
      "tokens_seen": 326041600
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009102908726178536,
      "loss": 1.6628,
      "theoretical_loss": 4.134762014178559,
      "tokens_seen": 326107136
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009102708124373119,
      "loss": 1.8064,
      "theoretical_loss": 4.13466651368304,
      "tokens_seen": 326172672
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009102507522567704,
      "loss": 1.748,
      "theoretical_loss": 4.134571037745471,
      "tokens_seen": 326238208
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009102306920762286,
      "loss": 1.6509,
      "theoretical_loss": 4.134475586354606,
      "tokens_seen": 326303744
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009102106318956871,
      "loss": 1.7845,
      "theoretical_loss": 4.134380159499204,
      "tokens_seen": 326369280
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009101905717151455,
      "loss": 1.783,
      "theoretical_loss": 4.134284757168037,
      "tokens_seen": 326434816
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009101705115346038,
      "loss": 1.6616,
      "theoretical_loss": 4.134189379349877,
      "tokens_seen": 326500352
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009101504513540622,
      "loss": 1.7684,
      "theoretical_loss": 4.13409402603351,
      "tokens_seen": 326565888
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009101303911735205,
      "loss": 1.8304,
      "theoretical_loss": 4.1339986972077245,
      "tokens_seen": 326631424
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009101103309929789,
      "loss": 1.6475,
      "theoretical_loss": 4.13390339286132,
      "tokens_seen": 326696960
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009100902708124374,
      "loss": 1.7013,
      "theoretical_loss": 4.1338081129831,
      "tokens_seen": 326762496
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009100702106318956,
      "loss": 1.7019,
      "theoretical_loss": 4.1337128575618785,
      "tokens_seen": 326828032
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009100501504513541,
      "loss": 1.6178,
      "theoretical_loss": 4.133617626586475,
      "tokens_seen": 326893568
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009100300902708125,
      "loss": 1.6345,
      "theoretical_loss": 4.1335224200457175,
      "tokens_seen": 326959104
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009100100300902708,
      "loss": 1.7228,
      "theoretical_loss": 4.133427237928439,
      "tokens_seen": 327024640
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009099899699097292,
      "loss": 1.7379,
      "theoretical_loss": 4.133332080223483,
      "tokens_seen": 327090176
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009099699097291875,
      "loss": 1.8473,
      "theoretical_loss": 4.133236946919698,
      "tokens_seen": 327155712
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009099498495486459,
      "loss": 2.0025,
      "theoretical_loss": 4.133141838005942,
      "tokens_seen": 327221248
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009099297893681044,
      "loss": 1.7865,
      "theoretical_loss": 4.133046753471077,
      "tokens_seen": 327286784
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009099097291875627,
      "loss": 1.7868,
      "theoretical_loss": 4.132951693303976,
      "tokens_seen": 327352320
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009098896690070211,
      "loss": 1.7852,
      "theoretical_loss": 4.132856657493516,
      "tokens_seen": 327417856
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009098696088264795,
      "loss": 1.8734,
      "theoretical_loss": 4.132761646028585,
      "tokens_seen": 327483392
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009098495486459378,
      "loss": 1.7436,
      "theoretical_loss": 4.132666658898074,
      "tokens_seen": 327548928
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009098294884653963,
      "loss": 1.7704,
      "theoretical_loss": 4.132571696090885,
      "tokens_seen": 327614464
    },
    {
      "debugging/Compilability": 1.0,
      "debugging/distinct-1-grams": 0.7246169962973871,
      "debugging/entropy-1-grams": 5.13190351820591,
      "debugging/length": 472.8,
      "debugging/num_segments": 10,
      "debugging/raw_token_scores_avg": 0.008663635700941086,
      "debugging/raw_token_scores_std": 0.023182114586234093,
      "debugging/score": 0.0033104506818385365,
      "debugging/score_std": 0.0017105448100298692,
      "epoch": 0.1,
      "objective/train/advantage_avg": 0.4913388192653656,
      "objective/train/docs_used": 126013,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.739636778831482,
      "objective/train/original_loss": 1.7396366596221924,
      "objective/train/theoretical_loss": 4.132476757595925,
      "objective/train/tokens_used": 348140000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24195149540901184,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503638982772827,
      "objective/train/weighted_lm_loss": 1.8274545669555664,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.132476757595925,
      "tokens_seen": 327680000
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009098094282848546,
      "loss": 1.8537,
      "theoretical_loss": 4.132476757595925,
      "tokens_seen": 327680000
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000909789368104313,
      "loss": 1.7564,
      "theoretical_loss": 4.13238184340211,
      "tokens_seen": 327745536
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009097693079237713,
      "loss": 1.8456,
      "theoretical_loss": 4.13228695349836,
      "tokens_seen": 327811072
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009097492477432297,
      "loss": 1.836,
      "theoretical_loss": 4.132192087873608,
      "tokens_seen": 327876608
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009097291875626881,
      "loss": 1.7369,
      "theoretical_loss": 4.132097246516788,
      "tokens_seen": 327942144
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009097091273821465,
      "loss": 1.8652,
      "theoretical_loss": 4.132002429416845,
      "tokens_seen": 328007680
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009096890672016048,
      "loss": 1.7203,
      "theoretical_loss": 4.131907636562731,
      "tokens_seen": 328073216
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009096690070210633,
      "loss": 1.8587,
      "theoretical_loss": 4.131812867943403,
      "tokens_seen": 328138752
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009096489468405216,
      "loss": 1.7171,
      "theoretical_loss": 4.131718123547829,
      "tokens_seen": 328204288
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00090962888665998,
      "loss": 1.6613,
      "theoretical_loss": 4.131623403364981,
      "tokens_seen": 328269824
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009096088264794383,
      "loss": 1.9833,
      "theoretical_loss": 4.131528707383838,
      "tokens_seen": 328335360
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009095887662988968,
      "loss": 1.7528,
      "theoretical_loss": 4.131434035593388,
      "tokens_seen": 328400896
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009095687061183551,
      "loss": 1.6557,
      "theoretical_loss": 4.131339387982628,
      "tokens_seen": 328466432
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009095486459378135,
      "loss": 1.6133,
      "theoretical_loss": 4.131244764540557,
      "tokens_seen": 328531968
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009095285857572718,
      "loss": 1.7815,
      "theoretical_loss": 4.1311501652561855,
      "tokens_seen": 328597504
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009095085255767302,
      "loss": 1.909,
      "theoretical_loss": 4.131055590118528,
      "tokens_seen": 328663040
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009094884653961886,
      "loss": 1.6056,
      "theoretical_loss": 4.1309610391166105,
      "tokens_seen": 328728576
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000909468405215647,
      "loss": 1.7617,
      "theoretical_loss": 4.1308665122394626,
      "tokens_seen": 328794112
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009094483450351053,
      "loss": 1.6625,
      "theoretical_loss": 4.1307720094761216,
      "tokens_seen": 328859648
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009094282848545638,
      "loss": 1.6576,
      "theoretical_loss": 4.130677530815632,
      "tokens_seen": 328925184
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000909408224674022,
      "loss": 1.7591,
      "theoretical_loss": 4.1305830762470475,
      "tokens_seen": 328990720
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009093881644934805,
      "loss": 1.8302,
      "theoretical_loss": 4.130488645759426,
      "tokens_seen": 329056256
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009093681043129388,
      "loss": 1.7202,
      "theoretical_loss": 4.130394239341835,
      "tokens_seen": 329121792
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009093480441323972,
      "loss": 1.7775,
      "theoretical_loss": 4.1302998569833465,
      "tokens_seen": 329187328
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009093279839518556,
      "loss": 1.7116,
      "theoretical_loss": 4.1302054986730425,
      "tokens_seen": 329252864
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000909307923771314,
      "loss": 1.7708,
      "theoretical_loss": 4.130111164400011,
      "tokens_seen": 329318400
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009092878635907723,
      "loss": 1.7346,
      "theoretical_loss": 4.130016854153346,
      "tokens_seen": 329383936
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009092678034102308,
      "loss": 1.7316,
      "theoretical_loss": 4.129922567922149,
      "tokens_seen": 329449472
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000909247743229689,
      "loss": 1.7205,
      "theoretical_loss": 4.129828305695531,
      "tokens_seen": 329515008
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009092276830491475,
      "loss": 1.688,
      "theoretical_loss": 4.129734067462607,
      "tokens_seen": 329580544
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009092076228686059,
      "loss": 1.7054,
      "theoretical_loss": 4.129639853212501,
      "tokens_seen": 329646080
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009091875626880642,
      "loss": 1.8751,
      "theoretical_loss": 4.129545662934341,
      "tokens_seen": 329711616
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009091675025075226,
      "loss": 1.7017,
      "theoretical_loss": 4.129451496617269,
      "tokens_seen": 329777152
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009091474423269809,
      "loss": 1.7523,
      "theoretical_loss": 4.129357354250426,
      "tokens_seen": 329842688
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009091273821464393,
      "loss": 1.7279,
      "theoretical_loss": 4.1292632358229655,
      "tokens_seen": 329908224
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009091073219658978,
      "loss": 1.7621,
      "theoretical_loss": 4.129169141324045,
      "tokens_seen": 329973760
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000909087261785356,
      "loss": 1.8877,
      "theoretical_loss": 4.129075070742831,
      "tokens_seen": 330039296
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009090672016048145,
      "loss": 1.7733,
      "theoretical_loss": 4.128981024068497,
      "tokens_seen": 330104832
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009090471414242729,
      "loss": 1.7082,
      "theoretical_loss": 4.128887001290222,
      "tokens_seen": 330170368
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009090270812437312,
      "loss": 1.761,
      "theoretical_loss": 4.1287930023971935,
      "tokens_seen": 330235904
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009090070210631896,
      "loss": 1.83,
      "theoretical_loss": 4.128699027378604,
      "tokens_seen": 330301440
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009089869608826479,
      "loss": 1.8233,
      "theoretical_loss": 4.128605076223657,
      "tokens_seen": 330366976
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009089669007021063,
      "loss": 1.6636,
      "theoretical_loss": 4.128511148921558,
      "tokens_seen": 330432512
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009089468405215648,
      "loss": 1.7057,
      "theoretical_loss": 4.128417245461525,
      "tokens_seen": 330498048
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000908926780341023,
      "loss": 1.8225,
      "theoretical_loss": 4.128323365832777,
      "tokens_seen": 330563584
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009089067201604815,
      "loss": 1.8282,
      "theoretical_loss": 4.1282295100245445,
      "tokens_seen": 330629120
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009088866599799398,
      "loss": 1.8193,
      "theoretical_loss": 4.128135678026064,
      "tokens_seen": 330694656
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009088665997993982,
      "loss": 1.7495,
      "theoretical_loss": 4.128041869826578,
      "tokens_seen": 330760192
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009088465396188566,
      "loss": 1.7241,
      "theoretical_loss": 4.127948085415338,
      "tokens_seen": 330825728
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000908826479438315,
      "loss": 1.8882,
      "theoretical_loss": 4.127854324781599,
      "tokens_seen": 330891264
    },
    {
      "epoch": 0.1,
      "objective/train/advantage_avg": 0.4939263164997101,
      "objective/train/docs_used": 127130,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7381356954574585,
      "objective/train/original_loss": 1.738135576248169,
      "objective/train/theoretical_loss": 4.127760587914626,
      "objective/train/tokens_used": 351416800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24445585906505585,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506354570388794,
      "objective/train/weighted_lm_loss": 1.8260606527328491,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.127760587914626,
      "tokens_seen": 330956800
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009088064192577733,
      "loss": 1.8542,
      "theoretical_loss": 4.127760587914626,
      "tokens_seen": 330956800
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009087863590772317,
      "loss": 1.715,
      "theoretical_loss": 4.12766687480369,
      "tokens_seen": 331022336
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00090876629889669,
      "loss": 1.6645,
      "theoretical_loss": 4.127573185438068,
      "tokens_seen": 331087872
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009087462387161485,
      "loss": 1.9245,
      "theoretical_loss": 4.127479519807047,
      "tokens_seen": 331153408
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009087261785356068,
      "loss": 1.8333,
      "theoretical_loss": 4.1273858778999175,
      "tokens_seen": 331218944
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009087061183550652,
      "loss": 1.7053,
      "theoretical_loss": 4.127292259705978,
      "tokens_seen": 331284480
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009086860581745236,
      "loss": 1.8399,
      "theoretical_loss": 4.127198665214536,
      "tokens_seen": 331350016
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000908665997993982,
      "loss": 1.7369,
      "theoretical_loss": 4.127105094414903,
      "tokens_seen": 331415552
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009086459378134403,
      "loss": 1.7215,
      "theoretical_loss": 4.127011547296398,
      "tokens_seen": 331481088
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009086258776328987,
      "loss": 1.6885,
      "theoretical_loss": 4.126918023848349,
      "tokens_seen": 331546624
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000908605817452357,
      "loss": 1.6354,
      "theoretical_loss": 4.126824524060088,
      "tokens_seen": 331612160
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009085857572718155,
      "loss": 1.7601,
      "theoretical_loss": 4.1267310479209565,
      "tokens_seen": 331677696
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009085656970912738,
      "loss": 1.7374,
      "theoretical_loss": 4.126637595420302,
      "tokens_seen": 331743232
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009085456369107322,
      "loss": 1.75,
      "theoretical_loss": 4.126544166547479,
      "tokens_seen": 331808768
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009085255767301905,
      "loss": 1.7956,
      "theoretical_loss": 4.126450761291847,
      "tokens_seen": 331874304
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000908505516549649,
      "loss": 1.7369,
      "theoretical_loss": 4.126357379642775,
      "tokens_seen": 331939840
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009084854563691073,
      "loss": 1.8566,
      "theoretical_loss": 4.126264021589638,
      "tokens_seen": 332005376
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009084653961885657,
      "loss": 1.7816,
      "theoretical_loss": 4.126170687121817,
      "tokens_seen": 332070912
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000908445336008024,
      "loss": 1.8452,
      "theoretical_loss": 4.126077376228702,
      "tokens_seen": 332136448
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009084252758274824,
      "loss": 1.6802,
      "theoretical_loss": 4.125984088899688,
      "tokens_seen": 332201984
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009084052156469409,
      "loss": 1.64,
      "theoretical_loss": 4.125890825124177,
      "tokens_seen": 332267520
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009083851554663992,
      "loss": 1.6285,
      "theoretical_loss": 4.125797584891578,
      "tokens_seen": 332333056
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009083650952858576,
      "loss": 1.699,
      "theoretical_loss": 4.1257043681913075,
      "tokens_seen": 332398592
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000908345035105316,
      "loss": 1.6606,
      "theoretical_loss": 4.125611175012789,
      "tokens_seen": 332464128
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009083249749247744,
      "loss": 1.6762,
      "theoretical_loss": 4.125518005345452,
      "tokens_seen": 332529664
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009083049147442327,
      "loss": 1.8607,
      "theoretical_loss": 4.1254248591787315,
      "tokens_seen": 332595200
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009082848545636912,
      "loss": 1.7981,
      "theoretical_loss": 4.125331736502073,
      "tokens_seen": 332660736
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009082647943831494,
      "loss": 1.8264,
      "theoretical_loss": 4.125238637304925,
      "tokens_seen": 332726272
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009082447342026079,
      "loss": 1.8123,
      "theoretical_loss": 4.125145561576746,
      "tokens_seen": 332791808
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009082246740220663,
      "loss": 1.8439,
      "theoretical_loss": 4.125052509306999,
      "tokens_seen": 332857344
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009082046138415246,
      "loss": 1.6351,
      "theoretical_loss": 4.124959480485156,
      "tokens_seen": 332922880
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000908184553660983,
      "loss": 1.7231,
      "theoretical_loss": 4.124866475100693,
      "tokens_seen": 332988416
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009081644934804413,
      "loss": 1.7502,
      "theoretical_loss": 4.124773493143094,
      "tokens_seen": 333053952
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009081444332998997,
      "loss": 1.7452,
      "theoretical_loss": 4.124680534601851,
      "tokens_seen": 333119488
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009081243731193582,
      "loss": 1.6137,
      "theoretical_loss": 4.124587599466462,
      "tokens_seen": 333185024
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009081043129388164,
      "loss": 1.7891,
      "theoretical_loss": 4.124494687726431,
      "tokens_seen": 333250560
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009080842527582749,
      "loss": 1.6966,
      "theoretical_loss": 4.124401799371269,
      "tokens_seen": 333316096
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009080641925777333,
      "loss": 1.8163,
      "theoretical_loss": 4.124308934390496,
      "tokens_seen": 333381632
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009080441323971916,
      "loss": 1.8086,
      "theoretical_loss": 4.124216092773635,
      "tokens_seen": 333447168
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00090802407221665,
      "loss": 1.712,
      "theoretical_loss": 4.124123274510217,
      "tokens_seen": 333512704
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009080040120361083,
      "loss": 1.6625,
      "theoretical_loss": 4.124030479589782,
      "tokens_seen": 333578240
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009079839518555667,
      "loss": 1.6341,
      "theoretical_loss": 4.1239377080018755,
      "tokens_seen": 333643776
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009079638916750252,
      "loss": 1.7989,
      "theoretical_loss": 4.123844959736049,
      "tokens_seen": 333709312
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009079438314944834,
      "loss": 1.7309,
      "theoretical_loss": 4.123752234781859,
      "tokens_seen": 333774848
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009079237713139419,
      "loss": 1.8247,
      "theoretical_loss": 4.1236595331288735,
      "tokens_seen": 333840384
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009079037111334002,
      "loss": 1.6904,
      "theoretical_loss": 4.123566854766663,
      "tokens_seen": 333905920
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009078836509528586,
      "loss": 1.6505,
      "theoretical_loss": 4.123474199684807,
      "tokens_seen": 333971456
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000907863590772317,
      "loss": 1.8607,
      "theoretical_loss": 4.123381567872891,
      "tokens_seen": 334036992
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009078435305917754,
      "loss": 1.8399,
      "theoretical_loss": 4.123288959320506,
      "tokens_seen": 334102528
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009078234704112337,
      "loss": 1.763,
      "theoretical_loss": 4.123196374017253,
      "tokens_seen": 334168064
    },
    {
      "epoch": 0.1,
      "objective/train/advantage_avg": 0.4906438887119293,
      "objective/train/docs_used": 128373,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.830473780632019,
      "objective/train/original_loss": 1.8304736614227295,
      "objective/train/theoretical_loss": 4.123103811952736,
      "objective/train/tokens_used": 354693600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24106791615486145,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502898693084717,
      "objective/train/weighted_lm_loss": 1.9224073886871338,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.123103811952736,
      "tokens_seen": 334233600
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009078034102306921,
      "loss": 1.8137,
      "theoretical_loss": 4.123103811952736,
      "tokens_seen": 334233600
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009077833500501504,
      "loss": 1.6342,
      "theoretical_loss": 4.123011273116566,
      "tokens_seen": 334299136
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009077632898696089,
      "loss": 1.8001,
      "theoretical_loss": 4.122918757498365,
      "tokens_seen": 334364672
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009077432296890672,
      "loss": 1.6724,
      "theoretical_loss": 4.122826265087756,
      "tokens_seen": 334430208
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009077231695085256,
      "loss": 1.6699,
      "theoretical_loss": 4.122733795874372,
      "tokens_seen": 334495744
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000907703109327984,
      "loss": 1.8302,
      "theoretical_loss": 4.122641349847852,
      "tokens_seen": 334561280
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009076830491474424,
      "loss": 1.6683,
      "theoretical_loss": 4.122548926997842,
      "tokens_seen": 334626816
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009076629889669007,
      "loss": 1.8169,
      "theoretical_loss": 4.122456527313994,
      "tokens_seen": 334692352
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009076429287863591,
      "loss": 1.8627,
      "theoretical_loss": 4.122364150785966,
      "tokens_seen": 334757888
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009076228686058174,
      "loss": 1.7632,
      "theoretical_loss": 4.1222717974034255,
      "tokens_seen": 334823424
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009076028084252759,
      "loss": 1.7083,
      "theoretical_loss": 4.122179467156043,
      "tokens_seen": 334888960
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009075827482447342,
      "loss": 1.8839,
      "theoretical_loss": 4.122087160033496,
      "tokens_seen": 334954496
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009075626880641926,
      "loss": 1.8024,
      "theoretical_loss": 4.1219948760254725,
      "tokens_seen": 335020032
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009075426278836509,
      "loss": 1.6807,
      "theoretical_loss": 4.121902615121663,
      "tokens_seen": 335085568
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009075225677031094,
      "loss": 1.8905,
      "theoretical_loss": 4.121810377311768,
      "tokens_seen": 335151104
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009075025075225677,
      "loss": 1.7158,
      "theoretical_loss": 4.1217181625854895,
      "tokens_seen": 335216640
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009074824473420261,
      "loss": 1.7767,
      "theoretical_loss": 4.121625970932542,
      "tokens_seen": 335282176
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009074623871614845,
      "loss": 1.8351,
      "theoretical_loss": 4.121533802342643,
      "tokens_seen": 335347712
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009074423269809428,
      "loss": 1.6733,
      "theoretical_loss": 4.121441656805517,
      "tokens_seen": 335413248
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009074222668004012,
      "loss": 1.7529,
      "theoretical_loss": 4.121349534310896,
      "tokens_seen": 335478784
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009074022066198596,
      "loss": 1.7908,
      "theoretical_loss": 4.121257434848519,
      "tokens_seen": 335544320
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009073821464393179,
      "loss": 1.7714,
      "theoretical_loss": 4.121165358408129,
      "tokens_seen": 335609856
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009073620862587764,
      "loss": 1.8105,
      "theoretical_loss": 4.121073304979479,
      "tokens_seen": 335675392
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009073420260782347,
      "loss": 1.6567,
      "theoretical_loss": 4.120981274552326,
      "tokens_seen": 335740928
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009073219658976931,
      "loss": 1.793,
      "theoretical_loss": 4.120889267116435,
      "tokens_seen": 335806464
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009073019057171515,
      "loss": 1.7369,
      "theoretical_loss": 4.1207972826615755,
      "tokens_seen": 335872000
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009072818455366098,
      "loss": 1.8146,
      "theoretical_loss": 4.120705321177526,
      "tokens_seen": 335937536
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009072617853560682,
      "loss": 1.7543,
      "theoretical_loss": 4.120613382654072,
      "tokens_seen": 336003072
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009072417251755267,
      "loss": 1.7126,
      "theoretical_loss": 4.1205214670810015,
      "tokens_seen": 336068608
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009072216649949849,
      "loss": 1.7503,
      "theoretical_loss": 4.120429574448114,
      "tokens_seen": 336134144
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009072016048144434,
      "loss": 1.6967,
      "theoretical_loss": 4.120337704745211,
      "tokens_seen": 336199680
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009071815446339016,
      "loss": 1.7868,
      "theoretical_loss": 4.120245857962104,
      "tokens_seen": 336265216
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009071614844533601,
      "loss": 1.7849,
      "theoretical_loss": 4.120154034088609,
      "tokens_seen": 336330752
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009071414242728185,
      "loss": 1.7815,
      "theoretical_loss": 4.120062233114549,
      "tokens_seen": 336396288
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009071213640922768,
      "loss": 1.8475,
      "theoretical_loss": 4.119970455029755,
      "tokens_seen": 336461824
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009071013039117352,
      "loss": 1.6997,
      "theoretical_loss": 4.119878699824063,
      "tokens_seen": 336527360
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009070812437311937,
      "loss": 1.7022,
      "theoretical_loss": 4.119786967487314,
      "tokens_seen": 336592896
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009070611835506519,
      "loss": 1.8374,
      "theoretical_loss": 4.119695258009358,
      "tokens_seen": 336658432
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009070411233701104,
      "loss": 1.7614,
      "theoretical_loss": 4.119603571380052,
      "tokens_seen": 336723968
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009070210631895686,
      "loss": 1.7412,
      "theoretical_loss": 4.119511907589256,
      "tokens_seen": 336789504
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009070010030090271,
      "loss": 1.7961,
      "theoretical_loss": 4.11942026662684,
      "tokens_seen": 336855040
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009069809428284855,
      "loss": 1.7222,
      "theoretical_loss": 4.1193286484826785,
      "tokens_seen": 336920576
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009069608826479438,
      "loss": 1.7334,
      "theoretical_loss": 4.119237053146653,
      "tokens_seen": 336986112
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009069408224674022,
      "loss": 1.7046,
      "theoretical_loss": 4.119145480608653,
      "tokens_seen": 337051648
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009069207622868606,
      "loss": 1.6898,
      "theoretical_loss": 4.11905393085857,
      "tokens_seen": 337117184
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000906900702106319,
      "loss": 1.8415,
      "theoretical_loss": 4.118962403886306,
      "tokens_seen": 337182720
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009068806419257774,
      "loss": 1.7412,
      "theoretical_loss": 4.11887089968177,
      "tokens_seen": 337248256
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009068605817452358,
      "loss": 1.8208,
      "theoretical_loss": 4.118779418234874,
      "tokens_seen": 337313792
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009068405215646941,
      "loss": 1.6441,
      "theoretical_loss": 4.118687959535539,
      "tokens_seen": 337379328
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009068204613841525,
      "loss": 1.8413,
      "theoretical_loss": 4.118596523573689,
      "tokens_seen": 337444864
    },
    {
      "epoch": 0.1,
      "objective/train/advantage_avg": 0.49176985025405884,
      "objective/train/docs_used": 129675,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5919711589813232,
      "objective/train/original_loss": 1.5919709205627441,
      "objective/train/theoretical_loss": 4.118505110339262,
      "objective/train/tokens_used": 357970400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24228759109973907,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504087209701538,
      "objective/train/weighted_lm_loss": 1.6722182035446167,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.118505110339262,
      "tokens_seen": 337510400
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009068004012036108,
      "loss": 1.6432,
      "theoretical_loss": 4.118505110339262,
      "tokens_seen": 337510400
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009067803410230693,
      "loss": 1.7769,
      "theoretical_loss": 4.118413719822192,
      "tokens_seen": 337575936
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009067602808425276,
      "loss": 1.6643,
      "theoretical_loss": 4.118322352012429,
      "tokens_seen": 337641472
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000906740220661986,
      "loss": 1.8076,
      "theoretical_loss": 4.118231006899924,
      "tokens_seen": 337707008
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009067201604814444,
      "loss": 1.6354,
      "theoretical_loss": 4.118139684474634,
      "tokens_seen": 337772544
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009067001003009028,
      "loss": 1.6821,
      "theoretical_loss": 4.118048384726525,
      "tokens_seen": 337838080
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009066800401203611,
      "loss": 1.7323,
      "theoretical_loss": 4.117957107645569,
      "tokens_seen": 337903616
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009066599799398195,
      "loss": 1.8233,
      "theoretical_loss": 4.117865853221743,
      "tokens_seen": 337969152
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009066399197592778,
      "loss": 1.7854,
      "theoretical_loss": 4.117774621445031,
      "tokens_seen": 338034688
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009066198595787363,
      "loss": 1.6559,
      "theoretical_loss": 4.117683412305425,
      "tokens_seen": 338100224
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009065997993981946,
      "loss": 1.807,
      "theoretical_loss": 4.1175922257929205,
      "tokens_seen": 338165760
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000906579739217653,
      "loss": 1.664,
      "theoretical_loss": 4.117501061897521,
      "tokens_seen": 338231296
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009065596790371113,
      "loss": 1.7741,
      "theoretical_loss": 4.117409920609235,
      "tokens_seen": 338296832
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009065396188565698,
      "loss": 1.9005,
      "theoretical_loss": 4.11731880191808,
      "tokens_seen": 338362368
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009065195586760281,
      "loss": 1.8514,
      "theoretical_loss": 4.117227705814078,
      "tokens_seen": 338427904
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009064994984954865,
      "loss": 1.8802,
      "theoretical_loss": 4.117136632287258,
      "tokens_seen": 338493440
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009064794383149449,
      "loss": 1.7452,
      "theoretical_loss": 4.117045581327653,
      "tokens_seen": 338558976
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009064593781344032,
      "loss": 1.8935,
      "theoretical_loss": 4.116954552925305,
      "tokens_seen": 338624512
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009064393179538616,
      "loss": 1.9351,
      "theoretical_loss": 4.116863547070264,
      "tokens_seen": 338690048
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00090641925777332,
      "loss": 1.8515,
      "theoretical_loss": 4.116772563752581,
      "tokens_seen": 338755584
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009063991975927783,
      "loss": 1.7043,
      "theoretical_loss": 4.116681602962317,
      "tokens_seen": 338821120
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009063791374122368,
      "loss": 1.7002,
      "theoretical_loss": 4.116590664689539,
      "tokens_seen": 338886656
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009063590772316951,
      "loss": 1.8262,
      "theoretical_loss": 4.116499748924319,
      "tokens_seen": 338952192
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009063390170511535,
      "loss": 1.768,
      "theoretical_loss": 4.116408855656737,
      "tokens_seen": 339017728
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009063189568706119,
      "loss": 1.7185,
      "theoretical_loss": 4.116317984876878,
      "tokens_seen": 339083264
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009062988966900702,
      "loss": 1.7734,
      "theoretical_loss": 4.1162271365748335,
      "tokens_seen": 339148800
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009062788365095286,
      "loss": 1.8295,
      "theoretical_loss": 4.116136310740702,
      "tokens_seen": 339214336
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000906258776328987,
      "loss": 1.6748,
      "theoretical_loss": 4.1160455073645865,
      "tokens_seen": 339279872
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009062387161484453,
      "loss": 1.5677,
      "theoretical_loss": 4.115954726436599,
      "tokens_seen": 339345408
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009062186559679038,
      "loss": 1.6308,
      "theoretical_loss": 4.115863967946856,
      "tokens_seen": 339410944
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000906198595787362,
      "loss": 1.7707,
      "theoretical_loss": 4.115773231885479,
      "tokens_seen": 339476480
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009061785356068205,
      "loss": 1.7975,
      "theoretical_loss": 4.115682518242599,
      "tokens_seen": 339542016
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009061584754262789,
      "loss": 1.6861,
      "theoretical_loss": 4.115591827008351,
      "tokens_seen": 339607552
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009061384152457372,
      "loss": 1.6768,
      "theoretical_loss": 4.115501158172876,
      "tokens_seen": 339673088
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009061183550651956,
      "loss": 1.7246,
      "theoretical_loss": 4.115410511726323,
      "tokens_seen": 339738624
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009060982948846541,
      "loss": 1.8952,
      "theoretical_loss": 4.115319887658846,
      "tokens_seen": 339804160
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009060782347041123,
      "loss": 1.6856,
      "theoretical_loss": 4.1152292859606066,
      "tokens_seen": 339869696
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009060581745235708,
      "loss": 1.6744,
      "theoretical_loss": 4.115138706621769,
      "tokens_seen": 339935232
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000906038114343029,
      "loss": 1.7339,
      "theoretical_loss": 4.115048149632507,
      "tokens_seen": 340000768
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009060180541624875,
      "loss": 1.8115,
      "theoretical_loss": 4.114957614983001,
      "tokens_seen": 340066304
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009059979939819459,
      "loss": 1.6957,
      "theoretical_loss": 4.114867102663435,
      "tokens_seen": 340131840
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009059779338014042,
      "loss": 1.8678,
      "theoretical_loss": 4.114776612664,
      "tokens_seen": 340197376
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009059578736208626,
      "loss": 1.7384,
      "theoretical_loss": 4.114686144974897,
      "tokens_seen": 340262912
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000905937813440321,
      "loss": 1.7461,
      "theoretical_loss": 4.114595699586326,
      "tokens_seen": 340328448
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009059177532597793,
      "loss": 1.7942,
      "theoretical_loss": 4.114505276488499,
      "tokens_seen": 340393984
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009058976930792378,
      "loss": 1.7345,
      "theoretical_loss": 4.114414875671631,
      "tokens_seen": 340459520
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000905877632898696,
      "loss": 1.7869,
      "theoretical_loss": 4.114324497125947,
      "tokens_seen": 340525056
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009058575727181545,
      "loss": 1.6033,
      "theoretical_loss": 4.114234140841673,
      "tokens_seen": 340590592
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009058375125376128,
      "loss": 1.8136,
      "theoretical_loss": 4.114143806809045,
      "tokens_seen": 340656128
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009058174523570712,
      "loss": 1.8029,
      "theoretical_loss": 4.114053495018304,
      "tokens_seen": 340721664
    },
    {
      "epoch": 0.1,
      "objective/train/advantage_avg": 0.49362632632255554,
      "objective/train/docs_used": 130755,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7257463932037354,
      "objective/train/original_loss": 1.7257463932037354,
      "objective/train/theoretical_loss": 4.113963205459697,
      "objective/train/tokens_used": 361247200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24409334361553192,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506036281585693,
      "objective/train/weighted_lm_loss": 1.8130598068237305,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.113963205459697,
      "tokens_seen": 340787200
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009057973921765296,
      "loss": 1.7073,
      "theoretical_loss": 4.113963205459697,
      "tokens_seen": 340787200
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000905777331995988,
      "loss": 1.6982,
      "theoretical_loss": 4.113872938123477,
      "tokens_seen": 340852736
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009057572718154463,
      "loss": 1.8377,
      "theoretical_loss": 4.113782692999903,
      "tokens_seen": 340918272
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009057372116349048,
      "loss": 1.5813,
      "theoretical_loss": 4.1136924700792425,
      "tokens_seen": 340983808
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000905717151454363,
      "loss": 1.8254,
      "theoretical_loss": 4.113602269351765,
      "tokens_seen": 341049344
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009056970912738215,
      "loss": 1.7014,
      "theoretical_loss": 4.113512090807749,
      "tokens_seen": 341114880
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009056770310932798,
      "loss": 1.6438,
      "theoretical_loss": 4.113421934437479,
      "tokens_seen": 341180416
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009056569709127382,
      "loss": 1.8645,
      "theoretical_loss": 4.1133318002312445,
      "tokens_seen": 341245952
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009056369107321966,
      "loss": 1.6754,
      "theoretical_loss": 4.113241688179341,
      "tokens_seen": 341311488
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000905616850551655,
      "loss": 1.8817,
      "theoretical_loss": 4.113151598272072,
      "tokens_seen": 341377024
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009055967903711133,
      "loss": 1.8039,
      "theoretical_loss": 4.113061530499746,
      "tokens_seen": 341442560
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009055767301905717,
      "loss": 1.6546,
      "theoretical_loss": 4.112971484852677,
      "tokens_seen": 341508096
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009055566700100301,
      "loss": 1.7393,
      "theoretical_loss": 4.1128814613211855,
      "tokens_seen": 341573632
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009055366098294885,
      "loss": 1.7555,
      "theoretical_loss": 4.112791459895599,
      "tokens_seen": 341639168
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009055165496489468,
      "loss": 1.6987,
      "theoretical_loss": 4.112701480566248,
      "tokens_seen": 341704704
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009054964894684053,
      "loss": 1.7557,
      "theoretical_loss": 4.112611523323474,
      "tokens_seen": 341770240
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009054764292878635,
      "loss": 1.6285,
      "theoretical_loss": 4.1125215881576205,
      "tokens_seen": 341835776
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000905456369107322,
      "loss": 1.7713,
      "theoretical_loss": 4.112431675059039,
      "tokens_seen": 341901312
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009054363089267803,
      "loss": 1.7883,
      "theoretical_loss": 4.112341784018086,
      "tokens_seen": 341966848
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009054162487462387,
      "loss": 1.552,
      "theoretical_loss": 4.112251915025125,
      "tokens_seen": 342032384
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009053961885656972,
      "loss": 1.7311,
      "theoretical_loss": 4.112162068070525,
      "tokens_seen": 342097920
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009053761283851555,
      "loss": 1.6875,
      "theoretical_loss": 4.112072243144662,
      "tokens_seen": 342163456
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009053560682046139,
      "loss": 1.7803,
      "theoretical_loss": 4.111982440237917,
      "tokens_seen": 342228992
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009053360080240723,
      "loss": 1.7886,
      "theoretical_loss": 4.1118926593406755,
      "tokens_seen": 342294528
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009053159478435306,
      "loss": 1.7322,
      "theoretical_loss": 4.111802900443333,
      "tokens_seen": 342360064
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000905295887662989,
      "loss": 1.5982,
      "theoretical_loss": 4.111713163536289,
      "tokens_seen": 342425600
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009052758274824475,
      "loss": 1.7256,
      "theoretical_loss": 4.111623448609948,
      "tokens_seen": 342491136
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009052557673019057,
      "loss": 1.8172,
      "theoretical_loss": 4.111533755654721,
      "tokens_seen": 342556672
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009052357071213642,
      "loss": 1.6669,
      "theoretical_loss": 4.111444084661026,
      "tokens_seen": 342622208
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009052156469408224,
      "loss": 1.7317,
      "theoretical_loss": 4.111354435619287,
      "tokens_seen": 342687744
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009051955867602809,
      "loss": 1.7397,
      "theoretical_loss": 4.111264808519933,
      "tokens_seen": 342753280
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009051755265797393,
      "loss": 1.6709,
      "theoretical_loss": 4.111175203353399,
      "tokens_seen": 342818816
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009051554663991976,
      "loss": 1.6312,
      "theoretical_loss": 4.111085620110127,
      "tokens_seen": 342884352
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000905135406218656,
      "loss": 1.8186,
      "theoretical_loss": 4.1109960587805645,
      "tokens_seen": 342949888
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009051153460381145,
      "loss": 1.7051,
      "theoretical_loss": 4.110906519355165,
      "tokens_seen": 343015424
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009050952858575727,
      "loss": 1.7722,
      "theoretical_loss": 4.110817001824387,
      "tokens_seen": 343080960
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009050752256770312,
      "loss": 1.6726,
      "theoretical_loss": 4.110727506178697,
      "tokens_seen": 343146496
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009050551654964894,
      "loss": 1.8694,
      "theoretical_loss": 4.110638032408565,
      "tokens_seen": 343212032
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009050351053159479,
      "loss": 1.6671,
      "theoretical_loss": 4.110548580504471,
      "tokens_seen": 343277568
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009050150451354063,
      "loss": 1.8713,
      "theoretical_loss": 4.1104591504568955,
      "tokens_seen": 343343104
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009049949849548646,
      "loss": 1.7105,
      "theoretical_loss": 4.110369742256329,
      "tokens_seen": 343408640
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000904974924774323,
      "loss": 1.7494,
      "theoretical_loss": 4.110280355893266,
      "tokens_seen": 343474176
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009049548645937814,
      "loss": 1.7071,
      "theoretical_loss": 4.110190991358209,
      "tokens_seen": 343539712
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009049348044132397,
      "loss": 1.7533,
      "theoretical_loss": 4.110101648641665,
      "tokens_seen": 343605248
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009049147442326982,
      "loss": 1.6532,
      "theoretical_loss": 4.110012327734145,
      "tokens_seen": 343670784
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009048946840521564,
      "loss": 1.7937,
      "theoretical_loss": 4.10992302862617,
      "tokens_seen": 343736320
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009048746238716149,
      "loss": 1.7139,
      "theoretical_loss": 4.109833751308264,
      "tokens_seen": 343801856
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009048545636910732,
      "loss": 1.683,
      "theoretical_loss": 4.109744495770959,
      "tokens_seen": 343867392
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009048345035105316,
      "loss": 1.746,
      "theoretical_loss": 4.1096552620047895,
      "tokens_seen": 343932928
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00090481444332999,
      "loss": 1.7218,
      "theoretical_loss": 4.1095660500003,
      "tokens_seen": 343998464
    },
    {
      "epoch": 0.1,
      "objective/train/advantage_avg": 0.49333980679512024,
      "objective/train/docs_used": 131936,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7032973766326904,
      "objective/train/original_loss": 1.7032976150512695,
      "objective/train/theoretical_loss": 4.109476859748038,
      "objective/train/tokens_used": 364524000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24367128312587738,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505727529525757,
      "objective/train/weighted_lm_loss": 1.789353609085083,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 4.109476859748038,
      "tokens_seen": 344064000
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009047943831494484,
      "loss": 1.9069,
      "theoretical_loss": 4.109476859748038,
      "tokens_seen": 344064000
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009047743229689067,
      "loss": 1.6547,
      "theoretical_loss": 4.109387691238559,
      "tokens_seen": 344129536
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009047542627883652,
      "loss": 1.736,
      "theoretical_loss": 4.109298544462423,
      "tokens_seen": 344195072
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009047342026078235,
      "loss": 1.6227,
      "theoretical_loss": 4.109209419410195,
      "tokens_seen": 344260608
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009047141424272819,
      "loss": 1.6497,
      "theoretical_loss": 4.10912031607245,
      "tokens_seen": 344326144
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009046940822467402,
      "loss": 1.6495,
      "theoretical_loss": 4.1090312344397635,
      "tokens_seen": 344391680
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009046740220661986,
      "loss": 1.6644,
      "theoretical_loss": 4.108942174502721,
      "tokens_seen": 344457216
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000904653961885657,
      "loss": 1.7992,
      "theoretical_loss": 4.10885313625191,
      "tokens_seen": 344522752
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009046339017051154,
      "loss": 1.7062,
      "theoretical_loss": 4.108764119677929,
      "tokens_seen": 344588288
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009046138415245737,
      "loss": 1.6949,
      "theoretical_loss": 4.108675124771377,
      "tokens_seen": 344653824
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009045937813440321,
      "loss": 1.7387,
      "theoretical_loss": 4.108586151522863,
      "tokens_seen": 344719360
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009045737211634905,
      "loss": 1.866,
      "theoretical_loss": 4.108497199923001,
      "tokens_seen": 344784896
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009045536609829489,
      "loss": 1.7702,
      "theoretical_loss": 4.108408269962408,
      "tokens_seen": 344850432
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009045336008024072,
      "loss": 1.7425,
      "theoretical_loss": 4.108319361631709,
      "tokens_seen": 344915968
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009045135406218657,
      "loss": 1.773,
      "theoretical_loss": 4.1082304749215375,
      "tokens_seen": 344981504
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009044934804413239,
      "loss": 1.9007,
      "theoretical_loss": 4.108141609822527,
      "tokens_seen": 345047040
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009044734202607824,
      "loss": 1.6881,
      "theoretical_loss": 4.108052766325322,
      "tokens_seen": 345112576
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009044533600802407,
      "loss": 1.686,
      "theoretical_loss": 4.10796394442057,
      "tokens_seen": 345178112
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009044332998996991,
      "loss": 1.6889,
      "theoretical_loss": 4.107875144098925,
      "tokens_seen": 345243648
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009044132397191575,
      "loss": 1.7391,
      "theoretical_loss": 4.107786365351047,
      "tokens_seen": 345309184
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009043931795386159,
      "loss": 1.7828,
      "theoretical_loss": 4.107697608167602,
      "tokens_seen": 345374720
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009043731193580742,
      "loss": 1.8004,
      "theoretical_loss": 4.107608872539261,
      "tokens_seen": 345440256
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009043530591775327,
      "loss": 1.9854,
      "theoretical_loss": 4.107520158456703,
      "tokens_seen": 345505792
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009043329989969909,
      "loss": 1.6611,
      "theoretical_loss": 4.107431465910609,
      "tokens_seen": 345571328
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009043129388164494,
      "loss": 1.6514,
      "theoretical_loss": 4.107342794891668,
      "tokens_seen": 345636864
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009042928786359077,
      "loss": 1.7724,
      "theoretical_loss": 4.107254145390577,
      "tokens_seen": 345702400
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009042728184553661,
      "loss": 1.8089,
      "theoretical_loss": 4.107165517398034,
      "tokens_seen": 345767936
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009042527582748245,
      "loss": 1.8605,
      "theoretical_loss": 4.107076910904747,
      "tokens_seen": 345833472
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009042326980942828,
      "loss": 1.7117,
      "theoretical_loss": 4.106988325901428,
      "tokens_seen": 345899008
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009042126379137412,
      "loss": 1.7907,
      "theoretical_loss": 4.106899762378793,
      "tokens_seen": 345964544
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009041925777331997,
      "loss": 1.7211,
      "theoretical_loss": 4.106811220327568,
      "tokens_seen": 346030080
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009041725175526579,
      "loss": 1.7923,
      "theoretical_loss": 4.106722699738481,
      "tokens_seen": 346095616
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009041524573721164,
      "loss": 1.8162,
      "theoretical_loss": 4.106634200602267,
      "tokens_seen": 346161152
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009041323971915746,
      "loss": 1.7074,
      "theoretical_loss": 4.106545722909668,
      "tokens_seen": 346226688
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009041123370110331,
      "loss": 1.6258,
      "theoretical_loss": 4.10645726665143,
      "tokens_seen": 346292224
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009040922768304915,
      "loss": 1.6748,
      "theoretical_loss": 4.106368831818305,
      "tokens_seen": 346357760
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009040722166499498,
      "loss": 1.8313,
      "theoretical_loss": 4.1062804184010515,
      "tokens_seen": 346423296
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0009040521564694082,
      "loss": 1.8808,
      "theoretical_loss": 4.106192026390433,
      "tokens_seen": 346488832
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009040320962888667,
      "loss": 1.7614,
      "theoretical_loss": 4.10610365577722,
      "tokens_seen": 346554368
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009040120361083249,
      "loss": 1.6404,
      "theoretical_loss": 4.106015306552187,
      "tokens_seen": 346619904
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009039919759277834,
      "loss": 1.8777,
      "theoretical_loss": 4.1059269787061154,
      "tokens_seen": 346685440
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009039719157472416,
      "loss": 1.6846,
      "theoretical_loss": 4.105838672229792,
      "tokens_seen": 346750976
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009039518555667001,
      "loss": 1.7727,
      "theoretical_loss": 4.105750387114009,
      "tokens_seen": 346816512
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009039317953861585,
      "loss": 1.6671,
      "theoretical_loss": 4.105662123349564,
      "tokens_seen": 346882048
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009039117352056168,
      "loss": 1.7073,
      "theoretical_loss": 4.105573880927262,
      "tokens_seen": 346947584
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009038916750250753,
      "loss": 1.7958,
      "theoretical_loss": 4.105485659837912,
      "tokens_seen": 347013120
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009038716148445336,
      "loss": 1.7587,
      "theoretical_loss": 4.105397460072329,
      "tokens_seen": 347078656
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000903851554663992,
      "loss": 1.6212,
      "theoretical_loss": 4.105309281621334,
      "tokens_seen": 347144192
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009038314944834504,
      "loss": 1.619,
      "theoretical_loss": 4.105221124475754,
      "tokens_seen": 347209728
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009038114343029088,
      "loss": 1.7397,
      "theoretical_loss": 4.105132988626422,
      "tokens_seen": 347275264
    },
    {
      "epoch": 0.11,
      "objective/train/advantage_avg": 0.49424776434898376,
      "objective/train/docs_used": 133174,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6256812810897827,
      "objective/train/original_loss": 1.6256814002990723,
      "objective/train/theoretical_loss": 4.105044874064174,
      "objective/train/tokens_used": 367800800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24457493424415588,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506681203842163,
      "objective/train/weighted_lm_loss": 1.7079211473464966,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.105044874064174,
      "tokens_seen": 347340800
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009037913741223671,
      "loss": 1.6559,
      "theoretical_loss": 4.105044874064174,
      "tokens_seen": 347340800
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009037713139418256,
      "loss": 1.7065,
      "theoretical_loss": 4.1049567807798555,
      "tokens_seen": 347406336
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009037512537612839,
      "loss": 1.7464,
      "theoretical_loss": 4.1048687087643145,
      "tokens_seen": 347471872
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009037311935807423,
      "loss": 1.7268,
      "theoretical_loss": 4.104780658008407,
      "tokens_seen": 347537408
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009037111334002006,
      "loss": 1.7817,
      "theoretical_loss": 4.104692628502993,
      "tokens_seen": 347602944
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000903691073219659,
      "loss": 1.6799,
      "theoretical_loss": 4.104604620238939,
      "tokens_seen": 347668480
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009036710130391174,
      "loss": 1.7885,
      "theoretical_loss": 4.104516633207116,
      "tokens_seen": 347734016
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009036509528585758,
      "loss": 1.6622,
      "theoretical_loss": 4.104428667398404,
      "tokens_seen": 347799552
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009036308926780341,
      "loss": 1.7008,
      "theoretical_loss": 4.104340722803683,
      "tokens_seen": 347865088
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009036108324974925,
      "loss": 1.6566,
      "theoretical_loss": 4.104252799413844,
      "tokens_seen": 347930624
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009035907723169509,
      "loss": 1.7029,
      "theoretical_loss": 4.104164897219781,
      "tokens_seen": 347996160
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009035707121364093,
      "loss": 1.7565,
      "theoretical_loss": 4.104077016212393,
      "tokens_seen": 348061696
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009035506519558676,
      "loss": 1.7934,
      "theoretical_loss": 4.103989156382589,
      "tokens_seen": 348127232
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000903530591775326,
      "loss": 1.7331,
      "theoretical_loss": 4.103901317721276,
      "tokens_seen": 348192768
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009035105315947843,
      "loss": 1.6297,
      "theoretical_loss": 4.103813500219374,
      "tokens_seen": 348258304
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009034904714142428,
      "loss": 1.7116,
      "theoretical_loss": 4.103725703867804,
      "tokens_seen": 348323840
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009034704112337011,
      "loss": 1.5938,
      "theoretical_loss": 4.103637928657495,
      "tokens_seen": 348389376
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009034503510531595,
      "loss": 1.8078,
      "theoretical_loss": 4.10355017457938,
      "tokens_seen": 348454912
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009034302908726179,
      "loss": 1.8058,
      "theoretical_loss": 4.103462441624398,
      "tokens_seen": 348520448
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009034102306920763,
      "loss": 1.7124,
      "theoretical_loss": 4.1033747297834955,
      "tokens_seen": 348585984
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009033901705115346,
      "loss": 1.9676,
      "theoretical_loss": 4.103287039047622,
      "tokens_seen": 348651520
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009033701103309931,
      "loss": 1.8403,
      "theoretical_loss": 4.103199369407734,
      "tokens_seen": 348717056
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009033500501504513,
      "loss": 1.7325,
      "theoretical_loss": 4.103111720854792,
      "tokens_seen": 348782592
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009033299899699098,
      "loss": 1.6372,
      "theoretical_loss": 4.103024093379764,
      "tokens_seen": 348848128
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009033099297893681,
      "loss": 1.7101,
      "theoretical_loss": 4.102936486973624,
      "tokens_seen": 348913664
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009032898696088265,
      "loss": 1.639,
      "theoretical_loss": 4.102848901627348,
      "tokens_seen": 348979200
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009032698094282849,
      "loss": 1.7128,
      "theoretical_loss": 4.102761337331921,
      "tokens_seen": 349044736
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009032497492477432,
      "loss": 1.7724,
      "theoretical_loss": 4.102673794078333,
      "tokens_seen": 349110272
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009032296890672016,
      "loss": 1.7385,
      "theoretical_loss": 4.102586271857579,
      "tokens_seen": 349175808
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009032096288866601,
      "loss": 1.6569,
      "theoretical_loss": 4.102498770660658,
      "tokens_seen": 349241344
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009031895687061183,
      "loss": 1.7852,
      "theoretical_loss": 4.102411290478577,
      "tokens_seen": 349306880
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009031695085255768,
      "loss": 1.6594,
      "theoretical_loss": 4.102323831302349,
      "tokens_seen": 349372416
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000903149448345035,
      "loss": 1.6494,
      "theoretical_loss": 4.102236393122989,
      "tokens_seen": 349437952
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009031293881644935,
      "loss": 1.7227,
      "theoretical_loss": 4.102148975931521,
      "tokens_seen": 349503488
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009031093279839519,
      "loss": 1.89,
      "theoretical_loss": 4.102061579718972,
      "tokens_seen": 349569024
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009030892678034102,
      "loss": 1.669,
      "theoretical_loss": 4.101974204476377,
      "tokens_seen": 349634560
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009030692076228686,
      "loss": 1.7407,
      "theoretical_loss": 4.101886850194775,
      "tokens_seen": 349700096
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009030491474423271,
      "loss": 1.7601,
      "theoretical_loss": 4.10179951686521,
      "tokens_seen": 349765632
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009030290872617853,
      "loss": 1.7153,
      "theoretical_loss": 4.1017122044787335,
      "tokens_seen": 349831168
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009030090270812438,
      "loss": 1.5519,
      "theoretical_loss": 4.1016249130263995,
      "tokens_seen": 349896704
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000902988966900702,
      "loss": 1.704,
      "theoretical_loss": 4.10153764249927,
      "tokens_seen": 349962240
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009029689067201605,
      "loss": 1.6163,
      "theoretical_loss": 4.101450392888412,
      "tokens_seen": 350027776
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009029488465396189,
      "loss": 1.7973,
      "theoretical_loss": 4.101363164184898,
      "tokens_seen": 350093312
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009029287863590772,
      "loss": 1.7025,
      "theoretical_loss": 4.101275956379805,
      "tokens_seen": 350158848
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009029087261785356,
      "loss": 1.6458,
      "theoretical_loss": 4.1011887694642155,
      "tokens_seen": 350224384
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000902888665997994,
      "loss": 1.7947,
      "theoretical_loss": 4.10110160342922,
      "tokens_seen": 350289920
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009028686058174523,
      "loss": 1.6099,
      "theoretical_loss": 4.101014458265912,
      "tokens_seen": 350355456
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009028485456369108,
      "loss": 1.7333,
      "theoretical_loss": 4.100927333965389,
      "tokens_seen": 350420992
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009028284854563691,
      "loss": 1.738,
      "theoretical_loss": 4.100840230518759,
      "tokens_seen": 350486528
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009028084252758275,
      "loss": 1.7421,
      "theoretical_loss": 4.10075314791713,
      "tokens_seen": 350552064
    },
    {
      "epoch": 0.11,
      "objective/train/advantage_avg": 0.4893075227737427,
      "objective/train/docs_used": 134382,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6907188892364502,
      "objective/train/original_loss": 1.6907188892364502,
      "objective/train/theoretical_loss": 4.10066608615162,
      "objective/train/tokens_used": 371077600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24007414281368256,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501511096954346,
      "objective/train/weighted_lm_loss": 1.7755835056304932,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 4.10066608615162,
      "tokens_seen": 350617600
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009027883650952859,
      "loss": 1.7911,
      "theoretical_loss": 4.10066608615162,
      "tokens_seen": 350617600
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009027683049147443,
      "loss": 1.707,
      "theoretical_loss": 4.100579045213349,
      "tokens_seen": 350683136
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009027482447342026,
      "loss": 1.7838,
      "theoretical_loss": 4.100492025093445,
      "tokens_seen": 350748672
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000902728184553661,
      "loss": 1.8857,
      "theoretical_loss": 4.100405025783039,
      "tokens_seen": 350814208
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009027081243731193,
      "loss": 1.7437,
      "theoretical_loss": 4.100318047273269,
      "tokens_seen": 350879744
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009026880641925778,
      "loss": 1.7306,
      "theoretical_loss": 4.100231089555279,
      "tokens_seen": 350945280
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009026680040120361,
      "loss": 1.6636,
      "theoretical_loss": 4.100144152620215,
      "tokens_seen": 351010816
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009026479438314945,
      "loss": 1.675,
      "theoretical_loss": 4.100057236459235,
      "tokens_seen": 351076352
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009026278836509528,
      "loss": 1.7887,
      "theoretical_loss": 4.099970341063495,
      "tokens_seen": 351141888
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009026078234704113,
      "loss": 1.835,
      "theoretical_loss": 4.099883466424161,
      "tokens_seen": 351207424
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009025877632898696,
      "loss": 1.6373,
      "theoretical_loss": 4.099796612532403,
      "tokens_seen": 351272960
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000902567703109328,
      "loss": 1.6707,
      "theoretical_loss": 4.099709779379397,
      "tokens_seen": 351338496
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009025476429287863,
      "loss": 1.757,
      "theoretical_loss": 4.099622966956324,
      "tokens_seen": 351404032
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009025275827482447,
      "loss": 1.688,
      "theoretical_loss": 4.09953617525437,
      "tokens_seen": 351469568
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009025075225677031,
      "loss": 1.7404,
      "theoretical_loss": 4.0994494042647265,
      "tokens_seen": 351535104
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009024874623871615,
      "loss": 1.648,
      "theoretical_loss": 4.099362653978591,
      "tokens_seen": 351600640
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009024674022066198,
      "loss": 1.8621,
      "theoretical_loss": 4.0992759243871655,
      "tokens_seen": 351666176
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009024473420260783,
      "loss": 1.6942,
      "theoretical_loss": 4.099189215481659,
      "tokens_seen": 351731712
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009024272818455366,
      "loss": 1.7197,
      "theoretical_loss": 4.099102527253285,
      "tokens_seen": 351797248
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000902407221664995,
      "loss": 1.7993,
      "theoretical_loss": 4.09901585969326,
      "tokens_seen": 351862784
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009023871614844535,
      "loss": 1.7709,
      "theoretical_loss": 4.098929212792811,
      "tokens_seen": 351928320
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009023671013039117,
      "loss": 1.7283,
      "theoretical_loss": 4.098842586543165,
      "tokens_seen": 351993856
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009023470411233702,
      "loss": 1.7918,
      "theoretical_loss": 4.098755980935557,
      "tokens_seen": 352059392
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009023269809428285,
      "loss": 1.7396,
      "theoretical_loss": 4.098669395961229,
      "tokens_seen": 352124928
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009023069207622869,
      "loss": 1.7039,
      "theoretical_loss": 4.098582831611425,
      "tokens_seen": 352190464
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009022868605817453,
      "loss": 1.6457,
      "theoretical_loss": 4.098496287877395,
      "tokens_seen": 352256000
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009022668004012036,
      "loss": 1.7406,
      "theoretical_loss": 4.0984097647503965,
      "tokens_seen": 352321536
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000902246740220662,
      "loss": 1.6913,
      "theoretical_loss": 4.09832326222169,
      "tokens_seen": 352387072
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009022266800401205,
      "loss": 1.8149,
      "theoretical_loss": 4.098236780282543,
      "tokens_seen": 352452608
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009022066198595787,
      "loss": 1.6853,
      "theoretical_loss": 4.098150318924227,
      "tokens_seen": 352518144
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009021865596790372,
      "loss": 1.7264,
      "theoretical_loss": 4.09806387813802,
      "tokens_seen": 352583680
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009021664994984954,
      "loss": 1.7176,
      "theoretical_loss": 4.097977457915205,
      "tokens_seen": 352649216
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009021464393179539,
      "loss": 1.7579,
      "theoretical_loss": 4.097891058247068,
      "tokens_seen": 352714752
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009021263791374123,
      "loss": 1.6048,
      "theoretical_loss": 4.097804679124906,
      "tokens_seen": 352780288
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009021063189568706,
      "loss": 1.6215,
      "theoretical_loss": 4.0977183205400145,
      "tokens_seen": 352845824
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000902086258776329,
      "loss": 1.6109,
      "theoretical_loss": 4.097631982483699,
      "tokens_seen": 352911360
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009020661985957875,
      "loss": 1.7666,
      "theoretical_loss": 4.097545664947268,
      "tokens_seen": 352976896
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009020461384152457,
      "loss": 1.6963,
      "theoretical_loss": 4.097459367922037,
      "tokens_seen": 353042432
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009020260782347042,
      "loss": 1.7839,
      "theoretical_loss": 4.097373091399324,
      "tokens_seen": 353107968
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009020060180541625,
      "loss": 1.7953,
      "theoretical_loss": 4.097286835370456,
      "tokens_seen": 353173504
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009019859578736209,
      "loss": 1.5052,
      "theoretical_loss": 4.097200599826763,
      "tokens_seen": 353239040
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009019658976930793,
      "loss": 1.8183,
      "theoretical_loss": 4.09711438475958,
      "tokens_seen": 353304576
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009019458375125376,
      "loss": 1.7284,
      "theoretical_loss": 4.097028190160249,
      "tokens_seen": 353370112
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000901925777331996,
      "loss": 1.7696,
      "theoretical_loss": 4.096942016020115,
      "tokens_seen": 353435648
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009019057171514544,
      "loss": 1.8074,
      "theoretical_loss": 4.096855862330531,
      "tokens_seen": 353501184
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009018856569709127,
      "loss": 1.6989,
      "theoretical_loss": 4.096769729082853,
      "tokens_seen": 353566720
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009018655967903712,
      "loss": 1.7032,
      "theoretical_loss": 4.096683616268442,
      "tokens_seen": 353632256
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009018455366098295,
      "loss": 1.7195,
      "theoretical_loss": 4.096597523878667,
      "tokens_seen": 353697792
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009018254764292879,
      "loss": 1.8013,
      "theoretical_loss": 4.096511451904899,
      "tokens_seen": 353763328
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009018054162487463,
      "loss": 1.687,
      "theoretical_loss": 4.096425400338516,
      "tokens_seen": 353828864
    },
    {
      "epoch": 0.11,
      "objective/train/advantage_avg": 0.49126100540161133,
      "objective/train/docs_used": 135614,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4353805780410767,
      "objective/train/original_loss": 1.435380458831787,
      "objective/train/theoretical_loss": 4.096339369170902,
      "objective/train/tokens_used": 374354400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24204039573669434,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503566265106201,
      "objective/train/weighted_lm_loss": 1.5078595876693726,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.096339369170902,
      "tokens_seen": 353894400
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009017853560682047,
      "loss": 1.669,
      "theoretical_loss": 4.096339369170902,
      "tokens_seen": 353894400
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000901765295887663,
      "loss": 1.7519,
      "theoretical_loss": 4.096253358393445,
      "tokens_seen": 353959936
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009017452357071214,
      "loss": 1.7505,
      "theoretical_loss": 4.096167367997538,
      "tokens_seen": 354025472
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009017251755265797,
      "loss": 1.731,
      "theoretical_loss": 4.0960813979745785,
      "tokens_seen": 354091008
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009017051153460382,
      "loss": 1.749,
      "theoretical_loss": 4.095995448315972,
      "tokens_seen": 354156544
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009016850551654965,
      "loss": 1.7191,
      "theoretical_loss": 4.095909519013127,
      "tokens_seen": 354222080
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009016649949849549,
      "loss": 1.7147,
      "theoretical_loss": 4.095823610057458,
      "tokens_seen": 354287616
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009016449348044132,
      "loss": 1.7708,
      "theoretical_loss": 4.0957377214403845,
      "tokens_seen": 354353152
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009016248746238717,
      "loss": 1.6917,
      "theoretical_loss": 4.095651853153331,
      "tokens_seen": 354418688
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00090160481444333,
      "loss": 1.7711,
      "theoretical_loss": 4.095566005187729,
      "tokens_seen": 354484224
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009015847542627884,
      "loss": 1.7752,
      "theoretical_loss": 4.095480177535011,
      "tokens_seen": 354549760
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009015646940822467,
      "loss": 1.7135,
      "theoretical_loss": 4.095394370186619,
      "tokens_seen": 354615296
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009015446339017051,
      "loss": 1.7425,
      "theoretical_loss": 4.095308583133997,
      "tokens_seen": 354680832
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009015245737211635,
      "loss": 1.7808,
      "theoretical_loss": 4.095222816368597,
      "tokens_seen": 354746368
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009015045135406219,
      "loss": 1.8011,
      "theoretical_loss": 4.095137069881876,
      "tokens_seen": 354811904
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009014844533600802,
      "loss": 1.736,
      "theoretical_loss": 4.095051343665292,
      "tokens_seen": 354877440
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009014643931795387,
      "loss": 1.7026,
      "theoretical_loss": 4.094965637710314,
      "tokens_seen": 354942976
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000901444332998997,
      "loss": 1.7112,
      "theoretical_loss": 4.094879952008412,
      "tokens_seen": 355008512
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009014242728184554,
      "loss": 1.8059,
      "theoretical_loss": 4.094794286551062,
      "tokens_seen": 355074048
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009014042126379138,
      "loss": 1.6808,
      "theoretical_loss": 4.094708641329748,
      "tokens_seen": 355139584
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009013841524573721,
      "loss": 1.7377,
      "theoretical_loss": 4.094623016335954,
      "tokens_seen": 355205120
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009013640922768305,
      "loss": 1.7313,
      "theoretical_loss": 4.094537411561174,
      "tokens_seen": 355270656
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000901344032096289,
      "loss": 1.6661,
      "theoretical_loss": 4.094451826996904,
      "tokens_seen": 355336192
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009013239719157472,
      "loss": 1.702,
      "theoretical_loss": 4.094366262634647,
      "tokens_seen": 355401728
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009013039117352057,
      "loss": 1.8119,
      "theoretical_loss": 4.094280718465911,
      "tokens_seen": 355467264
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009012838515546639,
      "loss": 1.7697,
      "theoretical_loss": 4.094195194482207,
      "tokens_seen": 355532800
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009012637913741224,
      "loss": 1.7872,
      "theoretical_loss": 4.0941096906750545,
      "tokens_seen": 355598336
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009012437311935808,
      "loss": 1.7232,
      "theoretical_loss": 4.094024207035974,
      "tokens_seen": 355663872
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009012236710130391,
      "loss": 1.6203,
      "theoretical_loss": 4.093938743556496,
      "tokens_seen": 355729408
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009012036108324975,
      "loss": 1.6374,
      "theoretical_loss": 4.093853300228152,
      "tokens_seen": 355794944
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009011835506519558,
      "loss": 1.7938,
      "theoretical_loss": 4.09376787704248,
      "tokens_seen": 355860480
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009011634904714142,
      "loss": 1.7368,
      "theoretical_loss": 4.0936824739910245,
      "tokens_seen": 355926016
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009011434302908727,
      "loss": 1.6517,
      "theoretical_loss": 4.093597091065333,
      "tokens_seen": 355991552
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009011233701103309,
      "loss": 1.7932,
      "theoretical_loss": 4.09351172825696,
      "tokens_seen": 356057088
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009011033099297894,
      "loss": 1.6453,
      "theoretical_loss": 4.093426385557462,
      "tokens_seen": 356122624
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009010832497492478,
      "loss": 1.7598,
      "theoretical_loss": 4.093341062958405,
      "tokens_seen": 356188160
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009010631895687061,
      "loss": 1.6493,
      "theoretical_loss": 4.093255760451357,
      "tokens_seen": 356253696
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009010431293881645,
      "loss": 1.7055,
      "theoretical_loss": 4.093170478027892,
      "tokens_seen": 356319232
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009010230692076229,
      "loss": 1.698,
      "theoretical_loss": 4.093085215679588,
      "tokens_seen": 356384768
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009010030090270812,
      "loss": 1.7086,
      "theoretical_loss": 4.0929999733980305,
      "tokens_seen": 356450304
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009009829488465397,
      "loss": 1.6501,
      "theoretical_loss": 4.092914751174808,
      "tokens_seen": 356515840
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009009628886659979,
      "loss": 1.7705,
      "theoretical_loss": 4.092829549001514,
      "tokens_seen": 356581376
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009009428284854564,
      "loss": 1.7081,
      "theoretical_loss": 4.0927443668697485,
      "tokens_seen": 356646912
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009009227683049147,
      "loss": 1.7028,
      "theoretical_loss": 4.092659204771116,
      "tokens_seen": 356712448
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009009027081243731,
      "loss": 1.7112,
      "theoretical_loss": 4.092574062697225,
      "tokens_seen": 356777984
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009008826479438316,
      "loss": 1.6097,
      "theoretical_loss": 4.092488940639691,
      "tokens_seen": 356843520
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009008625877632899,
      "loss": 1.705,
      "theoretical_loss": 4.0924038385901325,
      "tokens_seen": 356909056
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009008425275827483,
      "loss": 1.7643,
      "theoretical_loss": 4.092318756540174,
      "tokens_seen": 356974592
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009008224674022067,
      "loss": 1.6911,
      "theoretical_loss": 4.092233694481447,
      "tokens_seen": 357040128
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000900802407221665,
      "loss": 1.6974,
      "theoretical_loss": 4.092148652405584,
      "tokens_seen": 357105664
    },
    {
      "epoch": 0.11,
      "objective/train/advantage_avg": 0.4929529130458832,
      "objective/train/docs_used": 136331,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7153905630111694,
      "objective/train/original_loss": 1.715390682220459,
      "objective/train/theoretical_loss": 4.092063630304224,
      "objective/train/tokens_used": 377631200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24330100417137146,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505321025848389,
      "objective/train/weighted_lm_loss": 1.8019219636917114,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.092063630304224,
      "tokens_seen": 357171200
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009007823470411234,
      "loss": 1.803,
      "theoretical_loss": 4.092063630304224,
      "tokens_seen": 357171200
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009007622868605818,
      "loss": 1.8053,
      "theoretical_loss": 4.091978628169015,
      "tokens_seen": 357236736
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009007422266800401,
      "loss": 1.799,
      "theoretical_loss": 4.091893645991604,
      "tokens_seen": 357302272
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009007221664994986,
      "loss": 1.7151,
      "theoretical_loss": 4.091808683763646,
      "tokens_seen": 357367808
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009007021063189569,
      "loss": 1.8246,
      "theoretical_loss": 4.091723741476802,
      "tokens_seen": 357433344
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009006820461384153,
      "loss": 1.7746,
      "theoretical_loss": 4.091638819122735,
      "tokens_seen": 357498880
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009006619859578736,
      "loss": 1.6607,
      "theoretical_loss": 4.091553916693115,
      "tokens_seen": 357564416
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009006419257773321,
      "loss": 1.6609,
      "theoretical_loss": 4.091469034179619,
      "tokens_seen": 357629952
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009006218655967904,
      "loss": 1.5002,
      "theoretical_loss": 4.0913841715739245,
      "tokens_seen": 357695488
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009006018054162488,
      "loss": 1.6997,
      "theoretical_loss": 4.091299328867717,
      "tokens_seen": 357761024
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009005817452357071,
      "loss": 1.7176,
      "theoretical_loss": 4.091214506052687,
      "tokens_seen": 357826560
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009005616850551655,
      "loss": 1.5455,
      "theoretical_loss": 4.091129703120528,
      "tokens_seen": 357892096
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009005416248746239,
      "loss": 1.6436,
      "theoretical_loss": 4.09104492006294,
      "tokens_seen": 357957632
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009005215646940823,
      "loss": 1.6813,
      "theoretical_loss": 4.090960156871628,
      "tokens_seen": 358023168
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009005015045135406,
      "loss": 1.5725,
      "theoretical_loss": 4.090875413538302,
      "tokens_seen": 358088704
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009004814443329991,
      "loss": 1.8228,
      "theoretical_loss": 4.090790690054677,
      "tokens_seen": 358154240
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009004613841524574,
      "loss": 1.6769,
      "theoretical_loss": 4.0907059864124715,
      "tokens_seen": 358219776
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009004413239719158,
      "loss": 1.815,
      "theoretical_loss": 4.090621302603411,
      "tokens_seen": 358285312
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009004212637913742,
      "loss": 1.7456,
      "theoretical_loss": 4.090536638619224,
      "tokens_seen": 358350848
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009004012036108325,
      "loss": 1.6841,
      "theoretical_loss": 4.090451994451647,
      "tokens_seen": 358416384
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009003811434302909,
      "loss": 1.7235,
      "theoretical_loss": 4.090367370092419,
      "tokens_seen": 358481920
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009003610832497493,
      "loss": 1.6922,
      "theoretical_loss": 4.090282765533283,
      "tokens_seen": 358547456
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009003410230692076,
      "loss": 1.5614,
      "theoretical_loss": 4.090198180765989,
      "tokens_seen": 358612992
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009003209628886661,
      "loss": 1.738,
      "theoretical_loss": 4.090113615782292,
      "tokens_seen": 358678528
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009003009027081243,
      "loss": 1.6384,
      "theoretical_loss": 4.0900290705739515,
      "tokens_seen": 358744064
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009002808425275828,
      "loss": 1.8113,
      "theoretical_loss": 4.089944545132731,
      "tokens_seen": 358809600
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009002607823470412,
      "loss": 1.667,
      "theoretical_loss": 4.0898600394504,
      "tokens_seen": 358875136
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009002407221664995,
      "loss": 1.7378,
      "theoretical_loss": 4.089775553518733,
      "tokens_seen": 358940672
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009002206619859579,
      "loss": 1.7439,
      "theoretical_loss": 4.089691087329507,
      "tokens_seen": 359006208
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009002006018054162,
      "loss": 1.6211,
      "theoretical_loss": 4.089606640874509,
      "tokens_seen": 359071744
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009001805416248746,
      "loss": 1.688,
      "theoretical_loss": 4.089522214145525,
      "tokens_seen": 359137280
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009001604814443331,
      "loss": 1.7826,
      "theoretical_loss": 4.089437807134351,
      "tokens_seen": 359202816
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009001404212637913,
      "loss": 1.8018,
      "theoretical_loss": 4.0893534198327846,
      "tokens_seen": 359268352
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009001203610832498,
      "loss": 1.6606,
      "theoretical_loss": 4.0892690522326305,
      "tokens_seen": 359333888
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009001003009027082,
      "loss": 1.7576,
      "theoretical_loss": 4.089184704325695,
      "tokens_seen": 359399424
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009000802407221665,
      "loss": 1.6624,
      "theoretical_loss": 4.089100376103794,
      "tokens_seen": 359464960
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009000601805416249,
      "loss": 1.7172,
      "theoretical_loss": 4.0890160675587435,
      "tokens_seen": 359530496
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009000401203610833,
      "loss": 1.7355,
      "theoretical_loss": 4.088931778682369,
      "tokens_seen": 359596032
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009000200601805416,
      "loss": 1.6584,
      "theoretical_loss": 4.088847509466497,
      "tokens_seen": 359661568
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0009000000000000001,
      "loss": 1.6495,
      "theoretical_loss": 4.088763259902961,
      "tokens_seen": 359727104
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008999799398194583,
      "loss": 1.6659,
      "theoretical_loss": 4.088679029983599,
      "tokens_seen": 359792640
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008999598796389168,
      "loss": 1.6201,
      "theoretical_loss": 4.088594819700253,
      "tokens_seen": 359858176
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008999398194583751,
      "loss": 1.8343,
      "theoretical_loss": 4.088510629044771,
      "tokens_seen": 359923712
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008999197592778335,
      "loss": 1.6573,
      "theoretical_loss": 4.088426458009005,
      "tokens_seen": 359989248
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008998996990972919,
      "loss": 1.7296,
      "theoretical_loss": 4.0883423065848135,
      "tokens_seen": 360054784
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008998796389167503,
      "loss": 1.7597,
      "theoretical_loss": 4.088258174764058,
      "tokens_seen": 360120320
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008998595787362086,
      "loss": 1.62,
      "theoretical_loss": 4.088174062538605,
      "tokens_seen": 360185856
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008998395185556671,
      "loss": 1.7132,
      "theoretical_loss": 4.088089969900327,
      "tokens_seen": 360251392
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008998194583751253,
      "loss": 1.8231,
      "theoretical_loss": 4.088005896841102,
      "tokens_seen": 360316928
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008997993981945838,
      "loss": 1.77,
      "theoretical_loss": 4.087921843352808,
      "tokens_seen": 360382464
    },
    {
      "epoch": 0.11,
      "objective/train/advantage_avg": 0.49181121587753296,
      "objective/train/docs_used": 137426,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.762984037399292,
      "objective/train/original_loss": 1.7629839181900024,
      "objective/train/theoretical_loss": 4.087837809427334,
      "objective/train/tokens_used": 380908000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2422368824481964,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504125356674194,
      "objective/train/weighted_lm_loss": 1.8516653776168823,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.087837809427334,
      "tokens_seen": 360448000
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008997793380140421,
      "loss": 1.8454,
      "theoretical_loss": 4.087837809427334,
      "tokens_seen": 360448000
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008997592778335005,
      "loss": 1.7136,
      "theoretical_loss": 4.087753795056573,
      "tokens_seen": 360513536
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008997392176529589,
      "loss": 1.6479,
      "theoretical_loss": 4.087669800232417,
      "tokens_seen": 360579072
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008997191574724173,
      "loss": 1.6049,
      "theoretical_loss": 4.08758582494677,
      "tokens_seen": 360644608
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008996990972918756,
      "loss": 1.7941,
      "theoretical_loss": 4.087501869191536,
      "tokens_seen": 360710144
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000899679037111334,
      "loss": 1.793,
      "theoretical_loss": 4.087417932958627,
      "tokens_seen": 360775680
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008996589769307924,
      "loss": 1.6668,
      "theoretical_loss": 4.087334016239957,
      "tokens_seen": 360841216
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008996389167502508,
      "loss": 1.5709,
      "theoretical_loss": 4.0872501190274475,
      "tokens_seen": 360906752
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008996188565697091,
      "loss": 1.6844,
      "theoretical_loss": 4.087166241313023,
      "tokens_seen": 360972288
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008995987963891675,
      "loss": 1.7546,
      "theoretical_loss": 4.087082383088613,
      "tokens_seen": 361037824
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008995787362086258,
      "loss": 1.7811,
      "theoretical_loss": 4.086998544346152,
      "tokens_seen": 361103360
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008995586760280843,
      "loss": 1.845,
      "theoretical_loss": 4.08691472507758,
      "tokens_seen": 361168896
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008995386158475426,
      "loss": 1.744,
      "theoretical_loss": 4.086830925274842,
      "tokens_seen": 361234432
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000899518555667001,
      "loss": 1.7291,
      "theoretical_loss": 4.086747144929885,
      "tokens_seen": 361299968
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008994984954864594,
      "loss": 1.6854,
      "theoretical_loss": 4.086663384034664,
      "tokens_seen": 361365504
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008994784353059178,
      "loss": 1.712,
      "theoretical_loss": 4.086579642581138,
      "tokens_seen": 361431040
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008994583751253761,
      "loss": 1.6976,
      "theoretical_loss": 4.08649592056127,
      "tokens_seen": 361496576
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008994383149448346,
      "loss": 1.8639,
      "theoretical_loss": 4.086412217967028,
      "tokens_seen": 361562112
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008994182547642928,
      "loss": 1.8171,
      "theoretical_loss": 4.086328534790384,
      "tokens_seen": 361627648
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008993981945837513,
      "loss": 1.6998,
      "theoretical_loss": 4.086244871023318,
      "tokens_seen": 361693184
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008993781344032097,
      "loss": 1.7485,
      "theoretical_loss": 4.086161226657811,
      "tokens_seen": 361758720
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000899358074222668,
      "loss": 1.6268,
      "theoretical_loss": 4.086077601685851,
      "tokens_seen": 361824256
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008993380140421265,
      "loss": 1.6268,
      "theoretical_loss": 4.085993996099428,
      "tokens_seen": 361889792
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008993179538615847,
      "loss": 1.5707,
      "theoretical_loss": 4.085910409890541,
      "tokens_seen": 361955328
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008992978936810432,
      "loss": 1.7354,
      "theoretical_loss": 4.085826843051191,
      "tokens_seen": 362020864
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008992778335005016,
      "loss": 1.7003,
      "theoretical_loss": 4.0857432955733834,
      "tokens_seen": 362086400
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008992577733199599,
      "loss": 1.794,
      "theoretical_loss": 4.085659767449131,
      "tokens_seen": 362151936
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008992377131394183,
      "loss": 1.832,
      "theoretical_loss": 4.0855762586704465,
      "tokens_seen": 362217472
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008992176529588766,
      "loss": 1.7118,
      "theoretical_loss": 4.0854927692293534,
      "tokens_seen": 362283008
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000899197592778335,
      "loss": 1.661,
      "theoretical_loss": 4.085409299117876,
      "tokens_seen": 362348544
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008991775325977935,
      "loss": 1.7882,
      "theoretical_loss": 4.0853258483280435,
      "tokens_seen": 362414080
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008991574724172517,
      "loss": 1.6931,
      "theoretical_loss": 4.085242416851891,
      "tokens_seen": 362479616
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008991374122367102,
      "loss": 1.6965,
      "theoretical_loss": 4.085159004681458,
      "tokens_seen": 362545152
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008991173520561686,
      "loss": 1.7359,
      "theoretical_loss": 4.085075611808788,
      "tokens_seen": 362610688
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008990972918756269,
      "loss": 1.6945,
      "theoretical_loss": 4.084992238225931,
      "tokens_seen": 362676224
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008990772316950853,
      "loss": 1.7613,
      "theoretical_loss": 4.08490888392494,
      "tokens_seen": 362741760
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008990571715145437,
      "loss": 1.7596,
      "theoretical_loss": 4.084825548897873,
      "tokens_seen": 362807296
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000899037111334002,
      "loss": 1.7875,
      "theoretical_loss": 4.084742233136794,
      "tokens_seen": 362872832
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008990170511534605,
      "loss": 1.7156,
      "theoretical_loss": 4.084658936633769,
      "tokens_seen": 362938368
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008989969909729187,
      "loss": 1.8888,
      "theoretical_loss": 4.084575659380871,
      "tokens_seen": 363003904
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008989769307923772,
      "loss": 1.7725,
      "theoretical_loss": 4.084492401370177,
      "tokens_seen": 363069440
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008989568706118355,
      "loss": 1.796,
      "theoretical_loss": 4.0844091625937695,
      "tokens_seen": 363134976
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008989368104312939,
      "loss": 1.7835,
      "theoretical_loss": 4.084325943043734,
      "tokens_seen": 363200512
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008989167502507523,
      "loss": 1.7005,
      "theoretical_loss": 4.084242742712163,
      "tokens_seen": 363266048
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008988966900702107,
      "loss": 1.7358,
      "theoretical_loss": 4.08415956159115,
      "tokens_seen": 363331584
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000898876629889669,
      "loss": 1.7673,
      "theoretical_loss": 4.084076399672798,
      "tokens_seen": 363397120
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008988565697091275,
      "loss": 1.6401,
      "theoretical_loss": 4.08399325694921,
      "tokens_seen": 363462656
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008988365095285857,
      "loss": 1.7859,
      "theoretical_loss": 4.0839101334124965,
      "tokens_seen": 363528192
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008988164493480442,
      "loss": 1.6849,
      "theoretical_loss": 4.083827029054773,
      "tokens_seen": 363593728
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008987963891675025,
      "loss": 1.7831,
      "theoretical_loss": 4.083743943868157,
      "tokens_seen": 363659264
    },
    {
      "epoch": 0.11,
      "objective/train/advantage_avg": 0.49359139800071716,
      "objective/train/docs_used": 138686,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7015258073806763,
      "objective/train/original_loss": 1.7015259265899658,
      "objective/train/theoretical_loss": 4.083660877844774,
      "objective/train/tokens_used": 384184800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2438880205154419,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505989789962769,
      "objective/train/weighted_lm_loss": 1.7876434326171875,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.083660877844774,
      "tokens_seen": 363724800
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008987763289869609,
      "loss": 1.7741,
      "theoretical_loss": 4.083660877844774,
      "tokens_seen": 363724800
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008987562688064193,
      "loss": 1.7944,
      "theoretical_loss": 4.0835778309767505,
      "tokens_seen": 363790336
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008987362086258777,
      "loss": 1.6954,
      "theoretical_loss": 4.0834948032562215,
      "tokens_seen": 363855872
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000898716148445336,
      "loss": 1.8153,
      "theoretical_loss": 4.083411794675324,
      "tokens_seen": 363921408
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008986960882647944,
      "loss": 1.7651,
      "theoretical_loss": 4.083328805226199,
      "tokens_seen": 363986944
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008986760280842528,
      "loss": 1.6811,
      "theoretical_loss": 4.083245834900996,
      "tokens_seen": 364052480
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008986559679037112,
      "loss": 1.755,
      "theoretical_loss": 4.083162883691864,
      "tokens_seen": 364118016
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008986359077231695,
      "loss": 1.7445,
      "theoretical_loss": 4.083079951590962,
      "tokens_seen": 364183552
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008986158475426279,
      "loss": 1.8182,
      "theoretical_loss": 4.08299703859045,
      "tokens_seen": 364249088
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008985957873620862,
      "loss": 1.6772,
      "theoretical_loss": 4.082914144682492,
      "tokens_seen": 364314624
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008985757271815447,
      "loss": 1.7932,
      "theoretical_loss": 4.082831269859261,
      "tokens_seen": 364380160
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000898555667001003,
      "loss": 1.8315,
      "theoretical_loss": 4.082748414112929,
      "tokens_seen": 364445696
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008985356068204614,
      "loss": 1.6778,
      "theoretical_loss": 4.082665577435677,
      "tokens_seen": 364511232
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008985155466399198,
      "loss": 1.7543,
      "theoretical_loss": 4.082582759819688,
      "tokens_seen": 364576768
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008984954864593782,
      "loss": 1.6663,
      "theoretical_loss": 4.0824999612571515,
      "tokens_seen": 364642304
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008984754262788365,
      "loss": 1.7071,
      "theoretical_loss": 4.082417181740261,
      "tokens_seen": 364707840
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000898455366098295,
      "loss": 1.6863,
      "theoretical_loss": 4.082334421261213,
      "tokens_seen": 364773376
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008984353059177532,
      "loss": 1.6753,
      "theoretical_loss": 4.082251679812211,
      "tokens_seen": 364838912
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008984152457372117,
      "loss": 1.7689,
      "theoretical_loss": 4.082168957385462,
      "tokens_seen": 364904448
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00089839518555667,
      "loss": 1.8623,
      "theoretical_loss": 4.082086253973177,
      "tokens_seen": 364969984
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008983751253761284,
      "loss": 1.78,
      "theoretical_loss": 4.082003569567573,
      "tokens_seen": 365035520
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008983550651955868,
      "loss": 1.6577,
      "theoretical_loss": 4.08192090416087,
      "tokens_seen": 365101056
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008983350050150451,
      "loss": 1.7802,
      "theoretical_loss": 4.081838257745293,
      "tokens_seen": 365166592
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008983149448345035,
      "loss": 1.7077,
      "theoretical_loss": 4.081755630313074,
      "tokens_seen": 365232128
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000898294884653962,
      "loss": 1.6648,
      "theoretical_loss": 4.081673021856445,
      "tokens_seen": 365297664
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008982748244734202,
      "loss": 1.7078,
      "theoretical_loss": 4.081590432367647,
      "tokens_seen": 365363200
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008982547642928787,
      "loss": 1.6293,
      "theoretical_loss": 4.081507861838922,
      "tokens_seen": 365428736
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008982347041123369,
      "loss": 1.6485,
      "theoretical_loss": 4.08142531026252,
      "tokens_seen": 365494272
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008982146439317954,
      "loss": 1.7936,
      "theoretical_loss": 4.081342777630692,
      "tokens_seen": 365559808
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008981945837512538,
      "loss": 1.6933,
      "theoretical_loss": 4.081260263935697,
      "tokens_seen": 365625344
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008981745235707121,
      "loss": 1.7872,
      "theoretical_loss": 4.081177769169795,
      "tokens_seen": 365690880
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008981544633901705,
      "loss": 1.6915,
      "theoretical_loss": 4.081095293325253,
      "tokens_seen": 365756416
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000898134403209629,
      "loss": 1.6843,
      "theoretical_loss": 4.081012836394344,
      "tokens_seen": 365821952
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008981143430290872,
      "loss": 1.7095,
      "theoretical_loss": 4.0809303983693415,
      "tokens_seen": 365887488
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008980942828485457,
      "loss": 1.6318,
      "theoretical_loss": 4.080847979242526,
      "tokens_seen": 365953024
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008980742226680039,
      "loss": 1.7328,
      "theoretical_loss": 4.080765579006181,
      "tokens_seen": 366018560
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008980541624874624,
      "loss": 1.7227,
      "theoretical_loss": 4.080683197652597,
      "tokens_seen": 366084096
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008980341023069208,
      "loss": 1.7973,
      "theoretical_loss": 4.080600835174067,
      "tokens_seen": 366149632
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008980140421263791,
      "loss": 1.7067,
      "theoretical_loss": 4.0805184915628905,
      "tokens_seen": 366215168
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008979939819458375,
      "loss": 1.724,
      "theoretical_loss": 4.080436166811368,
      "tokens_seen": 366280704
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008979739217652959,
      "loss": 1.6432,
      "theoretical_loss": 4.080353860911808,
      "tokens_seen": 366346240
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008979538615847542,
      "loss": 1.7096,
      "theoretical_loss": 4.080271573856523,
      "tokens_seen": 366411776
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008979338014042127,
      "loss": 1.7991,
      "theoretical_loss": 4.080189305637827,
      "tokens_seen": 366477312
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000897913741223671,
      "loss": 1.6541,
      "theoretical_loss": 4.080107056248043,
      "tokens_seen": 366542848
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008978936810431294,
      "loss": 1.7085,
      "theoretical_loss": 4.0800248256794935,
      "tokens_seen": 366608384
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008978736208625879,
      "loss": 1.7117,
      "theoretical_loss": 4.07994261392451,
      "tokens_seen": 366673920
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008978535606820461,
      "loss": 1.8144,
      "theoretical_loss": 4.079860420975429,
      "tokens_seen": 366739456
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008978335005015046,
      "loss": 1.8416,
      "theoretical_loss": 4.079778246824585,
      "tokens_seen": 366804992
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008978134403209629,
      "loss": 1.7117,
      "theoretical_loss": 4.0796960914643225,
      "tokens_seen": 366870528
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008977933801404213,
      "loss": 1.9746,
      "theoretical_loss": 4.079613954886991,
      "tokens_seen": 366936064
    },
    {
      "epoch": 0.11,
      "objective/train/advantage_avg": 0.4951712489128113,
      "objective/train/docs_used": 139849,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7049927711486816,
      "objective/train/original_loss": 1.7049928903579712,
      "objective/train/theoretical_loss": 4.07953183708494,
      "objective/train/tokens_used": 387461600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2454361617565155,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507649183273315,
      "objective/train/weighted_lm_loss": 1.7913212776184082,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.07953183708494,
      "tokens_seen": 367001600
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008977733199598797,
      "loss": 1.7663,
      "theoretical_loss": 4.07953183708494,
      "tokens_seen": 367001600
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008977532597793381,
      "loss": 1.7864,
      "theoretical_loss": 4.0794497380505295,
      "tokens_seen": 367067136
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008977331995987964,
      "loss": 1.7711,
      "theoretical_loss": 4.0793676577761175,
      "tokens_seen": 367132672
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008977131394182548,
      "loss": 1.7033,
      "theoretical_loss": 4.079285596254071,
      "tokens_seen": 367198208
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008976930792377132,
      "loss": 1.7523,
      "theoretical_loss": 4.079203553476759,
      "tokens_seen": 367263744
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008976730190571716,
      "loss": 1.7355,
      "theoretical_loss": 4.0791215294365575,
      "tokens_seen": 367329280
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008976529588766299,
      "loss": 1.7067,
      "theoretical_loss": 4.079039524125845,
      "tokens_seen": 367394816
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008976328986960883,
      "loss": 1.8486,
      "theoretical_loss": 4.078957537537004,
      "tokens_seen": 367460352
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008976128385155466,
      "loss": 1.7477,
      "theoretical_loss": 4.078875569662424,
      "tokens_seen": 367525888
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008975927783350051,
      "loss": 1.6189,
      "theoretical_loss": 4.078793620494497,
      "tokens_seen": 367591424
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008975727181544634,
      "loss": 1.6504,
      "theoretical_loss": 4.078711690025619,
      "tokens_seen": 367656960
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008975526579739218,
      "loss": 1.8155,
      "theoretical_loss": 4.0786297782481915,
      "tokens_seen": 367722496
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008975325977933802,
      "loss": 1.7302,
      "theoretical_loss": 4.07854788515462,
      "tokens_seen": 367788032
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008975125376128386,
      "loss": 1.6967,
      "theoretical_loss": 4.078466010737316,
      "tokens_seen": 367853568
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008974924774322969,
      "loss": 1.6561,
      "theoretical_loss": 4.078384154988692,
      "tokens_seen": 367919104
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008974724172517554,
      "loss": 1.7334,
      "theoretical_loss": 4.078302317901168,
      "tokens_seen": 367984640
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008974523570712136,
      "loss": 1.7801,
      "theoretical_loss": 4.078220499467168,
      "tokens_seen": 368050176
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008974322968906721,
      "loss": 1.7574,
      "theoretical_loss": 4.078138699679119,
      "tokens_seen": 368115712
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008974122367101304,
      "loss": 1.7281,
      "theoretical_loss": 4.0780569185294535,
      "tokens_seen": 368181248
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008973921765295888,
      "loss": 1.5919,
      "theoretical_loss": 4.077975156010609,
      "tokens_seen": 368246784
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008973721163490472,
      "loss": 1.8051,
      "theoretical_loss": 4.077893412115025,
      "tokens_seen": 368312320
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008973520561685055,
      "loss": 1.7476,
      "theoretical_loss": 4.077811686835148,
      "tokens_seen": 368377856
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008973319959879639,
      "loss": 1.6414,
      "theoretical_loss": 4.077729980163428,
      "tokens_seen": 368443392
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008973119358074224,
      "loss": 1.6917,
      "theoretical_loss": 4.07764829209232,
      "tokens_seen": 368508928
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008972918756268806,
      "loss": 1.6509,
      "theoretical_loss": 4.077566622614281,
      "tokens_seen": 368574464
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008972718154463391,
      "loss": 1.5947,
      "theoretical_loss": 4.077484971721775,
      "tokens_seen": 368640000
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008972517552657973,
      "loss": 1.6623,
      "theoretical_loss": 4.07740333940727,
      "tokens_seen": 368705536
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008972316950852558,
      "loss": 1.8069,
      "theoretical_loss": 4.077321725663237,
      "tokens_seen": 368771072
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008972116349047142,
      "loss": 1.7939,
      "theoretical_loss": 4.077240130482153,
      "tokens_seen": 368836608
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008971915747241725,
      "loss": 1.7251,
      "theoretical_loss": 4.077158553856497,
      "tokens_seen": 368902144
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008971715145436309,
      "loss": 1.7161,
      "theoretical_loss": 4.077076995778757,
      "tokens_seen": 368967680
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008971514543630894,
      "loss": 1.8689,
      "theoretical_loss": 4.076995456241421,
      "tokens_seen": 369033216
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008971313941825476,
      "loss": 1.7164,
      "theoretical_loss": 4.076913935236982,
      "tokens_seen": 369098752
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008971113340020061,
      "loss": 1.8138,
      "theoretical_loss": 4.076832432757939,
      "tokens_seen": 369164288
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008970912738214643,
      "loss": 1.703,
      "theoretical_loss": 4.0767509487967954,
      "tokens_seen": 369229824
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008970712136409228,
      "loss": 1.6488,
      "theoretical_loss": 4.076669483346056,
      "tokens_seen": 369295360
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008970511534603812,
      "loss": 1.6958,
      "theoretical_loss": 4.076588036398235,
      "tokens_seen": 369360896
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008970310932798395,
      "loss": 1.6487,
      "theoretical_loss": 4.076506607945845,
      "tokens_seen": 369426432
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008970110330992979,
      "loss": 1.7663,
      "theoretical_loss": 4.076425197981408,
      "tokens_seen": 369491968
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008969909729187563,
      "loss": 1.7376,
      "theoretical_loss": 4.076343806497446,
      "tokens_seen": 369557504
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008969709127382146,
      "loss": 1.6712,
      "theoretical_loss": 4.076262433486491,
      "tokens_seen": 369623040
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008969508525576731,
      "loss": 1.79,
      "theoretical_loss": 4.076181078941074,
      "tokens_seen": 369688576
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008969307923771313,
      "loss": 1.8104,
      "theoretical_loss": 4.076099742853732,
      "tokens_seen": 369754112
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008969107321965898,
      "loss": 1.7615,
      "theoretical_loss": 4.076018425217009,
      "tokens_seen": 369819648
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008968906720160481,
      "loss": 1.6527,
      "theoretical_loss": 4.075937126023448,
      "tokens_seen": 369885184
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008968706118355065,
      "loss": 1.6824,
      "theoretical_loss": 4.075855845265602,
      "tokens_seen": 369950720
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008968505516549649,
      "loss": 1.6784,
      "theoretical_loss": 4.075774582936024,
      "tokens_seen": 370016256
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008968304914744233,
      "loss": 1.737,
      "theoretical_loss": 4.075693339027274,
      "tokens_seen": 370081792
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008968104312938816,
      "loss": 1.7294,
      "theoretical_loss": 4.075612113531915,
      "tokens_seen": 370147328
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008967903711133401,
      "loss": 1.7808,
      "theoretical_loss": 4.0755309064425145,
      "tokens_seen": 370212864
    },
    {
      "epoch": 0.11,
      "objective/train/advantage_avg": 0.48967820405960083,
      "objective/train/docs_used": 141078,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6121673583984375,
      "objective/train/original_loss": 1.6121673583984375,
      "objective/train/theoretical_loss": 4.0754497177516456,
      "objective/train/tokens_used": 390738400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24023215472698212,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501888990402222,
      "objective/train/weighted_lm_loss": 1.6929486989974976,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.0754497177516456,
      "tokens_seen": 370278400
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008967703109327984,
      "loss": 1.6428,
      "theoretical_loss": 4.0754497177516456,
      "tokens_seen": 370278400
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008967502507522568,
      "loss": 1.7115,
      "theoretical_loss": 4.075368547451882,
      "tokens_seen": 370343936
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008967301905717151,
      "loss": 1.6581,
      "theoretical_loss": 4.075287395535807,
      "tokens_seen": 370409472
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008967101303911736,
      "loss": 1.6987,
      "theoretical_loss": 4.075206261996005,
      "tokens_seen": 370475008
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008966900702106319,
      "loss": 1.8119,
      "theoretical_loss": 4.075125146825064,
      "tokens_seen": 370540544
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008966700100300903,
      "loss": 1.6419,
      "theoretical_loss": 4.0750440500155785,
      "tokens_seen": 370606080
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008966499498495486,
      "loss": 1.7411,
      "theoretical_loss": 4.074962971560145,
      "tokens_seen": 370671616
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000896629889669007,
      "loss": 1.6313,
      "theoretical_loss": 4.074881911451368,
      "tokens_seen": 370737152
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008966098294884654,
      "loss": 1.7241,
      "theoretical_loss": 4.074800869681852,
      "tokens_seen": 370802688
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008965897693079238,
      "loss": 1.7189,
      "theoretical_loss": 4.074719846244209,
      "tokens_seen": 370868224
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008965697091273821,
      "loss": 1.6872,
      "theoretical_loss": 4.074638841131052,
      "tokens_seen": 370933760
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008965496489468406,
      "loss": 1.7023,
      "theoretical_loss": 4.074557854335002,
      "tokens_seen": 370999296
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008965295887662988,
      "loss": 1.7682,
      "theoretical_loss": 4.074476885848682,
      "tokens_seen": 371064832
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008965095285857573,
      "loss": 1.7406,
      "theoretical_loss": 4.07439593566472,
      "tokens_seen": 371130368
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008964894684052156,
      "loss": 1.7485,
      "theoretical_loss": 4.074315003775746,
      "tokens_seen": 371195904
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000896469408224674,
      "loss": 1.7826,
      "theoretical_loss": 4.0742340901744,
      "tokens_seen": 371261440
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008964493480441324,
      "loss": 1.7554,
      "theoretical_loss": 4.07415319485332,
      "tokens_seen": 371326976
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008964292878635908,
      "loss": 1.701,
      "theoretical_loss": 4.074072317805152,
      "tokens_seen": 371392512
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008964092276830491,
      "loss": 1.7843,
      "theoretical_loss": 4.073991459022544,
      "tokens_seen": 371458048
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008963891675025076,
      "loss": 1.6254,
      "theoretical_loss": 4.07391061849815,
      "tokens_seen": 371523584
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008963691073219659,
      "loss": 1.6961,
      "theoretical_loss": 4.073829796224628,
      "tokens_seen": 371589120
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008963490471414243,
      "loss": 1.6637,
      "theoretical_loss": 4.0737489921946395,
      "tokens_seen": 371654656
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008963289869608828,
      "loss": 1.6752,
      "theoretical_loss": 4.073668206400851,
      "tokens_seen": 371720192
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000896308926780341,
      "loss": 1.8764,
      "theoretical_loss": 4.073587438835932,
      "tokens_seen": 371785728
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008962888665997995,
      "loss": 1.8139,
      "theoretical_loss": 4.073506689492557,
      "tokens_seen": 371851264
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008962688064192577,
      "loss": 1.7659,
      "theoretical_loss": 4.073425958363406,
      "tokens_seen": 371916800
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008962487462387162,
      "loss": 1.8909,
      "theoretical_loss": 4.0733452454411605,
      "tokens_seen": 371982336
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008962286860581746,
      "loss": 1.6557,
      "theoretical_loss": 4.073264550718509,
      "tokens_seen": 372047872
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008962086258776329,
      "loss": 1.6719,
      "theoretical_loss": 4.0731838741881425,
      "tokens_seen": 372113408
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008961885656970913,
      "loss": 1.7788,
      "theoretical_loss": 4.073103215842757,
      "tokens_seen": 372178944
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008961685055165498,
      "loss": 1.8669,
      "theoretical_loss": 4.0730225756750515,
      "tokens_seen": 372244480
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000896148445336008,
      "loss": 1.7435,
      "theoretical_loss": 4.072941953677732,
      "tokens_seen": 372310016
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008961283851554665,
      "loss": 1.5605,
      "theoretical_loss": 4.072861349843505,
      "tokens_seen": 372375552
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008961083249749247,
      "loss": 1.7418,
      "theoretical_loss": 4.072780764165083,
      "tokens_seen": 372441088
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008960882647943832,
      "loss": 1.7104,
      "theoretical_loss": 4.072700196635185,
      "tokens_seen": 372506624
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008960682046138416,
      "loss": 1.8206,
      "theoretical_loss": 4.07261964724653,
      "tokens_seen": 372572160
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008960481444332999,
      "loss": 1.7754,
      "theoretical_loss": 4.072539115991844,
      "tokens_seen": 372637696
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008960280842527583,
      "loss": 1.6862,
      "theoretical_loss": 4.072458602863856,
      "tokens_seen": 372703232
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008960080240722167,
      "loss": 1.7193,
      "theoretical_loss": 4.072378107855299,
      "tokens_seen": 372768768
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000895987963891675,
      "loss": 1.6753,
      "theoretical_loss": 4.0722976309589125,
      "tokens_seen": 372834304
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008959679037111335,
      "loss": 1.705,
      "theoretical_loss": 4.072217172167436,
      "tokens_seen": 372899840
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008959478435305918,
      "loss": 1.5978,
      "theoretical_loss": 4.072136731473618,
      "tokens_seen": 372965376
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008959277833500502,
      "loss": 1.6881,
      "theoretical_loss": 4.0720563088702075,
      "tokens_seen": 373030912
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008959077231695085,
      "loss": 1.8104,
      "theoretical_loss": 4.0719759043499595,
      "tokens_seen": 373096448
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008958876629889669,
      "loss": 1.6941,
      "theoretical_loss": 4.071895517905633,
      "tokens_seen": 373161984
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008958676028084253,
      "loss": 1.6762,
      "theoretical_loss": 4.071815149529988,
      "tokens_seen": 373227520
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008958475426278837,
      "loss": 1.6635,
      "theoretical_loss": 4.071734799215796,
      "tokens_seen": 373293056
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000895827482447342,
      "loss": 1.6936,
      "theoretical_loss": 4.071654466955825,
      "tokens_seen": 373358592
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008958074222668005,
      "loss": 1.698,
      "theoretical_loss": 4.071574152742851,
      "tokens_seen": 373424128
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008957873620862588,
      "loss": 1.5491,
      "theoretical_loss": 4.071493856569655,
      "tokens_seen": 373489664
    },
    {
      "epoch": 0.11,
      "objective/train/advantage_avg": 0.4907689690589905,
      "objective/train/docs_used": 142333,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7129801511764526,
      "objective/train/original_loss": 1.7129802703857422,
      "objective/train/theoretical_loss": 4.071413578429017,
      "objective/train/tokens_used": 394015200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24147610366344452,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503044128417969,
      "objective/train/weighted_lm_loss": 1.7990201711654663,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.071413578429017,
      "tokens_seen": 373555200
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008957673019057172,
      "loss": 1.7972,
      "theoretical_loss": 4.071413578429017,
      "tokens_seen": 373555200
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008957472417251755,
      "loss": 1.6007,
      "theoretical_loss": 4.071333318313728,
      "tokens_seen": 373620736
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000895727181544634,
      "loss": 1.7788,
      "theoretical_loss": 4.07125307621658,
      "tokens_seen": 373686272
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008957071213640923,
      "loss": 1.8137,
      "theoretical_loss": 4.071172852130368,
      "tokens_seen": 373751808
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008956870611835507,
      "loss": 1.7266,
      "theoretical_loss": 4.071092646047892,
      "tokens_seen": 373817344
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000895667001003009,
      "loss": 1.7166,
      "theoretical_loss": 4.071012457961956,
      "tokens_seen": 373882880
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008956469408224674,
      "loss": 1.5693,
      "theoretical_loss": 4.07093228786537,
      "tokens_seen": 373948416
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008956268806419258,
      "loss": 1.7416,
      "theoretical_loss": 4.070852135750946,
      "tokens_seen": 374013952
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008956068204613842,
      "loss": 1.7768,
      "theoretical_loss": 4.0707720016115,
      "tokens_seen": 374079488
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008955867602808425,
      "loss": 1.6761,
      "theoretical_loss": 4.070691885439854,
      "tokens_seen": 374145024
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000895566700100301,
      "loss": 1.6422,
      "theoretical_loss": 4.070611787228832,
      "tokens_seen": 374210560
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008955466399197592,
      "loss": 1.7156,
      "theoretical_loss": 4.0705317069712645,
      "tokens_seen": 374276096
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008955265797392177,
      "loss": 1.6769,
      "theoretical_loss": 4.0704516446599825,
      "tokens_seen": 374341632
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000895506519558676,
      "loss": 1.7857,
      "theoretical_loss": 4.070371600287826,
      "tokens_seen": 374407168
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008954864593781344,
      "loss": 1.647,
      "theoretical_loss": 4.070291573847635,
      "tokens_seen": 374472704
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008954663991975928,
      "loss": 1.7466,
      "theoretical_loss": 4.070211565332255,
      "tokens_seen": 374538240
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008954463390170512,
      "loss": 1.8851,
      "theoretical_loss": 4.070131574734536,
      "tokens_seen": 374603776
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008954262788365095,
      "loss": 1.7502,
      "theoretical_loss": 4.070051602047332,
      "tokens_seen": 374669312
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000895406218655968,
      "loss": 1.7069,
      "theoretical_loss": 4.069971647263501,
      "tokens_seen": 374734848
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008953861584754262,
      "loss": 1.743,
      "theoretical_loss": 4.069891710375904,
      "tokens_seen": 374800384
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008953660982948847,
      "loss": 1.7041,
      "theoretical_loss": 4.069811791377409,
      "tokens_seen": 374865920
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000895346038114343,
      "loss": 1.6945,
      "theoretical_loss": 4.0697318902608846,
      "tokens_seen": 374931456
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008953259779338014,
      "loss": 1.685,
      "theoretical_loss": 4.069652007019204,
      "tokens_seen": 374996992
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008953059177532598,
      "loss": 1.6655,
      "theoretical_loss": 4.069572141645249,
      "tokens_seen": 375062528
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008952858575727181,
      "loss": 1.743,
      "theoretical_loss": 4.0694922941319,
      "tokens_seen": 375128064
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008952657973921765,
      "loss": 1.743,
      "theoretical_loss": 4.069412464472042,
      "tokens_seen": 375193600
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000895245737211635,
      "loss": 1.7983,
      "theoretical_loss": 4.069332652658567,
      "tokens_seen": 375259136
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008952256770310932,
      "loss": 1.6777,
      "theoretical_loss": 4.069252858684371,
      "tokens_seen": 375324672
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008952056168505517,
      "loss": 1.8022,
      "theoretical_loss": 4.069173082542351,
      "tokens_seen": 375390208
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008951855566700101,
      "loss": 1.8357,
      "theoretical_loss": 4.069093324225411,
      "tokens_seen": 375455744
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008951654964894684,
      "loss": 1.7366,
      "theoretical_loss": 4.0690135837264565,
      "tokens_seen": 375521280
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008951454363089268,
      "loss": 1.8672,
      "theoretical_loss": 4.068933861038399,
      "tokens_seen": 375586816
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008951253761283851,
      "loss": 1.7984,
      "theoretical_loss": 4.068854156154154,
      "tokens_seen": 375652352
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008951053159478435,
      "loss": 1.8244,
      "theoretical_loss": 4.068774469066639,
      "tokens_seen": 375717888
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000895085255767302,
      "loss": 1.6594,
      "theoretical_loss": 4.068694799768779,
      "tokens_seen": 375783424
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008950651955867602,
      "loss": 1.7258,
      "theoretical_loss": 4.0686151482535,
      "tokens_seen": 375848960
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008950451354062187,
      "loss": 1.7146,
      "theoretical_loss": 4.068535514513734,
      "tokens_seen": 375914496
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000895025075225677,
      "loss": 1.6562,
      "theoretical_loss": 4.068455898542414,
      "tokens_seen": 375980032
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008950050150451354,
      "loss": 1.8663,
      "theoretical_loss": 4.068376300332482,
      "tokens_seen": 376045568
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008949849548645938,
      "loss": 1.8702,
      "theoretical_loss": 4.06829671987688,
      "tokens_seen": 376111104
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008949648946840522,
      "loss": 1.6686,
      "theoretical_loss": 4.068217157168556,
      "tokens_seen": 376176640
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008949448345035105,
      "loss": 1.7105,
      "theoretical_loss": 4.06813761220046,
      "tokens_seen": 376242176
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008949247743229689,
      "loss": 1.7407,
      "theoretical_loss": 4.068058084965549,
      "tokens_seen": 376307712
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008949047141424272,
      "loss": 1.6749,
      "theoretical_loss": 4.067978575456781,
      "tokens_seen": 376373248
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008948846539618857,
      "loss": 1.7597,
      "theoretical_loss": 4.06789908366712,
      "tokens_seen": 376438784
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008948645937813441,
      "loss": 1.7951,
      "theoretical_loss": 4.067819609589533,
      "tokens_seen": 376504320
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008948445336008024,
      "loss": 1.7017,
      "theoretical_loss": 4.067740153216993,
      "tokens_seen": 376569856
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008948244734202609,
      "loss": 1.6655,
      "theoretical_loss": 4.067660714542473,
      "tokens_seen": 376635392
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008948044132397192,
      "loss": 1.9055,
      "theoretical_loss": 4.067581293558955,
      "tokens_seen": 376700928
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008947843530591776,
      "loss": 1.7544,
      "theoretical_loss": 4.06750189025942,
      "tokens_seen": 376766464
    },
    {
      "epoch": 0.11,
      "objective/train/advantage_avg": 0.49255064129829407,
      "objective/train/docs_used": 143470,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6976510286331177,
      "objective/train/original_loss": 1.6976510286331177,
      "objective/train/theoretical_loss": 4.067422504636857,
      "objective/train/tokens_used": 397292000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24296031892299652,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050490140914917,
      "objective/train/weighted_lm_loss": 1.7832144498825073,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.067422504636857,
      "tokens_seen": 376832000
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008947642928786359,
      "loss": 1.7729,
      "theoretical_loss": 4.067422504636857,
      "tokens_seen": 376832000
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008947442326980944,
      "loss": 1.8275,
      "theoretical_loss": 4.067343136684259,
      "tokens_seen": 376897536
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008947241725175527,
      "loss": 1.6847,
      "theoretical_loss": 4.0672637863946175,
      "tokens_seen": 376963072
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008947041123370111,
      "loss": 1.6855,
      "theoretical_loss": 4.067184453760936,
      "tokens_seen": 377028608
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008946840521564694,
      "loss": 1.7641,
      "theoretical_loss": 4.067105138776214,
      "tokens_seen": 377094144
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008946639919759278,
      "loss": 1.6745,
      "theoretical_loss": 4.067025841433463,
      "tokens_seen": 377159680
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008946439317953862,
      "loss": 1.6787,
      "theoretical_loss": 4.0669465617256915,
      "tokens_seen": 377225216
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008946238716148446,
      "loss": 1.6579,
      "theoretical_loss": 4.066867299645916,
      "tokens_seen": 377290752
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008946038114343029,
      "loss": 1.6609,
      "theoretical_loss": 4.066788055187156,
      "tokens_seen": 377356288
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008945837512537614,
      "loss": 1.6764,
      "theoretical_loss": 4.066708828342435,
      "tokens_seen": 377421824
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008945636910732196,
      "loss": 1.7242,
      "theoretical_loss": 4.06662961910478,
      "tokens_seen": 377487360
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008945436308926781,
      "loss": 1.7348,
      "theoretical_loss": 4.066550427467221,
      "tokens_seen": 377552896
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008945235707121364,
      "loss": 1.5869,
      "theoretical_loss": 4.066471253422796,
      "tokens_seen": 377618432
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008945035105315948,
      "loss": 1.6576,
      "theoretical_loss": 4.066392096964542,
      "tokens_seen": 377683968
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008944834503510532,
      "loss": 1.7433,
      "theoretical_loss": 4.066312958085503,
      "tokens_seen": 377749504
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008944633901705116,
      "loss": 1.5885,
      "theoretical_loss": 4.066233836778727,
      "tokens_seen": 377815040
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008944433299899699,
      "loss": 1.656,
      "theoretical_loss": 4.066154733037264,
      "tokens_seen": 377880576
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008944232698094284,
      "loss": 1.6272,
      "theoretical_loss": 4.066075646854169,
      "tokens_seen": 377946112
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008944032096288866,
      "loss": 1.8592,
      "theoretical_loss": 4.065996578222502,
      "tokens_seen": 378011648
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008943831494483451,
      "loss": 1.7786,
      "theoretical_loss": 4.065917527135325,
      "tokens_seen": 378077184
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008943630892678035,
      "loss": 1.6858,
      "theoretical_loss": 4.065838493585705,
      "tokens_seen": 378142720
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008943430290872618,
      "loss": 1.9012,
      "theoretical_loss": 4.065759477566714,
      "tokens_seen": 378208256
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008943229689067202,
      "loss": 1.604,
      "theoretical_loss": 4.065680479071426,
      "tokens_seen": 378273792
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008943029087261785,
      "loss": 1.7117,
      "theoretical_loss": 4.065601498092919,
      "tokens_seen": 378339328
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008942828485456369,
      "loss": 1.6169,
      "theoretical_loss": 4.065522534624276,
      "tokens_seen": 378404864
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008942627883650954,
      "loss": 1.6878,
      "theoretical_loss": 4.065443588658584,
      "tokens_seen": 378470400
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008942427281845536,
      "loss": 1.6946,
      "theoretical_loss": 4.0653646601889335,
      "tokens_seen": 378535936
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008942226680040121,
      "loss": 1.8113,
      "theoretical_loss": 4.065285749208419,
      "tokens_seen": 378601472
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008942026078234705,
      "loss": 1.7916,
      "theoretical_loss": 4.065206855710139,
      "tokens_seen": 378667008
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008941825476429288,
      "loss": 1.6653,
      "theoretical_loss": 4.065127979687195,
      "tokens_seen": 378732544
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008941624874623872,
      "loss": 1.5783,
      "theoretical_loss": 4.065049121132693,
      "tokens_seen": 378798080
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008941424272818455,
      "loss": 1.6457,
      "theoretical_loss": 4.064970280039745,
      "tokens_seen": 378863616
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008941223671013039,
      "loss": 1.7093,
      "theoretical_loss": 4.0648914564014635,
      "tokens_seen": 378929152
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008941023069207624,
      "loss": 1.6528,
      "theoretical_loss": 4.064812650210966,
      "tokens_seen": 378994688
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008940822467402206,
      "loss": 1.753,
      "theoretical_loss": 4.0647338614613755,
      "tokens_seen": 379060224
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008940621865596791,
      "loss": 1.6148,
      "theoretical_loss": 4.064655090145817,
      "tokens_seen": 379125760
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008940421263791374,
      "loss": 1.6967,
      "theoretical_loss": 4.064576336257421,
      "tokens_seen": 379191296
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008940220661985958,
      "loss": 1.6283,
      "theoretical_loss": 4.064497599789319,
      "tokens_seen": 379256832
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008940020060180542,
      "loss": 1.6937,
      "theoretical_loss": 4.0644188807346495,
      "tokens_seen": 379322368
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008939819458375126,
      "loss": 1.6933,
      "theoretical_loss": 4.064340179086555,
      "tokens_seen": 379387904
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0008939618856569709,
      "loss": 1.7946,
      "theoretical_loss": 4.0642614948381794,
      "tokens_seen": 379453440
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008939418254764293,
      "loss": 1.8811,
      "theoretical_loss": 4.064182827982672,
      "tokens_seen": 379518976
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008939217652958876,
      "loss": 1.6261,
      "theoretical_loss": 4.064104178513186,
      "tokens_seen": 379584512
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008939017051153461,
      "loss": 1.624,
      "theoretical_loss": 4.064025546422878,
      "tokens_seen": 379650048
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008938816449348044,
      "loss": 1.717,
      "theoretical_loss": 4.063946931704908,
      "tokens_seen": 379715584
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008938615847542628,
      "loss": 1.8072,
      "theoretical_loss": 4.063868334352442,
      "tokens_seen": 379781120
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008938415245737212,
      "loss": 1.7027,
      "theoretical_loss": 4.0637897543586465,
      "tokens_seen": 379846656
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008938214643931796,
      "loss": 1.712,
      "theoretical_loss": 4.0637111917166955,
      "tokens_seen": 379912192
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008938014042126379,
      "loss": 1.644,
      "theoretical_loss": 4.063632646419765,
      "tokens_seen": 379977728
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008937813440320963,
      "loss": 1.7294,
      "theoretical_loss": 4.063554118461034,
      "tokens_seen": 380043264
    },
    {
      "epoch": 0.12,
      "objective/train/advantage_avg": 0.4942775368690491,
      "objective/train/docs_used": 144698,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5595202445983887,
      "objective/train/original_loss": 1.5595201253890991,
      "objective/train/theoretical_loss": 4.063475607833687,
      "objective/train/tokens_used": 400568800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2447211593389511,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050671935081482,
      "objective/train/weighted_lm_loss": 1.6384665966033936,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.063475607833687,
      "tokens_seen": 380108800
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008937612838515546,
      "loss": 1.6146,
      "theoretical_loss": 4.063475607833687,
      "tokens_seen": 380108800
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008937412236710131,
      "loss": 1.7456,
      "theoretical_loss": 4.063397114530912,
      "tokens_seen": 380174336
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008937211634904714,
      "loss": 1.69,
      "theoretical_loss": 4.0633186385458995,
      "tokens_seen": 380239872
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008937011033099298,
      "loss": 1.687,
      "theoretical_loss": 4.063240179871846,
      "tokens_seen": 380305408
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008936810431293881,
      "loss": 1.7677,
      "theoretical_loss": 4.063161738501951,
      "tokens_seen": 380370944
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008936609829488466,
      "loss": 1.7485,
      "theoretical_loss": 4.0630833144294165,
      "tokens_seen": 380436480
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008936409227683049,
      "loss": 1.6952,
      "theoretical_loss": 4.06300490764745,
      "tokens_seen": 380502016
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008936208625877633,
      "loss": 1.7707,
      "theoretical_loss": 4.062926518149263,
      "tokens_seen": 380567552
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008936008024072217,
      "loss": 1.6512,
      "theoretical_loss": 4.0628481459280685,
      "tokens_seen": 380633088
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00089358074222668,
      "loss": 1.5992,
      "theoretical_loss": 4.062769790977086,
      "tokens_seen": 380698624
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008935606820461384,
      "loss": 1.6339,
      "theoretical_loss": 4.062691453289538,
      "tokens_seen": 380764160
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008935406218655968,
      "loss": 1.7438,
      "theoretical_loss": 4.062613132858651,
      "tokens_seen": 380829696
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008935205616850551,
      "loss": 1.7412,
      "theoretical_loss": 4.062534829677653,
      "tokens_seen": 380895232
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008935005015045136,
      "loss": 1.6496,
      "theoretical_loss": 4.0624565437397795,
      "tokens_seen": 380960768
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008934804413239719,
      "loss": 1.8742,
      "theoretical_loss": 4.0623782750382675,
      "tokens_seen": 381026304
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008934603811434303,
      "loss": 1.761,
      "theoretical_loss": 4.062300023566358,
      "tokens_seen": 381091840
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008934403209628887,
      "loss": 1.5984,
      "theoretical_loss": 4.062221789317297,
      "tokens_seen": 381157376
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000893420260782347,
      "loss": 1.68,
      "theoretical_loss": 4.0621435722843335,
      "tokens_seen": 381222912
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008934002006018054,
      "loss": 1.6351,
      "theoretical_loss": 4.06206537246072,
      "tokens_seen": 381288448
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008933801404212639,
      "loss": 1.6759,
      "theoretical_loss": 4.061987189839712,
      "tokens_seen": 381353984
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008933600802407222,
      "loss": 1.7164,
      "theoretical_loss": 4.061909024414572,
      "tokens_seen": 381419520
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008933400200601806,
      "loss": 1.8483,
      "theoretical_loss": 4.061830876178562,
      "tokens_seen": 381485056
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008933199598796389,
      "loss": 1.6855,
      "theoretical_loss": 4.061752745124952,
      "tokens_seen": 381550592
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008932998996990973,
      "loss": 1.5888,
      "theoretical_loss": 4.061674631247014,
      "tokens_seen": 381616128
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008932798395185558,
      "loss": 1.6356,
      "theoretical_loss": 4.061596534538021,
      "tokens_seen": 381681664
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000893259779338014,
      "loss": 1.7905,
      "theoretical_loss": 4.061518454991255,
      "tokens_seen": 381747200
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008932397191574725,
      "loss": 1.7029,
      "theoretical_loss": 4.061440392599997,
      "tokens_seen": 381812736
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008932196589769309,
      "loss": 1.7284,
      "theoretical_loss": 4.061362347357536,
      "tokens_seen": 381878272
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008931995987963892,
      "loss": 1.8156,
      "theoretical_loss": 4.061284319257162,
      "tokens_seen": 381943808
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008931795386158476,
      "loss": 1.8333,
      "theoretical_loss": 4.061206308292169,
      "tokens_seen": 382009344
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008931594784353059,
      "loss": 1.7346,
      "theoretical_loss": 4.061128314455855,
      "tokens_seen": 382074880
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008931394182547643,
      "loss": 1.6804,
      "theoretical_loss": 4.0610503377415235,
      "tokens_seen": 382140416
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008931193580742228,
      "loss": 1.7114,
      "theoretical_loss": 4.060972378142479,
      "tokens_seen": 382205952
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000893099297893681,
      "loss": 1.6782,
      "theoretical_loss": 4.060894435652031,
      "tokens_seen": 382271488
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008930792377131395,
      "loss": 1.8034,
      "theoretical_loss": 4.060816510263494,
      "tokens_seen": 382337024
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008930591775325978,
      "loss": 1.79,
      "theoretical_loss": 4.0607386019701845,
      "tokens_seen": 382402560
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008930391173520562,
      "loss": 1.6879,
      "theoretical_loss": 4.060660710765423,
      "tokens_seen": 382468096
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008930190571715146,
      "loss": 1.8295,
      "theoretical_loss": 4.060582836642534,
      "tokens_seen": 382533632
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000892998996990973,
      "loss": 1.6777,
      "theoretical_loss": 4.060504979594847,
      "tokens_seen": 382599168
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008929789368104313,
      "loss": 1.7917,
      "theoretical_loss": 4.060427139615693,
      "tokens_seen": 382664704
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008929588766298897,
      "loss": 1.7472,
      "theoretical_loss": 4.060349316698408,
      "tokens_seen": 382730240
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000892938816449348,
      "loss": 1.613,
      "theoretical_loss": 4.060271510836332,
      "tokens_seen": 382795776
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008929187562688065,
      "loss": 1.7459,
      "theoretical_loss": 4.060193722022808,
      "tokens_seen": 382861312
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008928986960882648,
      "loss": 1.6282,
      "theoretical_loss": 4.060115950251183,
      "tokens_seen": 382926848
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008928786359077232,
      "loss": 1.7833,
      "theoretical_loss": 4.060038195514808,
      "tokens_seen": 382992384
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008928585757271816,
      "loss": 1.6383,
      "theoretical_loss": 4.059960457807038,
      "tokens_seen": 383057920
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00089283851554664,
      "loss": 1.6525,
      "theoretical_loss": 4.05988273712123,
      "tokens_seen": 383123456
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008928184553660983,
      "loss": 1.7811,
      "theoretical_loss": 4.059805033450747,
      "tokens_seen": 383188992
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008927983951855567,
      "loss": 1.6016,
      "theoretical_loss": 4.059727346788955,
      "tokens_seen": 383254528
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000892778335005015,
      "loss": 1.7204,
      "theoretical_loss": 4.059649677129222,
      "tokens_seen": 383320064
    },
    {
      "epoch": 0.12,
      "objective/train/advantage_avg": 0.49355781078338623,
      "objective/train/docs_used": 145827,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6360312700271606,
      "objective/train/original_loss": 1.6360312700271606,
      "objective/train/theoretical_loss": 4.059572024464923,
      "objective/train/tokens_used": 403845600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24408653378486633,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505965948104858,
      "objective/train/weighted_lm_loss": 1.7188907861709595,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.059572024464923,
      "tokens_seen": 383385600
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008927582748244735,
      "loss": 1.6921,
      "theoretical_loss": 4.059572024464923,
      "tokens_seen": 383385600
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008927382146439318,
      "loss": 1.6756,
      "theoretical_loss": 4.0594943887894335,
      "tokens_seen": 383451136
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008927181544633902,
      "loss": 1.657,
      "theoretical_loss": 4.059416770096134,
      "tokens_seen": 383516672
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008926980942828485,
      "loss": 1.7426,
      "theoretical_loss": 4.0593391683784095,
      "tokens_seen": 383582208
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000892678034102307,
      "loss": 1.6976,
      "theoretical_loss": 4.059261583629648,
      "tokens_seen": 383647744
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008926579739217653,
      "loss": 1.5842,
      "theoretical_loss": 4.059184015843241,
      "tokens_seen": 383713280
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008926379137412237,
      "loss": 1.7801,
      "theoretical_loss": 4.059106465012583,
      "tokens_seen": 383778816
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000892617853560682,
      "loss": 1.7129,
      "theoretical_loss": 4.059028931131075,
      "tokens_seen": 383844352
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008925977933801404,
      "loss": 1.6377,
      "theoretical_loss": 4.058951414192118,
      "tokens_seen": 383909888
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008925777331995988,
      "loss": 1.7708,
      "theoretical_loss": 4.058873914189119,
      "tokens_seen": 383975424
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008925576730190572,
      "loss": 1.6659,
      "theoretical_loss": 4.058796431115489,
      "tokens_seen": 384040960
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008925376128385155,
      "loss": 1.7973,
      "theoretical_loss": 4.058718964964641,
      "tokens_seen": 384106496
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000892517552657974,
      "loss": 1.699,
      "theoretical_loss": 4.0586415157299935,
      "tokens_seen": 384172032
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008924974924774323,
      "loss": 1.6343,
      "theoretical_loss": 4.058564083404967,
      "tokens_seen": 384237568
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008924774322968907,
      "loss": 1.7538,
      "theoretical_loss": 4.058486667982986,
      "tokens_seen": 384303104
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008924573721163491,
      "loss": 1.6043,
      "theoretical_loss": 4.05840926945748,
      "tokens_seen": 384368640
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008924373119358074,
      "loss": 1.7165,
      "theoretical_loss": 4.058331887821879,
      "tokens_seen": 384434176
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008924172517552658,
      "loss": 1.6977,
      "theoretical_loss": 4.058254523069623,
      "tokens_seen": 384499712
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008923971915747243,
      "loss": 1.7091,
      "theoretical_loss": 4.058177175194148,
      "tokens_seen": 384565248
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008923771313941825,
      "loss": 1.6789,
      "theoretical_loss": 4.058099844188899,
      "tokens_seen": 384630784
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000892357071213641,
      "loss": 1.7138,
      "theoretical_loss": 4.058022530047323,
      "tokens_seen": 384696320
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008923370110330992,
      "loss": 1.6221,
      "theoretical_loss": 4.05794523276287,
      "tokens_seen": 384761856
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008923169508525577,
      "loss": 1.8069,
      "theoretical_loss": 4.057867952328994,
      "tokens_seen": 384827392
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008922968906720161,
      "loss": 1.7147,
      "theoretical_loss": 4.0577906887391535,
      "tokens_seen": 384892928
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008922768304914744,
      "loss": 1.83,
      "theoretical_loss": 4.05771344198681,
      "tokens_seen": 384958464
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008922567703109328,
      "loss": 1.6993,
      "theoretical_loss": 4.057636212065429,
      "tokens_seen": 385024000
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008922367101303913,
      "loss": 1.6941,
      "theoretical_loss": 4.057558998968479,
      "tokens_seen": 385089536
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008922166499498495,
      "loss": 1.7719,
      "theoretical_loss": 4.0574818026894315,
      "tokens_seen": 385155072
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000892196589769308,
      "loss": 1.757,
      "theoretical_loss": 4.057404623221765,
      "tokens_seen": 385220608
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008921765295887662,
      "loss": 1.723,
      "theoretical_loss": 4.057327460558957,
      "tokens_seen": 385286144
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008921564694082247,
      "loss": 1.7299,
      "theoretical_loss": 4.0572503146944925,
      "tokens_seen": 385351680
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008921364092276831,
      "loss": 1.7218,
      "theoretical_loss": 4.057173185621857,
      "tokens_seen": 385417216
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008921163490471414,
      "loss": 1.706,
      "theoretical_loss": 4.057096073334543,
      "tokens_seen": 385482752
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008920962888665998,
      "loss": 1.7013,
      "theoretical_loss": 4.057018977826044,
      "tokens_seen": 385548288
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008920762286860582,
      "loss": 1.7012,
      "theoretical_loss": 4.056941899089858,
      "tokens_seen": 385613824
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008920561685055165,
      "loss": 1.646,
      "theoretical_loss": 4.056864837119486,
      "tokens_seen": 385679360
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000892036108324975,
      "loss": 1.7769,
      "theoretical_loss": 4.056787791908434,
      "tokens_seen": 385744896
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008920160481444332,
      "loss": 1.6899,
      "theoretical_loss": 4.05671076345021,
      "tokens_seen": 385810432
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008919959879638917,
      "loss": 1.6936,
      "theoretical_loss": 4.056633751738328,
      "tokens_seen": 385875968
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00089197592778335,
      "loss": 1.8251,
      "theoretical_loss": 4.056556756766303,
      "tokens_seen": 385941504
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008919558676028084,
      "loss": 1.6149,
      "theoretical_loss": 4.056479778527653,
      "tokens_seen": 386007040
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008919358074222668,
      "loss": 1.669,
      "theoretical_loss": 4.056402817015905,
      "tokens_seen": 386072576
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008919157472417252,
      "loss": 1.6661,
      "theoretical_loss": 4.0563258722245825,
      "tokens_seen": 386138112
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008918956870611835,
      "loss": 1.6326,
      "theoretical_loss": 4.056248944147217,
      "tokens_seen": 386203648
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000891875626880642,
      "loss": 1.7272,
      "theoretical_loss": 4.0561720327773445,
      "tokens_seen": 386269184
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008918555667001004,
      "loss": 1.7083,
      "theoretical_loss": 4.0560951381085,
      "tokens_seen": 386334720
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008918355065195587,
      "loss": 1.6469,
      "theoretical_loss": 4.056018260134226,
      "tokens_seen": 386400256
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008918154463390171,
      "loss": 1.6153,
      "theoretical_loss": 4.055941398848067,
      "tokens_seen": 386465792
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008917953861584754,
      "loss": 1.6674,
      "theoretical_loss": 4.055864554243572,
      "tokens_seen": 386531328
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008917753259779339,
      "loss": 1.7016,
      "theoretical_loss": 4.055787726314292,
      "tokens_seen": 386596864
    },
    {
      "epoch": 0.12,
      "objective/train/advantage_avg": 0.49315446615219116,
      "objective/train/docs_used": 146981,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7423198223114014,
      "objective/train/original_loss": 1.7423198223114014,
      "objective/train/theoretical_loss": 4.055710915053783,
      "objective/train/tokens_used": 407122400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24348418414592743,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505532026290894,
      "objective/train/weighted_lm_loss": 1.8302770853042603,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 4.055710915053783,
      "tokens_seen": 386662400
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008917552657973922,
      "loss": 1.8378,
      "theoretical_loss": 4.055710915053783,
      "tokens_seen": 386662400
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008917352056168506,
      "loss": 1.745,
      "theoretical_loss": 4.055634120455606,
      "tokens_seen": 386727936
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008917151454363089,
      "loss": 1.6814,
      "theoretical_loss": 4.055557342513321,
      "tokens_seen": 386793472
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008916950852557674,
      "loss": 1.6931,
      "theoretical_loss": 4.055480581220497,
      "tokens_seen": 386859008
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008916750250752257,
      "loss": 1.565,
      "theoretical_loss": 4.055403836570701,
      "tokens_seen": 386924544
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008916549648946841,
      "loss": 1.6658,
      "theoretical_loss": 4.055327108557509,
      "tokens_seen": 386990080
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008916349047141425,
      "loss": 1.7828,
      "theoretical_loss": 4.055250397174497,
      "tokens_seen": 387055616
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008916148445336008,
      "loss": 1.7989,
      "theoretical_loss": 4.055173702415246,
      "tokens_seen": 387121152
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008915947843530592,
      "loss": 1.8036,
      "theoretical_loss": 4.05509702427334,
      "tokens_seen": 387186688
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008915747241725176,
      "loss": 1.6286,
      "theoretical_loss": 4.055020362742367,
      "tokens_seen": 387252224
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008915546639919759,
      "loss": 1.6351,
      "theoretical_loss": 4.054943717815917,
      "tokens_seen": 387317760
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008915346038114344,
      "loss": 1.7932,
      "theoretical_loss": 4.054867089487587,
      "tokens_seen": 387383296
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008915145436308927,
      "loss": 1.6156,
      "theoretical_loss": 4.054790477750974,
      "tokens_seen": 387448832
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008914944834503511,
      "loss": 1.6245,
      "theoretical_loss": 4.054713882599681,
      "tokens_seen": 387514368
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008914744232698095,
      "loss": 1.7205,
      "theoretical_loss": 4.054637304027312,
      "tokens_seen": 387579904
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008914543630892678,
      "loss": 1.706,
      "theoretical_loss": 4.054560742027478,
      "tokens_seen": 387645440
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008914343029087262,
      "loss": 1.7275,
      "theoretical_loss": 4.054484196593791,
      "tokens_seen": 387710976
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008914142427281847,
      "loss": 1.8033,
      "theoretical_loss": 4.054407667719866,
      "tokens_seen": 387776512
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008913941825476429,
      "loss": 1.7001,
      "theoretical_loss": 4.054331155399325,
      "tokens_seen": 387842048
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008913741223671014,
      "loss": 1.664,
      "theoretical_loss": 4.054254659625788,
      "tokens_seen": 387907584
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008913540621865596,
      "loss": 1.6946,
      "theoretical_loss": 4.054178180392885,
      "tokens_seen": 387973120
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008913340020060181,
      "loss": 1.654,
      "theoretical_loss": 4.054101717694244,
      "tokens_seen": 388038656
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008913139418254765,
      "loss": 1.5095,
      "theoretical_loss": 4.0540252715235,
      "tokens_seen": 388104192
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008912938816449348,
      "loss": 1.5775,
      "theoretical_loss": 4.05394884187429,
      "tokens_seen": 388169728
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008912738214643932,
      "loss": 1.6788,
      "theoretical_loss": 4.053872428740256,
      "tokens_seen": 388235264
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008912537612838517,
      "loss": 1.8278,
      "theoretical_loss": 4.053796032115041,
      "tokens_seen": 388300800
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008912337011033099,
      "loss": 1.7206,
      "theoretical_loss": 4.053719651992293,
      "tokens_seen": 388366336
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008912136409227684,
      "loss": 1.7555,
      "theoretical_loss": 4.053643288365664,
      "tokens_seen": 388431872
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008911935807422266,
      "loss": 1.5701,
      "theoretical_loss": 4.053566941228809,
      "tokens_seen": 388497408
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008911735205616851,
      "loss": 1.6847,
      "theoretical_loss": 4.053490610575386,
      "tokens_seen": 388562944
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008911534603811435,
      "loss": 1.7782,
      "theoretical_loss": 4.0534142963990565,
      "tokens_seen": 388628480
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008911334002006018,
      "loss": 1.7133,
      "theoretical_loss": 4.053337998693488,
      "tokens_seen": 388694016
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008911133400200602,
      "loss": 1.7067,
      "theoretical_loss": 4.053261717452346,
      "tokens_seen": 388759552
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008910932798395186,
      "loss": 1.6116,
      "theoretical_loss": 4.053185452669307,
      "tokens_seen": 388825088
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008910732196589769,
      "loss": 1.7676,
      "theoretical_loss": 4.053109204338045,
      "tokens_seen": 388890624
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008910531594784354,
      "loss": 1.579,
      "theoretical_loss": 4.05303297245224,
      "tokens_seen": 388956160
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008910330992978936,
      "loss": 1.7036,
      "theoretical_loss": 4.052956757005573,
      "tokens_seen": 389021696
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008910130391173521,
      "loss": 1.7259,
      "theoretical_loss": 4.052880557991733,
      "tokens_seen": 389087232
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008909929789368104,
      "loss": 1.5456,
      "theoretical_loss": 4.052804375404408,
      "tokens_seen": 389152768
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008909729187562688,
      "loss": 1.7429,
      "theoretical_loss": 4.052728209237293,
      "tokens_seen": 389218304
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008909528585757272,
      "loss": 1.6288,
      "theoretical_loss": 4.0526520594840845,
      "tokens_seen": 389283840
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008909327983951856,
      "loss": 1.743,
      "theoretical_loss": 4.052575926138482,
      "tokens_seen": 389349376
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008909127382146439,
      "loss": 1.7862,
      "theoretical_loss": 4.052499809194191,
      "tokens_seen": 389414912
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008908926780341024,
      "loss": 1.7565,
      "theoretical_loss": 4.052423708644917,
      "tokens_seen": 389480448
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008908726178535606,
      "loss": 1.7019,
      "theoretical_loss": 4.052347624484373,
      "tokens_seen": 389545984
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008908525576730191,
      "loss": 1.5311,
      "theoretical_loss": 4.052271556706271,
      "tokens_seen": 389611520
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008908324974924774,
      "loss": 1.6859,
      "theoretical_loss": 4.052195505304329,
      "tokens_seen": 389677056
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008908124373119358,
      "loss": 1.7945,
      "theoretical_loss": 4.05211947027227,
      "tokens_seen": 389742592
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008907923771313942,
      "loss": 1.7262,
      "theoretical_loss": 4.052043451603818,
      "tokens_seen": 389808128
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008907723169508526,
      "loss": 1.6705,
      "theoretical_loss": 4.051967449292699,
      "tokens_seen": 389873664
    },
    {
      "epoch": 0.12,
      "objective/train/advantage_avg": 0.4954656660556793,
      "objective/train/docs_used": 148229,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7292380332946777,
      "objective/train/original_loss": 1.7292380332946777,
      "objective/train/theoretical_loss": 4.051891463332648,
      "objective/train/tokens_used": 410399200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2457554042339325,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507960319519043,
      "objective/train/weighted_lm_loss": 1.817002534866333,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.051891463332648,
      "tokens_seen": 389939200
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008907522567703109,
      "loss": 1.7942,
      "theoretical_loss": 4.051891463332648,
      "tokens_seen": 389939200
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008907321965897693,
      "loss": 1.7186,
      "theoretical_loss": 4.051815493717399,
      "tokens_seen": 390004736
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008907121364092277,
      "loss": 1.7202,
      "theoretical_loss": 4.051739540440688,
      "tokens_seen": 390070272
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008906920762286861,
      "loss": 1.7049,
      "theoretical_loss": 4.0516636034962605,
      "tokens_seen": 390135808
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008906720160481444,
      "loss": 1.6774,
      "theoretical_loss": 4.051587682877859,
      "tokens_seen": 390201344
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008906519558676029,
      "loss": 1.6931,
      "theoretical_loss": 4.0515117785792345,
      "tokens_seen": 390266880
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008906318956870611,
      "loss": 1.5631,
      "theoretical_loss": 4.0514358905941386,
      "tokens_seen": 390332416
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008906118355065196,
      "loss": 1.6724,
      "theoretical_loss": 4.051360018916327,
      "tokens_seen": 390397952
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008905917753259779,
      "loss": 1.7544,
      "theoretical_loss": 4.051284163539558,
      "tokens_seen": 390463488
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008905717151454363,
      "loss": 1.7298,
      "theoretical_loss": 4.051208324457595,
      "tokens_seen": 390529024
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008905516549648947,
      "loss": 1.628,
      "theoretical_loss": 4.051132501664204,
      "tokens_seen": 390594560
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008905315947843531,
      "loss": 1.8071,
      "theoretical_loss": 4.0510566951531555,
      "tokens_seen": 390660096
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008905115346038114,
      "loss": 1.6248,
      "theoretical_loss": 4.05098090491822,
      "tokens_seen": 390725632
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008904914744232699,
      "loss": 1.6874,
      "theoretical_loss": 4.0509051309531765,
      "tokens_seen": 390791168
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008904714142427281,
      "loss": 1.8073,
      "theoretical_loss": 4.050829373251803,
      "tokens_seen": 390856704
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008904513540621866,
      "loss": 1.7927,
      "theoretical_loss": 4.050753631807883,
      "tokens_seen": 390922240
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008904312938816449,
      "loss": 1.7519,
      "theoretical_loss": 4.050677906615203,
      "tokens_seen": 390987776
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008904112337011033,
      "loss": 1.751,
      "theoretical_loss": 4.050602197667553,
      "tokens_seen": 391053312
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008903911735205617,
      "loss": 1.6699,
      "theoretical_loss": 4.050526504958727,
      "tokens_seen": 391118848
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00089037111334002,
      "loss": 1.6841,
      "theoretical_loss": 4.050450828482521,
      "tokens_seen": 391184384
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008903510531594785,
      "loss": 1.4873,
      "theoretical_loss": 4.050375168232737,
      "tokens_seen": 391249920
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008903309929789369,
      "loss": 1.6055,
      "theoretical_loss": 4.050299524203176,
      "tokens_seen": 391315456
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008903109327983952,
      "loss": 1.772,
      "theoretical_loss": 4.050223896387647,
      "tokens_seen": 391380992
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008902908726178536,
      "loss": 1.7945,
      "theoretical_loss": 4.05014828477996,
      "tokens_seen": 391446528
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008902708124373121,
      "loss": 1.6458,
      "theoretical_loss": 4.050072689373928,
      "tokens_seen": 391512064
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008902507522567703,
      "loss": 1.6568,
      "theoretical_loss": 4.049997110163369,
      "tokens_seen": 391577600
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008902306920762288,
      "loss": 1.6862,
      "theoretical_loss": 4.0499215471421035,
      "tokens_seen": 391643136
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000890210631895687,
      "loss": 1.7335,
      "theoretical_loss": 4.049846000303956,
      "tokens_seen": 391708672
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008901905717151455,
      "loss": 1.6957,
      "theoretical_loss": 4.049770469642753,
      "tokens_seen": 391774208
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008901705115346039,
      "loss": 1.6915,
      "theoretical_loss": 4.049694955152326,
      "tokens_seen": 391839744
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008901504513540622,
      "loss": 1.717,
      "theoretical_loss": 4.049619456826508,
      "tokens_seen": 391905280
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008901303911735206,
      "loss": 1.664,
      "theoretical_loss": 4.049543974659139,
      "tokens_seen": 391970816
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000890110330992979,
      "loss": 1.55,
      "theoretical_loss": 4.049468508644058,
      "tokens_seen": 392036352
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008900902708124373,
      "loss": 1.7676,
      "theoretical_loss": 4.049393058775109,
      "tokens_seen": 392101888
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008900702106318958,
      "loss": 1.7747,
      "theoretical_loss": 4.0493176250461405,
      "tokens_seen": 392167424
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000890050150451354,
      "loss": 1.6398,
      "theoretical_loss": 4.0492422074510035,
      "tokens_seen": 392232960
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008900300902708125,
      "loss": 1.778,
      "theoretical_loss": 4.049166805983553,
      "tokens_seen": 392298496
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008900100300902708,
      "loss": 1.716,
      "theoretical_loss": 4.049091420637646,
      "tokens_seen": 392364032
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008899899699097292,
      "loss": 1.645,
      "theoretical_loss": 4.049016051407145,
      "tokens_seen": 392429568
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008899699097291876,
      "loss": 1.5746,
      "theoretical_loss": 4.048940698285913,
      "tokens_seen": 392495104
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000889949849548646,
      "loss": 1.625,
      "theoretical_loss": 4.048865361267817,
      "tokens_seen": 392560640
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008899297893681043,
      "loss": 1.7186,
      "theoretical_loss": 4.048790040346731,
      "tokens_seen": 392626176
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008899097291875628,
      "loss": 1.7445,
      "theoretical_loss": 4.048714735516527,
      "tokens_seen": 392691712
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000889889669007021,
      "loss": 1.8012,
      "theoretical_loss": 4.048639446771085,
      "tokens_seen": 392757248
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008898696088264795,
      "loss": 1.7188,
      "theoretical_loss": 4.048564174104285,
      "tokens_seen": 392822784
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008898495486459378,
      "loss": 1.7577,
      "theoretical_loss": 4.048488917510012,
      "tokens_seen": 392888320
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008898294884653962,
      "loss": 1.7189,
      "theoretical_loss": 4.048413676982155,
      "tokens_seen": 392953856
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008898094282848546,
      "loss": 1.6715,
      "theoretical_loss": 4.048338452514604,
      "tokens_seen": 393019392
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000889789368104313,
      "loss": 1.7221,
      "theoretical_loss": 4.048263244101253,
      "tokens_seen": 393084928
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008897693079237713,
      "loss": 1.6308,
      "theoretical_loss": 4.048188051736003,
      "tokens_seen": 393150464
    },
    {
      "epoch": 0.12,
      "objective/train/advantage_avg": 0.49091511964797974,
      "objective/train/docs_used": 149327,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7502202987670898,
      "objective/train/original_loss": 1.7502202987670898,
      "objective/train/theoretical_loss": 4.048112875412752,
      "objective/train/tokens_used": 413676000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24176764488220215,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503205060958862,
      "objective/train/weighted_lm_loss": 1.838315725326538,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.048112875412752,
      "tokens_seen": 393216000
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008897492477432297,
      "loss": 1.7159,
      "theoretical_loss": 4.048112875412752,
      "tokens_seen": 393216000
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008897291875626881,
      "loss": 1.7339,
      "theoretical_loss": 4.048037715125407,
      "tokens_seen": 393281536
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008897091273821465,
      "loss": 1.6511,
      "theoretical_loss": 4.047962570867876,
      "tokens_seen": 393347072
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008896890672016048,
      "loss": 1.7161,
      "theoretical_loss": 4.047887442634067,
      "tokens_seen": 393412608
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008896690070210633,
      "loss": 1.7808,
      "theoretical_loss": 4.0478123304179,
      "tokens_seen": 393478144
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008896489468405215,
      "loss": 1.7958,
      "theoretical_loss": 4.0477372342132885,
      "tokens_seen": 393543680
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00088962888665998,
      "loss": 1.7283,
      "theoretical_loss": 4.047662154014157,
      "tokens_seen": 393609216
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008896088264794383,
      "loss": 1.7392,
      "theoretical_loss": 4.047587089814427,
      "tokens_seen": 393674752
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008895887662988967,
      "loss": 1.7529,
      "theoretical_loss": 4.047512041608029,
      "tokens_seen": 393740288
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008895687061183551,
      "loss": 1.7174,
      "theoretical_loss": 4.047437009388894,
      "tokens_seen": 393805824
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008895486459378135,
      "loss": 1.7715,
      "theoretical_loss": 4.047361993150955,
      "tokens_seen": 393871360
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008895285857572718,
      "loss": 1.6531,
      "theoretical_loss": 4.047286992888151,
      "tokens_seen": 393936896
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008895085255767303,
      "loss": 1.7352,
      "theoretical_loss": 4.047212008594424,
      "tokens_seen": 394002432
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008894884653961885,
      "loss": 1.5426,
      "theoretical_loss": 4.047137040263717,
      "tokens_seen": 394067968
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000889468405215647,
      "loss": 1.7159,
      "theoretical_loss": 4.047062087889978,
      "tokens_seen": 394133504
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008894483450351053,
      "loss": 1.6867,
      "theoretical_loss": 4.046987151467159,
      "tokens_seen": 394199040
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008894282848545637,
      "loss": 1.7914,
      "theoretical_loss": 4.046912230989214,
      "tokens_seen": 394264576
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008894082246740221,
      "loss": 1.7054,
      "theoretical_loss": 4.046837326450102,
      "tokens_seen": 394330112
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008893881644934804,
      "loss": 1.7448,
      "theoretical_loss": 4.046762437843781,
      "tokens_seen": 394395648
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008893681043129388,
      "loss": 1.7828,
      "theoretical_loss": 4.046687565164217,
      "tokens_seen": 394461184
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008893480441323973,
      "loss": 1.6893,
      "theoretical_loss": 4.0466127084053785,
      "tokens_seen": 394526720
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008893279839518555,
      "loss": 1.7992,
      "theoretical_loss": 4.046537867561234,
      "tokens_seen": 394592256
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000889307923771314,
      "loss": 1.8085,
      "theoretical_loss": 4.04646304262576,
      "tokens_seen": 394657792
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008892878635907722,
      "loss": 1.6054,
      "theoretical_loss": 4.046388233592933,
      "tokens_seen": 394723328
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008892678034102307,
      "loss": 1.6152,
      "theoretical_loss": 4.046313440456733,
      "tokens_seen": 394788864
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008892477432296891,
      "loss": 1.6747,
      "theoretical_loss": 4.046238663211145,
      "tokens_seen": 394854400
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008892276830491474,
      "loss": 1.7601,
      "theoretical_loss": 4.046163901850156,
      "tokens_seen": 394919936
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008892076228686058,
      "loss": 1.7451,
      "theoretical_loss": 4.046089156367756,
      "tokens_seen": 394985472
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008891875626880643,
      "loss": 1.7268,
      "theoretical_loss": 4.0460144267579405,
      "tokens_seen": 395051008
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008891675025075225,
      "loss": 1.7339,
      "theoretical_loss": 4.045939713014705,
      "tokens_seen": 395116544
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000889147442326981,
      "loss": 1.7601,
      "theoretical_loss": 4.045865015132049,
      "tokens_seen": 395182080
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008891273821464392,
      "loss": 1.7139,
      "theoretical_loss": 4.045790333103978,
      "tokens_seen": 395247616
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008891073219658977,
      "loss": 1.7569,
      "theoretical_loss": 4.045715666924499,
      "tokens_seen": 395313152
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008890872617853561,
      "loss": 1.6522,
      "theoretical_loss": 4.04564101658762,
      "tokens_seen": 395378688
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008890672016048144,
      "loss": 1.7247,
      "theoretical_loss": 4.045566382087356,
      "tokens_seen": 395444224
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008890471414242728,
      "loss": 1.6179,
      "theoretical_loss": 4.045491763417724,
      "tokens_seen": 395509760
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008890270812437312,
      "loss": 1.8097,
      "theoretical_loss": 4.045417160572743,
      "tokens_seen": 395575296
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008890070210631895,
      "loss": 1.6343,
      "theoretical_loss": 4.045342573546435,
      "tokens_seen": 395640832
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000888986960882648,
      "loss": 1.7958,
      "theoretical_loss": 4.045268002332829,
      "tokens_seen": 395706368
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008889669007021063,
      "loss": 1.755,
      "theoretical_loss": 4.0451934469259525,
      "tokens_seen": 395771904
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008889468405215647,
      "loss": 1.787,
      "theoretical_loss": 4.045118907319839,
      "tokens_seen": 395837440
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000888926780341023,
      "loss": 1.6481,
      "theoretical_loss": 4.045044383508526,
      "tokens_seen": 395902976
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008889067201604815,
      "loss": 1.7161,
      "theoretical_loss": 4.044969875486051,
      "tokens_seen": 395968512
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008888866599799399,
      "loss": 1.6305,
      "theoretical_loss": 4.044895383246457,
      "tokens_seen": 396034048
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008888665997993982,
      "loss": 1.706,
      "theoretical_loss": 4.04482090678379,
      "tokens_seen": 396099584
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008888465396188566,
      "loss": 1.6584,
      "theoretical_loss": 4.044746446092098,
      "tokens_seen": 396165120
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000888826479438315,
      "loss": 1.7699,
      "theoretical_loss": 4.044672001165436,
      "tokens_seen": 396230656
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008888064192577734,
      "loss": 1.8087,
      "theoretical_loss": 4.0445975719978575,
      "tokens_seen": 396296192
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008887863590772317,
      "loss": 1.8132,
      "theoretical_loss": 4.044523158583421,
      "tokens_seen": 396361728
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008887662988966901,
      "loss": 1.8177,
      "theoretical_loss": 4.044448760916189,
      "tokens_seen": 396427264
    },
    {
      "epoch": 0.12,
      "objective/train/advantage_avg": 0.4924962520599365,
      "objective/train/docs_used": 150450,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.605733036994934,
      "objective/train/original_loss": 1.6057329177856445,
      "objective/train/theoretical_loss": 4.044374378990227,
      "objective/train/tokens_used": 416952800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24324040114879608,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504863262176514,
      "objective/train/weighted_lm_loss": 1.6867058277130127,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.044374378990227,
      "tokens_seen": 396492800
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008887462387161485,
      "loss": 1.6627,
      "theoretical_loss": 4.044374378990227,
      "tokens_seen": 396492800
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008887261785356069,
      "loss": 1.7032,
      "theoretical_loss": 4.044300012799603,
      "tokens_seen": 396558336
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008887061183550652,
      "loss": 1.7416,
      "theoretical_loss": 4.044225662338388,
      "tokens_seen": 396623872
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008886860581745237,
      "loss": 1.8403,
      "theoretical_loss": 4.044151327600657,
      "tokens_seen": 396689408
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008886659979939819,
      "loss": 1.6319,
      "theoretical_loss": 4.04407700858049,
      "tokens_seen": 396754944
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008886459378134404,
      "loss": 1.759,
      "theoretical_loss": 4.044002705271966,
      "tokens_seen": 396820480
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008886258776328987,
      "loss": 1.6112,
      "theoretical_loss": 4.04392841766917,
      "tokens_seen": 396886016
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008886058174523571,
      "loss": 1.5804,
      "theoretical_loss": 4.04385414576619,
      "tokens_seen": 396951552
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008885857572718155,
      "loss": 1.5721,
      "theoretical_loss": 4.043779889557116,
      "tokens_seen": 397017088
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008885656970912739,
      "loss": 1.7787,
      "theoretical_loss": 4.043705649036043,
      "tokens_seen": 397082624
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008885456369107322,
      "loss": 1.7641,
      "theoretical_loss": 4.043631424197067,
      "tokens_seen": 397148160
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008885255767301907,
      "loss": 1.6464,
      "theoretical_loss": 4.043557215034291,
      "tokens_seen": 397213696
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008885055165496489,
      "loss": 1.6706,
      "theoretical_loss": 4.043483021541815,
      "tokens_seen": 397279232
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008884854563691074,
      "loss": 1.8047,
      "theoretical_loss": 4.043408843713749,
      "tokens_seen": 397344768
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008884653961885657,
      "loss": 1.8259,
      "theoretical_loss": 4.0433346815442,
      "tokens_seen": 397410304
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008884453360080241,
      "loss": 1.778,
      "theoretical_loss": 4.043260535027285,
      "tokens_seen": 397475840
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008884252758274825,
      "loss": 1.7848,
      "theoretical_loss": 4.043186404157117,
      "tokens_seen": 397541376
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008884052156469408,
      "loss": 1.7881,
      "theoretical_loss": 4.0431122889278175,
      "tokens_seen": 397606912
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008883851554663992,
      "loss": 1.7349,
      "theoretical_loss": 4.043038189333508,
      "tokens_seen": 397672448
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008883650952858577,
      "loss": 1.8538,
      "theoretical_loss": 4.0429641053683145,
      "tokens_seen": 397737984
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008883450351053159,
      "loss": 1.7809,
      "theoretical_loss": 4.042890037026368,
      "tokens_seen": 397803520
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008883249749247744,
      "loss": 1.7622,
      "theoretical_loss": 4.042815984301798,
      "tokens_seen": 397869056
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008883049147442326,
      "loss": 1.7507,
      "theoretical_loss": 4.042741947188741,
      "tokens_seen": 397934592
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008882848545636911,
      "loss": 1.7028,
      "theoretical_loss": 4.042667925681337,
      "tokens_seen": 398000128
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008882647943831495,
      "loss": 1.7455,
      "theoretical_loss": 4.042593919773726,
      "tokens_seen": 398065664
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008882447342026078,
      "loss": 1.59,
      "theoretical_loss": 4.0425199294600525,
      "tokens_seen": 398131200
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008882246740220662,
      "loss": 1.7077,
      "theoretical_loss": 4.042445954734466,
      "tokens_seen": 398196736
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008882046138415247,
      "loss": 1.8407,
      "theoretical_loss": 4.042371995591118,
      "tokens_seen": 398262272
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008881845536609829,
      "loss": 1.8003,
      "theoretical_loss": 4.042298052024162,
      "tokens_seen": 398327808
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008881644934804414,
      "loss": 1.8477,
      "theoretical_loss": 4.042224124027754,
      "tokens_seen": 398393344
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008881444332998996,
      "loss": 1.7051,
      "theoretical_loss": 4.0421502115960575,
      "tokens_seen": 398458880
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008881243731193581,
      "loss": 1.5807,
      "theoretical_loss": 4.042076314723236,
      "tokens_seen": 398524416
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008881043129388165,
      "loss": 1.6232,
      "theoretical_loss": 4.0420024334034546,
      "tokens_seen": 398589952
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008880842527582748,
      "loss": 1.7288,
      "theoretical_loss": 4.041928567630885,
      "tokens_seen": 398655488
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008880641925777332,
      "loss": 1.6554,
      "theoretical_loss": 4.0418547173997,
      "tokens_seen": 398721024
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008880441323971916,
      "loss": 1.7113,
      "theoretical_loss": 4.041780882704076,
      "tokens_seen": 398786560
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008880240722166499,
      "loss": 1.7606,
      "theoretical_loss": 4.041707063538193,
      "tokens_seen": 398852096
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008880040120361084,
      "loss": 1.7467,
      "theoretical_loss": 4.0416332598962335,
      "tokens_seen": 398917632
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008879839518555667,
      "loss": 1.6222,
      "theoretical_loss": 4.041559471772382,
      "tokens_seen": 398983168
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008879638916750251,
      "loss": 1.77,
      "theoretical_loss": 4.0414856991608294,
      "tokens_seen": 399048704
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008879438314944834,
      "loss": 1.6854,
      "theoretical_loss": 4.041411942055767,
      "tokens_seen": 399114240
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008879237713139419,
      "loss": 1.7659,
      "theoretical_loss": 4.04133820045139,
      "tokens_seen": 399179776
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008879037111334002,
      "loss": 1.8196,
      "theoretical_loss": 4.041264474341896,
      "tokens_seen": 399245312
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008878836509528586,
      "loss": 1.8015,
      "theoretical_loss": 4.041190763721488,
      "tokens_seen": 399310848
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008878635907723169,
      "loss": 1.7079,
      "theoretical_loss": 4.04111706858437,
      "tokens_seen": 399376384
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008878435305917754,
      "loss": 1.7024,
      "theoretical_loss": 4.04104338892475,
      "tokens_seen": 399441920
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008878234704112337,
      "loss": 1.6853,
      "theoretical_loss": 4.040969724736838,
      "tokens_seen": 399507456
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008878034102306921,
      "loss": 1.6133,
      "theoretical_loss": 4.040896076014848,
      "tokens_seen": 399572992
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008877833500501504,
      "loss": 1.7696,
      "theoretical_loss": 4.040822442752997,
      "tokens_seen": 399638528
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008877632898696089,
      "loss": 1.6849,
      "theoretical_loss": 4.040748824945507,
      "tokens_seen": 399704064
    },
    {
      "epoch": 0.12,
      "objective/train/advantage_avg": 0.4920065701007843,
      "objective/train/docs_used": 151634,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6224642992019653,
      "objective/train/original_loss": 1.6224645376205444,
      "objective/train/theoretical_loss": 4.040675222586599,
      "objective/train/tokens_used": 420229600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24278749525547028,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050434947013855,
      "objective/train/weighted_lm_loss": 1.7043081521987915,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.040675222586599,
      "tokens_seen": 399769600
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008877432296890672,
      "loss": 1.6058,
      "theoretical_loss": 4.040675222586599,
      "tokens_seen": 399769600
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008877231695085256,
      "loss": 1.5935,
      "theoretical_loss": 4.040601635670501,
      "tokens_seen": 399835136
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008877031093279839,
      "loss": 1.6786,
      "theoretical_loss": 4.040528064191441,
      "tokens_seen": 399900672
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008876830491474423,
      "loss": 1.6391,
      "theoretical_loss": 4.040454508143652,
      "tokens_seen": 399966208
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008876629889669007,
      "loss": 1.6811,
      "theoretical_loss": 4.04038096752137,
      "tokens_seen": 400031744
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008876429287863591,
      "loss": 1.6204,
      "theoretical_loss": 4.040307442318834,
      "tokens_seen": 400097280
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008876228686058174,
      "loss": 1.7193,
      "theoretical_loss": 4.0402339325302865,
      "tokens_seen": 400162816
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008876028084252759,
      "loss": 1.772,
      "theoretical_loss": 4.040160438149971,
      "tokens_seen": 400228352
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008875827482447342,
      "loss": 1.7396,
      "theoretical_loss": 4.040086959172136,
      "tokens_seen": 400293888
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008875626880641926,
      "loss": 1.6233,
      "theoretical_loss": 4.040013495591033,
      "tokens_seen": 400359424
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000887542627883651,
      "loss": 1.6648,
      "theoretical_loss": 4.039940047400916,
      "tokens_seen": 400424960
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008875225677031093,
      "loss": 1.6596,
      "theoretical_loss": 4.0398666145960425,
      "tokens_seen": 400490496
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008875025075225677,
      "loss": 1.5847,
      "theoretical_loss": 4.039793197170672,
      "tokens_seen": 400556032
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008874824473420261,
      "loss": 1.6665,
      "theoretical_loss": 4.03971979511907,
      "tokens_seen": 400621568
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008874623871614844,
      "loss": 1.8048,
      "theoretical_loss": 4.039646408435502,
      "tokens_seen": 400687104
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008874423269809429,
      "loss": 1.7428,
      "theoretical_loss": 4.039573037114236,
      "tokens_seen": 400752640
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008874222668004011,
      "loss": 1.737,
      "theoretical_loss": 4.039499681149547,
      "tokens_seen": 400818176
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008874022066198596,
      "loss": 1.7766,
      "theoretical_loss": 4.0394263405357105,
      "tokens_seen": 400883712
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008873821464393181,
      "loss": 1.7353,
      "theoretical_loss": 4.039353015267004,
      "tokens_seen": 400949248
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008873620862587763,
      "loss": 1.6646,
      "theoretical_loss": 4.0392797053377105,
      "tokens_seen": 401014784
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008873420260782348,
      "loss": 1.7196,
      "theoretical_loss": 4.039206410742114,
      "tokens_seen": 401080320
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000887321965897693,
      "loss": 1.7357,
      "theoretical_loss": 4.039133131474505,
      "tokens_seen": 401145856
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008873019057171515,
      "loss": 1.5398,
      "theoretical_loss": 4.039059867529171,
      "tokens_seen": 401211392
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008872818455366099,
      "loss": 1.7319,
      "theoretical_loss": 4.0389866189004096,
      "tokens_seen": 401276928
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008872617853560682,
      "loss": 1.6942,
      "theoretical_loss": 4.038913385582515,
      "tokens_seen": 401342464
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008872417251755266,
      "loss": 1.6616,
      "theoretical_loss": 4.03884016756979,
      "tokens_seen": 401408000
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008872216649949851,
      "loss": 1.6534,
      "theoretical_loss": 4.038766964856537,
      "tokens_seen": 401473536
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008872016048144433,
      "loss": 1.675,
      "theoretical_loss": 4.038693777437061,
      "tokens_seen": 401539072
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008871815446339018,
      "loss": 1.7144,
      "theoretical_loss": 4.038620605305673,
      "tokens_seen": 401604608
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00088716148445336,
      "loss": 1.6308,
      "theoretical_loss": 4.0385474484566855,
      "tokens_seen": 401670144
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008871414242728185,
      "loss": 1.6743,
      "theoretical_loss": 4.038474306884414,
      "tokens_seen": 401735680
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008871213640922769,
      "loss": 1.7202,
      "theoretical_loss": 4.038401180583175,
      "tokens_seen": 401801216
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008871013039117352,
      "loss": 1.5014,
      "theoretical_loss": 4.038328069547293,
      "tokens_seen": 401866752
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008870812437311936,
      "loss": 1.7556,
      "theoretical_loss": 4.0382549737710915,
      "tokens_seen": 401932288
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000887061183550652,
      "loss": 1.6331,
      "theoretical_loss": 4.038181893248898,
      "tokens_seen": 401997824
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008870411233701103,
      "loss": 1.7542,
      "theoretical_loss": 4.038108827975043,
      "tokens_seen": 402063360
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008870210631895688,
      "loss": 1.6324,
      "theoretical_loss": 4.03803577794386,
      "tokens_seen": 402128896
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008870010030090271,
      "loss": 1.6813,
      "theoretical_loss": 4.037962743149687,
      "tokens_seen": 402194432
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008869809428284855,
      "loss": 1.6299,
      "theoretical_loss": 4.037889723586863,
      "tokens_seen": 402259968
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008869608826479438,
      "loss": 1.8496,
      "theoretical_loss": 4.03781671924973,
      "tokens_seen": 402325504
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008869408224674023,
      "loss": 1.6649,
      "theoretical_loss": 4.037743730132635,
      "tokens_seen": 402391040
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008869207622868606,
      "loss": 1.6666,
      "theoretical_loss": 4.037670756229927,
      "tokens_seen": 402456576
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000886900702106319,
      "loss": 1.5544,
      "theoretical_loss": 4.037597797535957,
      "tokens_seen": 402522112
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008868806419257773,
      "loss": 1.6848,
      "theoretical_loss": 4.03752485404508,
      "tokens_seen": 402587648
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008868605817452358,
      "loss": 1.6837,
      "theoretical_loss": 4.037451925751654,
      "tokens_seen": 402653184
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008868405215646941,
      "loss": 1.7491,
      "theoretical_loss": 4.037379012650041,
      "tokens_seen": 402718720
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008868204613841525,
      "loss": 1.6797,
      "theoretical_loss": 4.037306114734603,
      "tokens_seen": 402784256
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008868004012036108,
      "loss": 1.7346,
      "theoretical_loss": 4.037233231999708,
      "tokens_seen": 402849792
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008867803410230693,
      "loss": 1.6458,
      "theoretical_loss": 4.0371603644397265,
      "tokens_seen": 402915328
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008867602808425276,
      "loss": 1.7762,
      "theoretical_loss": 4.03708751204903,
      "tokens_seen": 402980864
    },
    {
      "epoch": 0.12,
      "objective/train/advantage_avg": 0.4884590804576874,
      "objective/train/docs_used": 152845,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.555906891822815,
      "objective/train/original_loss": 1.5559066534042358,
      "objective/train/theoretical_loss": 4.037014674821996,
      "objective/train/tokens_used": 423506400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23926691710948944,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500620603561401,
      "objective/train/weighted_lm_loss": 1.6336621046066284,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.037014674821996,
      "tokens_seen": 403046400
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000886740220661986,
      "loss": 1.7122,
      "theoretical_loss": 4.037014674821996,
      "tokens_seen": 403046400
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008867201604814443,
      "loss": 1.781,
      "theoretical_loss": 4.036941852753001,
      "tokens_seen": 403111936
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008867001003009027,
      "loss": 1.6895,
      "theoretical_loss": 4.03686904583643,
      "tokens_seen": 403177472
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008866800401203611,
      "loss": 1.7167,
      "theoretical_loss": 4.036796254066666,
      "tokens_seen": 403243008
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008866599799398195,
      "loss": 1.6275,
      "theoretical_loss": 4.036723477438097,
      "tokens_seen": 403308544
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008866399197592778,
      "loss": 1.6639,
      "theoretical_loss": 4.0366507159451155,
      "tokens_seen": 403374080
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008866198595787363,
      "loss": 1.7336,
      "theoretical_loss": 4.036577969582114,
      "tokens_seen": 403439616
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008865997993981946,
      "loss": 1.7358,
      "theoretical_loss": 4.036505238343488,
      "tokens_seen": 403505152
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000886579739217653,
      "loss": 1.6951,
      "theoretical_loss": 4.03643252222364,
      "tokens_seen": 403570688
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008865596790371114,
      "loss": 1.5281,
      "theoretical_loss": 4.036359821216973,
      "tokens_seen": 403636224
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008865396188565697,
      "loss": 1.6679,
      "theoretical_loss": 4.03628713531789,
      "tokens_seen": 403701760
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008865195586760281,
      "loss": 1.6987,
      "theoretical_loss": 4.036214464520802,
      "tokens_seen": 403767296
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008864994984954865,
      "loss": 1.7154,
      "theoretical_loss": 4.03614180882012,
      "tokens_seen": 403832832
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008864794383149448,
      "loss": 1.7357,
      "theoretical_loss": 4.036069168210259,
      "tokens_seen": 403898368
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008864593781344033,
      "loss": 1.7683,
      "theoretical_loss": 4.035996542685638,
      "tokens_seen": 403963904
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008864393179538615,
      "loss": 1.5992,
      "theoretical_loss": 4.035923932240674,
      "tokens_seen": 404029440
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00088641925777332,
      "loss": 1.6875,
      "theoretical_loss": 4.035851336869795,
      "tokens_seen": 404094976
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008863991975927784,
      "loss": 1.6608,
      "theoretical_loss": 4.035778756567426,
      "tokens_seen": 404160512
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008863791374122367,
      "loss": 1.7379,
      "theoretical_loss": 4.0357061913279955,
      "tokens_seen": 404226048
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008863590772316951,
      "loss": 1.687,
      "theoretical_loss": 4.035633641145937,
      "tokens_seen": 404291584
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008863390170511534,
      "loss": 1.6801,
      "theoretical_loss": 4.035561106015686,
      "tokens_seen": 404357120
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008863189568706118,
      "loss": 1.7063,
      "theoretical_loss": 4.035488585931681,
      "tokens_seen": 404422656
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008862988966900703,
      "loss": 1.6407,
      "theoretical_loss": 4.035416080888364,
      "tokens_seen": 404488192
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008862788365095285,
      "loss": 1.7144,
      "theoretical_loss": 4.035343590880178,
      "tokens_seen": 404553728
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000886258776328987,
      "loss": 1.6793,
      "theoretical_loss": 4.035271115901571,
      "tokens_seen": 404619264
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008862387161484454,
      "loss": 1.8213,
      "theoretical_loss": 4.035198655946994,
      "tokens_seen": 404684800
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008862186559679037,
      "loss": 1.7371,
      "theoretical_loss": 4.035126211010899,
      "tokens_seen": 404750336
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008861985957873621,
      "loss": 1.5614,
      "theoretical_loss": 4.035053781087743,
      "tokens_seen": 404815872
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008861785356068204,
      "loss": 1.6326,
      "theoretical_loss": 4.034981366171985,
      "tokens_seen": 404881408
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008861584754262788,
      "loss": 1.6813,
      "theoretical_loss": 4.0349089662580875,
      "tokens_seen": 404946944
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008861384152457373,
      "loss": 1.7213,
      "theoretical_loss": 4.034836581340515,
      "tokens_seen": 405012480
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008861183550651955,
      "loss": 1.6701,
      "theoretical_loss": 4.034764211413735,
      "tokens_seen": 405078016
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000886098294884654,
      "loss": 1.6912,
      "theoretical_loss": 4.034691856472219,
      "tokens_seen": 405143552
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008860782347041123,
      "loss": 1.6378,
      "theoretical_loss": 4.034619516510441,
      "tokens_seen": 405209088
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008860581745235707,
      "loss": 1.7264,
      "theoretical_loss": 4.034547191522877,
      "tokens_seen": 405274624
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008860381143430291,
      "loss": 1.8048,
      "theoretical_loss": 4.034474881504007,
      "tokens_seen": 405340160
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008860180541624875,
      "loss": 1.743,
      "theoretical_loss": 4.034402586448315,
      "tokens_seen": 405405696
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008859979939819458,
      "loss": 1.7185,
      "theoretical_loss": 4.0343303063502844,
      "tokens_seen": 405471232
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008859779338014042,
      "loss": 1.9438,
      "theoretical_loss": 4.034258041204404,
      "tokens_seen": 405536768
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008859578736208625,
      "loss": 1.6551,
      "theoretical_loss": 4.034185791005166,
      "tokens_seen": 405602304
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000885937813440321,
      "loss": 1.703,
      "theoretical_loss": 4.034113555747065,
      "tokens_seen": 405667840
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008859177532597793,
      "loss": 1.73,
      "theoretical_loss": 4.034041335424597,
      "tokens_seen": 405733376
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008858976930792377,
      "loss": 1.6462,
      "theoretical_loss": 4.033969130032263,
      "tokens_seen": 405798912
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008858776328986962,
      "loss": 1.6174,
      "theoretical_loss": 4.033896939564565,
      "tokens_seen": 405864448
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008858575727181545,
      "loss": 1.6147,
      "theoretical_loss": 4.0338247640160105,
      "tokens_seen": 405929984
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008858375125376129,
      "loss": 1.8021,
      "theoretical_loss": 4.0337526033811075,
      "tokens_seen": 405995520
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008858174523570712,
      "loss": 1.6092,
      "theoretical_loss": 4.033680457654368,
      "tokens_seen": 406061056
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008857973921765297,
      "loss": 1.576,
      "theoretical_loss": 4.033608326830306,
      "tokens_seen": 406126592
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000885777331995988,
      "loss": 1.7152,
      "theoretical_loss": 4.033536210903439,
      "tokens_seen": 406192128
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008857572718154464,
      "loss": 1.7216,
      "theoretical_loss": 4.033464109868289,
      "tokens_seen": 406257664
    },
    {
      "epoch": 0.12,
      "objective/train/advantage_avg": 0.49587690830230713,
      "objective/train/docs_used": 154148,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5138962268829346,
      "objective/train/original_loss": 1.5138959884643555,
      "objective/train/theoretical_loss": 4.0333920237193785,
      "objective/train/tokens_used": 426783200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24605120718479156,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508387088775635,
      "objective/train/weighted_lm_loss": 1.5908278226852417,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.0333920237193785,
      "tokens_seen": 406323200
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008857372116349047,
      "loss": 1.614,
      "theoretical_loss": 4.0333920237193785,
      "tokens_seen": 406323200
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008857171514543631,
      "loss": 1.5859,
      "theoretical_loss": 4.033319952451233,
      "tokens_seen": 406388736
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008856970912738215,
      "loss": 1.5933,
      "theoretical_loss": 4.033247896058382,
      "tokens_seen": 406454272
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008856770310932799,
      "loss": 1.6813,
      "theoretical_loss": 4.03317585453536,
      "tokens_seen": 406519808
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008856569709127382,
      "loss": 1.4754,
      "theoretical_loss": 4.0331038278766975,
      "tokens_seen": 406585344
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008856369107321967,
      "loss": 1.6915,
      "theoretical_loss": 4.033031816076936,
      "tokens_seen": 406650880
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000885616850551655,
      "loss": 1.7038,
      "theoretical_loss": 4.032959819130615,
      "tokens_seen": 406716416
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008855967903711134,
      "loss": 1.7263,
      "theoretical_loss": 4.032887837032277,
      "tokens_seen": 406781952
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008855767301905718,
      "loss": 1.8097,
      "theoretical_loss": 4.032815869776471,
      "tokens_seen": 406847488
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008855566700100301,
      "loss": 1.7635,
      "theoretical_loss": 4.0327439173577435,
      "tokens_seen": 406913024
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008855366098294885,
      "loss": 1.7598,
      "theoretical_loss": 4.032671979770649,
      "tokens_seen": 406978560
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008855165496489469,
      "loss": 1.6579,
      "theoretical_loss": 4.032600057009741,
      "tokens_seen": 407044096
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008854964894684052,
      "loss": 1.6921,
      "theoretical_loss": 4.032528149069579,
      "tokens_seen": 407109632
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008854764292878637,
      "loss": 1.7444,
      "theoretical_loss": 4.032456255944723,
      "tokens_seen": 407175168
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008854563691073219,
      "loss": 1.6354,
      "theoretical_loss": 4.032384377629737,
      "tokens_seen": 407240704
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008854363089267804,
      "loss": 1.84,
      "theoretical_loss": 4.032312514119187,
      "tokens_seen": 407306240
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008854162487462388,
      "loss": 1.6797,
      "theoretical_loss": 4.0322406654076435,
      "tokens_seen": 407371776
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008853961885656971,
      "loss": 1.7291,
      "theoretical_loss": 4.032168831489678,
      "tokens_seen": 407437312
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008853761283851555,
      "loss": 1.7825,
      "theoretical_loss": 4.032097012359867,
      "tokens_seen": 407502848
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008853560682046138,
      "loss": 1.6212,
      "theoretical_loss": 4.0320252080127865,
      "tokens_seen": 407568384
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008853360080240722,
      "loss": 1.7734,
      "theoretical_loss": 4.0319534184430195,
      "tokens_seen": 407633920
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008853159478435307,
      "loss": 1.6859,
      "theoretical_loss": 4.031881643645149,
      "tokens_seen": 407699456
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008852958876629889,
      "loss": 1.7826,
      "theoretical_loss": 4.031809883613761,
      "tokens_seen": 407764992
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008852758274824474,
      "loss": 1.5211,
      "theoretical_loss": 4.031738138343446,
      "tokens_seen": 407830528
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008852557673019058,
      "loss": 1.7162,
      "theoretical_loss": 4.031666407828795,
      "tokens_seen": 407896064
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008852357071213641,
      "loss": 1.7394,
      "theoretical_loss": 4.031594692064405,
      "tokens_seen": 407961600
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008852156469408225,
      "loss": 1.6073,
      "theoretical_loss": 4.031522991044873,
      "tokens_seen": 408027136
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008851955867602808,
      "loss": 1.7767,
      "theoretical_loss": 4.0314513047648,
      "tokens_seen": 408092672
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008851755265797392,
      "loss": 1.5802,
      "theoretical_loss": 4.03137963321879,
      "tokens_seen": 408158208
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008851554663991977,
      "loss": 1.5961,
      "theoretical_loss": 4.031307976401449,
      "tokens_seen": 408223744
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008851354062186559,
      "loss": 1.7465,
      "theoretical_loss": 4.031236334307387,
      "tokens_seen": 408289280
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008851153460381144,
      "loss": 1.8301,
      "theoretical_loss": 4.031164706931215,
      "tokens_seen": 408354816
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008850952858575727,
      "loss": 1.6845,
      "theoretical_loss": 4.0310930942675505,
      "tokens_seen": 408420352
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008850752256770311,
      "loss": 1.7814,
      "theoretical_loss": 4.03102149631101,
      "tokens_seen": 408485888
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008850551654964895,
      "loss": 1.5873,
      "theoretical_loss": 4.030949913056213,
      "tokens_seen": 408551424
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008850351053159479,
      "loss": 1.7496,
      "theoretical_loss": 4.030878344497786,
      "tokens_seen": 408616960
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008850150451354062,
      "loss": 1.6609,
      "theoretical_loss": 4.030806790630353,
      "tokens_seen": 408682496
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008849949849548646,
      "loss": 1.6315,
      "theoretical_loss": 4.030735251448545,
      "tokens_seen": 408748032
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008849749247743229,
      "loss": 1.6608,
      "theoretical_loss": 4.030663726946992,
      "tokens_seen": 408813568
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008849548645937814,
      "loss": 1.7091,
      "theoretical_loss": 4.030592217120331,
      "tokens_seen": 408879104
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008849348044132397,
      "loss": 1.605,
      "theoretical_loss": 4.030520721963199,
      "tokens_seen": 408944640
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008849147442326981,
      "loss": 1.6562,
      "theoretical_loss": 4.030449241470235,
      "tokens_seen": 409010176
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008848946840521565,
      "loss": 1.6737,
      "theoretical_loss": 4.030377775636084,
      "tokens_seen": 409075712
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008848746238716149,
      "loss": 1.6781,
      "theoretical_loss": 4.030306324455392,
      "tokens_seen": 409141248
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008848545636910732,
      "loss": 1.5943,
      "theoretical_loss": 4.030234887922808,
      "tokens_seen": 409206784
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008848345035105316,
      "loss": 1.6954,
      "theoretical_loss": 4.030163466032984,
      "tokens_seen": 409272320
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00088481444332999,
      "loss": 1.6496,
      "theoretical_loss": 4.030092058780573,
      "tokens_seen": 409337856
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008847943831494484,
      "loss": 1.8028,
      "theoretical_loss": 4.030020666160234,
      "tokens_seen": 409403392
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008847743229689067,
      "loss": 1.6431,
      "theoretical_loss": 4.029949288166627,
      "tokens_seen": 409468928
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008847542627883651,
      "loss": 1.5435,
      "theoretical_loss": 4.029877924794415,
      "tokens_seen": 409534464
    },
    {
      "epoch": 0.12,
      "objective/train/advantage_avg": 0.49289989471435547,
      "objective/train/docs_used": 155224,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.73809814453125,
      "objective/train/original_loss": 1.738098382949829,
      "objective/train/theoretical_loss": 4.029806576038263,
      "objective/train/tokens_used": 430060000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24331925809383392,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505269765853882,
      "objective/train/weighted_lm_loss": 1.8255553245544434,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.029806576038263,
      "tokens_seen": 409600000
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008847342026078234,
      "loss": 1.6939,
      "theoretical_loss": 4.029806576038263,
      "tokens_seen": 409600000
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008847141424272819,
      "loss": 1.5399,
      "theoretical_loss": 4.0297352418928405,
      "tokens_seen": 409665536
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008846940822467402,
      "loss": 1.5897,
      "theoretical_loss": 4.0296639223528175,
      "tokens_seen": 409731072
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008846740220661986,
      "loss": 1.7601,
      "theoretical_loss": 4.02959261741287,
      "tokens_seen": 409796608
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000884653961885657,
      "loss": 1.4896,
      "theoretical_loss": 4.029521327067675,
      "tokens_seen": 409862144
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008846339017051154,
      "loss": 1.6404,
      "theoretical_loss": 4.029450051311912,
      "tokens_seen": 409927680
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008846138415245737,
      "loss": 1.7851,
      "theoretical_loss": 4.029378790140261,
      "tokens_seen": 409993216
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008845937813440322,
      "loss": 1.6865,
      "theoretical_loss": 4.029307543547411,
      "tokens_seen": 410058752
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008845737211634904,
      "loss": 1.6179,
      "theoretical_loss": 4.029236311528049,
      "tokens_seen": 410124288
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008845536609829489,
      "loss": 1.6504,
      "theoretical_loss": 4.029165094076864,
      "tokens_seen": 410189824
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008845336008024072,
      "loss": 1.784,
      "theoretical_loss": 4.029093891188552,
      "tokens_seen": 410255360
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008845135406218656,
      "loss": 1.7336,
      "theoretical_loss": 4.029022702857809,
      "tokens_seen": 410320896
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000884493480441324,
      "loss": 1.7137,
      "theoretical_loss": 4.028951529079334,
      "tokens_seen": 410386432
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008844734202607823,
      "loss": 1.6833,
      "theoretical_loss": 4.028880369847829,
      "tokens_seen": 410451968
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008844533600802407,
      "loss": 1.6997,
      "theoretical_loss": 4.028809225158,
      "tokens_seen": 410517504
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008844332998996992,
      "loss": 1.7912,
      "theoretical_loss": 4.0287380950045515,
      "tokens_seen": 410583040
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008844132397191574,
      "loss": 1.5385,
      "theoretical_loss": 4.028666979382197,
      "tokens_seen": 410648576
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008843931795386159,
      "loss": 1.6839,
      "theoretical_loss": 4.028595878285648,
      "tokens_seen": 410714112
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008843731193580742,
      "loss": 1.6628,
      "theoretical_loss": 4.028524791709621,
      "tokens_seen": 410779648
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008843530591775326,
      "loss": 1.6706,
      "theoretical_loss": 4.0284537196488355,
      "tokens_seen": 410845184
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008843329989969911,
      "loss": 1.6575,
      "theoretical_loss": 4.028382662098011,
      "tokens_seen": 410910720
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008843129388164493,
      "loss": 1.6833,
      "theoretical_loss": 4.028311619051873,
      "tokens_seen": 410976256
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008842928786359078,
      "loss": 1.7105,
      "theoretical_loss": 4.028240590505148,
      "tokens_seen": 411041792
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008842728184553662,
      "loss": 1.6891,
      "theoretical_loss": 4.028169576452566,
      "tokens_seen": 411107328
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008842527582748245,
      "loss": 1.6028,
      "theoretical_loss": 4.028098576888859,
      "tokens_seen": 411172864
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008842326980942829,
      "loss": 1.8081,
      "theoretical_loss": 4.028027591808764,
      "tokens_seen": 411238400
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008842126379137413,
      "loss": 1.6304,
      "theoretical_loss": 4.027956621207015,
      "tokens_seen": 411303936
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008841925777331996,
      "loss": 1.7428,
      "theoretical_loss": 4.027885665078356,
      "tokens_seen": 411369472
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008841725175526581,
      "loss": 1.6705,
      "theoretical_loss": 4.0278147234175306,
      "tokens_seen": 411435008
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008841524573721163,
      "loss": 1.6065,
      "theoretical_loss": 4.0277437962192835,
      "tokens_seen": 411500544
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008841323971915748,
      "loss": 1.6634,
      "theoretical_loss": 4.027672883478364,
      "tokens_seen": 411566080
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008841123370110331,
      "loss": 1.8244,
      "theoretical_loss": 4.027601985189524,
      "tokens_seen": 411631616
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008840922768304915,
      "loss": 1.6679,
      "theoretical_loss": 4.027531101347519,
      "tokens_seen": 411697152
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008840722166499499,
      "loss": 1.7921,
      "theoretical_loss": 4.027460231947105,
      "tokens_seen": 411762688
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008840521564694083,
      "loss": 1.6681,
      "theoretical_loss": 4.027389376983041,
      "tokens_seen": 411828224
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008840320962888666,
      "loss": 1.5784,
      "theoretical_loss": 4.027318536450091,
      "tokens_seen": 411893760
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000884012036108325,
      "loss": 1.8254,
      "theoretical_loss": 4.02724771034302,
      "tokens_seen": 411959296
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008839919759277833,
      "loss": 1.8391,
      "theoretical_loss": 4.027176898656596,
      "tokens_seen": 412024832
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008839719157472418,
      "loss": 1.716,
      "theoretical_loss": 4.02710610138559,
      "tokens_seen": 412090368
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008839518555667001,
      "loss": 1.7174,
      "theoretical_loss": 4.027035318524776,
      "tokens_seen": 412155904
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008839317953861585,
      "loss": 1.6868,
      "theoretical_loss": 4.02696455006893,
      "tokens_seen": 412221440
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008839117352056169,
      "loss": 1.6135,
      "theoretical_loss": 4.026893796012831,
      "tokens_seen": 412286976
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008838916750250753,
      "loss": 1.6772,
      "theoretical_loss": 4.02682305635126,
      "tokens_seen": 412352512
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0008838716148445336,
      "loss": 1.7418,
      "theoretical_loss": 4.0267523310790025,
      "tokens_seen": 412418048
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000883851554663992,
      "loss": 1.5971,
      "theoretical_loss": 4.026681620190845,
      "tokens_seen": 412483584
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008838314944834503,
      "loss": 1.7827,
      "theoretical_loss": 4.026610923681578,
      "tokens_seen": 412549120
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008838114343029088,
      "loss": 1.6737,
      "theoretical_loss": 4.026540241545994,
      "tokens_seen": 412614656
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008837913741223671,
      "loss": 1.663,
      "theoretical_loss": 4.026469573778888,
      "tokens_seen": 412680192
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008837713139418255,
      "loss": 1.8313,
      "theoretical_loss": 4.026398920375057,
      "tokens_seen": 412745728
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008837512537612838,
      "loss": 1.5836,
      "theoretical_loss": 4.026328281329304,
      "tokens_seen": 412811264
    },
    {
      "epoch": 0.13,
      "objective/train/advantage_avg": 0.49095261096954346,
      "objective/train/docs_used": 156351,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5056899785995483,
      "objective/train/original_loss": 1.5056899785995483,
      "objective/train/theoretical_loss": 4.026257656636431,
      "objective/train/tokens_used": 433336800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24246512353420258,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503277778625488,
      "objective/train/weighted_lm_loss": 1.5813159942626953,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.026257656636431,
      "tokens_seen": 412876800
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008837311935807423,
      "loss": 1.5957,
      "theoretical_loss": 4.026257656636431,
      "tokens_seen": 412876800
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008837111334002006,
      "loss": 1.7631,
      "theoretical_loss": 4.026187046291244,
      "tokens_seen": 412942336
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000883691073219659,
      "loss": 1.6013,
      "theoretical_loss": 4.026116450288553,
      "tokens_seen": 413007872
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008836710130391174,
      "loss": 1.7958,
      "theoretical_loss": 4.026045868623169,
      "tokens_seen": 413073408
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008836509528585758,
      "loss": 1.6075,
      "theoretical_loss": 4.025975301289906,
      "tokens_seen": 413138944
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008836308926780341,
      "loss": 1.6469,
      "theoretical_loss": 4.02590474828358,
      "tokens_seen": 413204480
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008836108324974926,
      "loss": 1.6781,
      "theoretical_loss": 4.0258342095990125,
      "tokens_seen": 413270016
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008835907723169508,
      "loss": 1.645,
      "theoretical_loss": 4.025763685231025,
      "tokens_seen": 413335552
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008835707121364093,
      "loss": 1.769,
      "theoretical_loss": 4.025693175174443,
      "tokens_seen": 413401088
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008835506519558676,
      "loss": 1.631,
      "theoretical_loss": 4.025622679424094,
      "tokens_seen": 413466624
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000883530591775326,
      "loss": 1.5995,
      "theoretical_loss": 4.0255521979748075,
      "tokens_seen": 413532160
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008835105315947844,
      "loss": 1.612,
      "theoretical_loss": 4.025481730821418,
      "tokens_seen": 413597696
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008834904714142427,
      "loss": 1.6699,
      "theoretical_loss": 4.02541127795876,
      "tokens_seen": 413663232
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008834704112337011,
      "loss": 1.7716,
      "theoretical_loss": 4.025340839381673,
      "tokens_seen": 413728768
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008834503510531596,
      "loss": 1.6416,
      "theoretical_loss": 4.025270415084998,
      "tokens_seen": 413794304
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008834302908726178,
      "loss": 1.5975,
      "theoretical_loss": 4.025200005063578,
      "tokens_seen": 413859840
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008834102306920763,
      "loss": 1.6981,
      "theoretical_loss": 4.02512960931226,
      "tokens_seen": 413925376
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008833901705115345,
      "loss": 1.536,
      "theoretical_loss": 4.025059227825894,
      "tokens_seen": 413990912
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000883370110330993,
      "loss": 1.6279,
      "theoretical_loss": 4.024988860599331,
      "tokens_seen": 414056448
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008833500501504514,
      "loss": 1.5693,
      "theoretical_loss": 4.024918507627426,
      "tokens_seen": 414121984
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008833299899699097,
      "loss": 1.7256,
      "theoretical_loss": 4.024848168905035,
      "tokens_seen": 414187520
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008833099297893681,
      "loss": 1.7206,
      "theoretical_loss": 4.024777844427019,
      "tokens_seen": 414253056
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008832898696088266,
      "loss": 1.6334,
      "theoretical_loss": 4.024707534188241,
      "tokens_seen": 414318592
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008832698094282848,
      "loss": 1.6603,
      "theoretical_loss": 4.024637238183566,
      "tokens_seen": 414384128
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008832497492477433,
      "loss": 1.7197,
      "theoretical_loss": 4.02456695640786,
      "tokens_seen": 414449664
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008832296890672015,
      "loss": 1.5665,
      "theoretical_loss": 4.024496688855995,
      "tokens_seen": 414515200
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00088320962888666,
      "loss": 1.6843,
      "theoretical_loss": 4.0244264355228445,
      "tokens_seen": 414580736
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008831895687061184,
      "loss": 1.7103,
      "theoretical_loss": 4.024356196403284,
      "tokens_seen": 414646272
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008831695085255767,
      "loss": 1.6974,
      "theoretical_loss": 4.0242859714921915,
      "tokens_seen": 414711808
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008831494483450351,
      "loss": 1.657,
      "theoretical_loss": 4.0242157607844495,
      "tokens_seen": 414777344
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008831293881644935,
      "loss": 1.657,
      "theoretical_loss": 4.02414556427494,
      "tokens_seen": 414842880
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008831093279839518,
      "loss": 1.8203,
      "theoretical_loss": 4.024075381958551,
      "tokens_seen": 414908416
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008830892678034103,
      "loss": 1.7671,
      "theoretical_loss": 4.024005213830171,
      "tokens_seen": 414973952
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008830692076228685,
      "loss": 1.6444,
      "theoretical_loss": 4.023935059884692,
      "tokens_seen": 415039488
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000883049147442327,
      "loss": 1.7232,
      "theoretical_loss": 4.023864920117007,
      "tokens_seen": 415105024
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008830290872617853,
      "loss": 1.6406,
      "theoretical_loss": 4.023794794522015,
      "tokens_seen": 415170560
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008830090270812437,
      "loss": 1.471,
      "theoretical_loss": 4.023724683094615,
      "tokens_seen": 415236096
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008829889669007021,
      "loss": 1.4989,
      "theoretical_loss": 4.023654585829709,
      "tokens_seen": 415301632
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008829689067201605,
      "loss": 1.6739,
      "theoretical_loss": 4.023584502722203,
      "tokens_seen": 415367168
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008829488465396188,
      "loss": 1.6447,
      "theoretical_loss": 4.023514433767002,
      "tokens_seen": 415432704
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008829287863590773,
      "loss": 1.6639,
      "theoretical_loss": 4.023444378959019,
      "tokens_seen": 415498240
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008829087261785356,
      "loss": 1.6446,
      "theoretical_loss": 4.023374338293166,
      "tokens_seen": 415563776
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000882888665997994,
      "loss": 1.5449,
      "theoretical_loss": 4.023304311764359,
      "tokens_seen": 415629312
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008828686058174524,
      "loss": 1.7832,
      "theoretical_loss": 4.023234299367514,
      "tokens_seen": 415694848
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008828485456369108,
      "loss": 1.6643,
      "theoretical_loss": 4.023164301097555,
      "tokens_seen": 415760384
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008828284854563692,
      "loss": 1.76,
      "theoretical_loss": 4.023094316949403,
      "tokens_seen": 415825920
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008828084252758275,
      "loss": 1.5454,
      "theoretical_loss": 4.0230243469179845,
      "tokens_seen": 415891456
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008827883650952859,
      "loss": 1.6855,
      "theoretical_loss": 4.022954390998228,
      "tokens_seen": 415956992
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008827683049147442,
      "loss": 1.666,
      "theoretical_loss": 4.0228844491850655,
      "tokens_seen": 416022528
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008827482447342027,
      "loss": 1.6823,
      "theoretical_loss": 4.02281452147343,
      "tokens_seen": 416088064
    },
    {
      "epoch": 0.13,
      "objective/train/advantage_avg": 0.49329674243927,
      "objective/train/docs_used": 157543,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.461356520652771,
      "objective/train/original_loss": 1.461356520652771,
      "objective/train/theoretical_loss": 4.022744607858259,
      "objective/train/tokens_used": 436613600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2439008206129074,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505695343017578,
      "objective/train/weighted_lm_loss": 1.535351037979126,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.022744607858259,
      "tokens_seen": 416153600
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000882728184553661,
      "loss": 1.6574,
      "theoretical_loss": 4.022744607858259,
      "tokens_seen": 416153600
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008827081243731194,
      "loss": 1.6683,
      "theoretical_loss": 4.022674708334492,
      "tokens_seen": 416219136
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008826880641925778,
      "loss": 1.8037,
      "theoretical_loss": 4.022604822897068,
      "tokens_seen": 416284672
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008826680040120362,
      "loss": 1.7671,
      "theoretical_loss": 4.022534951540933,
      "tokens_seen": 416350208
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008826479438314945,
      "loss": 1.7737,
      "theoretical_loss": 4.022465094261035,
      "tokens_seen": 416415744
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000882627883650953,
      "loss": 1.5922,
      "theoretical_loss": 4.022395251052322,
      "tokens_seen": 416481280
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008826078234704112,
      "loss": 1.6788,
      "theoretical_loss": 4.0223254219097475,
      "tokens_seen": 416546816
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008825877632898697,
      "loss": 1.6164,
      "theoretical_loss": 4.022255606828264,
      "tokens_seen": 416612352
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000882567703109328,
      "loss": 1.7734,
      "theoretical_loss": 4.022185805802832,
      "tokens_seen": 416677888
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008825476429287864,
      "loss": 1.6331,
      "theoretical_loss": 4.022116018828409,
      "tokens_seen": 416743424
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008825275827482448,
      "loss": 1.7915,
      "theoretical_loss": 4.022046245899958,
      "tokens_seen": 416808960
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008825075225677031,
      "loss": 1.591,
      "theoretical_loss": 4.021976487012445,
      "tokens_seen": 416874496
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008824874623871615,
      "loss": 1.821,
      "theoretical_loss": 4.021906742160837,
      "tokens_seen": 416940032
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00088246740220662,
      "loss": 1.7072,
      "theoretical_loss": 4.0218370113401045,
      "tokens_seen": 417005568
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008824473420260782,
      "loss": 1.5424,
      "theoretical_loss": 4.021767294545221,
      "tokens_seen": 417071104
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008824272818455367,
      "loss": 1.5978,
      "theoretical_loss": 4.0216975917711615,
      "tokens_seen": 417136640
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008824072216649949,
      "loss": 1.771,
      "theoretical_loss": 4.021627903012904,
      "tokens_seen": 417202176
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008823871614844534,
      "loss": 1.6915,
      "theoretical_loss": 4.021558228265429,
      "tokens_seen": 417267712
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008823671013039118,
      "loss": 1.7399,
      "theoretical_loss": 4.021488567523721,
      "tokens_seen": 417333248
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008823470411233701,
      "loss": 1.6824,
      "theoretical_loss": 4.021418920782764,
      "tokens_seen": 417398784
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008823269809428285,
      "loss": 1.7192,
      "theoretical_loss": 4.021349288037549,
      "tokens_seen": 417464320
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000882306920762287,
      "loss": 1.6438,
      "theoretical_loss": 4.021279669283064,
      "tokens_seen": 417529856
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008822868605817452,
      "loss": 1.7288,
      "theoretical_loss": 4.021210064514305,
      "tokens_seen": 417595392
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008822668004012037,
      "loss": 1.6427,
      "theoretical_loss": 4.021140473726268,
      "tokens_seen": 417660928
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008822467402206619,
      "loss": 1.6762,
      "theoretical_loss": 4.021070896913951,
      "tokens_seen": 417726464
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008822266800401204,
      "loss": 1.551,
      "theoretical_loss": 4.021001334072355,
      "tokens_seen": 417792000
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008822066198595788,
      "loss": 1.6138,
      "theoretical_loss": 4.020931785196484,
      "tokens_seen": 417857536
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008821865596790371,
      "loss": 1.673,
      "theoretical_loss": 4.020862250281345,
      "tokens_seen": 417923072
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008821664994984955,
      "loss": 1.5015,
      "theoretical_loss": 4.020792729321947,
      "tokens_seen": 417988608
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008821464393179539,
      "loss": 1.6645,
      "theoretical_loss": 4.020723222313302,
      "tokens_seen": 418054144
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008821263791374122,
      "loss": 1.7655,
      "theoretical_loss": 4.020653729250424,
      "tokens_seen": 418119680
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008821063189568707,
      "loss": 1.6516,
      "theoretical_loss": 4.020584250128329,
      "tokens_seen": 418185216
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000882086258776329,
      "loss": 1.7177,
      "theoretical_loss": 4.020514784942037,
      "tokens_seen": 418250752
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008820661985957874,
      "loss": 1.5534,
      "theoretical_loss": 4.0204453336865695,
      "tokens_seen": 418316288
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008820461384152457,
      "loss": 1.6524,
      "theoretical_loss": 4.020375896356951,
      "tokens_seen": 418381824
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008820260782347041,
      "loss": 1.7709,
      "theoretical_loss": 4.020306472948208,
      "tokens_seen": 418447360
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008820060180541625,
      "loss": 1.8081,
      "theoretical_loss": 4.020237063455372,
      "tokens_seen": 418512896
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008819859578736209,
      "loss": 1.765,
      "theoretical_loss": 4.020167667873473,
      "tokens_seen": 418578432
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008819658976930792,
      "loss": 1.73,
      "theoretical_loss": 4.0200982861975465,
      "tokens_seen": 418643968
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008819458375125377,
      "loss": 1.7852,
      "theoretical_loss": 4.020028918422629,
      "tokens_seen": 418709504
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000881925777331996,
      "loss": 1.7194,
      "theoretical_loss": 4.019959564543761,
      "tokens_seen": 418775040
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008819057171514544,
      "loss": 1.686,
      "theoretical_loss": 4.019890224555985,
      "tokens_seen": 418840576
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008818856569709127,
      "loss": 1.6681,
      "theoretical_loss": 4.019820898454345,
      "tokens_seen": 418906112
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008818655967903712,
      "loss": 1.6145,
      "theoretical_loss": 4.019751586233889,
      "tokens_seen": 418971648
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008818455366098295,
      "loss": 1.6892,
      "theoretical_loss": 4.019682287889666,
      "tokens_seen": 419037184
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008818254764292879,
      "loss": 1.5315,
      "theoretical_loss": 4.01961300341673,
      "tokens_seen": 419102720
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008818054162487462,
      "loss": 1.7103,
      "theoretical_loss": 4.019543732810134,
      "tokens_seen": 419168256
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008817853560682046,
      "loss": 1.654,
      "theoretical_loss": 4.0194744760649375,
      "tokens_seen": 419233792
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000881765295887663,
      "loss": 1.6612,
      "theoretical_loss": 4.019405233176199,
      "tokens_seen": 419299328
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008817452357071214,
      "loss": 1.7786,
      "theoretical_loss": 4.019336004138982,
      "tokens_seen": 419364864
    },
    {
      "epoch": 0.13,
      "objective/train/advantage_avg": 0.4930272698402405,
      "objective/train/docs_used": 158751,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7433662414550781,
      "objective/train/original_loss": 1.7433662414550781,
      "objective/train/theoretical_loss": 4.019266788948352,
      "objective/train/tokens_used": 439890400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24367401003837585,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505414009094238,
      "objective/train/weighted_lm_loss": 1.8314003944396973,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.019266788948352,
      "tokens_seen": 419430400
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008817251755265797,
      "loss": 1.6892,
      "theoretical_loss": 4.019266788948352,
      "tokens_seen": 419430400
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008817051153460382,
      "loss": 1.6757,
      "theoretical_loss": 4.0191975875993755,
      "tokens_seen": 419495936
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008816850551654964,
      "loss": 1.8176,
      "theoretical_loss": 4.019128400087123,
      "tokens_seen": 419561472
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008816649949849549,
      "loss": 1.7537,
      "theoretical_loss": 4.019059226406667,
      "tokens_seen": 419627008
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008816449348044132,
      "loss": 1.7643,
      "theoretical_loss": 4.0189900665530836,
      "tokens_seen": 419692544
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008816248746238716,
      "loss": 1.7258,
      "theoretical_loss": 4.018920920521451,
      "tokens_seen": 419758080
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00088160481444333,
      "loss": 1.7165,
      "theoretical_loss": 4.018851788306847,
      "tokens_seen": 419823616
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008815847542627884,
      "loss": 1.6832,
      "theoretical_loss": 4.018782669904357,
      "tokens_seen": 419889152
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008815646940822467,
      "loss": 1.7228,
      "theoretical_loss": 4.0187135653090635,
      "tokens_seen": 419954688
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008815446339017052,
      "loss": 1.5663,
      "theoretical_loss": 4.018644474516058,
      "tokens_seen": 420020224
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008815245737211634,
      "loss": 1.515,
      "theoretical_loss": 4.0185753975204275,
      "tokens_seen": 420085760
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008815045135406219,
      "loss": 1.6648,
      "theoretical_loss": 4.018506334317267,
      "tokens_seen": 420151296
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008814844533600802,
      "loss": 1.6975,
      "theoretical_loss": 4.018437284901671,
      "tokens_seen": 420216832
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008814643931795386,
      "loss": 1.7339,
      "theoretical_loss": 4.018368249268737,
      "tokens_seen": 420282368
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000881444332998997,
      "loss": 1.7528,
      "theoretical_loss": 4.018299227413566,
      "tokens_seen": 420347904
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008814242728184553,
      "loss": 1.7641,
      "theoretical_loss": 4.018230219331261,
      "tokens_seen": 420413440
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008814042126379137,
      "loss": 1.7406,
      "theoretical_loss": 4.018161225016926,
      "tokens_seen": 420478976
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008813841524573722,
      "loss": 1.5882,
      "theoretical_loss": 4.01809224446567,
      "tokens_seen": 420544512
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008813640922768305,
      "loss": 1.7368,
      "theoretical_loss": 4.018023277672604,
      "tokens_seen": 420610048
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008813440320962889,
      "loss": 1.7926,
      "theoretical_loss": 4.01795432463284,
      "tokens_seen": 420675584
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008813239719157474,
      "loss": 1.7206,
      "theoretical_loss": 4.0178853853414935,
      "tokens_seen": 420741120
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008813039117352056,
      "loss": 1.7873,
      "theoretical_loss": 4.017816459793683,
      "tokens_seen": 420806656
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008812838515546641,
      "loss": 1.6735,
      "theoretical_loss": 4.017747547984528,
      "tokens_seen": 420872192
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008812637913741223,
      "loss": 1.605,
      "theoretical_loss": 4.0176786499091515,
      "tokens_seen": 420937728
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008812437311935808,
      "loss": 1.7147,
      "theoretical_loss": 4.017609765562678,
      "tokens_seen": 421003264
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008812236710130392,
      "loss": 1.7636,
      "theoretical_loss": 4.017540894940238,
      "tokens_seen": 421068800
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008812036108324975,
      "loss": 1.7168,
      "theoretical_loss": 4.017472038036959,
      "tokens_seen": 421134336
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008811835506519559,
      "loss": 1.7088,
      "theoretical_loss": 4.017403194847975,
      "tokens_seen": 421199872
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008811634904714143,
      "loss": 1.6162,
      "theoretical_loss": 4.017334365368422,
      "tokens_seen": 421265408
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008811434302908726,
      "loss": 1.7104,
      "theoretical_loss": 4.017265549593437,
      "tokens_seen": 421330944
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008811233701103311,
      "loss": 1.7413,
      "theoretical_loss": 4.01719674751816,
      "tokens_seen": 421396480
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008811033099297893,
      "loss": 1.7278,
      "theoretical_loss": 4.017127959137733,
      "tokens_seen": 421462016
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008810832497492478,
      "loss": 1.7578,
      "theoretical_loss": 4.017059184447303,
      "tokens_seen": 421527552
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008810631895687061,
      "loss": 1.6661,
      "theoretical_loss": 4.016990423442016,
      "tokens_seen": 421593088
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008810431293881645,
      "loss": 1.7898,
      "theoretical_loss": 4.0169216761170246,
      "tokens_seen": 421658624
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008810230692076229,
      "loss": 1.6991,
      "theoretical_loss": 4.016852942467479,
      "tokens_seen": 421724160
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008810030090270813,
      "loss": 1.6974,
      "theoretical_loss": 4.0167842224885355,
      "tokens_seen": 421789696
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008809829488465396,
      "loss": 1.6438,
      "theoretical_loss": 4.016715516175351,
      "tokens_seen": 421855232
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008809628886659981,
      "loss": 1.6117,
      "theoretical_loss": 4.016646823523087,
      "tokens_seen": 421920768
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008809428284854564,
      "loss": 1.5419,
      "theoretical_loss": 4.016578144526903,
      "tokens_seen": 421986304
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008809227683049148,
      "loss": 1.6463,
      "theoretical_loss": 4.016509479181968,
      "tokens_seen": 422051840
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008809027081243731,
      "loss": 1.739,
      "theoretical_loss": 4.0164408274834456,
      "tokens_seen": 422117376
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008808826479438316,
      "loss": 1.6391,
      "theoretical_loss": 4.016372189426509,
      "tokens_seen": 422182912
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008808625877632899,
      "loss": 1.7629,
      "theoretical_loss": 4.016303565006328,
      "tokens_seen": 422248448
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008808425275827483,
      "loss": 1.5975,
      "theoretical_loss": 4.016234954218078,
      "tokens_seen": 422313984
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008808224674022066,
      "loss": 1.7056,
      "theoretical_loss": 4.016166357056937,
      "tokens_seen": 422379520
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000880802407221665,
      "loss": 1.7482,
      "theoretical_loss": 4.016097773518084,
      "tokens_seen": 422445056
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008807823470411234,
      "loss": 1.6291,
      "theoretical_loss": 4.016029203596702,
      "tokens_seen": 422510592
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008807622868605818,
      "loss": 1.5805,
      "theoretical_loss": 4.015960647287975,
      "tokens_seen": 422576128
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008807422266800401,
      "loss": 1.6762,
      "theoretical_loss": 4.01589210458709,
      "tokens_seen": 422641664
    },
    {
      "epoch": 0.13,
      "objective/train/advantage_avg": 0.49287301301956177,
      "objective/train/docs_used": 159824,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5584911108016968,
      "objective/train/original_loss": 1.5584911108016968,
      "objective/train/theoretical_loss": 4.015823575489237,
      "objective/train/tokens_used": 443167200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24386560916900635,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505269765853882,
      "objective/train/weighted_lm_loss": 1.6371653079986572,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.015823575489237,
      "tokens_seen": 422707200
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008807221664994986,
      "loss": 1.6797,
      "theoretical_loss": 4.015823575489237,
      "tokens_seen": 422707200
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008807021063189568,
      "loss": 1.768,
      "theoretical_loss": 4.0157550599896075,
      "tokens_seen": 422772736
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008806820461384153,
      "loss": 1.7092,
      "theoretical_loss": 4.015686558083396,
      "tokens_seen": 422838272
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008806619859578736,
      "loss": 1.7596,
      "theoretical_loss": 4.015618069765799,
      "tokens_seen": 422903808
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000880641925777332,
      "loss": 1.6727,
      "theoretical_loss": 4.015549595032017,
      "tokens_seen": 422969344
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008806218655967904,
      "loss": 1.6833,
      "theoretical_loss": 4.015481133877251,
      "tokens_seen": 423034880
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008806018054162488,
      "loss": 1.7244,
      "theoretical_loss": 4.015412686296704,
      "tokens_seen": 423100416
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008805817452357071,
      "loss": 1.6397,
      "theoretical_loss": 4.015344252285585,
      "tokens_seen": 423165952
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008805616850551656,
      "loss": 1.665,
      "theoretical_loss": 4.0152758318391015,
      "tokens_seen": 423231488
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008805416248746238,
      "loss": 1.8013,
      "theoretical_loss": 4.0152074249524645,
      "tokens_seen": 423297024
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008805215646940823,
      "loss": 1.6566,
      "theoretical_loss": 4.0151390316208895,
      "tokens_seen": 423362560
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008805015045135407,
      "loss": 1.7786,
      "theoretical_loss": 4.015070651839592,
      "tokens_seen": 423428096
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000880481444332999,
      "loss": 1.5864,
      "theoretical_loss": 4.01500228560379,
      "tokens_seen": 423493632
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008804613841524574,
      "loss": 1.7001,
      "theoretical_loss": 4.014933932908706,
      "tokens_seen": 423559168
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008804413239719157,
      "loss": 1.7395,
      "theoretical_loss": 4.014865593749563,
      "tokens_seen": 423624704
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008804212637913741,
      "loss": 1.7334,
      "theoretical_loss": 4.014797268121587,
      "tokens_seen": 423690240
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008804012036108326,
      "loss": 1.6614,
      "theoretical_loss": 4.014728956020006,
      "tokens_seen": 423755776
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008803811434302908,
      "loss": 1.6281,
      "theoretical_loss": 4.014660657440052,
      "tokens_seen": 423821312
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008803610832497493,
      "loss": 1.6153,
      "theoretical_loss": 4.014592372376958,
      "tokens_seen": 423886848
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008803410230692077,
      "loss": 1.5827,
      "theoretical_loss": 4.014524100825959,
      "tokens_seen": 423952384
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000880320962888666,
      "loss": 1.7422,
      "theoretical_loss": 4.014455842782294,
      "tokens_seen": 424017920
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008803009027081244,
      "loss": 1.787,
      "theoretical_loss": 4.0143875982412025,
      "tokens_seen": 424083456
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008802808425275827,
      "loss": 1.6612,
      "theoretical_loss": 4.014319367197928,
      "tokens_seen": 424148992
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008802607823470411,
      "loss": 1.6953,
      "theoretical_loss": 4.014251149647716,
      "tokens_seen": 424214528
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008802407221664996,
      "loss": 1.7357,
      "theoretical_loss": 4.014182945585815,
      "tokens_seen": 424280064
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008802206619859578,
      "loss": 1.7909,
      "theoretical_loss": 4.014114755007474,
      "tokens_seen": 424345600
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008802006018054163,
      "loss": 1.8349,
      "theoretical_loss": 4.014046577907946,
      "tokens_seen": 424411136
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008801805416248746,
      "loss": 1.7697,
      "theoretical_loss": 4.0139784142824855,
      "tokens_seen": 424476672
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000880160481444333,
      "loss": 1.7789,
      "theoretical_loss": 4.01391026412635,
      "tokens_seen": 424542208
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008801404212637914,
      "loss": 1.7296,
      "theoretical_loss": 4.013842127434801,
      "tokens_seen": 424607744
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008801203610832497,
      "loss": 1.6642,
      "theoretical_loss": 4.013774004203099,
      "tokens_seen": 424673280
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008801003009027081,
      "loss": 1.7023,
      "theoretical_loss": 4.013705894426509,
      "tokens_seen": 424738816
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008800802407221665,
      "loss": 1.6199,
      "theoretical_loss": 4.013637798100298,
      "tokens_seen": 424804352
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008800601805416248,
      "loss": 1.7503,
      "theoretical_loss": 4.0135697152197345,
      "tokens_seen": 424869888
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008800401203610833,
      "loss": 1.7619,
      "theoretical_loss": 4.013501645780092,
      "tokens_seen": 424935424
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008800200601805416,
      "loss": 1.6706,
      "theoretical_loss": 4.0134335897766436,
      "tokens_seen": 425000960
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00088,
      "loss": 1.5715,
      "theoretical_loss": 4.013365547204666,
      "tokens_seen": 425066496
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008799799398194584,
      "loss": 1.6421,
      "theoretical_loss": 4.013297518059439,
      "tokens_seen": 425132032
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008799598796389168,
      "loss": 1.6172,
      "theoretical_loss": 4.013229502336242,
      "tokens_seen": 425197568
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008799398194583751,
      "loss": 1.5202,
      "theoretical_loss": 4.01316150003036,
      "tokens_seen": 425263104
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008799197592778335,
      "loss": 1.7793,
      "theoretical_loss": 4.013093511137079,
      "tokens_seen": 425328640
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008798996990972918,
      "loss": 1.8113,
      "theoretical_loss": 4.013025535651688,
      "tokens_seen": 425394176
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008798796389167503,
      "loss": 1.6631,
      "theoretical_loss": 4.0129575735694765,
      "tokens_seen": 425459712
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008798595787362087,
      "loss": 1.7407,
      "theoretical_loss": 4.012889624885739,
      "tokens_seen": 425525248
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000879839518555667,
      "loss": 1.7129,
      "theoretical_loss": 4.012821689595771,
      "tokens_seen": 425590784
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008798194583751254,
      "loss": 1.8266,
      "theoretical_loss": 4.01275376769487,
      "tokens_seen": 425656320
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008797993981945838,
      "loss": 1.692,
      "theoretical_loss": 4.012685859178337,
      "tokens_seen": 425721856
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008797793380140422,
      "loss": 1.6876,
      "theoretical_loss": 4.012617964041474,
      "tokens_seen": 425787392
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008797592778335005,
      "loss": 1.6699,
      "theoretical_loss": 4.012550082279587,
      "tokens_seen": 425852928
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000879739217652959,
      "loss": 1.544,
      "theoretical_loss": 4.012482213887981,
      "tokens_seen": 425918464
    },
    {
      "epoch": 0.13,
      "objective/train/advantage_avg": 0.4950158894062042,
      "objective/train/docs_used": 161088,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.462551474571228,
      "objective/train/original_loss": 1.4625513553619385,
      "objective/train/theoretical_loss": 4.012414358861969,
      "objective/train/tokens_used": 446444000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24526995420455933,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050748586654663,
      "objective/train/weighted_lm_loss": 1.5366774797439575,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.012414358861969,
      "tokens_seen": 425984000
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008797191574724172,
      "loss": 1.6405,
      "theoretical_loss": 4.012414358861969,
      "tokens_seen": 425984000
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008796990972918757,
      "loss": 1.7246,
      "theoretical_loss": 4.012346517196862,
      "tokens_seen": 426049536
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000879679037111334,
      "loss": 1.5951,
      "theoretical_loss": 4.012278688887975,
      "tokens_seen": 426115072
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008796589769307924,
      "loss": 1.7454,
      "theoretical_loss": 4.012210873930625,
      "tokens_seen": 426180608
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008796389167502508,
      "loss": 1.7153,
      "theoretical_loss": 4.01214307232013,
      "tokens_seen": 426246144
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008796188565697092,
      "loss": 1.6092,
      "theoretical_loss": 4.012075284051813,
      "tokens_seen": 426311680
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008795987963891675,
      "loss": 1.7099,
      "theoretical_loss": 4.012007509120997,
      "tokens_seen": 426377216
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000879578736208626,
      "loss": 1.7302,
      "theoretical_loss": 4.011939747523009,
      "tokens_seen": 426442752
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008795586760280842,
      "loss": 1.6965,
      "theoretical_loss": 4.011871999253178,
      "tokens_seen": 426508288
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008795386158475427,
      "loss": 1.8017,
      "theoretical_loss": 4.011804264306835,
      "tokens_seen": 426573824
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000879518555667001,
      "loss": 1.7581,
      "theoretical_loss": 4.011736542679313,
      "tokens_seen": 426639360
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008794984954864594,
      "loss": 1.7327,
      "theoretical_loss": 4.011668834365948,
      "tokens_seen": 426704896
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008794784353059178,
      "loss": 1.6182,
      "theoretical_loss": 4.011601139362078,
      "tokens_seen": 426770432
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008794583751253761,
      "loss": 1.691,
      "theoretical_loss": 4.011533457663044,
      "tokens_seen": 426835968
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008794383149448345,
      "loss": 1.6054,
      "theoretical_loss": 4.0114657892641885,
      "tokens_seen": 426901504
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000879418254764293,
      "loss": 1.698,
      "theoretical_loss": 4.011398134160857,
      "tokens_seen": 426967040
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008793981945837512,
      "loss": 1.6539,
      "theoretical_loss": 4.011330492348397,
      "tokens_seen": 427032576
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008793781344032097,
      "loss": 1.7655,
      "theoretical_loss": 4.011262863822157,
      "tokens_seen": 427098112
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008793580742226681,
      "loss": 1.6945,
      "theoretical_loss": 4.011195248577492,
      "tokens_seen": 427163648
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008793380140421264,
      "loss": 1.6915,
      "theoretical_loss": 4.011127646609753,
      "tokens_seen": 427229184
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008793179538615848,
      "loss": 1.8055,
      "theoretical_loss": 4.0110600579143,
      "tokens_seen": 427294720
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008792978936810431,
      "loss": 1.6929,
      "theoretical_loss": 4.0109924824864915,
      "tokens_seen": 427360256
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008792778335005015,
      "loss": 1.72,
      "theoretical_loss": 4.0109249203216875,
      "tokens_seen": 427425792
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00087925777331996,
      "loss": 1.7089,
      "theoretical_loss": 4.010857371415254,
      "tokens_seen": 427491328
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008792377131394182,
      "loss": 1.6548,
      "theoretical_loss": 4.010789835762555,
      "tokens_seen": 427556864
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008792176529588767,
      "loss": 1.7562,
      "theoretical_loss": 4.010722313358961,
      "tokens_seen": 427622400
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000879197592778335,
      "loss": 1.7734,
      "theoretical_loss": 4.010654804199842,
      "tokens_seen": 427687936
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008791775325977934,
      "loss": 1.664,
      "theoretical_loss": 4.010587308280572,
      "tokens_seen": 427753472
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008791574724172518,
      "loss": 1.7622,
      "theoretical_loss": 4.010519825596525,
      "tokens_seen": 427819008
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008791374122367101,
      "loss": 1.7167,
      "theoretical_loss": 4.01045235614308,
      "tokens_seen": 427884544
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008791173520561685,
      "loss": 1.684,
      "theoretical_loss": 4.010384899915618,
      "tokens_seen": 427950080
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008790972918756269,
      "loss": 1.6483,
      "theoretical_loss": 4.010317456909519,
      "tokens_seen": 428015616
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008790772316950852,
      "loss": 1.6849,
      "theoretical_loss": 4.010250027120169,
      "tokens_seen": 428081152
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008790571715145437,
      "loss": 1.557,
      "theoretical_loss": 4.010182610542957,
      "tokens_seen": 428146688
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000879037111334002,
      "loss": 1.6306,
      "theoretical_loss": 4.0101152071732695,
      "tokens_seen": 428212224
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008790170511534604,
      "loss": 1.7604,
      "theoretical_loss": 4.0100478170065,
      "tokens_seen": 428277760
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008789969909729188,
      "loss": 1.6982,
      "theoretical_loss": 4.009980440038043,
      "tokens_seen": 428343296
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008789769307923772,
      "loss": 1.7026,
      "theoretical_loss": 4.0099130762632935,
      "tokens_seen": 428408832
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008789568706118355,
      "loss": 1.749,
      "theoretical_loss": 4.00984572567765,
      "tokens_seen": 428474368
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008789368104312939,
      "loss": 1.8499,
      "theoretical_loss": 4.009778388276516,
      "tokens_seen": 428539904
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008789167502507522,
      "loss": 1.6362,
      "theoretical_loss": 4.009711064055291,
      "tokens_seen": 428605440
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008788966900702107,
      "loss": 1.6866,
      "theoretical_loss": 4.009643753009383,
      "tokens_seen": 428670976
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000878876629889669,
      "loss": 1.5747,
      "theoretical_loss": 4.0095764551342015,
      "tokens_seen": 428736512
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008788565697091274,
      "loss": 1.6387,
      "theoretical_loss": 4.0095091704251535,
      "tokens_seen": 428802048
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008788365095285857,
      "loss": 1.7684,
      "theoretical_loss": 4.009441898877652,
      "tokens_seen": 428867584
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008788164493480442,
      "loss": 1.6421,
      "theoretical_loss": 4.009374640487114,
      "tokens_seen": 428933120
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008787963891675025,
      "loss": 1.6364,
      "theoretical_loss": 4.009307395248955,
      "tokens_seen": 428998656
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008787763289869609,
      "loss": 1.6676,
      "theoretical_loss": 4.009240163158595,
      "tokens_seen": 429064192
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008787562688064192,
      "loss": 1.73,
      "theoretical_loss": 4.009172944211455,
      "tokens_seen": 429129728
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008787362086258776,
      "loss": 1.597,
      "theoretical_loss": 4.00910573840296,
      "tokens_seen": 429195264
    },
    {
      "epoch": 0.13,
      "objective/train/advantage_avg": 0.49417299032211304,
      "objective/train/docs_used": 162190,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6557865142822266,
      "objective/train/original_loss": 1.6557865142822266,
      "objective/train/theoretical_loss": 4.009038545728536,
      "objective/train/tokens_used": 449720800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24450218677520752,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506603717803955,
      "objective/train/weighted_lm_loss": 1.7396289110183716,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.009038545728536,
      "tokens_seen": 429260800
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000878716148445336,
      "loss": 1.7554,
      "theoretical_loss": 4.009038545728536,
      "tokens_seen": 429260800
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008786960882647944,
      "loss": 1.6362,
      "theoretical_loss": 4.00897136618361,
      "tokens_seen": 429326336
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008786760280842527,
      "loss": 1.8445,
      "theoretical_loss": 4.008904199763615,
      "tokens_seen": 429391872
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008786559679037112,
      "loss": 1.7453,
      "theoretical_loss": 4.0088370464639835,
      "tokens_seen": 429457408
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008786359077231695,
      "loss": 1.5949,
      "theoretical_loss": 4.0087699062801505,
      "tokens_seen": 429522944
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008786158475426279,
      "loss": 1.6461,
      "theoretical_loss": 4.008702779207555,
      "tokens_seen": 429588480
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008785957873620863,
      "loss": 1.7022,
      "theoretical_loss": 4.008635665241635,
      "tokens_seen": 429654016
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008785757271815446,
      "loss": 1.7671,
      "theoretical_loss": 4.008568564377834,
      "tokens_seen": 429719552
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000878555667001003,
      "loss": 1.6948,
      "theoretical_loss": 4.008501476611597,
      "tokens_seen": 429785088
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008785356068204615,
      "loss": 1.7342,
      "theoretical_loss": 4.00843440193837,
      "tokens_seen": 429850624
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008785155466399197,
      "loss": 1.7383,
      "theoretical_loss": 4.008367340353602,
      "tokens_seen": 429916160
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008784954864593782,
      "loss": 1.5975,
      "theoretical_loss": 4.008300291852745,
      "tokens_seen": 429981696
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008784754262788364,
      "loss": 1.7106,
      "theoretical_loss": 4.008233256431253,
      "tokens_seen": 430047232
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008784553660982949,
      "loss": 1.7632,
      "theoretical_loss": 4.00816623408458,
      "tokens_seen": 430112768
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008784353059177533,
      "loss": 1.7036,
      "theoretical_loss": 4.008099224808188,
      "tokens_seen": 430178304
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008784152457372116,
      "loss": 1.58,
      "theoretical_loss": 4.008032228597534,
      "tokens_seen": 430243840
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00087839518555667,
      "loss": 1.7657,
      "theoretical_loss": 4.007965245448082,
      "tokens_seen": 430309376
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008783751253761285,
      "loss": 1.5189,
      "theoretical_loss": 4.007898275355297,
      "tokens_seen": 430374912
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008783550651955868,
      "loss": 1.6877,
      "theoretical_loss": 4.007831318314645,
      "tokens_seen": 430440448
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008783350050150452,
      "loss": 1.6408,
      "theoretical_loss": 4.007764374321599,
      "tokens_seen": 430505984
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008783149448345035,
      "loss": 1.604,
      "theoretical_loss": 4.007697443371628,
      "tokens_seen": 430571520
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008782948846539619,
      "loss": 1.7851,
      "theoretical_loss": 4.007630525460206,
      "tokens_seen": 430637056
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008782748244734204,
      "loss": 1.5834,
      "theoretical_loss": 4.00756362058281,
      "tokens_seen": 430702592
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008782547642928786,
      "loss": 1.7205,
      "theoretical_loss": 4.007496728734919,
      "tokens_seen": 430768128
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008782347041123371,
      "loss": 1.6197,
      "theoretical_loss": 4.007429849912014,
      "tokens_seen": 430833664
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008782146439317954,
      "loss": 1.5597,
      "theoretical_loss": 4.007362984109577,
      "tokens_seen": 430899200
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008781945837512538,
      "loss": 1.6667,
      "theoretical_loss": 4.007296131323094,
      "tokens_seen": 430964736
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008781745235707122,
      "loss": 1.657,
      "theoretical_loss": 4.007229291548052,
      "tokens_seen": 431030272
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008781544633901706,
      "loss": 1.6742,
      "theoretical_loss": 4.007162464779943,
      "tokens_seen": 431095808
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008781344032096289,
      "loss": 1.7575,
      "theoretical_loss": 4.0070956510142555,
      "tokens_seen": 431161344
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008781143430290873,
      "loss": 1.6987,
      "theoretical_loss": 4.007028850246487,
      "tokens_seen": 431226880
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008780942828485456,
      "loss": 1.8072,
      "theoretical_loss": 4.006962062472133,
      "tokens_seen": 431292416
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008780742226680041,
      "loss": 1.6421,
      "theoretical_loss": 4.006895287686692,
      "tokens_seen": 431357952
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008780541624874624,
      "loss": 1.6763,
      "theoretical_loss": 4.006828525885666,
      "tokens_seen": 431423488
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008780341023069208,
      "loss": 1.6193,
      "theoretical_loss": 4.006761777064557,
      "tokens_seen": 431489024
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008780140421263792,
      "loss": 1.6439,
      "theoretical_loss": 4.0066950412188715,
      "tokens_seen": 431554560
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008779939819458376,
      "loss": 1.6766,
      "theoretical_loss": 4.006628318344117,
      "tokens_seen": 431620096
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008779739217652959,
      "loss": 1.6931,
      "theoretical_loss": 4.0065616084358044,
      "tokens_seen": 431685632
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008779538615847543,
      "loss": 1.6771,
      "theoretical_loss": 4.006494911489444,
      "tokens_seen": 431751168
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008779338014042126,
      "loss": 1.6814,
      "theoretical_loss": 4.006428227500552,
      "tokens_seen": 431816704
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008779137412236711,
      "loss": 1.6386,
      "theoretical_loss": 4.006361556464646,
      "tokens_seen": 431882240
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008778936810431294,
      "loss": 1.7042,
      "theoretical_loss": 4.006294898377242,
      "tokens_seen": 431947776
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008778736208625878,
      "loss": 1.735,
      "theoretical_loss": 4.006228253233864,
      "tokens_seen": 432013312
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008778535606820461,
      "loss": 1.8019,
      "theoretical_loss": 4.006161621030033,
      "tokens_seen": 432078848
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008778335005015046,
      "loss": 1.6797,
      "theoretical_loss": 4.0060950017612775,
      "tokens_seen": 432144384
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008778134403209629,
      "loss": 1.6797,
      "theoretical_loss": 4.006028395423122,
      "tokens_seen": 432209920
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008777933801404213,
      "loss": 1.7321,
      "theoretical_loss": 4.0059618020111,
      "tokens_seen": 432275456
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008777733199598796,
      "loss": 1.6531,
      "theoretical_loss": 4.005895221520742,
      "tokens_seen": 432340992
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000877753259779338,
      "loss": 1.7324,
      "theoretical_loss": 4.005828653947583,
      "tokens_seen": 432406528
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008777331995987964,
      "loss": 1.6861,
      "theoretical_loss": 4.00576209928716,
      "tokens_seen": 432472064
    },
    {
      "epoch": 0.13,
      "objective/train/advantage_avg": 0.4946593642234802,
      "objective/train/docs_used": 163281,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5753886699676514,
      "objective/train/original_loss": 1.5753885507583618,
      "objective/train/theoretical_loss": 4.00569555753501,
      "objective/train/tokens_used": 452997600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2450278401374817,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507116317749023,
      "objective/train/weighted_lm_loss": 1.6551384925842285,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.00569555753501,
      "tokens_seen": 432537600
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008777131394182548,
      "loss": 1.7133,
      "theoretical_loss": 4.00569555753501,
      "tokens_seen": 432537600
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008776930792377131,
      "loss": 1.7018,
      "theoretical_loss": 4.005629028686678,
      "tokens_seen": 432603136
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008776730190571716,
      "loss": 1.7188,
      "theoretical_loss": 4.005562512737704,
      "tokens_seen": 432668672
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008776529588766299,
      "loss": 1.6953,
      "theoretical_loss": 4.005496009683635,
      "tokens_seen": 432734208
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008776328986960883,
      "loss": 1.5622,
      "theoretical_loss": 4.0054295195200185,
      "tokens_seen": 432799744
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008776128385155467,
      "loss": 1.6209,
      "theoretical_loss": 4.005363042242405,
      "tokens_seen": 432865280
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000877592778335005,
      "loss": 1.7354,
      "theoretical_loss": 4.005296577846347,
      "tokens_seen": 432930816
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008775727181544634,
      "loss": 1.6755,
      "theoretical_loss": 4.005230126327398,
      "tokens_seen": 432996352
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008775526579739219,
      "loss": 1.6384,
      "theoretical_loss": 4.005163687681116,
      "tokens_seen": 433061888
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008775325977933801,
      "loss": 1.6189,
      "theoretical_loss": 4.0050972619030585,
      "tokens_seen": 433127424
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008775125376128386,
      "loss": 1.7411,
      "theoretical_loss": 4.005030848988788,
      "tokens_seen": 433192960
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008774924774322968,
      "loss": 1.6849,
      "theoretical_loss": 4.004964448933866,
      "tokens_seen": 433258496
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008774724172517553,
      "loss": 1.6512,
      "theoretical_loss": 4.00489806173386,
      "tokens_seen": 433324032
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008774523570712137,
      "loss": 1.7051,
      "theoretical_loss": 4.004831687384336,
      "tokens_seen": 433389568
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000877432296890672,
      "loss": 1.7895,
      "theoretical_loss": 4.004765325880865,
      "tokens_seen": 433455104
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008774122367101304,
      "loss": 1.8067,
      "theoretical_loss": 4.00469897721902,
      "tokens_seen": 433520640
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008773921765295889,
      "loss": 1.7438,
      "theoretical_loss": 4.004632641394372,
      "tokens_seen": 433586176
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008773721163490471,
      "loss": 1.8116,
      "theoretical_loss": 4.0045663184025,
      "tokens_seen": 433651712
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008773520561685056,
      "loss": 1.6301,
      "theoretical_loss": 4.004500008238983,
      "tokens_seen": 433717248
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008773319959879638,
      "loss": 1.6341,
      "theoretical_loss": 4.0044337108994,
      "tokens_seen": 433782784
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008773119358074223,
      "loss": 1.7843,
      "theoretical_loss": 4.0043674263793365,
      "tokens_seen": 433848320
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008772918756268807,
      "loss": 1.584,
      "theoretical_loss": 4.004301154674375,
      "tokens_seen": 433913856
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000877271815446339,
      "loss": 1.6833,
      "theoretical_loss": 4.0042348957801055,
      "tokens_seen": 433979392
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008772517552657974,
      "loss": 1.6795,
      "theoretical_loss": 4.004168649692115,
      "tokens_seen": 434044928
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008772316950852558,
      "loss": 1.6922,
      "theoretical_loss": 4.004102416405998,
      "tokens_seen": 434110464
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008772116349047141,
      "loss": 1.687,
      "theoretical_loss": 4.004036195917347,
      "tokens_seen": 434176000
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008771915747241726,
      "loss": 1.6465,
      "theoretical_loss": 4.003969988221758,
      "tokens_seen": 434241536
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008771715145436308,
      "loss": 1.7862,
      "theoretical_loss": 4.00390379331483,
      "tokens_seen": 434307072
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008771514543630893,
      "loss": 1.5721,
      "theoretical_loss": 4.0038376111921625,
      "tokens_seen": 434372608
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008771313941825476,
      "loss": 1.746,
      "theoretical_loss": 4.00377144184936,
      "tokens_seen": 434438144
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000877111334002006,
      "loss": 1.5794,
      "theoretical_loss": 4.003705285282026,
      "tokens_seen": 434503680
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008770912738214644,
      "loss": 1.8313,
      "theoretical_loss": 4.003639141485767,
      "tokens_seen": 434569216
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008770712136409228,
      "loss": 1.7165,
      "theoretical_loss": 4.0035730104561935,
      "tokens_seen": 434634752
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008770511534603811,
      "loss": 1.752,
      "theoretical_loss": 4.003506892188917,
      "tokens_seen": 434700288
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008770310932798396,
      "loss": 1.7615,
      "theoretical_loss": 4.00344078667955,
      "tokens_seen": 434765824
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008770110330992978,
      "loss": 1.7861,
      "theoretical_loss": 4.00337469392371,
      "tokens_seen": 434831360
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008769909729187563,
      "loss": 1.7035,
      "theoretical_loss": 4.003308613917012,
      "tokens_seen": 434896896
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008769709127382146,
      "loss": 1.713,
      "theoretical_loss": 4.003242546655079,
      "tokens_seen": 434962432
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000876950852557673,
      "loss": 1.7261,
      "theoretical_loss": 4.003176492133531,
      "tokens_seen": 435027968
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008769307923771314,
      "loss": 1.8141,
      "theoretical_loss": 4.003110450347994,
      "tokens_seen": 435093504
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008769107321965898,
      "loss": 1.6346,
      "theoretical_loss": 4.003044421294094,
      "tokens_seen": 435159040
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008768906720160481,
      "loss": 1.7725,
      "theoretical_loss": 4.002978404967459,
      "tokens_seen": 435224576
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008768706118355065,
      "loss": 1.5136,
      "theoretical_loss": 4.002912401363721,
      "tokens_seen": 435290112
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000876850551654965,
      "loss": 1.6824,
      "theoretical_loss": 4.0028464104785115,
      "tokens_seen": 435355648
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008768304914744233,
      "loss": 1.7163,
      "theoretical_loss": 4.002780432307468,
      "tokens_seen": 435421184
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008768104312938817,
      "loss": 1.795,
      "theoretical_loss": 4.002714466846227,
      "tokens_seen": 435486720
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00087679037111334,
      "loss": 1.5681,
      "theoretical_loss": 4.002648514090426,
      "tokens_seen": 435552256
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008767703109327984,
      "loss": 1.6463,
      "theoretical_loss": 4.002582574035709,
      "tokens_seen": 435617792
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008767502507522568,
      "loss": 1.702,
      "theoretical_loss": 4.0025166466777184,
      "tokens_seen": 435683328
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008767301905717152,
      "loss": 1.7896,
      "theoretical_loss": 4.002450732012102,
      "tokens_seen": 435748864
    },
    {
      "epoch": 0.13,
      "objective/train/advantage_avg": 0.49324312806129456,
      "objective/train/docs_used": 164592,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.526324987411499,
      "objective/train/original_loss": 1.526324987411499,
      "objective/train/theoretical_loss": 4.002384830034506,
      "objective/train/tokens_used": 456274400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24383056163787842,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505638122558594,
      "objective/train/weighted_lm_loss": 1.6033390760421753,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 4.002384830034506,
      "tokens_seen": 435814400
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008767101303911735,
      "loss": 1.7303,
      "theoretical_loss": 4.002384830034506,
      "tokens_seen": 435814400
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000876690070210632,
      "loss": 1.6407,
      "theoretical_loss": 4.002318940740581,
      "tokens_seen": 435879936
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008766700100300903,
      "loss": 1.6781,
      "theoretical_loss": 4.00225306412598,
      "tokens_seen": 435945472
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008766499498495487,
      "loss": 1.7347,
      "theoretical_loss": 4.002187200186357,
      "tokens_seen": 436011008
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008766298896690071,
      "loss": 1.776,
      "theoretical_loss": 4.002121348917369,
      "tokens_seen": 436076544
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008766098294884654,
      "loss": 1.6167,
      "theoretical_loss": 4.002055510314674,
      "tokens_seen": 436142080
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008765897693079238,
      "loss": 1.7234,
      "theoretical_loss": 4.001989684373934,
      "tokens_seen": 436207616
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008765697091273823,
      "loss": 1.6475,
      "theoretical_loss": 4.0019238710908125,
      "tokens_seen": 436273152
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008765496489468405,
      "loss": 1.759,
      "theoretical_loss": 4.001858070460973,
      "tokens_seen": 436338688
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000876529588766299,
      "loss": 1.6858,
      "theoretical_loss": 4.001792282480085,
      "tokens_seen": 436404224
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008765095285857572,
      "loss": 1.6897,
      "theoretical_loss": 4.0017265071438155,
      "tokens_seen": 436469760
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008764894684052157,
      "loss": 1.7926,
      "theoretical_loss": 4.001660744447838,
      "tokens_seen": 436535296
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008764694082246741,
      "loss": 1.5353,
      "theoretical_loss": 4.001594994387825,
      "tokens_seen": 436600832
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008764493480441324,
      "loss": 1.6745,
      "theoretical_loss": 4.001529256959454,
      "tokens_seen": 436666368
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008764292878635908,
      "loss": 1.744,
      "theoretical_loss": 4.001463532158402,
      "tokens_seen": 436731904
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008764092276830493,
      "loss": 1.7642,
      "theoretical_loss": 4.001397819980349,
      "tokens_seen": 436797440
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008763891675025075,
      "loss": 1.807,
      "theoretical_loss": 4.001332120420978,
      "tokens_seen": 436862976
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000876369107321966,
      "loss": 1.6098,
      "theoretical_loss": 4.001266433475973,
      "tokens_seen": 436928512
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008763490471414242,
      "loss": 1.616,
      "theoretical_loss": 4.001200759141019,
      "tokens_seen": 436994048
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008763289869608827,
      "loss": 1.7438,
      "theoretical_loss": 4.001135097411807,
      "tokens_seen": 437059584
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008763089267803411,
      "loss": 1.7516,
      "theoretical_loss": 4.001069448284027,
      "tokens_seen": 437125120
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008762888665997994,
      "loss": 1.6991,
      "theoretical_loss": 4.001003811753371,
      "tokens_seen": 437190656
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008762688064192578,
      "loss": 1.551,
      "theoretical_loss": 4.000938187815535,
      "tokens_seen": 437256192
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008762487462387162,
      "loss": 1.6185,
      "theoretical_loss": 4.000872576466216,
      "tokens_seen": 437321728
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008762286860581745,
      "loss": 1.6573,
      "theoretical_loss": 4.000806977701113,
      "tokens_seen": 437387264
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000876208625877633,
      "loss": 1.7303,
      "theoretical_loss": 4.000741391515927,
      "tokens_seen": 437452800
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008761885656970912,
      "loss": 1.6935,
      "theoretical_loss": 4.000675817906362,
      "tokens_seen": 437518336
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008761685055165497,
      "loss": 1.5171,
      "theoretical_loss": 4.000610256868123,
      "tokens_seen": 437583872
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000876148445336008,
      "loss": 1.8138,
      "theoretical_loss": 4.000544708396919,
      "tokens_seen": 437649408
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008761283851554664,
      "loss": 1.6239,
      "theoretical_loss": 4.000479172488458,
      "tokens_seen": 437714944
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008761083249749248,
      "loss": 1.7383,
      "theoretical_loss": 4.000413649138453,
      "tokens_seen": 437780480
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008760882647943832,
      "loss": 1.7627,
      "theoretical_loss": 4.000348138342618,
      "tokens_seen": 437846016
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008760682046138415,
      "loss": 1.7317,
      "theoretical_loss": 4.0002826400966685,
      "tokens_seen": 437911552
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008760481444333,
      "loss": 1.6734,
      "theoretical_loss": 4.0002171543963225,
      "tokens_seen": 437977088
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008760280842527582,
      "loss": 1.6857,
      "theoretical_loss": 4.000151681237301,
      "tokens_seen": 438042624
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008760080240722167,
      "loss": 1.5352,
      "theoretical_loss": 4.0000862206153265,
      "tokens_seen": 438108160
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000875987963891675,
      "loss": 1.6934,
      "theoretical_loss": 4.000020772526123,
      "tokens_seen": 438173696
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008759679037111334,
      "loss": 1.7256,
      "theoretical_loss": 3.999955336965418,
      "tokens_seen": 438239232
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008759478435305918,
      "loss": 1.718,
      "theoretical_loss": 3.9998899139289392,
      "tokens_seen": 438304768
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008759277833500502,
      "loss": 1.6726,
      "theoretical_loss": 3.999824503412417,
      "tokens_seen": 438370304
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008759077231695085,
      "loss": 1.5715,
      "theoretical_loss": 3.9997591054115853,
      "tokens_seen": 438435840
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008758876629889669,
      "loss": 1.6995,
      "theoretical_loss": 3.9996937199221785,
      "tokens_seen": 438501376
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008758676028084253,
      "loss": 1.7369,
      "theoretical_loss": 3.999628346939934,
      "tokens_seen": 438566912
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008758475426278837,
      "loss": 1.7391,
      "theoretical_loss": 3.999562986460591,
      "tokens_seen": 438632448
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000875827482447342,
      "loss": 1.7917,
      "theoretical_loss": 3.9994976384798906,
      "tokens_seen": 438697984
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008758074222668005,
      "loss": 1.6776,
      "theoretical_loss": 3.9994323029935757,
      "tokens_seen": 438763520
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008757873620862587,
      "loss": 1.7134,
      "theoretical_loss": 3.9993669799973928,
      "tokens_seen": 438829056
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008757673019057172,
      "loss": 1.6238,
      "theoretical_loss": 3.999301669487088,
      "tokens_seen": 438894592
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008757472417251755,
      "loss": 1.7006,
      "theoretical_loss": 3.9992363714584123,
      "tokens_seen": 438960128
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008757271815446339,
      "loss": 1.6622,
      "theoretical_loss": 3.9991710859071157,
      "tokens_seen": 439025664
    },
    {
      "epoch": 0.13,
      "objective/train/advantage_avg": 0.49032777547836304,
      "objective/train/docs_used": 165818,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6073089838027954,
      "objective/train/original_loss": 1.6073087453842163,
      "objective/train/theoretical_loss": 3.9991058128289536,
      "objective/train/tokens_used": 459551200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24182981252670288,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502620935440063,
      "objective/train/weighted_lm_loss": 1.688266634941101,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9991058128289536,
      "tokens_seen": 439091200
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008757071213640923,
      "loss": 1.6988,
      "theoretical_loss": 3.9991058128289536,
      "tokens_seen": 439091200
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008756870611835507,
      "loss": 1.7678,
      "theoretical_loss": 3.9990405522196806,
      "tokens_seen": 439156736
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000875667001003009,
      "loss": 1.7156,
      "theoretical_loss": 3.9989753040750555,
      "tokens_seen": 439222272
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008756469408224675,
      "loss": 1.8251,
      "theoretical_loss": 3.9989100683908374,
      "tokens_seen": 439287808
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008756268806419257,
      "loss": 1.7089,
      "theoretical_loss": 3.998844845162789,
      "tokens_seen": 439353344
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008756068204613842,
      "loss": 1.7908,
      "theoretical_loss": 3.9987796343866746,
      "tokens_seen": 439418880
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008755867602808425,
      "loss": 1.6618,
      "theoretical_loss": 3.9987144360582594,
      "tokens_seen": 439484416
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008755667001003009,
      "loss": 1.6314,
      "theoretical_loss": 3.9986492501733126,
      "tokens_seen": 439549952
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008755466399197593,
      "loss": 1.7443,
      "theoretical_loss": 3.998584076727604,
      "tokens_seen": 439615488
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008755265797392176,
      "loss": 1.8115,
      "theoretical_loss": 3.9985189157169065,
      "tokens_seen": 439681024
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000875506519558676,
      "loss": 1.7254,
      "theoretical_loss": 3.998453767136994,
      "tokens_seen": 439746560
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008754864593781345,
      "loss": 1.6744,
      "theoretical_loss": 3.9983886309836434,
      "tokens_seen": 439812096
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008754663991975927,
      "loss": 1.7076,
      "theoretical_loss": 3.998323507252633,
      "tokens_seen": 439877632
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008754463390170512,
      "loss": 1.6958,
      "theoretical_loss": 3.998258395939744,
      "tokens_seen": 439943168
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008754262788365094,
      "loss": 1.6202,
      "theoretical_loss": 3.998193297040758,
      "tokens_seen": 440008704
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008754062186559679,
      "loss": 1.673,
      "theoretical_loss": 3.998128210551461,
      "tokens_seen": 440074240
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008753861584754263,
      "loss": 1.6711,
      "theoretical_loss": 3.998063136467639,
      "tokens_seen": 440139776
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008753660982948846,
      "loss": 1.7422,
      "theoretical_loss": 3.997998074785082,
      "tokens_seen": 440205312
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008753460381143431,
      "loss": 1.7045,
      "theoretical_loss": 3.99793302549958,
      "tokens_seen": 440270848
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008753259779338015,
      "loss": 1.6724,
      "theoretical_loss": 3.997867988606926,
      "tokens_seen": 440336384
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008753059177532598,
      "loss": 1.7573,
      "theoretical_loss": 3.9978029641029154,
      "tokens_seen": 440401920
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008752858575727182,
      "loss": 1.6944,
      "theoretical_loss": 3.997737951983345,
      "tokens_seen": 440467456
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008752657973921766,
      "loss": 1.6073,
      "theoretical_loss": 3.997672952244014,
      "tokens_seen": 440532992
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008752457372116349,
      "loss": 1.8349,
      "theoretical_loss": 3.997607964880724,
      "tokens_seen": 440598528
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008752256770310934,
      "loss": 1.5962,
      "theoretical_loss": 3.9975429898892783,
      "tokens_seen": 440664064
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008752056168505516,
      "loss": 1.6863,
      "theoretical_loss": 3.997478027265481,
      "tokens_seen": 440729600
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008751855566700101,
      "loss": 1.7402,
      "theoretical_loss": 3.997413077005141,
      "tokens_seen": 440795136
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008751654964894684,
      "loss": 1.6486,
      "theoretical_loss": 3.9973481391040675,
      "tokens_seen": 440860672
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008751454363089268,
      "loss": 1.7126,
      "theoretical_loss": 3.9972832135580707,
      "tokens_seen": 440926208
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008751253761283852,
      "loss": 1.643,
      "theoretical_loss": 3.9972183003629653,
      "tokens_seen": 440991744
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008751053159478436,
      "loss": 1.7164,
      "theoretical_loss": 3.9971533995145663,
      "tokens_seen": 441057280
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008750852557673019,
      "loss": 1.7811,
      "theoretical_loss": 3.997088511008691,
      "tokens_seen": 441122816
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008750651955867604,
      "loss": 1.6393,
      "theoretical_loss": 3.9970236348411605,
      "tokens_seen": 441188352
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008750451354062186,
      "loss": 1.6637,
      "theoretical_loss": 3.996958771007794,
      "tokens_seen": 441253888
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008750250752256771,
      "loss": 1.7993,
      "theoretical_loss": 3.996893919504417,
      "tokens_seen": 441319424
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008750050150451354,
      "loss": 1.6861,
      "theoretical_loss": 3.996829080326855,
      "tokens_seen": 441384960
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008749849548645938,
      "loss": 1.615,
      "theoretical_loss": 3.996764253470935,
      "tokens_seen": 441450496
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008749648946840522,
      "loss": 1.6733,
      "theoretical_loss": 3.9966994389324872,
      "tokens_seen": 441516032
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008749448345035106,
      "loss": 1.6116,
      "theoretical_loss": 3.996634636707344,
      "tokens_seen": 441581568
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008749247743229689,
      "loss": 1.6706,
      "theoretical_loss": 3.9965698467913375,
      "tokens_seen": 441647104
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008749047141424273,
      "loss": 1.4712,
      "theoretical_loss": 3.996505069180306,
      "tokens_seen": 441712640
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008748846539618857,
      "loss": 1.73,
      "theoretical_loss": 3.9964403038700844,
      "tokens_seen": 441778176
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008748645937813441,
      "loss": 1.7456,
      "theoretical_loss": 3.9963755508565155,
      "tokens_seen": 441843712
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008748445336008024,
      "loss": 1.6791,
      "theoretical_loss": 3.99631081013544,
      "tokens_seen": 441909248
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008748244734202609,
      "loss": 1.6616,
      "theoretical_loss": 3.9962460817027017,
      "tokens_seen": 441974784
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008748044132397191,
      "loss": 1.6458,
      "theoretical_loss": 3.996181365554146,
      "tokens_seen": 442040320
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008747843530591776,
      "loss": 1.8686,
      "theoretical_loss": 3.9961166616856225,
      "tokens_seen": 442105856
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008747642928786359,
      "loss": 1.6062,
      "theoretical_loss": 3.9960519700929797,
      "tokens_seen": 442171392
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008747442326980943,
      "loss": 1.6539,
      "theoretical_loss": 3.995987290772071,
      "tokens_seen": 442236928
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008747241725175527,
      "loss": 1.7101,
      "theoretical_loss": 3.995922623718749,
      "tokens_seen": 442302464
    },
    {
      "epoch": 0.13,
      "objective/train/advantage_avg": 0.4943990111351013,
      "objective/train/docs_used": 166951,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5713313817977905,
      "objective/train/original_loss": 1.57133150100708,
      "objective/train/theoretical_loss": 3.9958579689288705,
      "objective/train/tokens_used": 462828000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24468739330768585,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050683856010437,
      "objective/train/weighted_lm_loss": 1.6510241031646729,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9958579689288705,
      "tokens_seen": 442368000
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008747041123370111,
      "loss": 1.719,
      "theoretical_loss": 3.9958579689288705,
      "tokens_seen": 442368000
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008746840521564694,
      "loss": 1.5793,
      "theoretical_loss": 3.995793326398293,
      "tokens_seen": 442433536
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008746639919759279,
      "loss": 1.6893,
      "theoretical_loss": 3.9957286961228786,
      "tokens_seen": 442499072
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008746439317953861,
      "loss": 1.6669,
      "theoretical_loss": 3.995664078098487,
      "tokens_seen": 442564608
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008746238716148446,
      "loss": 1.5583,
      "theoretical_loss": 3.9955994723209836,
      "tokens_seen": 442630144
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008746038114343029,
      "loss": 1.717,
      "theoretical_loss": 3.9955348787862337,
      "tokens_seen": 442695680
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008745837512537613,
      "loss": 1.5958,
      "theoretical_loss": 3.995470297490106,
      "tokens_seen": 442761216
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008745636910732197,
      "loss": 1.7998,
      "theoretical_loss": 3.9954057284284703,
      "tokens_seen": 442826752
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000874543630892678,
      "loss": 1.6256,
      "theoretical_loss": 3.9953411715971994,
      "tokens_seen": 442892288
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008745235707121364,
      "loss": 1.6284,
      "theoretical_loss": 3.9952766269921667,
      "tokens_seen": 442957824
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008745035105315949,
      "loss": 1.4994,
      "theoretical_loss": 3.995212094609249,
      "tokens_seen": 443023360
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008744834503510531,
      "loss": 1.8477,
      "theoretical_loss": 3.9951475744443234,
      "tokens_seen": 443088896
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008744633901705116,
      "loss": 1.5417,
      "theoretical_loss": 3.9950830664932706,
      "tokens_seen": 443154432
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008744433299899698,
      "loss": 1.6993,
      "theoretical_loss": 3.9950185707519736,
      "tokens_seen": 443219968
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008744232698094283,
      "loss": 1.5878,
      "theoretical_loss": 3.994954087216315,
      "tokens_seen": 443285504
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008744032096288867,
      "loss": 1.7611,
      "theoretical_loss": 3.994889615882182,
      "tokens_seen": 443351040
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000874383149448345,
      "loss": 1.6522,
      "theoretical_loss": 3.9948251567454625,
      "tokens_seen": 443416576
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008743630892678034,
      "loss": 1.7121,
      "theoretical_loss": 3.994760709802046,
      "tokens_seen": 443482112
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008743430290872619,
      "loss": 1.7826,
      "theoretical_loss": 3.994696275047825,
      "tokens_seen": 443547648
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008743229689067201,
      "loss": 1.6826,
      "theoretical_loss": 3.994631852478695,
      "tokens_seen": 443613184
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008743029087261786,
      "loss": 1.6519,
      "theoretical_loss": 3.99456744209055,
      "tokens_seen": 443678720
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008742828485456368,
      "loss": 1.6761,
      "theoretical_loss": 3.9945030438792886,
      "tokens_seen": 443744256
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008742627883650953,
      "loss": 1.7019,
      "theoretical_loss": 3.9944386578408113,
      "tokens_seen": 443809792
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008742427281845537,
      "loss": 1.638,
      "theoretical_loss": 3.99437428397102,
      "tokens_seen": 443875328
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000874222668004012,
      "loss": 1.673,
      "theoretical_loss": 3.9943099222658187,
      "tokens_seen": 443940864
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008742026078234704,
      "loss": 1.6107,
      "theoretical_loss": 3.9942455727211135,
      "tokens_seen": 444006400
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008741825476429288,
      "loss": 1.6643,
      "theoretical_loss": 3.9941812353328126,
      "tokens_seen": 444071936
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008741624874623871,
      "loss": 1.7482,
      "theoretical_loss": 3.9941169100968255,
      "tokens_seen": 444137472
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008741424272818456,
      "loss": 1.7802,
      "theoretical_loss": 3.9940525970090643,
      "tokens_seen": 444203008
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008741223671013039,
      "loss": 1.5553,
      "theoretical_loss": 3.993988296065443,
      "tokens_seen": 444268544
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008741023069207623,
      "loss": 1.5713,
      "theoretical_loss": 3.993924007261878,
      "tokens_seen": 444334080
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008740822467402206,
      "loss": 1.6922,
      "theoretical_loss": 3.993859730594286,
      "tokens_seen": 444399616
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000874062186559679,
      "loss": 1.8085,
      "theoretical_loss": 3.9937954660585877,
      "tokens_seen": 444465152
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008740421263791374,
      "loss": 1.6667,
      "theoretical_loss": 3.9937312136507055,
      "tokens_seen": 444530688
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008740220661985958,
      "loss": 1.717,
      "theoretical_loss": 3.9936669733665617,
      "tokens_seen": 444596224
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008740020060180541,
      "loss": 1.6322,
      "theoretical_loss": 3.9936027452020832,
      "tokens_seen": 444661760
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008739819458375126,
      "loss": 1.6617,
      "theoretical_loss": 3.9935385291531977,
      "tokens_seen": 444727296
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008739618856569709,
      "loss": 1.6524,
      "theoretical_loss": 3.9934743252158347,
      "tokens_seen": 444792832
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008739418254764293,
      "loss": 1.7594,
      "theoretical_loss": 3.9934101333859253,
      "tokens_seen": 444858368
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008739217652958876,
      "loss": 1.7404,
      "theoretical_loss": 3.9933459536594036,
      "tokens_seen": 444923904
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008739017051153461,
      "loss": 1.8118,
      "theoretical_loss": 3.993281786032206,
      "tokens_seen": 444989440
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008738816449348044,
      "loss": 1.6814,
      "theoretical_loss": 3.9932176305002693,
      "tokens_seen": 445054976
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008738615847542628,
      "loss": 1.6673,
      "theoretical_loss": 3.9931534870595327,
      "tokens_seen": 445120512
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008738415245737213,
      "loss": 1.6706,
      "theoretical_loss": 3.993089355705939,
      "tokens_seen": 445186048
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008738214643931795,
      "loss": 1.6417,
      "theoretical_loss": 3.9930252364354297,
      "tokens_seen": 445251584
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000873801404212638,
      "loss": 1.7252,
      "theoretical_loss": 3.9929611292439517,
      "tokens_seen": 445317120
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008737813440320963,
      "loss": 1.6898,
      "theoretical_loss": 3.9928970341274517,
      "tokens_seen": 445382656
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0008737612838515547,
      "loss": 1.6426,
      "theoretical_loss": 3.99283295108188,
      "tokens_seen": 445448192
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008737412236710131,
      "loss": 1.6633,
      "theoretical_loss": 3.9927688801031866,
      "tokens_seen": 445513728
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008737211634904715,
      "loss": 1.6642,
      "theoretical_loss": 3.992704821187325,
      "tokens_seen": 445579264
    },
    {
      "epoch": 0.14,
      "objective/train/advantage_avg": 0.4929441809654236,
      "objective/train/docs_used": 168097,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7667591571807861,
      "objective/train/original_loss": 1.7667592763900757,
      "objective/train/theoretical_loss": 3.992640774330251,
      "objective/train/tokens_used": 466104800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24337506294250488,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505316257476807,
      "objective/train/weighted_lm_loss": 1.855811595916748,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.992640774330251,
      "tokens_seen": 445644800
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008737011033099298,
      "loss": 1.7735,
      "theoretical_loss": 3.992640774330251,
      "tokens_seen": 445644800
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008736810431293883,
      "loss": 1.6244,
      "theoretical_loss": 3.992576739527922,
      "tokens_seen": 445710336
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008736609829488465,
      "loss": 1.7317,
      "theoretical_loss": 3.9925127167762966,
      "tokens_seen": 445775872
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000873640922768305,
      "loss": 1.6294,
      "theoretical_loss": 3.992448706071335,
      "tokens_seen": 445841408
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008736208625877633,
      "loss": 1.582,
      "theoretical_loss": 3.9923847074090015,
      "tokens_seen": 445906944
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008736008024072217,
      "loss": 1.658,
      "theoretical_loss": 3.9923207207852602,
      "tokens_seen": 445972480
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008735807422266801,
      "loss": 1.6315,
      "theoretical_loss": 3.9922567461960785,
      "tokens_seen": 446038016
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008735606820461384,
      "loss": 1.6296,
      "theoretical_loss": 3.9921927836374245,
      "tokens_seen": 446103552
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008735406218655968,
      "loss": 1.6206,
      "theoretical_loss": 3.9921288331052702,
      "tokens_seen": 446169088
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008735205616850553,
      "loss": 1.6181,
      "theoretical_loss": 3.992064894595587,
      "tokens_seen": 446234624
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008735005015045135,
      "loss": 1.6894,
      "theoretical_loss": 3.9920009681043505,
      "tokens_seen": 446300160
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000873480441323972,
      "loss": 1.7492,
      "theoretical_loss": 3.9919370536275367,
      "tokens_seen": 446365696
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008734603811434302,
      "loss": 1.635,
      "theoretical_loss": 3.991873151161124,
      "tokens_seen": 446431232
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008734403209628887,
      "loss": 1.6543,
      "theoretical_loss": 3.9918092607010935,
      "tokens_seen": 446496768
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008734202607823471,
      "loss": 1.7162,
      "theoretical_loss": 3.991745382243427,
      "tokens_seen": 446562304
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008734002006018054,
      "loss": 1.8018,
      "theoretical_loss": 3.9916815157841095,
      "tokens_seen": 446627840
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008733801404212638,
      "loss": 1.7373,
      "theoretical_loss": 3.9916176613191263,
      "tokens_seen": 446693376
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008733600802407223,
      "loss": 1.6995,
      "theoretical_loss": 3.9915538188444666,
      "tokens_seen": 446758912
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008733400200601805,
      "loss": 1.6555,
      "theoretical_loss": 3.9914899883561197,
      "tokens_seen": 446824448
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000873319959879639,
      "loss": 1.693,
      "theoretical_loss": 3.9914261698500777,
      "tokens_seen": 446889984
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008732998996990972,
      "loss": 1.6775,
      "theoretical_loss": 3.9913623633223354,
      "tokens_seen": 446955520
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008732798395185557,
      "loss": 1.5977,
      "theoretical_loss": 3.9912985687688876,
      "tokens_seen": 447021056
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008732597793380141,
      "loss": 1.6131,
      "theoretical_loss": 3.9912347861857334,
      "tokens_seen": 447086592
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008732397191574724,
      "loss": 1.8031,
      "theoretical_loss": 3.9911710155688715,
      "tokens_seen": 447152128
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008732196589769308,
      "loss": 1.4963,
      "theoretical_loss": 3.9911072569143036,
      "tokens_seen": 447217664
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008731995987963892,
      "loss": 1.7504,
      "theoretical_loss": 3.9910435102180344,
      "tokens_seen": 447283200
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008731795386158475,
      "loss": 1.6181,
      "theoretical_loss": 3.9909797754760676,
      "tokens_seen": 447348736
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000873159478435306,
      "loss": 1.5899,
      "theoretical_loss": 3.990916052684412,
      "tokens_seen": 447414272
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008731394182547643,
      "loss": 1.6269,
      "theoretical_loss": 3.9908523418390764,
      "tokens_seen": 447479808
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008731193580742227,
      "loss": 1.7889,
      "theoretical_loss": 3.9907886429360726,
      "tokens_seen": 447545344
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000873099297893681,
      "loss": 1.6514,
      "theoretical_loss": 3.9907249559714133,
      "tokens_seen": 447610880
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008730792377131394,
      "loss": 1.5434,
      "theoretical_loss": 3.9906612809411137,
      "tokens_seen": 447676416
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008730591775325978,
      "loss": 1.5948,
      "theoretical_loss": 3.990597617841191,
      "tokens_seen": 447741952
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008730391173520562,
      "loss": 1.6374,
      "theoretical_loss": 3.9905339666676642,
      "tokens_seen": 447807488
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008730190571715145,
      "loss": 1.5835,
      "theoretical_loss": 3.9904703274165527,
      "tokens_seen": 447873024
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000872998996990973,
      "loss": 1.7278,
      "theoretical_loss": 3.9904067000838817,
      "tokens_seen": 447938560
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008729789368104313,
      "loss": 1.6106,
      "theoretical_loss": 3.9903430846656742,
      "tokens_seen": 448004096
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008729588766298897,
      "loss": 1.7808,
      "theoretical_loss": 3.990279481157957,
      "tokens_seen": 448069632
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000872938816449348,
      "loss": 1.6589,
      "theoretical_loss": 3.990215889556759,
      "tokens_seen": 448135168
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008729187562688065,
      "loss": 1.6781,
      "theoretical_loss": 3.990152309858111,
      "tokens_seen": 448200704
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008728986960882648,
      "loss": 1.6864,
      "theoretical_loss": 3.990088742058043,
      "tokens_seen": 448266240
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008728786359077232,
      "loss": 1.7106,
      "theoretical_loss": 3.990025186152592,
      "tokens_seen": 448331776
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008728585757271815,
      "loss": 1.5557,
      "theoretical_loss": 3.989961642137792,
      "tokens_seen": 448397312
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008728385155466399,
      "loss": 1.6223,
      "theoretical_loss": 3.9898981100096824,
      "tokens_seen": 448462848
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008728184553660983,
      "loss": 1.6308,
      "theoretical_loss": 3.9898345897643024,
      "tokens_seen": 448528384
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008727983951855567,
      "loss": 1.7198,
      "theoretical_loss": 3.9897710813976937,
      "tokens_seen": 448593920
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000872778335005015,
      "loss": 1.5826,
      "theoretical_loss": 3.9897075849059003,
      "tokens_seen": 448659456
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008727582748244735,
      "loss": 1.5786,
      "theoretical_loss": 3.989644100284967,
      "tokens_seen": 448724992
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008727382146439318,
      "loss": 1.617,
      "theoretical_loss": 3.989580627530943,
      "tokens_seen": 448790528
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008727181544633902,
      "loss": 1.6481,
      "theoretical_loss": 3.989517166639876,
      "tokens_seen": 448856064
    },
    {
      "epoch": 0.14,
      "objective/train/advantage_avg": 0.49197646975517273,
      "objective/train/docs_used": 169309,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.678598403930664,
      "objective/train/original_loss": 1.678598403930664,
      "objective/train/theoretical_loss": 3.9894537176078178,
      "objective/train/tokens_used": 469381600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24267804622650146,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504313707351685,
      "objective/train/weighted_lm_loss": 1.7632415294647217,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9894537176078178,
      "tokens_seen": 448921600
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008726980942828485,
      "loss": 1.6911,
      "theoretical_loss": 3.9894537176078178,
      "tokens_seen": 448921600
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008726780341023069,
      "loss": 1.6666,
      "theoretical_loss": 3.989390280430821,
      "tokens_seen": 448987136
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008726579739217653,
      "loss": 1.5208,
      "theoretical_loss": 3.9893268551049417,
      "tokens_seen": 449052672
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008726379137412237,
      "loss": 1.755,
      "theoretical_loss": 3.989263441626236,
      "tokens_seen": 449118208
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000872617853560682,
      "loss": 1.6132,
      "theoretical_loss": 3.989200039990763,
      "tokens_seen": 449183744
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008725977933801405,
      "loss": 1.576,
      "theoretical_loss": 3.989136650194583,
      "tokens_seen": 449249280
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008725777331995987,
      "loss": 1.6667,
      "theoretical_loss": 3.9890732722337594,
      "tokens_seen": 449314816
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008725576730190572,
      "loss": 1.6623,
      "theoretical_loss": 3.9890099061043562,
      "tokens_seen": 449380352
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008725376128385156,
      "loss": 1.7523,
      "theoretical_loss": 3.9889465518024396,
      "tokens_seen": 449445888
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008725175526579739,
      "loss": 1.7429,
      "theoretical_loss": 3.9888832093240776,
      "tokens_seen": 449511424
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008724974924774323,
      "loss": 1.7226,
      "theoretical_loss": 3.988819878665341,
      "tokens_seen": 449576960
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008724774322968906,
      "loss": 1.59,
      "theoretical_loss": 3.988756559822301,
      "tokens_seen": 449642496
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000872457372116349,
      "loss": 1.6499,
      "theoretical_loss": 3.988693252791032,
      "tokens_seen": 449708032
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008724373119358075,
      "loss": 1.7629,
      "theoretical_loss": 3.9886299575676096,
      "tokens_seen": 449773568
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008724172517552657,
      "loss": 1.6231,
      "theoretical_loss": 3.988566674148111,
      "tokens_seen": 449839104
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008723971915747242,
      "loss": 1.6632,
      "theoretical_loss": 3.988503402528617,
      "tokens_seen": 449904640
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008723771313941826,
      "loss": 1.726,
      "theoretical_loss": 3.9884401427052074,
      "tokens_seen": 449970176
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008723570712136409,
      "loss": 1.6258,
      "theoretical_loss": 3.9883768946739657,
      "tokens_seen": 450035712
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008723370110330994,
      "loss": 1.6387,
      "theoretical_loss": 3.988313658430978,
      "tokens_seen": 450101248
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008723169508525576,
      "loss": 1.6668,
      "theoretical_loss": 3.988250433972331,
      "tokens_seen": 450166784
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008722968906720161,
      "loss": 1.7694,
      "theoretical_loss": 3.9881872212941123,
      "tokens_seen": 450232320
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008722768304914745,
      "loss": 1.7324,
      "theoretical_loss": 3.9881240203924135,
      "tokens_seen": 450297856
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008722567703109328,
      "loss": 1.6488,
      "theoretical_loss": 3.9880608312633274,
      "tokens_seen": 450363392
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008722367101303912,
      "loss": 1.6934,
      "theoretical_loss": 3.9879976539029482,
      "tokens_seen": 450428928
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008722166499498496,
      "loss": 1.7163,
      "theoretical_loss": 3.9879344883073724,
      "tokens_seen": 450494464
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008721965897693079,
      "loss": 1.7422,
      "theoretical_loss": 3.9878713344726973,
      "tokens_seen": 450560000
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008721765295887664,
      "loss": 1.7008,
      "theoretical_loss": 3.9878081923950237,
      "tokens_seen": 450625536
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008721564694082247,
      "loss": 1.6685,
      "theoretical_loss": 3.9877450620704535,
      "tokens_seen": 450691072
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008721364092276831,
      "loss": 1.6977,
      "theoretical_loss": 3.9876819434950903,
      "tokens_seen": 450756608
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008721163490471414,
      "loss": 1.6814,
      "theoretical_loss": 3.987618836665039,
      "tokens_seen": 450822144
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008720962888665998,
      "loss": 1.6058,
      "theoretical_loss": 3.9875557415764087,
      "tokens_seen": 450887680
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008720762286860582,
      "loss": 1.6729,
      "theoretical_loss": 3.987492658225307,
      "tokens_seen": 450953216
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008720561685055166,
      "loss": 1.6695,
      "theoretical_loss": 3.987429586607846,
      "tokens_seen": 451018752
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008720361083249749,
      "loss": 1.5177,
      "theoretical_loss": 3.9873665267201384,
      "tokens_seen": 451084288
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008720160481444334,
      "loss": 1.6466,
      "theoretical_loss": 3.9873034785582995,
      "tokens_seen": 451149824
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008719959879638917,
      "loss": 1.5169,
      "theoretical_loss": 3.9872404421184453,
      "tokens_seen": 451215360
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008719759277833501,
      "loss": 1.6245,
      "theoretical_loss": 3.987177417396695,
      "tokens_seen": 451280896
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008719558676028084,
      "loss": 1.6227,
      "theoretical_loss": 3.9871144043891684,
      "tokens_seen": 451346432
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008719358074222669,
      "loss": 1.6857,
      "theoretical_loss": 3.9870514030919884,
      "tokens_seen": 451411968
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008719157472417252,
      "loss": 1.672,
      "theoretical_loss": 3.9869884135012787,
      "tokens_seen": 451477504
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008718956870611836,
      "loss": 1.6045,
      "theoretical_loss": 3.9869254356131654,
      "tokens_seen": 451543040
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008718756268806419,
      "loss": 1.6892,
      "theoretical_loss": 3.9868624694237766,
      "tokens_seen": 451608576
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008718555667001003,
      "loss": 1.6541,
      "theoretical_loss": 3.986799514929242,
      "tokens_seen": 451674112
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008718355065195587,
      "loss": 1.6627,
      "theoretical_loss": 3.986736572125692,
      "tokens_seen": 451739648
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008718154463390171,
      "loss": 1.653,
      "theoretical_loss": 3.9866736410092614,
      "tokens_seen": 451805184
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008717953861584754,
      "loss": 1.7972,
      "theoretical_loss": 3.986610721576084,
      "tokens_seen": 451870720
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008717753259779339,
      "loss": 1.7353,
      "theoretical_loss": 3.9865478138222974,
      "tokens_seen": 451936256
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008717552657973922,
      "loss": 1.6426,
      "theoretical_loss": 3.9864849177440407,
      "tokens_seen": 452001792
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008717352056168506,
      "loss": 1.6807,
      "theoretical_loss": 3.9864220333374547,
      "tokens_seen": 452067328
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000871715145436309,
      "loss": 1.6885,
      "theoretical_loss": 3.9863591605986812,
      "tokens_seen": 452132864
    },
    {
      "epoch": 0.14,
      "objective/train/advantage_avg": 0.49392029643058777,
      "objective/train/docs_used": 170598,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.682446002960205,
      "objective/train/original_loss": 1.682446002960205,
      "objective/train/theoretical_loss": 3.9862962995238647,
      "objective/train/tokens_used": 472658400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2446342259645462,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506356954574585,
      "objective/train/weighted_lm_loss": 1.7676743268966675,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9862962995238647,
      "tokens_seen": 452198400
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008716950852557673,
      "loss": 1.7119,
      "theoretical_loss": 3.9862962995238647,
      "tokens_seen": 452198400
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008716750250752257,
      "loss": 1.7363,
      "theoretical_loss": 3.986233450109152,
      "tokens_seen": 452263936
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008716549648946841,
      "loss": 1.7325,
      "theoretical_loss": 3.9861706123506906,
      "tokens_seen": 452329472
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008716349047141424,
      "loss": 1.7326,
      "theoretical_loss": 3.986107786244631,
      "tokens_seen": 452395008
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008716148445336009,
      "loss": 1.7056,
      "theoretical_loss": 3.9860449717871234,
      "tokens_seen": 452460544
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008715947843530591,
      "loss": 1.6295,
      "theoretical_loss": 3.985982168974323,
      "tokens_seen": 452526080
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008715747241725176,
      "loss": 1.6789,
      "theoretical_loss": 3.985919377802384,
      "tokens_seen": 452591616
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000871554663991976,
      "loss": 1.5917,
      "theoretical_loss": 3.9858565982674636,
      "tokens_seen": 452657152
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008715346038114343,
      "loss": 1.7548,
      "theoretical_loss": 3.9857938303657217,
      "tokens_seen": 452722688
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008715145436308927,
      "loss": 1.7352,
      "theoretical_loss": 3.985731074093318,
      "tokens_seen": 452788224
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000871494483450351,
      "loss": 1.6529,
      "theoretical_loss": 3.985668329446416,
      "tokens_seen": 452853760
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008714744232698094,
      "loss": 1.7179,
      "theoretical_loss": 3.9856055964211796,
      "tokens_seen": 452919296
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008714543630892679,
      "loss": 1.8021,
      "theoretical_loss": 3.9855428750137754,
      "tokens_seen": 452984832
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008714343029087261,
      "loss": 1.7151,
      "theoretical_loss": 3.9854801652203706,
      "tokens_seen": 453050368
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008714142427281846,
      "loss": 1.672,
      "theoretical_loss": 3.985417467037137,
      "tokens_seen": 453115904
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000871394182547643,
      "loss": 1.7477,
      "theoretical_loss": 3.985354780460244,
      "tokens_seen": 453181440
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008713741223671013,
      "loss": 1.8086,
      "theoretical_loss": 3.9852921054858665,
      "tokens_seen": 453246976
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008713540621865597,
      "loss": 1.7011,
      "theoretical_loss": 3.98522944211018,
      "tokens_seen": 453312512
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000871334002006018,
      "loss": 1.6457,
      "theoretical_loss": 3.9851667903293615,
      "tokens_seen": 453378048
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008713139418254764,
      "loss": 1.6447,
      "theoretical_loss": 3.9851041501395894,
      "tokens_seen": 453443584
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008712938816449349,
      "loss": 1.6089,
      "theoretical_loss": 3.9850415215370445,
      "tokens_seen": 453509120
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008712738214643931,
      "loss": 1.578,
      "theoretical_loss": 3.9849789045179103,
      "tokens_seen": 453574656
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008712537612838516,
      "loss": 1.5896,
      "theoretical_loss": 3.98491629907837,
      "tokens_seen": 453640192
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008712337011033099,
      "loss": 1.6224,
      "theoretical_loss": 3.984853705214611,
      "tokens_seen": 453705728
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008712136409227683,
      "loss": 1.5858,
      "theoretical_loss": 3.984791122922821,
      "tokens_seen": 453771264
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008711935807422267,
      "loss": 1.7106,
      "theoretical_loss": 3.984728552199189,
      "tokens_seen": 453836800
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008711735205616851,
      "loss": 1.6795,
      "theoretical_loss": 3.984665993039907,
      "tokens_seen": 453902336
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008711534603811434,
      "loss": 1.6176,
      "theoretical_loss": 3.984603445441169,
      "tokens_seen": 453967872
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008711334002006018,
      "loss": 1.6872,
      "theoretical_loss": 3.98454090939917,
      "tokens_seen": 454033408
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008711133400200601,
      "loss": 1.6369,
      "theoretical_loss": 3.9844783849101075,
      "tokens_seen": 454098944
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008710932798395186,
      "loss": 1.6765,
      "theoretical_loss": 3.984415871970179,
      "tokens_seen": 454164480
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008710732196589769,
      "loss": 1.6515,
      "theoretical_loss": 3.984353370575586,
      "tokens_seen": 454230016
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008710531594784353,
      "loss": 1.7684,
      "theoretical_loss": 3.984290880722531,
      "tokens_seen": 454295552
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008710330992978937,
      "loss": 1.6004,
      "theoretical_loss": 3.9842284024072177,
      "tokens_seen": 454361088
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008710130391173521,
      "loss": 1.659,
      "theoretical_loss": 3.9841659356258523,
      "tokens_seen": 454426624
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008709929789368104,
      "loss": 1.621,
      "theoretical_loss": 3.984103480374644,
      "tokens_seen": 454492160
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008709729187562688,
      "loss": 1.6278,
      "theoretical_loss": 3.9840410366498,
      "tokens_seen": 454557696
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008709528585757271,
      "loss": 1.8119,
      "theoretical_loss": 3.9839786044475334,
      "tokens_seen": 454623232
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008709327983951856,
      "loss": 1.6085,
      "theoretical_loss": 3.983916183764057,
      "tokens_seen": 454688768
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008709127382146439,
      "loss": 1.6838,
      "theoretical_loss": 3.9838537745955858,
      "tokens_seen": 454754304
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008708926780341023,
      "loss": 1.6876,
      "theoretical_loss": 3.983791376938336,
      "tokens_seen": 454819840
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008708726178535606,
      "loss": 1.5865,
      "theoretical_loss": 3.9837289907885274,
      "tokens_seen": 454885376
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008708525576730191,
      "loss": 1.6709,
      "theoretical_loss": 3.9836666161423793,
      "tokens_seen": 454950912
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008708324974924775,
      "loss": 1.616,
      "theoretical_loss": 3.9836042529961135,
      "tokens_seen": 455016448
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008708124373119358,
      "loss": 1.6237,
      "theoretical_loss": 3.983541901345955,
      "tokens_seen": 455081984
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008707923771313943,
      "loss": 1.7014,
      "theoretical_loss": 3.9834795611881297,
      "tokens_seen": 455147520
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008707723169508526,
      "loss": 1.7364,
      "theoretical_loss": 3.983417232518864,
      "tokens_seen": 455213056
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000870752256770311,
      "loss": 1.6464,
      "theoretical_loss": 3.9833549153343872,
      "tokens_seen": 455278592
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008707321965897693,
      "loss": 1.7088,
      "theoretical_loss": 3.983292609630931,
      "tokens_seen": 455344128
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008707121364092277,
      "loss": 1.681,
      "theoretical_loss": 3.983230315404728,
      "tokens_seen": 455409664
    },
    {
      "epoch": 0.14,
      "objective/train/advantage_avg": 0.48992982506752014,
      "objective/train/docs_used": 171735,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6366153955459595,
      "objective/train/original_loss": 1.63661527633667,
      "objective/train/theoretical_loss": 3.983168032652013,
      "objective/train/tokens_used": 475935200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2419193685054779,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502227544784546,
      "objective/train/weighted_lm_loss": 1.718919277191162,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.983168032652013,
      "tokens_seen": 455475200
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008706920762286861,
      "loss": 1.7645,
      "theoretical_loss": 3.983168032652013,
      "tokens_seen": 455475200
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008706720160481445,
      "loss": 1.6411,
      "theoretical_loss": 3.983105761369022,
      "tokens_seen": 455540736
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008706519558676028,
      "loss": 1.7468,
      "theoretical_loss": 3.9830435015519936,
      "tokens_seen": 455606272
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008706318956870613,
      "loss": 1.6752,
      "theoretical_loss": 3.9829812531971673,
      "tokens_seen": 455671808
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008706118355065195,
      "loss": 1.6877,
      "theoretical_loss": 3.9829190163007846,
      "tokens_seen": 455737344
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000870591775325978,
      "loss": 1.6541,
      "theoretical_loss": 3.98285679085909,
      "tokens_seen": 455802880
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008705717151454364,
      "loss": 1.709,
      "theoretical_loss": 3.982794576868328,
      "tokens_seen": 455868416
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008705516549648947,
      "loss": 1.6451,
      "theoretical_loss": 3.9827323743247454,
      "tokens_seen": 455933952
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008705315947843531,
      "loss": 1.7059,
      "theoretical_loss": 3.9826701832245917,
      "tokens_seen": 455999488
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008705115346038114,
      "loss": 1.6,
      "theoretical_loss": 3.982608003564117,
      "tokens_seen": 456065024
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008704914744232698,
      "loss": 1.6583,
      "theoretical_loss": 3.982545835339573,
      "tokens_seen": 456130560
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008704714142427283,
      "loss": 1.6567,
      "theoretical_loss": 3.982483678547215,
      "tokens_seen": 456196096
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008704513540621865,
      "loss": 1.6338,
      "theoretical_loss": 3.982421533183299,
      "tokens_seen": 456261632
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000870431293881645,
      "loss": 1.7144,
      "theoretical_loss": 3.982359399244081,
      "tokens_seen": 456327168
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008704112337011034,
      "loss": 1.6345,
      "theoretical_loss": 3.982297276725822,
      "tokens_seen": 456392704
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008703911735205617,
      "loss": 1.6444,
      "theoretical_loss": 3.9822351656247816,
      "tokens_seen": 456458240
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008703711133400201,
      "loss": 1.5523,
      "theoretical_loss": 3.982173065937225,
      "tokens_seen": 456523776
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008703510531594784,
      "loss": 1.7206,
      "theoretical_loss": 3.9821109776594144,
      "tokens_seen": 456589312
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008703309929789368,
      "loss": 1.6512,
      "theoretical_loss": 3.9820489007876176,
      "tokens_seen": 456654848
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008703109327983953,
      "loss": 1.5267,
      "theoretical_loss": 3.9819868353181027,
      "tokens_seen": 456720384
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008702908726178535,
      "loss": 1.6197,
      "theoretical_loss": 3.981924781247139,
      "tokens_seen": 456785920
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000870270812437312,
      "loss": 1.5956,
      "theoretical_loss": 3.9818627385709986,
      "tokens_seen": 456851456
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008702507522567703,
      "loss": 1.5874,
      "theoretical_loss": 3.981800707285955,
      "tokens_seen": 456916992
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008702306920762287,
      "loss": 1.6294,
      "theoretical_loss": 3.9817386873882836,
      "tokens_seen": 456982528
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008702106318956871,
      "loss": 1.7186,
      "theoretical_loss": 3.981676678874261,
      "tokens_seen": 457048064
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008701905717151455,
      "loss": 1.6111,
      "theoretical_loss": 3.981614681740166,
      "tokens_seen": 457113600
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008701705115346038,
      "loss": 1.721,
      "theoretical_loss": 3.9815526959822787,
      "tokens_seen": 457179136
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008701504513540622,
      "loss": 1.7458,
      "theoretical_loss": 3.9814907215968818,
      "tokens_seen": 457244672
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008701303911735205,
      "loss": 1.6473,
      "theoretical_loss": 3.9814287585802592,
      "tokens_seen": 457310208
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000870110330992979,
      "loss": 1.6326,
      "theoretical_loss": 3.9813668069286967,
      "tokens_seen": 457375744
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008700902708124373,
      "loss": 1.4838,
      "theoretical_loss": 3.981304866638481,
      "tokens_seen": 457441280
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008700702106318957,
      "loss": 1.718,
      "theoretical_loss": 3.981242937705902,
      "tokens_seen": 457506816
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008700501504513541,
      "loss": 1.7773,
      "theoretical_loss": 3.9811810201272513,
      "tokens_seen": 457572352
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008700300902708125,
      "loss": 1.6464,
      "theoretical_loss": 3.98111911389882,
      "tokens_seen": 457637888
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008700100300902708,
      "loss": 1.6777,
      "theoretical_loss": 3.9810572190169027,
      "tokens_seen": 457703424
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008699899699097292,
      "loss": 1.6622,
      "theoretical_loss": 3.9809953354777967,
      "tokens_seen": 457768960
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008699699097291875,
      "loss": 1.7199,
      "theoretical_loss": 3.9809334632777986,
      "tokens_seen": 457834496
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000869949849548646,
      "loss": 1.7178,
      "theoretical_loss": 3.9808716024132096,
      "tokens_seen": 457900032
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008699297893681043,
      "loss": 1.6575,
      "theoretical_loss": 3.9808097528803295,
      "tokens_seen": 457965568
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008699097291875627,
      "loss": 1.6969,
      "theoretical_loss": 3.9807479146754625,
      "tokens_seen": 458031104
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000869889669007021,
      "loss": 1.5759,
      "theoretical_loss": 3.9806860877949126,
      "tokens_seen": 458096640
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008698696088264795,
      "loss": 1.6771,
      "theoretical_loss": 3.9806242722349867,
      "tokens_seen": 458162176
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008698495486459378,
      "loss": 1.7865,
      "theoretical_loss": 3.9805624679919935,
      "tokens_seen": 458227712
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008698294884653962,
      "loss": 1.8104,
      "theoretical_loss": 3.9805006750622427,
      "tokens_seen": 458293248
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008698094282848546,
      "loss": 1.695,
      "theoretical_loss": 3.980438893442046,
      "tokens_seen": 458358784
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000869789368104313,
      "loss": 1.7279,
      "theoretical_loss": 3.9803771231277167,
      "tokens_seen": 458424320
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008697693079237713,
      "loss": 1.6565,
      "theoretical_loss": 3.98031536411557,
      "tokens_seen": 458489856
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008697492477432297,
      "loss": 1.6428,
      "theoretical_loss": 3.9802536164019235,
      "tokens_seen": 458555392
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000869729187562688,
      "loss": 1.6321,
      "theoretical_loss": 3.9801918799830958,
      "tokens_seen": 458620928
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008697091273821465,
      "loss": 1.7232,
      "theoretical_loss": 3.9801301548554067,
      "tokens_seen": 458686464
    },
    {
      "epoch": 0.14,
      "objective/train/advantage_avg": 0.49433034658432007,
      "objective/train/docs_used": 173009,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5577558279037476,
      "objective/train/original_loss": 1.557755708694458,
      "objective/train/theoretical_loss": 3.9800684410151783,
      "objective/train/tokens_used": 479212000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2445380836725235,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506762266159058,
      "objective/train/weighted_lm_loss": 1.6367053985595703,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9800684410151783,
      "tokens_seen": 458752000
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008696890672016048,
      "loss": 1.6053,
      "theoretical_loss": 3.9800684410151783,
      "tokens_seen": 458752000
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008696690070210632,
      "loss": 1.7288,
      "theoretical_loss": 3.9800067384587354,
      "tokens_seen": 458817536
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008696489468405216,
      "loss": 1.5646,
      "theoretical_loss": 3.9799450471824023,
      "tokens_seen": 458883072
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008696288866599799,
      "loss": 1.5993,
      "theoretical_loss": 3.9798833671825076,
      "tokens_seen": 458948608
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008696088264794383,
      "loss": 1.5431,
      "theoretical_loss": 3.979821698455379,
      "tokens_seen": 459014144
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008695887662988968,
      "loss": 1.6967,
      "theoretical_loss": 3.9797600409973484,
      "tokens_seen": 459079680
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000869568706118355,
      "loss": 1.7241,
      "theoretical_loss": 3.979698394804747,
      "tokens_seen": 459145216
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008695486459378135,
      "loss": 1.6775,
      "theoretical_loss": 3.9796367598739106,
      "tokens_seen": 459210752
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008695285857572717,
      "loss": 1.6272,
      "theoretical_loss": 3.9795751362011735,
      "tokens_seen": 459276288
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008695085255767302,
      "loss": 1.6908,
      "theoretical_loss": 3.9795135237828747,
      "tokens_seen": 459341824
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008694884653961886,
      "loss": 1.6438,
      "theoretical_loss": 3.979451922615352,
      "tokens_seen": 459407360
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008694684052156469,
      "loss": 1.7366,
      "theoretical_loss": 3.9793903326949476,
      "tokens_seen": 459472896
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008694483450351053,
      "loss": 1.5684,
      "theoretical_loss": 3.979328754018004,
      "tokens_seen": 459538432
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008694282848545638,
      "loss": 1.6673,
      "theoretical_loss": 3.9792671865808646,
      "tokens_seen": 459603968
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000869408224674022,
      "loss": 1.609,
      "theoretical_loss": 3.9792056303798775,
      "tokens_seen": 459669504
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008693881644934805,
      "loss": 1.6739,
      "theoretical_loss": 3.9791440854113893,
      "tokens_seen": 459735040
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008693681043129388,
      "loss": 1.7139,
      "theoretical_loss": 3.979082551671749,
      "tokens_seen": 459800576
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008693480441323972,
      "loss": 1.7229,
      "theoretical_loss": 3.97902102915731,
      "tokens_seen": 459866112
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008693279839518557,
      "loss": 1.6994,
      "theoretical_loss": 3.9789595178644226,
      "tokens_seen": 459931648
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008693079237713139,
      "loss": 1.7112,
      "theoretical_loss": 3.9788980177894433,
      "tokens_seen": 459997184
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008692878635907724,
      "loss": 1.7255,
      "theoretical_loss": 3.9788365289287286,
      "tokens_seen": 460062720
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008692678034102307,
      "loss": 1.6869,
      "theoretical_loss": 3.9787750512786353,
      "tokens_seen": 460128256
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008692477432296891,
      "loss": 1.6763,
      "theoretical_loss": 3.978713584835524,
      "tokens_seen": 460193792
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008692276830491475,
      "loss": 1.7387,
      "theoretical_loss": 3.9786521295957558,
      "tokens_seen": 460259328
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008692076228686059,
      "loss": 1.7619,
      "theoretical_loss": 3.9785906855556945,
      "tokens_seen": 460324864
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008691875626880642,
      "loss": 1.6247,
      "theoretical_loss": 3.978529252711705,
      "tokens_seen": 460390400
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008691675025075226,
      "loss": 1.6519,
      "theoretical_loss": 3.9784678310601533,
      "tokens_seen": 460455936
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008691474423269809,
      "loss": 1.6962,
      "theoretical_loss": 3.9784064205974077,
      "tokens_seen": 460521472
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008691273821464394,
      "loss": 1.6269,
      "theoretical_loss": 3.9783450213198384,
      "tokens_seen": 460587008
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008691073219658977,
      "loss": 1.5943,
      "theoretical_loss": 3.978283633223817,
      "tokens_seen": 460652544
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008690872617853561,
      "loss": 1.64,
      "theoretical_loss": 3.9782222563057177,
      "tokens_seen": 460718080
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008690672016048145,
      "loss": 1.673,
      "theoretical_loss": 3.978160890561914,
      "tokens_seen": 460783616
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008690471414242729,
      "loss": 1.6254,
      "theoretical_loss": 3.9780995359887843,
      "tokens_seen": 460849152
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008690270812437312,
      "loss": 1.6223,
      "theoretical_loss": 3.978038192582706,
      "tokens_seen": 460914688
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008690070210631896,
      "loss": 1.7393,
      "theoretical_loss": 3.9779768603400587,
      "tokens_seen": 460980224
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000868986960882648,
      "loss": 1.6811,
      "theoretical_loss": 3.977915539257226,
      "tokens_seen": 461045760
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008689669007021064,
      "loss": 1.7563,
      "theoretical_loss": 3.9778542293305894,
      "tokens_seen": 461111296
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008689468405215647,
      "loss": 1.8148,
      "theoretical_loss": 3.977792930556536,
      "tokens_seen": 461176832
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008689267803410231,
      "loss": 1.5884,
      "theoretical_loss": 3.9777316429314507,
      "tokens_seen": 461242368
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008689067201604814,
      "loss": 1.6516,
      "theoretical_loss": 3.9776703664517234,
      "tokens_seen": 461307904
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008688866599799399,
      "loss": 1.7036,
      "theoretical_loss": 3.977609101113744,
      "tokens_seen": 461373440
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008688665997993982,
      "loss": 1.6015,
      "theoretical_loss": 3.9775478469139047,
      "tokens_seen": 461438976
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008688465396188566,
      "loss": 1.5942,
      "theoretical_loss": 3.9774866038485985,
      "tokens_seen": 461504512
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000868826479438315,
      "loss": 1.6146,
      "theoretical_loss": 3.9774253719142214,
      "tokens_seen": 461570048
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008688064192577734,
      "loss": 1.517,
      "theoretical_loss": 3.97736415110717,
      "tokens_seen": 461635584
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008687863590772317,
      "loss": 1.8449,
      "theoretical_loss": 3.9773029414238428,
      "tokens_seen": 461701120
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008687662988966902,
      "loss": 1.7828,
      "theoretical_loss": 3.97724174286064,
      "tokens_seen": 461766656
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008687462387161484,
      "loss": 1.698,
      "theoretical_loss": 3.977180555413964,
      "tokens_seen": 461832192
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008687261785356069,
      "loss": 1.7773,
      "theoretical_loss": 3.977119379080218,
      "tokens_seen": 461897728
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008687061183550652,
      "loss": 1.6191,
      "theoretical_loss": 3.9770582138558077,
      "tokens_seen": 461963264
    },
    {
      "epoch": 0.14,
      "objective/train/advantage_avg": 0.4929687976837158,
      "objective/train/docs_used": 174193,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7287741899490356,
      "objective/train/original_loss": 1.728774070739746,
      "objective/train/theoretical_loss": 3.9769970597371405,
      "objective/train/tokens_used": 482488800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2434602826833725,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505346059799194,
      "objective/train/weighted_lm_loss": 1.8161708116531372,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.9769970597371405,
      "tokens_seen": 462028800
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008686860581745236,
      "loss": 1.7777,
      "theoretical_loss": 3.9769970597371405,
      "tokens_seen": 462028800
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000868665997993982,
      "loss": 1.7605,
      "theoretical_loss": 3.976935916720624,
      "tokens_seen": 462094336
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008686459378134403,
      "loss": 1.808,
      "theoretical_loss": 3.9768747848026695,
      "tokens_seen": 462159872
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008686258776328987,
      "loss": 1.5267,
      "theoretical_loss": 3.976813663979689,
      "tokens_seen": 462225408
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008686058174523572,
      "loss": 1.6349,
      "theoretical_loss": 3.976752554248095,
      "tokens_seen": 462290944
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008685857572718154,
      "loss": 1.6544,
      "theoretical_loss": 3.9766914556043043,
      "tokens_seen": 462356480
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008685656970912739,
      "loss": 1.7583,
      "theoretical_loss": 3.9766303680447335,
      "tokens_seen": 462422016
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008685456369107321,
      "loss": 1.6585,
      "theoretical_loss": 3.976569291565801,
      "tokens_seen": 462487552
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008685255767301906,
      "loss": 1.6723,
      "theoretical_loss": 3.9765082261639275,
      "tokens_seen": 462553088
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000868505516549649,
      "loss": 1.6631,
      "theoretical_loss": 3.976447171835534,
      "tokens_seen": 462618624
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008684854563691073,
      "loss": 1.5455,
      "theoretical_loss": 3.9763861285770457,
      "tokens_seen": 462684160
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008684653961885657,
      "loss": 1.6465,
      "theoretical_loss": 3.9763250963848873,
      "tokens_seen": 462749696
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008684453360080242,
      "loss": 1.6204,
      "theoretical_loss": 3.976264075255485,
      "tokens_seen": 462815232
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008684252758274824,
      "loss": 1.6969,
      "theoretical_loss": 3.9762030651852687,
      "tokens_seen": 462880768
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008684052156469409,
      "loss": 1.8199,
      "theoretical_loss": 3.9761420661706683,
      "tokens_seen": 462946304
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008683851554663991,
      "loss": 1.6761,
      "theoretical_loss": 3.976081078208116,
      "tokens_seen": 463011840
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008683650952858576,
      "loss": 1.6621,
      "theoretical_loss": 3.976020101294045,
      "tokens_seen": 463077376
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000868345035105316,
      "loss": 1.6585,
      "theoretical_loss": 3.97595913542489,
      "tokens_seen": 463142912
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008683249749247743,
      "loss": 1.5819,
      "theoretical_loss": 3.975898180597089,
      "tokens_seen": 463208448
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008683049147442327,
      "loss": 1.5759,
      "theoretical_loss": 3.97583723680708,
      "tokens_seen": 463273984
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008682848545636911,
      "loss": 1.6291,
      "theoretical_loss": 3.975776304051304,
      "tokens_seen": 463339520
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008682647943831494,
      "loss": 1.774,
      "theoretical_loss": 3.975715382326203,
      "tokens_seen": 463405056
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008682447342026079,
      "loss": 1.7556,
      "theoretical_loss": 3.9756544716282187,
      "tokens_seen": 463470592
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008682246740220661,
      "loss": 1.5184,
      "theoretical_loss": 3.975593571953798,
      "tokens_seen": 463536128
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008682046138415246,
      "loss": 1.6399,
      "theoretical_loss": 3.975532683299387,
      "tokens_seen": 463601664
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008681845536609829,
      "loss": 1.6015,
      "theoretical_loss": 3.975471805661435,
      "tokens_seen": 463667200
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008681644934804413,
      "loss": 1.7948,
      "theoretical_loss": 3.975410939036392,
      "tokens_seen": 463732736
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008681444332998997,
      "loss": 1.6474,
      "theoretical_loss": 3.9753500834207083,
      "tokens_seen": 463798272
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008681243731193581,
      "loss": 1.6236,
      "theoretical_loss": 3.975289238810839,
      "tokens_seen": 463863808
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008681043129388164,
      "loss": 1.7626,
      "theoretical_loss": 3.975228405203239,
      "tokens_seen": 463929344
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008680842527582749,
      "loss": 1.6521,
      "theoretical_loss": 3.9751675825943638,
      "tokens_seen": 463994880
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008680641925777332,
      "loss": 1.8156,
      "theoretical_loss": 3.9751067709806733,
      "tokens_seen": 464060416
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008680441323971916,
      "loss": 1.8843,
      "theoretical_loss": 3.975045970358626,
      "tokens_seen": 464125952
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008680240722166499,
      "loss": 1.7733,
      "theoretical_loss": 3.974985180724685,
      "tokens_seen": 464191488
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008680040120361083,
      "loss": 1.5756,
      "theoretical_loss": 3.974924402075313,
      "tokens_seen": 464257024
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008679839518555667,
      "loss": 1.6463,
      "theoretical_loss": 3.9748636344069745,
      "tokens_seen": 464322560
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008679638916750251,
      "loss": 1.8061,
      "theoretical_loss": 3.9748028777161366,
      "tokens_seen": 464388096
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008679438314944834,
      "loss": 1.6805,
      "theoretical_loss": 3.9747421319992666,
      "tokens_seen": 464453632
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008679237713139418,
      "loss": 1.6201,
      "theoretical_loss": 3.9746813972528354,
      "tokens_seen": 464519168
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008679037111334002,
      "loss": 1.6282,
      "theoretical_loss": 3.9746206734733143,
      "tokens_seen": 464584704
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008678836509528586,
      "loss": 1.6675,
      "theoretical_loss": 3.974559960657176,
      "tokens_seen": 464650240
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000867863590772317,
      "loss": 1.7889,
      "theoretical_loss": 3.974499258800895,
      "tokens_seen": 464715776
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008678435305917754,
      "loss": 1.7932,
      "theoretical_loss": 3.9744385679009486,
      "tokens_seen": 464781312
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008678234704112338,
      "loss": 1.5776,
      "theoretical_loss": 3.974377887953813,
      "tokens_seen": 464846848
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008678034102306921,
      "loss": 1.6429,
      "theoretical_loss": 3.97431721895597,
      "tokens_seen": 464912384
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008677833500501506,
      "loss": 1.7277,
      "theoretical_loss": 3.9742565609039,
      "tokens_seen": 464977920
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008677632898696088,
      "loss": 1.5898,
      "theoretical_loss": 3.9741959137940848,
      "tokens_seen": 465043456
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008677432296890673,
      "loss": 1.5901,
      "theoretical_loss": 3.9741352776230103,
      "tokens_seen": 465108992
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008677231695085256,
      "loss": 1.6582,
      "theoretical_loss": 3.9740746523871615,
      "tokens_seen": 465174528
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000867703109327984,
      "loss": 1.6925,
      "theoretical_loss": 3.974014038083027,
      "tokens_seen": 465240064
    },
    {
      "epoch": 0.14,
      "objective/train/advantage_avg": 0.49394601583480835,
      "objective/train/docs_used": 175471,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3470487594604492,
      "objective/train/original_loss": 1.3470487594604492,
      "objective/train/theoretical_loss": 3.973953434707096,
      "objective/train/tokens_used": 485765600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2442466765642166,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506362915039062,
      "objective/train/weighted_lm_loss": 1.415244221687317,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.973953434707096,
      "tokens_seen": 465305600
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008676830491474424,
      "loss": 1.5362,
      "theoretical_loss": 3.973953434707096,
      "tokens_seen": 465305600
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008676629889669007,
      "loss": 1.6045,
      "theoretical_loss": 3.973892842255859,
      "tokens_seen": 465371136
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008676429287863591,
      "loss": 1.6473,
      "theoretical_loss": 3.9738322607258088,
      "tokens_seen": 465436672
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008676228686058176,
      "loss": 1.7065,
      "theoretical_loss": 3.97377169011344,
      "tokens_seen": 465502208
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008676028084252758,
      "loss": 1.6219,
      "theoretical_loss": 3.973711130415248,
      "tokens_seen": 465567744
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008675827482447343,
      "loss": 1.6834,
      "theoretical_loss": 3.9736505816277305,
      "tokens_seen": 465633280
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008675626880641925,
      "loss": 1.5651,
      "theoretical_loss": 3.973590043747387,
      "tokens_seen": 465698816
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000867542627883651,
      "loss": 1.7065,
      "theoretical_loss": 3.9735295167707165,
      "tokens_seen": 465764352
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008675225677031094,
      "loss": 1.7577,
      "theoretical_loss": 3.973469000694223,
      "tokens_seen": 465829888
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008675025075225677,
      "loss": 1.6708,
      "theoretical_loss": 3.9734084955144104,
      "tokens_seen": 465895424
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008674824473420261,
      "loss": 1.6733,
      "theoretical_loss": 3.9733480012277833,
      "tokens_seen": 465960960
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008674623871614846,
      "loss": 1.5407,
      "theoretical_loss": 3.9732875178308493,
      "tokens_seen": 466026496
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008674423269809428,
      "loss": 1.6824,
      "theoretical_loss": 3.973227045320117,
      "tokens_seen": 466092032
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008674222668004013,
      "loss": 1.4976,
      "theoretical_loss": 3.9731665836920973,
      "tokens_seen": 466157568
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008674022066198595,
      "loss": 1.6535,
      "theoretical_loss": 3.9731061329433013,
      "tokens_seen": 466223104
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000867382146439318,
      "loss": 1.8235,
      "theoretical_loss": 3.9730456930702434,
      "tokens_seen": 466288640
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008673620862587764,
      "loss": 1.7147,
      "theoretical_loss": 3.9729852640694383,
      "tokens_seen": 466354176
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008673420260782347,
      "loss": 1.736,
      "theoretical_loss": 3.9729248459374027,
      "tokens_seen": 466419712
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008673219658976931,
      "loss": 1.6526,
      "theoretical_loss": 3.9728644386706553,
      "tokens_seen": 466485248
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008673019057171515,
      "loss": 1.6138,
      "theoretical_loss": 3.9728040422657163,
      "tokens_seen": 466550784
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008672818455366098,
      "loss": 1.7119,
      "theoretical_loss": 3.972743656719107,
      "tokens_seen": 466616320
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008672617853560683,
      "loss": 1.5707,
      "theoretical_loss": 3.97268328202735,
      "tokens_seen": 466681856
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008672417251755265,
      "loss": 1.5306,
      "theoretical_loss": 3.972622918186971,
      "tokens_seen": 466747392
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000867221664994985,
      "loss": 1.6185,
      "theoretical_loss": 3.9725625651944965,
      "tokens_seen": 466812928
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008672016048144433,
      "loss": 1.6831,
      "theoretical_loss": 3.9725022230464537,
      "tokens_seen": 466878464
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008671815446339017,
      "loss": 1.7115,
      "theoretical_loss": 3.972441891739373,
      "tokens_seen": 466944000
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008671614844533601,
      "loss": 1.6376,
      "theoretical_loss": 3.9723815712697848,
      "tokens_seen": 467009536
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008671414242728185,
      "loss": 1.5643,
      "theoretical_loss": 3.9723212616342227,
      "tokens_seen": 467075072
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008671213640922768,
      "loss": 1.7625,
      "theoretical_loss": 3.9722609628292207,
      "tokens_seen": 467140608
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008671013039117353,
      "loss": 1.6438,
      "theoretical_loss": 3.9722006748513143,
      "tokens_seen": 467206144
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008670812437311936,
      "loss": 1.6106,
      "theoretical_loss": 3.9721403976970424,
      "tokens_seen": 467271680
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000867061183550652,
      "loss": 1.6831,
      "theoretical_loss": 3.972080131362943,
      "tokens_seen": 467337216
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008670411233701103,
      "loss": 1.6706,
      "theoretical_loss": 3.9720198758455574,
      "tokens_seen": 467402752
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008670210631895687,
      "loss": 1.6925,
      "theoretical_loss": 3.971959631141428,
      "tokens_seen": 467468288
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008670010030090271,
      "loss": 1.6552,
      "theoretical_loss": 3.971899397247099,
      "tokens_seen": 467533824
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008669809428284855,
      "loss": 1.6002,
      "theoretical_loss": 3.9718391741591144,
      "tokens_seen": 467599360
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008669608826479438,
      "loss": 1.7323,
      "theoretical_loss": 3.9717789618740227,
      "tokens_seen": 467664896
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008669408224674022,
      "loss": 1.7703,
      "theoretical_loss": 3.971718760388373,
      "tokens_seen": 467730432
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008669207622868606,
      "loss": 1.6947,
      "theoretical_loss": 3.971658569698714,
      "tokens_seen": 467795968
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000866900702106319,
      "loss": 1.5733,
      "theoretical_loss": 3.971598389801599,
      "tokens_seen": 467861504
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008668806419257773,
      "loss": 1.7538,
      "theoretical_loss": 3.9715382206935814,
      "tokens_seen": 467927040
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008668605817452358,
      "loss": 1.6245,
      "theoretical_loss": 3.971478062371215,
      "tokens_seen": 467992576
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000866840521564694,
      "loss": 1.5672,
      "theoretical_loss": 3.9714179148310587,
      "tokens_seen": 468058112
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008668204613841525,
      "loss": 1.59,
      "theoretical_loss": 3.971357778069668,
      "tokens_seen": 468123648
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008668004012036108,
      "loss": 1.6417,
      "theoretical_loss": 3.9712976520836043,
      "tokens_seen": 468189184
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008667803410230692,
      "loss": 1.6598,
      "theoretical_loss": 3.971237536869429,
      "tokens_seen": 468254720
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008667602808425276,
      "loss": 1.645,
      "theoretical_loss": 3.971177432423705,
      "tokens_seen": 468320256
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000866740220661986,
      "loss": 1.6404,
      "theoretical_loss": 3.9711173387429954,
      "tokens_seen": 468385792
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008667201604814443,
      "loss": 1.6616,
      "theoretical_loss": 3.971057255823868,
      "tokens_seen": 468451328
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008667001003009028,
      "loss": 1.5759,
      "theoretical_loss": 3.9709971836628903,
      "tokens_seen": 468516864
    },
    {
      "epoch": 0.14,
      "objective/train/advantage_avg": 0.49590790271759033,
      "objective/train/docs_used": 176758,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7070120573043823,
      "objective/train/original_loss": 1.7070120573043823,
      "objective/train/theoretical_loss": 3.9709371222566308,
      "objective/train/tokens_used": 489042400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24609065055847168,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508419275283813,
      "objective/train/weighted_lm_loss": 1.793755054473877,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9709371222566308,
      "tokens_seen": 468582400
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000866680040120361,
      "loss": 1.7491,
      "theoretical_loss": 3.9709371222566308,
      "tokens_seen": 468582400
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008666599799398195,
      "loss": 1.6753,
      "theoretical_loss": 3.9708770716016613,
      "tokens_seen": 468647936
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008666399197592778,
      "loss": 1.7408,
      "theoretical_loss": 3.9708170316945526,
      "tokens_seen": 468713472
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008666198595787362,
      "loss": 1.7188,
      "theoretical_loss": 3.9707570025318804,
      "tokens_seen": 468779008
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008665997993981946,
      "loss": 1.6429,
      "theoretical_loss": 3.9706969841102193,
      "tokens_seen": 468844544
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008665797392176529,
      "loss": 1.6667,
      "theoretical_loss": 3.970636976426147,
      "tokens_seen": 468910080
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008665596790371113,
      "loss": 1.6802,
      "theoretical_loss": 3.9705769794762418,
      "tokens_seen": 468975616
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008665396188565698,
      "loss": 1.719,
      "theoretical_loss": 3.9705169932570836,
      "tokens_seen": 469041152
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000866519558676028,
      "loss": 1.6971,
      "theoretical_loss": 3.9704570177652547,
      "tokens_seen": 469106688
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008664994984954865,
      "loss": 1.7641,
      "theoretical_loss": 3.9703970529973387,
      "tokens_seen": 469172224
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008664794383149449,
      "loss": 1.6095,
      "theoretical_loss": 3.97033709894992,
      "tokens_seen": 469237760
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008664593781344032,
      "loss": 1.678,
      "theoretical_loss": 3.970277155619586,
      "tokens_seen": 469303296
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008664393179538616,
      "loss": 1.6985,
      "theoretical_loss": 3.9702172230029236,
      "tokens_seen": 469368832
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008664192577733199,
      "loss": 1.686,
      "theoretical_loss": 3.970157301096523,
      "tokens_seen": 469434368
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008663991975927783,
      "loss": 1.5955,
      "theoretical_loss": 3.970097389896975,
      "tokens_seen": 469499904
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008663791374122368,
      "loss": 1.6344,
      "theoretical_loss": 3.9700374894008736,
      "tokens_seen": 469565440
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008663590772316951,
      "loss": 1.7155,
      "theoretical_loss": 3.969977599604812,
      "tokens_seen": 469630976
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008663390170511535,
      "loss": 1.7049,
      "theoretical_loss": 3.969917720505386,
      "tokens_seen": 469696512
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008663189568706119,
      "loss": 1.669,
      "theoretical_loss": 3.9698578520991936,
      "tokens_seen": 469762048
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008662988966900702,
      "loss": 1.7185,
      "theoretical_loss": 3.9697979943828337,
      "tokens_seen": 469827584
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008662788365095287,
      "loss": 1.6554,
      "theoretical_loss": 3.969738147352907,
      "tokens_seen": 469893120
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000866258776328987,
      "loss": 1.7524,
      "theoretical_loss": 3.9696783110060148,
      "tokens_seen": 469958656
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008662387161484454,
      "loss": 1.7266,
      "theoretical_loss": 3.9696184853387617,
      "tokens_seen": 470024192
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008662186559679037,
      "loss": 1.6859,
      "theoretical_loss": 3.9695586703477526,
      "tokens_seen": 470089728
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008661985957873621,
      "loss": 1.5995,
      "theoretical_loss": 3.9694988660295936,
      "tokens_seen": 470155264
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008661785356068205,
      "loss": 1.6525,
      "theoretical_loss": 3.9694390723808946,
      "tokens_seen": 470220800
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008661584754262789,
      "loss": 1.6127,
      "theoretical_loss": 3.9693792893982636,
      "tokens_seen": 470286336
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008661384152457372,
      "loss": 1.6209,
      "theoretical_loss": 3.969319517078313,
      "tokens_seen": 470351872
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008661183550651957,
      "loss": 1.7191,
      "theoretical_loss": 3.969259755417656,
      "tokens_seen": 470417408
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000866098294884654,
      "loss": 1.7615,
      "theoretical_loss": 3.969200004412907,
      "tokens_seen": 470482944
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008660782347041124,
      "loss": 1.6576,
      "theoretical_loss": 3.969140264060681,
      "tokens_seen": 470548480
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008660581745235707,
      "loss": 1.5596,
      "theoretical_loss": 3.9690805343575972,
      "tokens_seen": 470614016
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008660381143430291,
      "loss": 1.7647,
      "theoretical_loss": 3.969020815300274,
      "tokens_seen": 470679552
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008660180541624875,
      "loss": 1.6108,
      "theoretical_loss": 3.968961106885332,
      "tokens_seen": 470745088
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008659979939819459,
      "loss": 1.6862,
      "theoretical_loss": 3.9689014091093933,
      "tokens_seen": 470810624
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008659779338014042,
      "loss": 1.684,
      "theoretical_loss": 3.9688417219690817,
      "tokens_seen": 470876160
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008659578736208626,
      "loss": 1.6453,
      "theoretical_loss": 3.9687820454610234,
      "tokens_seen": 470941696
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000865937813440321,
      "loss": 1.6115,
      "theoretical_loss": 3.9687223795818443,
      "tokens_seen": 471007232
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008659177532597794,
      "loss": 1.6774,
      "theoretical_loss": 3.9686627243281727,
      "tokens_seen": 471072768
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008658976930792377,
      "loss": 1.7675,
      "theoretical_loss": 3.9686030796966394,
      "tokens_seen": 471138304
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008658776328986962,
      "loss": 1.7617,
      "theoretical_loss": 3.968543445683875,
      "tokens_seen": 471203840
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008658575727181544,
      "loss": 1.6725,
      "theoretical_loss": 3.9684838222865135,
      "tokens_seen": 471269376
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008658375125376129,
      "loss": 1.7021,
      "theoretical_loss": 3.9684242095011886,
      "tokens_seen": 471334912
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008658174523570712,
      "loss": 1.5613,
      "theoretical_loss": 3.9683646073245367,
      "tokens_seen": 471400448
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008657973921765296,
      "loss": 1.6433,
      "theoretical_loss": 3.968305015753195,
      "tokens_seen": 471465984
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000865777331995988,
      "loss": 1.7898,
      "theoretical_loss": 3.968245434783803,
      "tokens_seen": 471531520
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008657572718154464,
      "loss": 1.6126,
      "theoretical_loss": 3.9681858644130017,
      "tokens_seen": 471597056
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008657372116349047,
      "loss": 1.7728,
      "theoretical_loss": 3.9681263046374333,
      "tokens_seen": 471662592
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008657171514543632,
      "loss": 1.693,
      "theoretical_loss": 3.968066755453741,
      "tokens_seen": 471728128
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008656970912738214,
      "loss": 1.6555,
      "theoretical_loss": 3.9680072168585703,
      "tokens_seen": 471793664
    },
    {
      "epoch": 0.14,
      "objective/train/advantage_avg": 0.4907188415527344,
      "objective/train/docs_used": 177948,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6511681079864502,
      "objective/train/original_loss": 1.6511683464050293,
      "objective/train/theoretical_loss": 3.967947688848568,
      "objective/train/tokens_used": 492319200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2415388971567154,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502997636795044,
      "objective/train/weighted_lm_loss": 1.7343614101409912,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.967947688848568,
      "tokens_seen": 471859200
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008656770310932799,
      "loss": 1.6961,
      "theoretical_loss": 3.967947688848568,
      "tokens_seen": 471859200
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008656569709127382,
      "loss": 1.5168,
      "theoretical_loss": 3.967888171420382,
      "tokens_seen": 471924736
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008656369107321966,
      "loss": 1.6529,
      "theoretical_loss": 3.9678286645706637,
      "tokens_seen": 471990272
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000865616850551655,
      "loss": 1.5362,
      "theoretical_loss": 3.967769168296063,
      "tokens_seen": 472055808
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008655967903711133,
      "loss": 1.6113,
      "theoretical_loss": 3.9677096825932328,
      "tokens_seen": 472121344
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008655767301905717,
      "loss": 1.5861,
      "theoretical_loss": 3.967650207458828,
      "tokens_seen": 472186880
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008655566700100302,
      "loss": 1.7183,
      "theoretical_loss": 3.9675907428895045,
      "tokens_seen": 472252416
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008655366098294884,
      "loss": 1.678,
      "theoretical_loss": 3.9675312888819194,
      "tokens_seen": 472317952
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008655165496489469,
      "loss": 1.605,
      "theoretical_loss": 3.9674718454327325,
      "tokens_seen": 472383488
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008654964894684053,
      "loss": 1.7314,
      "theoretical_loss": 3.967412412538604,
      "tokens_seen": 472449024
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008654764292878636,
      "loss": 1.688,
      "theoretical_loss": 3.9673529901961957,
      "tokens_seen": 472514560
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000865456369107322,
      "loss": 1.6936,
      "theoretical_loss": 3.9672935784021712,
      "tokens_seen": 472580096
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008654363089267803,
      "loss": 1.6403,
      "theoretical_loss": 3.9672341771531956,
      "tokens_seen": 472645632
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008654162487462387,
      "loss": 1.7436,
      "theoretical_loss": 3.967174786445935,
      "tokens_seen": 472711168
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008653961885656972,
      "loss": 1.6937,
      "theoretical_loss": 3.967115406277059,
      "tokens_seen": 472776704
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008653761283851554,
      "loss": 1.5322,
      "theoretical_loss": 3.9670560366432355,
      "tokens_seen": 472842240
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008653560682046139,
      "loss": 1.8065,
      "theoretical_loss": 3.9669966775411365,
      "tokens_seen": 472907776
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008653360080240722,
      "loss": 1.6547,
      "theoretical_loss": 3.9669373289674343,
      "tokens_seen": 472973312
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008653159478435306,
      "loss": 1.6371,
      "theoretical_loss": 3.966877990918804,
      "tokens_seen": 473038848
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000865295887662989,
      "loss": 1.5539,
      "theoretical_loss": 3.96681866339192,
      "tokens_seen": 473104384
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008652758274824473,
      "loss": 1.747,
      "theoretical_loss": 3.96675934638346,
      "tokens_seen": 473169920
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008652557673019057,
      "loss": 1.6029,
      "theoretical_loss": 3.9667000398901022,
      "tokens_seen": 473235456
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008652357071213641,
      "loss": 1.7309,
      "theoretical_loss": 3.966640743908528,
      "tokens_seen": 473300992
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008652156469408224,
      "loss": 1.7987,
      "theoretical_loss": 3.966581458435418,
      "tokens_seen": 473366528
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008651955867602809,
      "loss": 1.6473,
      "theoretical_loss": 3.9665221834674558,
      "tokens_seen": 473432064
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008651755265797392,
      "loss": 1.6844,
      "theoretical_loss": 3.9664629190013265,
      "tokens_seen": 473497600
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008651554663991976,
      "loss": 1.7004,
      "theoretical_loss": 3.9664036650337158,
      "tokens_seen": 473563136
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000865135406218656,
      "loss": 1.5744,
      "theoretical_loss": 3.9663444215613115,
      "tokens_seen": 473628672
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008651153460381144,
      "loss": 1.6554,
      "theoretical_loss": 3.9662851885808026,
      "tokens_seen": 473694208
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008650952858575727,
      "loss": 1.6671,
      "theoretical_loss": 3.9662259660888806,
      "tokens_seen": 473759744
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008650752256770311,
      "loss": 1.7138,
      "theoretical_loss": 3.9661667540822365,
      "tokens_seen": 473825280
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008650551654964894,
      "loss": 1.606,
      "theoretical_loss": 3.9661075525575655,
      "tokens_seen": 473890816
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008650351053159479,
      "loss": 1.6872,
      "theoretical_loss": 3.966048361511562,
      "tokens_seen": 473956352
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008650150451354062,
      "loss": 1.5797,
      "theoretical_loss": 3.9659891809409222,
      "tokens_seen": 474021888
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008649949849548646,
      "loss": 1.704,
      "theoretical_loss": 3.9659300108423454,
      "tokens_seen": 474087424
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008649749247743229,
      "loss": 1.6559,
      "theoretical_loss": 3.965870851212531,
      "tokens_seen": 474152960
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008649548645937814,
      "loss": 1.5749,
      "theoretical_loss": 3.96581170204818,
      "tokens_seen": 474218496
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008649348044132397,
      "loss": 1.7093,
      "theoretical_loss": 3.965752563345995,
      "tokens_seen": 474284032
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008649147442326981,
      "loss": 1.7509,
      "theoretical_loss": 3.965693435102681,
      "tokens_seen": 474349568
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008648946840521564,
      "loss": 1.6831,
      "theoretical_loss": 3.9656343173149424,
      "tokens_seen": 474415104
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008648746238716148,
      "loss": 1.7082,
      "theoretical_loss": 3.9655752099794874,
      "tokens_seen": 474480640
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008648545636910733,
      "loss": 1.7715,
      "theoretical_loss": 3.9655161130930248,
      "tokens_seen": 474546176
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008648345035105316,
      "loss": 1.6436,
      "theoretical_loss": 3.965457026652264,
      "tokens_seen": 474611712
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00086481444332999,
      "loss": 1.8402,
      "theoretical_loss": 3.965397950653917,
      "tokens_seen": 474677248
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008647943831494484,
      "loss": 1.7173,
      "theoretical_loss": 3.9653388850946976,
      "tokens_seen": 474742784
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008647743229689068,
      "loss": 1.631,
      "theoretical_loss": 3.9652798299713194,
      "tokens_seen": 474808320
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008647542627883651,
      "loss": 1.5619,
      "theoretical_loss": 3.965220785280499,
      "tokens_seen": 474873856
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008647342026078236,
      "loss": 1.7494,
      "theoretical_loss": 3.9651617510189543,
      "tokens_seen": 474939392
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008647141424272818,
      "loss": 1.6467,
      "theoretical_loss": 3.965102727183404,
      "tokens_seen": 475004928
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008646940822467403,
      "loss": 1.7128,
      "theoretical_loss": 3.9650437137705685,
      "tokens_seen": 475070464
    },
    {
      "epoch": 0.14,
      "objective/train/advantage_avg": 0.4953663647174835,
      "objective/train/docs_used": 179087,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4755969047546387,
      "objective/train/original_loss": 1.4755969047546387,
      "objective/train/theoretical_loss": 3.9649847107771707,
      "objective/train/tokens_used": 495596000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24567000567913055,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507856607437134,
      "objective/train/weighted_lm_loss": 1.550560712814331,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9649847107771707,
      "tokens_seen": 475136000
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008646740220661986,
      "loss": 1.556,
      "theoretical_loss": 3.9649847107771707,
      "tokens_seen": 475136000
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000864653961885657,
      "loss": 1.625,
      "theoretical_loss": 3.9649257181999333,
      "tokens_seen": 475201536
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008646339017051154,
      "loss": 1.643,
      "theoretical_loss": 3.9648667360355816,
      "tokens_seen": 475267072
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008646138415245737,
      "loss": 1.6444,
      "theoretical_loss": 3.9648077642808426,
      "tokens_seen": 475332608
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008645937813440321,
      "loss": 1.6659,
      "theoretical_loss": 3.9647488029324434,
      "tokens_seen": 475398144
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008645737211634906,
      "loss": 1.8248,
      "theoretical_loss": 3.9646898519871145,
      "tokens_seen": 475463680
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008645536609829488,
      "loss": 1.6969,
      "theoretical_loss": 3.9646309114415863,
      "tokens_seen": 475529216
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008645336008024073,
      "loss": 1.7494,
      "theoretical_loss": 3.9645719812925906,
      "tokens_seen": 475594752
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008645135406218657,
      "loss": 1.6244,
      "theoretical_loss": 3.964513061536863,
      "tokens_seen": 475660288
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000864493480441324,
      "loss": 1.6211,
      "theoretical_loss": 3.964454152171137,
      "tokens_seen": 475725824
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008644734202607824,
      "loss": 1.5774,
      "theoretical_loss": 3.9643952531921505,
      "tokens_seen": 475791360
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008644533600802407,
      "loss": 1.7316,
      "theoretical_loss": 3.964336364596642,
      "tokens_seen": 475856896
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008644332998996991,
      "loss": 1.5758,
      "theoretical_loss": 3.964277486381351,
      "tokens_seen": 475922432
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008644132397191576,
      "loss": 1.5914,
      "theoretical_loss": 3.9642186185430184,
      "tokens_seen": 475987968
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008643931795386158,
      "loss": 1.4938,
      "theoretical_loss": 3.9641597610783874,
      "tokens_seen": 476053504
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008643731193580743,
      "loss": 1.6912,
      "theoretical_loss": 3.9641009139842023,
      "tokens_seen": 476119040
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008643530591775326,
      "loss": 1.647,
      "theoretical_loss": 3.9640420772572087,
      "tokens_seen": 476184576
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000864332998996991,
      "loss": 1.7011,
      "theoretical_loss": 3.9639832508941533,
      "tokens_seen": 476250112
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008643129388164494,
      "loss": 1.6084,
      "theoretical_loss": 3.9639244348917853,
      "tokens_seen": 476315648
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008642928786359077,
      "loss": 1.5955,
      "theoretical_loss": 3.963865629246855,
      "tokens_seen": 476381184
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008642728184553661,
      "loss": 1.5804,
      "theoretical_loss": 3.963806833956113,
      "tokens_seen": 476446720
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008642527582748245,
      "loss": 1.6083,
      "theoretical_loss": 3.9637480490163135,
      "tokens_seen": 476512256
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008642326980942828,
      "loss": 1.6791,
      "theoretical_loss": 3.96368927442421,
      "tokens_seen": 476577792
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008642126379137413,
      "loss": 1.8064,
      "theoretical_loss": 3.9636305101765585,
      "tokens_seen": 476643328
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008641925777331996,
      "loss": 1.5492,
      "theoretical_loss": 3.9635717562701176,
      "tokens_seen": 476708864
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000864172517552658,
      "loss": 1.7523,
      "theoretical_loss": 3.963513012701645,
      "tokens_seen": 476774400
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008641524573721164,
      "loss": 1.581,
      "theoretical_loss": 3.9634542794679013,
      "tokens_seen": 476839936
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008641323971915748,
      "loss": 1.667,
      "theoretical_loss": 3.963395556565649,
      "tokens_seen": 476905472
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008641123370110331,
      "loss": 1.5052,
      "theoretical_loss": 3.96333684399165,
      "tokens_seen": 476971008
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008640922768304915,
      "loss": 1.6253,
      "theoretical_loss": 3.9632781417426703,
      "tokens_seen": 477036544
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008640722166499498,
      "loss": 1.7274,
      "theoretical_loss": 3.963219449815475,
      "tokens_seen": 477102080
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008640521564694083,
      "loss": 1.4946,
      "theoretical_loss": 3.9631607682068335,
      "tokens_seen": 477167616
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008640320962888666,
      "loss": 1.6131,
      "theoretical_loss": 3.963102096913513,
      "tokens_seen": 477233152
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000864012036108325,
      "loss": 1.6323,
      "theoretical_loss": 3.963043435932285,
      "tokens_seen": 477298688
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008639919759277833,
      "loss": 1.5629,
      "theoretical_loss": 3.9629847852599207,
      "tokens_seen": 477364224
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008639719157472418,
      "loss": 1.6746,
      "theoretical_loss": 3.9629261448931943,
      "tokens_seen": 477429760
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008639518555667001,
      "loss": 1.7605,
      "theoretical_loss": 3.962867514828881,
      "tokens_seen": 477495296
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008639317953861585,
      "loss": 1.7014,
      "theoretical_loss": 3.9628088950637563,
      "tokens_seen": 477560832
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008639117352056168,
      "loss": 1.7168,
      "theoretical_loss": 3.9627502855945984,
      "tokens_seen": 477626368
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008638916750250752,
      "loss": 1.674,
      "theoretical_loss": 3.9626916864181867,
      "tokens_seen": 477691904
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008638716148445336,
      "loss": 1.6337,
      "theoretical_loss": 3.962633097531302,
      "tokens_seen": 477757440
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000863851554663992,
      "loss": 1.5767,
      "theoretical_loss": 3.962574518930726,
      "tokens_seen": 477822976
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008638314944834503,
      "loss": 1.6741,
      "theoretical_loss": 3.962515950613242,
      "tokens_seen": 477888512
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008638114343029088,
      "loss": 1.7077,
      "theoretical_loss": 3.9624573925756366,
      "tokens_seen": 477954048
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008637913741223671,
      "loss": 1.7455,
      "theoretical_loss": 3.962398844814695,
      "tokens_seen": 478019584
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008637713139418255,
      "loss": 1.6796,
      "theoretical_loss": 3.9623403073272048,
      "tokens_seen": 478085120
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008637512537612839,
      "loss": 1.6243,
      "theoretical_loss": 3.962281780109957,
      "tokens_seen": 478150656
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008637311935807422,
      "loss": 1.6601,
      "theoretical_loss": 3.962223263159741,
      "tokens_seen": 478216192
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008637111334002006,
      "loss": 1.5535,
      "theoretical_loss": 3.9621647564733493,
      "tokens_seen": 478281728
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000863691073219659,
      "loss": 1.6916,
      "theoretical_loss": 3.9621062600475763,
      "tokens_seen": 478347264
    },
    {
      "epoch": 0.14,
      "objective/train/advantage_avg": 0.49374765157699585,
      "objective/train/docs_used": 180329,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6273490190505981,
      "objective/train/original_loss": 1.627348780632019,
      "objective/train/theoretical_loss": 3.9620477738792164,
      "objective/train/tokens_used": 498872800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24403226375579834,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506154298782349,
      "objective/train/weighted_lm_loss": 1.7097764015197754,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.9620477738792164,
      "tokens_seen": 478412800
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008636710130391173,
      "loss": 1.7097,
      "theoretical_loss": 3.9620477738792164,
      "tokens_seen": 478412800
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0008636509528585758,
      "loss": 1.7818,
      "theoretical_loss": 3.9619892979650664,
      "tokens_seen": 478478336
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000863630892678034,
      "loss": 1.5919,
      "theoretical_loss": 3.9619308323019244,
      "tokens_seen": 478543872
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008636108324974925,
      "loss": 1.6456,
      "theoretical_loss": 3.96187237688659,
      "tokens_seen": 478609408
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008635907723169509,
      "loss": 1.6536,
      "theoretical_loss": 3.9618139317158647,
      "tokens_seen": 478674944
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008635707121364092,
      "loss": 1.7279,
      "theoretical_loss": 3.96175549678655,
      "tokens_seen": 478740480
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008635506519558676,
      "loss": 1.6284,
      "theoretical_loss": 3.9616970720954487,
      "tokens_seen": 478806016
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008635305917753261,
      "loss": 1.6787,
      "theoretical_loss": 3.9616386576393676,
      "tokens_seen": 478871552
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008635105315947843,
      "loss": 1.5603,
      "theoretical_loss": 3.961580253415114,
      "tokens_seen": 478937088
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008634904714142428,
      "loss": 1.5645,
      "theoretical_loss": 3.9615218594194936,
      "tokens_seen": 479002624
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000863470411233701,
      "loss": 1.7041,
      "theoretical_loss": 3.961463475649318,
      "tokens_seen": 479068160
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008634503510531595,
      "loss": 1.6872,
      "theoretical_loss": 3.9614051021013967,
      "tokens_seen": 479133696
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008634302908726179,
      "loss": 1.6363,
      "theoretical_loss": 3.9613467387725434,
      "tokens_seen": 479199232
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008634102306920762,
      "loss": 1.491,
      "theoretical_loss": 3.961288385659571,
      "tokens_seen": 479264768
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008633901705115346,
      "loss": 1.7415,
      "theoretical_loss": 3.961230042759295,
      "tokens_seen": 479330304
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000863370110330993,
      "loss": 1.6011,
      "theoretical_loss": 3.961171710068532,
      "tokens_seen": 479395840
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008633500501504514,
      "loss": 1.5091,
      "theoretical_loss": 3.9611133875841,
      "tokens_seen": 479461376
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008633299899699098,
      "loss": 1.669,
      "theoretical_loss": 3.961055075302819,
      "tokens_seen": 479526912
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008633099297893681,
      "loss": 1.644,
      "theoretical_loss": 3.9609967732215097,
      "tokens_seen": 479592448
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008632898696088265,
      "loss": 1.7018,
      "theoretical_loss": 3.960938481336994,
      "tokens_seen": 479657984
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008632698094282849,
      "loss": 1.6864,
      "theoretical_loss": 3.960880199646096,
      "tokens_seen": 479723520
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008632497492477432,
      "loss": 1.7271,
      "theoretical_loss": 3.9608219281456414,
      "tokens_seen": 479789056
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008632296890672017,
      "loss": 1.6468,
      "theoretical_loss": 3.960763666832456,
      "tokens_seen": 479854592
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00086320962888666,
      "loss": 1.5897,
      "theoretical_loss": 3.9607054157033685,
      "tokens_seen": 479920128
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008631895687061184,
      "loss": 1.8161,
      "theoretical_loss": 3.9606471747552083,
      "tokens_seen": 479985664
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008631695085255768,
      "loss": 1.7091,
      "theoretical_loss": 3.960588943984806,
      "tokens_seen": 480051200
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008631494483450352,
      "loss": 1.5867,
      "theoretical_loss": 3.960530723388994,
      "tokens_seen": 480116736
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008631293881644935,
      "loss": 1.6212,
      "theoretical_loss": 3.960472512964606,
      "tokens_seen": 480182272
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008631093279839519,
      "loss": 1.578,
      "theoretical_loss": 3.9604143127084774,
      "tokens_seen": 480247808
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008630892678034102,
      "loss": 1.6746,
      "theoretical_loss": 3.960356122617445,
      "tokens_seen": 480313344
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008630692076228687,
      "loss": 1.6223,
      "theoretical_loss": 3.9602979426883462,
      "tokens_seen": 480378880
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000863049147442327,
      "loss": 1.7078,
      "theoretical_loss": 3.9602397729180208,
      "tokens_seen": 480444416
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008630290872617854,
      "loss": 1.6483,
      "theoretical_loss": 3.960181613303309,
      "tokens_seen": 480509952
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008630090270812437,
      "loss": 1.7556,
      "theoretical_loss": 3.960123463841054,
      "tokens_seen": 480575488
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008629889669007022,
      "loss": 1.7327,
      "theoretical_loss": 3.9600653245280983,
      "tokens_seen": 480641024
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008629689067201605,
      "loss": 1.7339,
      "theoretical_loss": 3.960007195361288,
      "tokens_seen": 480706560
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008629488465396189,
      "loss": 1.6611,
      "theoretical_loss": 3.9599490763374687,
      "tokens_seen": 480772096
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008629287863590772,
      "loss": 1.7667,
      "theoretical_loss": 3.9598909674534895,
      "tokens_seen": 480837632
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008629087261785356,
      "loss": 1.7056,
      "theoretical_loss": 3.9598328687061985,
      "tokens_seen": 480903168
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000862888665997994,
      "loss": 1.6263,
      "theoretical_loss": 3.959774780092447,
      "tokens_seen": 480968704
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008628686058174524,
      "loss": 1.628,
      "theoretical_loss": 3.9597167016090866,
      "tokens_seen": 481034240
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008628485456369107,
      "loss": 1.5508,
      "theoretical_loss": 3.959658633252971,
      "tokens_seen": 481099776
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008628284854563692,
      "loss": 1.6159,
      "theoretical_loss": 3.9596005750209553,
      "tokens_seen": 481165312
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008628084252758275,
      "loss": 1.6213,
      "theoretical_loss": 3.9595425269098956,
      "tokens_seen": 481230848
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008627883650952859,
      "loss": 1.7252,
      "theoretical_loss": 3.95948448891665,
      "tokens_seen": 481296384
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008627683049147443,
      "loss": 1.6368,
      "theoretical_loss": 3.959426461038077,
      "tokens_seen": 481361920
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008627482447342026,
      "loss": 1.689,
      "theoretical_loss": 3.9593684432710377,
      "tokens_seen": 481427456
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000862728184553661,
      "loss": 1.6588,
      "theoretical_loss": 3.9593104356123936,
      "tokens_seen": 481492992
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008627081243731195,
      "loss": 1.6145,
      "theoretical_loss": 3.959252438059009,
      "tokens_seen": 481558528
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008626880641925777,
      "loss": 1.609,
      "theoretical_loss": 3.9591944506077468,
      "tokens_seen": 481624064
    },
    {
      "epoch": 0.15,
      "objective/train/advantage_avg": 0.492854505777359,
      "objective/train/docs_used": 181505,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5590598583221436,
      "objective/train/original_loss": 1.559059977531433,
      "objective/train/theoretical_loss": 3.959136473255475,
      "objective/train/tokens_used": 502149600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2436048984527588,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505238771438599,
      "objective/train/weighted_lm_loss": 1.6377798318862915,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.8607079386711121,
      "theoretical_loss": 3.959136473255475,
      "tokens_seen": 481689600
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008626680040120362,
      "loss": 1.6004,
      "theoretical_loss": 3.959136473255475,
      "tokens_seen": 481689600
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008626479438314944,
      "loss": 1.7456,
      "theoretical_loss": 3.95907850599906,
      "tokens_seen": 481755136
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008626278836509529,
      "loss": 1.6157,
      "theoretical_loss": 3.9590205488353707,
      "tokens_seen": 481820672
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008626078234704113,
      "loss": 1.673,
      "theoretical_loss": 3.9589626017612782,
      "tokens_seen": 481886208
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008625877632898696,
      "loss": 1.7322,
      "theoretical_loss": 3.9589046647736534,
      "tokens_seen": 481951744
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000862567703109328,
      "loss": 1.6756,
      "theoretical_loss": 3.9588467378693704,
      "tokens_seen": 482017280
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008625476429287865,
      "loss": 1.7392,
      "theoretical_loss": 3.958788821045302,
      "tokens_seen": 482082816
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008625275827482447,
      "loss": 1.6884,
      "theoretical_loss": 3.958730914298326,
      "tokens_seen": 482148352
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008625075225677032,
      "loss": 1.7627,
      "theoretical_loss": 3.958673017625319,
      "tokens_seen": 482213888
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008624874623871614,
      "loss": 1.6483,
      "theoretical_loss": 3.958615131023159,
      "tokens_seen": 482279424
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008624674022066199,
      "loss": 1.5586,
      "theoretical_loss": 3.958557254488727,
      "tokens_seen": 482344960
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008624473420260783,
      "loss": 1.6014,
      "theoretical_loss": 3.958499388018904,
      "tokens_seen": 482410496
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008624272818455366,
      "loss": 1.6735,
      "theoretical_loss": 3.958441531610573,
      "tokens_seen": 482476032
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000862407221664995,
      "loss": 1.6801,
      "theoretical_loss": 3.958383685260618,
      "tokens_seen": 482541568
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008623871614844534,
      "loss": 1.6305,
      "theoretical_loss": 3.958325848965925,
      "tokens_seen": 482607104
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008623671013039117,
      "loss": 1.683,
      "theoretical_loss": 3.958268022723381,
      "tokens_seen": 482672640
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008623470411233702,
      "loss": 1.6403,
      "theoretical_loss": 3.9582102065298734,
      "tokens_seen": 482738176
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008623269809428284,
      "loss": 1.5577,
      "theoretical_loss": 3.9581524003822937,
      "tokens_seen": 482803712
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008623069207622869,
      "loss": 1.64,
      "theoretical_loss": 3.958094604277532,
      "tokens_seen": 482869248
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008622868605817452,
      "loss": 1.6997,
      "theoretical_loss": 3.9580368182124808,
      "tokens_seen": 482934784
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008622668004012036,
      "loss": 1.7209,
      "theoretical_loss": 3.957979042184035,
      "tokens_seen": 483000320
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000862246740220662,
      "loss": 1.6709,
      "theoretical_loss": 3.9579212761890883,
      "tokens_seen": 483065856
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008622266800401204,
      "loss": 1.6907,
      "theoretical_loss": 3.9578635202245387,
      "tokens_seen": 483131392
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008622066198595787,
      "loss": 1.6686,
      "theoretical_loss": 3.957805774287284,
      "tokens_seen": 483196928
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008621865596790372,
      "loss": 1.4088,
      "theoretical_loss": 3.957748038374224,
      "tokens_seen": 483262464
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008621664994984954,
      "loss": 1.6271,
      "theoretical_loss": 3.957690312482258,
      "tokens_seen": 483328000
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008621464393179539,
      "loss": 1.6111,
      "theoretical_loss": 3.95763259660829,
      "tokens_seen": 483393536
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008621263791374122,
      "loss": 1.6555,
      "theoretical_loss": 3.957574890749223,
      "tokens_seen": 483459072
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008621063189568706,
      "loss": 1.5846,
      "theoretical_loss": 3.9575171949019623,
      "tokens_seen": 483524608
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000862086258776329,
      "loss": 1.5948,
      "theoretical_loss": 3.957459509063413,
      "tokens_seen": 483590144
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008620661985957874,
      "loss": 1.6619,
      "theoretical_loss": 3.957401833230484,
      "tokens_seen": 483655680
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008620461384152457,
      "loss": 1.6941,
      "theoretical_loss": 3.9573441674000844,
      "tokens_seen": 483721216
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008620260782347041,
      "loss": 1.6845,
      "theoretical_loss": 3.9572865115691243,
      "tokens_seen": 483786752
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008620060180541625,
      "loss": 1.5978,
      "theoretical_loss": 3.9572288657345154,
      "tokens_seen": 483852288
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008619859578736209,
      "loss": 1.5465,
      "theoretical_loss": 3.957171229893171,
      "tokens_seen": 483917824
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008619658976930792,
      "loss": 1.6417,
      "theoretical_loss": 3.9571136040420063,
      "tokens_seen": 483983360
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008619458375125376,
      "loss": 1.673,
      "theoretical_loss": 3.9570559881779364,
      "tokens_seen": 484048896
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008619257773319959,
      "loss": 1.7413,
      "theoretical_loss": 3.956998382297879,
      "tokens_seen": 484114432
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008619057171514544,
      "loss": 1.5772,
      "theoretical_loss": 3.956940786398753,
      "tokens_seen": 484179968
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008618856569709127,
      "loss": 1.6466,
      "theoretical_loss": 3.9568832004774785,
      "tokens_seen": 484245504
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008618655967903711,
      "loss": 1.5917,
      "theoretical_loss": 3.9568256245309765,
      "tokens_seen": 484311040
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008618455366098296,
      "loss": 1.6796,
      "theoretical_loss": 3.9567680585561695,
      "tokens_seen": 484376576
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008618254764292879,
      "loss": 1.7908,
      "theoretical_loss": 3.9567105025499827,
      "tokens_seen": 484442112
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008618054162487463,
      "loss": 1.5868,
      "theoretical_loss": 3.956652956509341,
      "tokens_seen": 484507648
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008617853560682047,
      "loss": 1.6694,
      "theoretical_loss": 3.9565954204311717,
      "tokens_seen": 484573184
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000861765295887663,
      "loss": 1.678,
      "theoretical_loss": 3.9565378943124023,
      "tokens_seen": 484638720
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008617452357071214,
      "loss": 1.6422,
      "theoretical_loss": 3.9564803781499633,
      "tokens_seen": 484704256
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008617251755265799,
      "loss": 1.6743,
      "theoretical_loss": 3.956422871940785,
      "tokens_seen": 484769792
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008617051153460381,
      "loss": 1.6426,
      "theoretical_loss": 3.9563653756818002,
      "tokens_seen": 484835328
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008616850551654966,
      "loss": 1.7416,
      "theoretical_loss": 3.956307889369942,
      "tokens_seen": 484900864
    },
    {
      "epoch": 0.15,
      "objective/train/advantage_avg": 0.49089962244033813,
      "objective/train/docs_used": 182723,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3967673778533936,
      "objective/train/original_loss": 1.3967671394348145,
      "objective/train/theoretical_loss": 3.956250413002146,
      "objective/train/tokens_used": 505426400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24230317771434784,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503216981887817,
      "objective/train/weighted_lm_loss": 1.4670437574386597,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.956250413002146,
      "tokens_seen": 484966400
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008616649949849548,
      "loss": 1.5712,
      "theoretical_loss": 3.956250413002146,
      "tokens_seen": 484966400
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008616449348044133,
      "loss": 1.6428,
      "theoretical_loss": 3.9561929465753485,
      "tokens_seen": 485031936
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008616248746238717,
      "loss": 1.7151,
      "theoretical_loss": 3.9561354900864876,
      "tokens_seen": 485097472
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00086160481444333,
      "loss": 1.6437,
      "theoretical_loss": 3.956078043532502,
      "tokens_seen": 485163008
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008615847542627884,
      "loss": 1.6562,
      "theoretical_loss": 3.9560206069103314,
      "tokens_seen": 485228544
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008615646940822469,
      "loss": 1.7284,
      "theoretical_loss": 3.9559631802169193,
      "tokens_seen": 485294080
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008615446339017051,
      "loss": 1.6371,
      "theoretical_loss": 3.955905763449208,
      "tokens_seen": 485359616
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008615245737211636,
      "loss": 1.7452,
      "theoretical_loss": 3.9558483566041422,
      "tokens_seen": 485425152
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008615045135406218,
      "loss": 1.6399,
      "theoretical_loss": 3.9557909596786676,
      "tokens_seen": 485490688
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008614844533600803,
      "loss": 1.6872,
      "theoretical_loss": 3.9557335726697316,
      "tokens_seen": 485556224
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008614643931795387,
      "loss": 1.6266,
      "theoretical_loss": 3.9556761955742834,
      "tokens_seen": 485621760
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000861444332998997,
      "loss": 1.6871,
      "theoretical_loss": 3.955618828389272,
      "tokens_seen": 485687296
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008614242728184554,
      "loss": 1.61,
      "theoretical_loss": 3.9555614711116487,
      "tokens_seen": 485752832
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008614042126379138,
      "loss": 1.7205,
      "theoretical_loss": 3.9555041237383675,
      "tokens_seen": 485818368
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008613841524573721,
      "loss": 1.6493,
      "theoretical_loss": 3.9554467862663807,
      "tokens_seen": 485883904
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008613640922768306,
      "loss": 1.5156,
      "theoretical_loss": 3.955389458692645,
      "tokens_seen": 485949440
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008613440320962888,
      "loss": 1.6917,
      "theoretical_loss": 3.9553321410141162,
      "tokens_seen": 486014976
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008613239719157473,
      "loss": 1.6528,
      "theoretical_loss": 3.9552748332277528,
      "tokens_seen": 486080512
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008613039117352056,
      "loss": 1.6826,
      "theoretical_loss": 3.9552175353305135,
      "tokens_seen": 486146048
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000861283851554664,
      "loss": 1.6517,
      "theoretical_loss": 3.9551602473193603,
      "tokens_seen": 486211584
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008612637913741224,
      "loss": 1.5281,
      "theoretical_loss": 3.9551029691912545,
      "tokens_seen": 486277120
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008612437311935808,
      "loss": 1.8018,
      "theoretical_loss": 3.955045700943159,
      "tokens_seen": 486342656
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008612236710130391,
      "loss": 1.6226,
      "theoretical_loss": 3.95498844257204,
      "tokens_seen": 486408192
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008612036108324976,
      "loss": 1.6285,
      "theoretical_loss": 3.9549311940748626,
      "tokens_seen": 486473728
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008611835506519558,
      "loss": 1.5453,
      "theoretical_loss": 3.954873955448594,
      "tokens_seen": 486539264
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008611634904714143,
      "loss": 1.6237,
      "theoretical_loss": 3.9548167266902037,
      "tokens_seen": 486604800
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008611434302908726,
      "loss": 1.6427,
      "theoretical_loss": 3.9547595077966617,
      "tokens_seen": 486670336
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000861123370110331,
      "loss": 1.643,
      "theoretical_loss": 3.954702298764939,
      "tokens_seen": 486735872
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008611033099297894,
      "loss": 1.7013,
      "theoretical_loss": 3.9546450995920086,
      "tokens_seen": 486801408
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008610832497492478,
      "loss": 1.6646,
      "theoretical_loss": 3.9545879102748454,
      "tokens_seen": 486866944
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008610631895687061,
      "loss": 1.6002,
      "theoretical_loss": 3.954530730810424,
      "tokens_seen": 486932480
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008610431293881645,
      "loss": 1.5922,
      "theoretical_loss": 3.954473561195721,
      "tokens_seen": 486998016
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008610230692076229,
      "loss": 1.6364,
      "theoretical_loss": 3.954416401427715,
      "tokens_seen": 487063552
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008610030090270813,
      "loss": 1.6667,
      "theoretical_loss": 3.9543592515033854,
      "tokens_seen": 487129088
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008609829488465396,
      "loss": 1.5311,
      "theoretical_loss": 3.9543021114197137,
      "tokens_seen": 487194624
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000860962888665998,
      "loss": 1.5933,
      "theoretical_loss": 3.9542449811736806,
      "tokens_seen": 487260160
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008609428284854563,
      "loss": 1.682,
      "theoretical_loss": 3.9541878607622705,
      "tokens_seen": 487325696
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008609227683049148,
      "loss": 1.5848,
      "theoretical_loss": 3.9541307501824683,
      "tokens_seen": 487391232
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008609027081243731,
      "loss": 1.6445,
      "theoretical_loss": 3.95407364943126,
      "tokens_seen": 487456768
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008608826479438315,
      "loss": 1.5271,
      "theoretical_loss": 3.954016558505633,
      "tokens_seen": 487522304
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008608625877632899,
      "loss": 1.5393,
      "theoretical_loss": 3.9539594774025755,
      "tokens_seen": 487587840
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008608425275827483,
      "loss": 1.5841,
      "theoretical_loss": 3.9539024061190786,
      "tokens_seen": 487653376
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008608224674022066,
      "loss": 1.57,
      "theoretical_loss": 3.953845344652133,
      "tokens_seen": 487718912
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008608024072216651,
      "loss": 1.6846,
      "theoretical_loss": 3.9537882929987322,
      "tokens_seen": 487784448
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008607823470411233,
      "loss": 1.7473,
      "theoretical_loss": 3.9537312511558698,
      "tokens_seen": 487849984
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008607622868605818,
      "loss": 1.5851,
      "theoretical_loss": 3.9536742191205407,
      "tokens_seen": 487915520
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008607422266800401,
      "loss": 1.6263,
      "theoretical_loss": 3.9536171968897427,
      "tokens_seen": 487981056
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008607221664994985,
      "loss": 1.6595,
      "theoretical_loss": 3.953560184460473,
      "tokens_seen": 488046592
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008607021063189569,
      "loss": 1.6056,
      "theoretical_loss": 3.953503181829732,
      "tokens_seen": 488112128
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008606820461384152,
      "loss": 1.6239,
      "theoretical_loss": 3.95344618899452,
      "tokens_seen": 488177664
    },
    {
      "epoch": 0.15,
      "objective/train/advantage_avg": 0.488285094499588,
      "objective/train/docs_used": 183865,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.634871244430542,
      "objective/train/original_loss": 1.634871244430542,
      "objective/train/theoretical_loss": 3.9533892059518383,
      "objective/train/tokens_used": 508703200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.240000382065773,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500484704971313,
      "objective/train/weighted_lm_loss": 1.7165744304656982,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9533892059518383,
      "tokens_seen": 488243200
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008606619859578736,
      "loss": 1.7296,
      "theoretical_loss": 3.9533892059518383,
      "tokens_seen": 488243200
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008606419257773321,
      "loss": 1.5925,
      "theoretical_loss": 3.95333223269869,
      "tokens_seen": 488308736
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008606218655967903,
      "loss": 1.658,
      "theoretical_loss": 3.9532752692320816,
      "tokens_seen": 488374272
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008606018054162488,
      "loss": 1.6043,
      "theoretical_loss": 3.9532183155490177,
      "tokens_seen": 488439808
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000860581745235707,
      "loss": 1.7125,
      "theoretical_loss": 3.953161371646506,
      "tokens_seen": 488505344
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008605616850551655,
      "loss": 1.4775,
      "theoretical_loss": 3.9531044375215547,
      "tokens_seen": 488570880
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008605416248746239,
      "loss": 1.5956,
      "theoretical_loss": 3.9530475131711746,
      "tokens_seen": 488636416
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008605215646940822,
      "loss": 1.6439,
      "theoretical_loss": 3.9529905985923763,
      "tokens_seen": 488701952
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008605015045135406,
      "loss": 1.5719,
      "theoretical_loss": 3.952933693782172,
      "tokens_seen": 488767488
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008604814443329991,
      "loss": 1.5986,
      "theoretical_loss": 3.9528767987375764,
      "tokens_seen": 488833024
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008604613841524573,
      "loss": 1.6765,
      "theoretical_loss": 3.9528199134556044,
      "tokens_seen": 488898560
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008604413239719158,
      "loss": 1.6545,
      "theoretical_loss": 3.9527630379332725,
      "tokens_seen": 488964096
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000860421263791374,
      "loss": 1.6569,
      "theoretical_loss": 3.9527061721675985,
      "tokens_seen": 489029632
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008604012036108325,
      "loss": 1.6657,
      "theoretical_loss": 3.9526493161556013,
      "tokens_seen": 489095168
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008603811434302909,
      "loss": 1.6495,
      "theoretical_loss": 3.9525924698943022,
      "tokens_seen": 489160704
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008603610832497492,
      "loss": 1.687,
      "theoretical_loss": 3.952535633380722,
      "tokens_seen": 489226240
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008603410230692077,
      "loss": 1.5027,
      "theoretical_loss": 3.9524788066118837,
      "tokens_seen": 489291776
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000860320962888666,
      "loss": 1.6547,
      "theoretical_loss": 3.9524219895848116,
      "tokens_seen": 489357312
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008603009027081244,
      "loss": 1.6899,
      "theoretical_loss": 3.952365182296533,
      "tokens_seen": 489422848
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008602808425275828,
      "loss": 1.6411,
      "theoretical_loss": 3.952308384744073,
      "tokens_seen": 489488384
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008602607823470412,
      "loss": 1.6792,
      "theoretical_loss": 3.9522515969244605,
      "tokens_seen": 489553920
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008602407221664995,
      "loss": 1.5841,
      "theoretical_loss": 3.952194818834725,
      "tokens_seen": 489619456
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000860220661985958,
      "loss": 1.7696,
      "theoretical_loss": 3.9521380504718975,
      "tokens_seen": 489684992
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008602006018054162,
      "loss": 1.7318,
      "theoretical_loss": 3.952081291833011,
      "tokens_seen": 489750528
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008601805416248747,
      "loss": 1.5766,
      "theoretical_loss": 3.9520245429150975,
      "tokens_seen": 489816064
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000860160481444333,
      "loss": 1.6828,
      "theoretical_loss": 3.951967803715193,
      "tokens_seen": 489881600
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008601404212637914,
      "loss": 1.7221,
      "theoretical_loss": 3.9519110742303325,
      "tokens_seen": 489947136
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008601203610832498,
      "loss": 1.7698,
      "theoretical_loss": 3.951854354457554,
      "tokens_seen": 490012672
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008601003009027082,
      "loss": 1.6475,
      "theoretical_loss": 3.951797644393897,
      "tokens_seen": 490078208
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008600802407221665,
      "loss": 1.5595,
      "theoretical_loss": 3.9517409440364,
      "tokens_seen": 490143744
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008600601805416249,
      "loss": 1.5627,
      "theoretical_loss": 3.951684253382105,
      "tokens_seen": 490209280
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008600401203610833,
      "loss": 1.6962,
      "theoretical_loss": 3.9516275724280545,
      "tokens_seen": 490274816
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008600200601805417,
      "loss": 1.6326,
      "theoretical_loss": 3.9515709011712934,
      "tokens_seen": 490340352
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00086,
      "loss": 1.6999,
      "theoretical_loss": 3.9515142396088647,
      "tokens_seen": 490405888
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008599799398194584,
      "loss": 1.5187,
      "theoretical_loss": 3.951457587737817,
      "tokens_seen": 490471424
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008599598796389167,
      "loss": 1.5951,
      "theoretical_loss": 3.9514009455551973,
      "tokens_seen": 490536960
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008599398194583752,
      "loss": 1.6903,
      "theoretical_loss": 3.951344313058054,
      "tokens_seen": 490602496
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008599197592778335,
      "loss": 1.6521,
      "theoretical_loss": 3.951287690243438,
      "tokens_seen": 490668032
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008598996990972919,
      "loss": 1.6147,
      "theoretical_loss": 3.9512310771084014,
      "tokens_seen": 490733568
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008598796389167503,
      "loss": 1.6771,
      "theoretical_loss": 3.9511744736499965,
      "tokens_seen": 490799104
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008598595787362087,
      "loss": 1.7336,
      "theoretical_loss": 3.951117879865278,
      "tokens_seen": 490864640
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000859839518555667,
      "loss": 1.5961,
      "theoretical_loss": 3.951061295751301,
      "tokens_seen": 490930176
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008598194583751255,
      "loss": 1.6592,
      "theoretical_loss": 3.951004721305123,
      "tokens_seen": 490995712
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008597993981945837,
      "loss": 1.6885,
      "theoretical_loss": 3.9509481565238014,
      "tokens_seen": 491061248
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008597793380140422,
      "loss": 1.655,
      "theoretical_loss": 3.950891601404395,
      "tokens_seen": 491126784
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008597592778335005,
      "loss": 1.5425,
      "theoretical_loss": 3.9508350559439664,
      "tokens_seen": 491192320
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008597392176529589,
      "loss": 1.6547,
      "theoretical_loss": 3.950778520139576,
      "tokens_seen": 491257856
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008597191574724173,
      "loss": 1.6784,
      "theoretical_loss": 3.9507219939882874,
      "tokens_seen": 491323392
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008596990972918756,
      "loss": 1.6538,
      "theoretical_loss": 3.9506654774871652,
      "tokens_seen": 491388928
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000859679037111334,
      "loss": 1.64,
      "theoretical_loss": 3.950608970633275,
      "tokens_seen": 491454464
    },
    {
      "epoch": 0.15,
      "objective/train/advantage_avg": 0.4913667142391205,
      "objective/train/docs_used": 185162,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5327297449111938,
      "objective/train/original_loss": 1.5327297449111938,
      "objective/train/theoretical_loss": 3.9505524734236848,
      "objective/train/tokens_used": 511980000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24189111590385437,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503664016723633,
      "objective/train/weighted_lm_loss": 1.609784483909607,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9753099083900452,
      "theoretical_loss": 3.9505524734236848,
      "tokens_seen": 491520000
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008596589769307925,
      "loss": 1.6483,
      "theoretical_loss": 3.9505524734236848,
      "tokens_seen": 491520000
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008596389167502507,
      "loss": 1.5931,
      "theoretical_loss": 3.9504959858554622,
      "tokens_seen": 491585536
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008596188565697092,
      "loss": 1.5681,
      "theoretical_loss": 3.950439507925677,
      "tokens_seen": 491651072
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008595987963891674,
      "loss": 1.5339,
      "theoretical_loss": 3.9503830396314,
      "tokens_seen": 491716608
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008595787362086259,
      "loss": 1.7121,
      "theoretical_loss": 3.950326580969703,
      "tokens_seen": 491782144
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008595586760280843,
      "loss": 1.6072,
      "theoretical_loss": 3.950270131937661,
      "tokens_seen": 491847680
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008595386158475426,
      "loss": 1.7294,
      "theoretical_loss": 3.9502136925323477,
      "tokens_seen": 491913216
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000859518555667001,
      "loss": 1.7427,
      "theoretical_loss": 3.950157262750839,
      "tokens_seen": 491978752
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008594984954864595,
      "loss": 1.6756,
      "theoretical_loss": 3.950100842590212,
      "tokens_seen": 492044288
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008594784353059177,
      "loss": 1.6451,
      "theoretical_loss": 3.9500444320475467,
      "tokens_seen": 492109824
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008594583751253762,
      "loss": 1.6721,
      "theoretical_loss": 3.949988031119922,
      "tokens_seen": 492175360
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008594383149448344,
      "loss": 1.6257,
      "theoretical_loss": 3.9499316398044195,
      "tokens_seen": 492240896
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008594182547642929,
      "loss": 1.7382,
      "theoretical_loss": 3.949875258098121,
      "tokens_seen": 492306432
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008593981945837513,
      "loss": 1.6505,
      "theoretical_loss": 3.949818885998111,
      "tokens_seen": 492371968
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008593781344032096,
      "loss": 1.7274,
      "theoretical_loss": 3.9497625235014735,
      "tokens_seen": 492437504
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000859358074222668,
      "loss": 1.6178,
      "theoretical_loss": 3.949706170605296,
      "tokens_seen": 492503040
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008593380140421264,
      "loss": 1.7184,
      "theoretical_loss": 3.949649827306665,
      "tokens_seen": 492568576
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008593179538615847,
      "loss": 1.7109,
      "theoretical_loss": 3.9495934936026695,
      "tokens_seen": 492634112
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008592978936810432,
      "loss": 1.6992,
      "theoretical_loss": 3.9495371694904002,
      "tokens_seen": 492699648
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008592778335005015,
      "loss": 1.5218,
      "theoretical_loss": 3.9494808549669473,
      "tokens_seen": 492765184
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008592577733199599,
      "loss": 1.704,
      "theoretical_loss": 3.9494245500294047,
      "tokens_seen": 492830720
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008592377131394182,
      "loss": 1.7143,
      "theoretical_loss": 3.949368254674866,
      "tokens_seen": 492896256
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008592176529588766,
      "loss": 1.6393,
      "theoretical_loss": 3.9493119689004255,
      "tokens_seen": 492961792
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000859197592778335,
      "loss": 1.665,
      "theoretical_loss": 3.9492556927031806,
      "tokens_seen": 493027328
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008591775325977934,
      "loss": 1.5913,
      "theoretical_loss": 3.949199426080228,
      "tokens_seen": 493092864
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008591574724172517,
      "loss": 1.6989,
      "theoretical_loss": 3.949143169028668,
      "tokens_seen": 493158400
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008591374122367102,
      "loss": 1.6804,
      "theoretical_loss": 3.9490869215455993,
      "tokens_seen": 493223936
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008591173520561685,
      "loss": 1.6273,
      "theoretical_loss": 3.9490306836281244,
      "tokens_seen": 493289472
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008590972918756269,
      "loss": 1.6909,
      "theoretical_loss": 3.9489744552733455,
      "tokens_seen": 493355008
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008590772316950852,
      "loss": 1.677,
      "theoretical_loss": 3.948918236478367,
      "tokens_seen": 493420544
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008590571715145437,
      "loss": 1.8298,
      "theoretical_loss": 3.9488620272402937,
      "tokens_seen": 493486080
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000859037111334002,
      "loss": 1.7091,
      "theoretical_loss": 3.9488058275562326,
      "tokens_seen": 493551616
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008590170511534604,
      "loss": 1.8155,
      "theoretical_loss": 3.9487496374232913,
      "tokens_seen": 493617152
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008589969909729187,
      "loss": 1.6607,
      "theoretical_loss": 3.948693456838579,
      "tokens_seen": 493682688
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008589769307923771,
      "loss": 1.5868,
      "theoretical_loss": 3.9486372857992054,
      "tokens_seen": 493748224
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008589568706118355,
      "loss": 1.6415,
      "theoretical_loss": 3.948581124302283,
      "tokens_seen": 493813760
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008589368104312939,
      "loss": 1.6505,
      "theoretical_loss": 3.9485249723449236,
      "tokens_seen": 493879296
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008589167502507522,
      "loss": 1.5553,
      "theoretical_loss": 3.9484688299242423,
      "tokens_seen": 493944832
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008588966900702107,
      "loss": 1.5332,
      "theoretical_loss": 3.9484126970373534,
      "tokens_seen": 494010368
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000858876629889669,
      "loss": 1.7607,
      "theoretical_loss": 3.948356573681374,
      "tokens_seen": 494075904
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008588565697091274,
      "loss": 1.6188,
      "theoretical_loss": 3.9483004598534217,
      "tokens_seen": 494141440
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008588365095285859,
      "loss": 1.7941,
      "theoretical_loss": 3.9482443555506164,
      "tokens_seen": 494206976
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008588164493480441,
      "loss": 1.6809,
      "theoretical_loss": 3.9481882607700776,
      "tokens_seen": 494272512
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008587963891675026,
      "loss": 1.6599,
      "theoretical_loss": 3.948132175508927,
      "tokens_seen": 494338048
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008587763289869609,
      "loss": 1.6267,
      "theoretical_loss": 3.948076099764288,
      "tokens_seen": 494403584
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008587562688064193,
      "loss": 1.7205,
      "theoretical_loss": 3.948020033533284,
      "tokens_seen": 494469120
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008587362086258777,
      "loss": 1.7124,
      "theoretical_loss": 3.9479639768130403,
      "tokens_seen": 494534656
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000858716148445336,
      "loss": 1.74,
      "theoretical_loss": 3.9479079296006843,
      "tokens_seen": 494600192
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008586960882647944,
      "loss": 1.8365,
      "theoretical_loss": 3.947851891893343,
      "tokens_seen": 494665728
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008586760280842529,
      "loss": 1.6611,
      "theoretical_loss": 3.947795863688146,
      "tokens_seen": 494731264
    },
    {
      "epoch": 0.15,
      "objective/train/advantage_avg": 0.4944104850292206,
      "objective/train/docs_used": 186433,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6233147382736206,
      "objective/train/original_loss": 1.6233150959014893,
      "objective/train/theoretical_loss": 3.947739844982224,
      "objective/train/tokens_used": 515256800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24478408694267273,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506855249404907,
      "objective/train/weighted_lm_loss": 1.7054861783981323,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.947739844982224,
      "tokens_seen": 494796800
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008586559679037111,
      "loss": 1.6021,
      "theoretical_loss": 3.947739844982224,
      "tokens_seen": 494796800
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008586359077231696,
      "loss": 1.5467,
      "theoretical_loss": 3.9476838357727075,
      "tokens_seen": 494862336
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008586158475426278,
      "loss": 1.5192,
      "theoretical_loss": 3.9476278360567303,
      "tokens_seen": 494927872
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008585957873620863,
      "loss": 1.6917,
      "theoretical_loss": 3.9475718458314257,
      "tokens_seen": 494993408
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008585757271815447,
      "loss": 1.6019,
      "theoretical_loss": 3.94751586509393,
      "tokens_seen": 495058944
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000858555667001003,
      "loss": 1.6493,
      "theoretical_loss": 3.9474598938413785,
      "tokens_seen": 495124480
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008585356068204614,
      "loss": 1.6765,
      "theoretical_loss": 3.9474039320709107,
      "tokens_seen": 495190016
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008585155466399199,
      "loss": 1.7863,
      "theoretical_loss": 3.9473479797796633,
      "tokens_seen": 495255552
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008584954864593781,
      "loss": 1.7943,
      "theoretical_loss": 3.947292036964779,
      "tokens_seen": 495321088
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008584754262788366,
      "loss": 1.6114,
      "theoretical_loss": 3.947236103623398,
      "tokens_seen": 495386624
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008584553660982948,
      "loss": 1.6652,
      "theoretical_loss": 3.9471801797526633,
      "tokens_seen": 495452160
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008584353059177533,
      "loss": 1.5683,
      "theoretical_loss": 3.947124265349719,
      "tokens_seen": 495517696
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008584152457372117,
      "loss": 1.5359,
      "theoretical_loss": 3.9470683604117105,
      "tokens_seen": 495583232
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00085839518555667,
      "loss": 1.7465,
      "theoretical_loss": 3.947012464935784,
      "tokens_seen": 495648768
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008583751253761284,
      "loss": 1.6241,
      "theoretical_loss": 3.946956578919088,
      "tokens_seen": 495714304
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008583550651955868,
      "loss": 1.5361,
      "theoretical_loss": 3.9469007023587697,
      "tokens_seen": 495779840
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008583350050150451,
      "loss": 1.7957,
      "theoretical_loss": 3.946844835251981,
      "tokens_seen": 495845376
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008583149448345036,
      "loss": 1.7185,
      "theoretical_loss": 3.946788977595873,
      "tokens_seen": 495910912
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008582948846539619,
      "loss": 1.6026,
      "theoretical_loss": 3.9467331293875976,
      "tokens_seen": 495976448
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008582748244734203,
      "loss": 1.708,
      "theoretical_loss": 3.9466772906243097,
      "tokens_seen": 496041984
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008582547642928786,
      "loss": 1.6868,
      "theoretical_loss": 3.946621461303164,
      "tokens_seen": 496107520
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000858234704112337,
      "loss": 1.6382,
      "theoretical_loss": 3.9465656414213166,
      "tokens_seen": 496173056
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008582146439317954,
      "loss": 1.6916,
      "theoretical_loss": 3.9465098309759252,
      "tokens_seen": 496238592
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008581945837512538,
      "loss": 1.691,
      "theoretical_loss": 3.9464540299641495,
      "tokens_seen": 496304128
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008581745235707121,
      "loss": 1.5692,
      "theoretical_loss": 3.946398238383148,
      "tokens_seen": 496369664
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008581544633901706,
      "loss": 1.6496,
      "theoretical_loss": 3.9463424562300835,
      "tokens_seen": 496435200
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008581344032096289,
      "loss": 1.7202,
      "theoretical_loss": 3.9462866835021178,
      "tokens_seen": 496500736
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008581143430290873,
      "loss": 1.6107,
      "theoretical_loss": 3.946230920196414,
      "tokens_seen": 496566272
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008580942828485456,
      "loss": 1.6035,
      "theoretical_loss": 3.9461751663101388,
      "tokens_seen": 496631808
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008580742226680041,
      "loss": 1.7093,
      "theoretical_loss": 3.946119421840457,
      "tokens_seen": 496697344
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008580541624874624,
      "loss": 1.7058,
      "theoretical_loss": 3.9460636867845365,
      "tokens_seen": 496762880
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008580341023069208,
      "loss": 1.6845,
      "theoretical_loss": 3.9460079611395456,
      "tokens_seen": 496828416
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008580140421263791,
      "loss": 1.6648,
      "theoretical_loss": 3.9459522449026547,
      "tokens_seen": 496893952
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008579939819458375,
      "loss": 1.6558,
      "theoretical_loss": 3.945896538071035,
      "tokens_seen": 496959488
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008579739217652959,
      "loss": 1.6979,
      "theoretical_loss": 3.9458408406418584,
      "tokens_seen": 497025024
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008579538615847543,
      "loss": 1.6416,
      "theoretical_loss": 3.9457851526122987,
      "tokens_seen": 497090560
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008579338014042126,
      "loss": 1.6168,
      "theoretical_loss": 3.9457294739795303,
      "tokens_seen": 497156096
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008579137412236711,
      "loss": 1.7117,
      "theoretical_loss": 3.9456738047407294,
      "tokens_seen": 497221632
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008578936810431294,
      "loss": 1.5237,
      "theoretical_loss": 3.945618144893073,
      "tokens_seen": 497287168
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008578736208625878,
      "loss": 1.7463,
      "theoretical_loss": 3.945562494433741,
      "tokens_seen": 497352704
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008578535606820461,
      "loss": 1.7703,
      "theoretical_loss": 3.9455068533599107,
      "tokens_seen": 497418240
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008578335005015045,
      "loss": 1.6323,
      "theoretical_loss": 3.9454512216687645,
      "tokens_seen": 497483776
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008578134403209629,
      "loss": 1.6856,
      "theoretical_loss": 3.9453955993574845,
      "tokens_seen": 497549312
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008577933801404213,
      "loss": 1.6327,
      "theoretical_loss": 3.945339986423254,
      "tokens_seen": 497614848
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008577733199598796,
      "loss": 1.6471,
      "theoretical_loss": 3.9452843828632562,
      "tokens_seen": 497680384
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008577532597793381,
      "loss": 1.6732,
      "theoretical_loss": 3.945228788674678,
      "tokens_seen": 497745920
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008577331995987963,
      "loss": 1.6235,
      "theoretical_loss": 3.945173203854707,
      "tokens_seen": 497811456
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008577131394182548,
      "loss": 1.6917,
      "theoretical_loss": 3.9451176284005296,
      "tokens_seen": 497876992
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008576930792377132,
      "loss": 1.6517,
      "theoretical_loss": 3.945062062309337,
      "tokens_seen": 497942528
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008576730190571715,
      "loss": 1.7057,
      "theoretical_loss": 3.9450065055783186,
      "tokens_seen": 498008064
    },
    {
      "epoch": 0.15,
      "objective/train/advantage_avg": 0.4921090602874756,
      "objective/train/docs_used": 187487,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.686283826828003,
      "objective/train/original_loss": 1.686283826828003,
      "objective/train/theoretical_loss": 3.9449509582046662,
      "objective/train/tokens_used": 518533600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2426682561635971,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050444483757019,
      "objective/train/weighted_lm_loss": 1.7715121507644653,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9449509582046662,
      "tokens_seen": 498073600
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008576529588766299,
      "loss": 1.7322,
      "theoretical_loss": 3.9449509582046662,
      "tokens_seen": 498073600
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008576328986960882,
      "loss": 1.622,
      "theoretical_loss": 3.944895420185574,
      "tokens_seen": 498139136
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008576128385155466,
      "loss": 1.7711,
      "theoretical_loss": 3.944839891518235,
      "tokens_seen": 498204672
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008575927783350051,
      "loss": 1.6856,
      "theoretical_loss": 3.9447843721998455,
      "tokens_seen": 498270208
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008575727181544633,
      "loss": 1.5376,
      "theoretical_loss": 3.944728862227601,
      "tokens_seen": 498335744
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008575526579739218,
      "loss": 1.5706,
      "theoretical_loss": 3.9446733615987,
      "tokens_seen": 498401280
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008575325977933802,
      "loss": 1.6647,
      "theoretical_loss": 3.944617870310343,
      "tokens_seen": 498466816
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008575125376128385,
      "loss": 1.6552,
      "theoretical_loss": 3.944562388359728,
      "tokens_seen": 498532352
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008574924774322969,
      "loss": 1.5662,
      "theoretical_loss": 3.9445069157440575,
      "tokens_seen": 498597888
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008574724172517552,
      "loss": 1.7122,
      "theoretical_loss": 3.9444514524605347,
      "tokens_seen": 498663424
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008574523570712136,
      "loss": 1.6933,
      "theoretical_loss": 3.9443959985063626,
      "tokens_seen": 498728960
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008574322968906721,
      "loss": 1.7744,
      "theoretical_loss": 3.9443405538787464,
      "tokens_seen": 498794496
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008574122367101303,
      "loss": 1.6706,
      "theoretical_loss": 3.944285118574893,
      "tokens_seen": 498860032
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008573921765295888,
      "loss": 1.5819,
      "theoretical_loss": 3.9442296925920095,
      "tokens_seen": 498925568
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008573721163490471,
      "loss": 1.8027,
      "theoretical_loss": 3.944174275927305,
      "tokens_seen": 498991104
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008573520561685055,
      "loss": 1.7501,
      "theoretical_loss": 3.9441188685779887,
      "tokens_seen": 499056640
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000857331995987964,
      "loss": 1.6712,
      "theoretical_loss": 3.9440634705412725,
      "tokens_seen": 499122176
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008573119358074223,
      "loss": 1.7242,
      "theoretical_loss": 3.9440080818143683,
      "tokens_seen": 499187712
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008572918756268807,
      "loss": 1.6799,
      "theoretical_loss": 3.9439527023944896,
      "tokens_seen": 499253248
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000857271815446339,
      "loss": 1.7857,
      "theoretical_loss": 3.9438973322788518,
      "tokens_seen": 499318784
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008572517552657974,
      "loss": 1.6998,
      "theoretical_loss": 3.9438419714646695,
      "tokens_seen": 499384320
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008572316950852558,
      "loss": 1.8287,
      "theoretical_loss": 3.943786619949161,
      "tokens_seen": 499449856
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008572116349047142,
      "loss": 1.6951,
      "theoretical_loss": 3.943731277729544,
      "tokens_seen": 499515392
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008571915747241725,
      "loss": 1.7141,
      "theoretical_loss": 3.9436759448030383,
      "tokens_seen": 499580928
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000857171514543631,
      "loss": 1.8381,
      "theoretical_loss": 3.9436206211668647,
      "tokens_seen": 499646464
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008571514543630893,
      "loss": 1.713,
      "theoretical_loss": 3.943565306818245,
      "tokens_seen": 499712000
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008571313941825477,
      "loss": 1.6078,
      "theoretical_loss": 3.943510001754402,
      "tokens_seen": 499777536
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000857111334002006,
      "loss": 1.6472,
      "theoretical_loss": 3.9434547059725604,
      "tokens_seen": 499843072
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008570912738214645,
      "loss": 1.7676,
      "theoretical_loss": 3.9433994194699453,
      "tokens_seen": 499908608
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008570712136409228,
      "loss": 1.844,
      "theoretical_loss": 3.9433441422437836,
      "tokens_seen": 499974144
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008570511534603812,
      "loss": 1.7302,
      "theoretical_loss": 3.9432888742913033,
      "tokens_seen": 500039680
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008570310932798395,
      "loss": 1.8328,
      "theoretical_loss": 3.943233615609733,
      "tokens_seen": 500105216
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008570110330992979,
      "loss": 1.8726,
      "theoretical_loss": 3.943178366196304,
      "tokens_seen": 500170752
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008569909729187563,
      "loss": 1.7599,
      "theoretical_loss": 3.9431231260482464,
      "tokens_seen": 500236288
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008569709127382147,
      "loss": 1.7863,
      "theoretical_loss": 3.9430678951627938,
      "tokens_seen": 500301824
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000856950852557673,
      "loss": 1.7858,
      "theoretical_loss": 3.94301267353718,
      "tokens_seen": 500367360
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008569307923771315,
      "loss": 1.7798,
      "theoretical_loss": 3.942957461168639,
      "tokens_seen": 500432896
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008569107321965898,
      "loss": 1.8921,
      "theoretical_loss": 3.942902258054408,
      "tokens_seen": 500498432
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008568906720160482,
      "loss": 1.7891,
      "theoretical_loss": 3.9428470641917244,
      "tokens_seen": 500563968
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008568706118355065,
      "loss": 1.8246,
      "theoretical_loss": 3.9427918795778267,
      "tokens_seen": 500629504
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008568505516549649,
      "loss": 1.7398,
      "theoretical_loss": 3.9427367042099544,
      "tokens_seen": 500695040
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008568304914744233,
      "loss": 1.6855,
      "theoretical_loss": 3.9426815380853477,
      "tokens_seen": 500760576
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008568104312938817,
      "loss": 1.6292,
      "theoretical_loss": 3.9426263812012503,
      "tokens_seen": 500826112
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00085679037111334,
      "loss": 1.716,
      "theoretical_loss": 3.9425712335549044,
      "tokens_seen": 500891648
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008567703109327985,
      "loss": 1.7085,
      "theoretical_loss": 3.942516095143555,
      "tokens_seen": 500957184
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008567502507522567,
      "loss": 1.7111,
      "theoretical_loss": 3.9424609659644476,
      "tokens_seen": 501022720
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008567301905717152,
      "loss": 1.8806,
      "theoretical_loss": 3.9424058460148292,
      "tokens_seen": 501088256
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008567101303911736,
      "loss": 1.7525,
      "theoretical_loss": 3.9423507352919476,
      "tokens_seen": 501153792
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008566900702106319,
      "loss": 1.8146,
      "theoretical_loss": 3.9422956337930524,
      "tokens_seen": 501219328
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008566700100300903,
      "loss": 1.7736,
      "theoretical_loss": 3.9422405415153934,
      "tokens_seen": 501284864
    },
    {
      "epoch": 0.15,
      "objective/train/advantage_avg": 0.49449148774147034,
      "objective/train/docs_used": 188727,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5148030519485474,
      "objective/train/original_loss": 1.5148030519485474,
      "objective/train/theoretical_loss": 3.9421854584562226,
      "objective/train/tokens_used": 521810400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24478769302368164,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506936311721802,
      "objective/train/weighted_lm_loss": 1.5914528369903564,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9421854584562226,
      "tokens_seen": 501350400
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008566499498495486,
      "loss": 1.7312,
      "theoretical_loss": 3.9421854584562226,
      "tokens_seen": 501350400
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000856629889669007,
      "loss": 1.6754,
      "theoretical_loss": 3.942130384612793,
      "tokens_seen": 501415936
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008566098294884655,
      "loss": 1.726,
      "theoretical_loss": 3.942075319982358,
      "tokens_seen": 501481472
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008565897693079237,
      "loss": 1.8364,
      "theoretical_loss": 3.9420202645621725,
      "tokens_seen": 501547008
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008565697091273822,
      "loss": 1.7069,
      "theoretical_loss": 3.9419652183494938,
      "tokens_seen": 501612544
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008565496489468406,
      "loss": 1.7698,
      "theoretical_loss": 3.9419101813415787,
      "tokens_seen": 501678080
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008565295887662989,
      "loss": 1.663,
      "theoretical_loss": 3.941855153535686,
      "tokens_seen": 501743616
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008565095285857573,
      "loss": 1.6876,
      "theoretical_loss": 3.941800134929075,
      "tokens_seen": 501809152
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008564894684052156,
      "loss": 1.7476,
      "theoretical_loss": 3.9417451255190077,
      "tokens_seen": 501874688
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000856469408224674,
      "loss": 1.808,
      "theoretical_loss": 3.9416901253027454,
      "tokens_seen": 501940224
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008564493480441325,
      "loss": 1.7675,
      "theoretical_loss": 3.9416351342775524,
      "tokens_seen": 502005760
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008564292878635907,
      "loss": 1.6907,
      "theoretical_loss": 3.9415801524406917,
      "tokens_seen": 502071296
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008564092276830492,
      "loss": 1.7417,
      "theoretical_loss": 3.94152517978943,
      "tokens_seen": 502136832
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008563891675025075,
      "loss": 1.588,
      "theoretical_loss": 3.941470216321034,
      "tokens_seen": 502202368
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008563691073219659,
      "loss": 1.8457,
      "theoretical_loss": 3.9414152620327716,
      "tokens_seen": 502267904
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008563490471414243,
      "loss": 1.6521,
      "theoretical_loss": 3.9413603169219122,
      "tokens_seen": 502333440
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008563289869608827,
      "loss": 1.6019,
      "theoretical_loss": 3.9413053809857264,
      "tokens_seen": 502398976
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000856308926780341,
      "loss": 1.6222,
      "theoretical_loss": 3.941250454221485,
      "tokens_seen": 502464512
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008562888665997994,
      "loss": 1.7308,
      "theoretical_loss": 3.941195536626461,
      "tokens_seen": 502530048
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008562688064192577,
      "loss": 1.7501,
      "theoretical_loss": 3.941140628197928,
      "tokens_seen": 502595584
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008562487462387162,
      "loss": 1.7315,
      "theoretical_loss": 3.941085728933162,
      "tokens_seen": 502661120
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008562286860581745,
      "loss": 1.7432,
      "theoretical_loss": 3.9410308388294384,
      "tokens_seen": 502726656
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008562086258776329,
      "loss": 1.6503,
      "theoretical_loss": 3.940975957884034,
      "tokens_seen": 502792192
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008561885656970913,
      "loss": 1.7376,
      "theoretical_loss": 3.9409210860942285,
      "tokens_seen": 502857728
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008561685055165497,
      "loss": 1.7768,
      "theoretical_loss": 3.9408662234573013,
      "tokens_seen": 502923264
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000856148445336008,
      "loss": 1.6886,
      "theoretical_loss": 3.9408113699705325,
      "tokens_seen": 502988800
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008561283851554664,
      "loss": 1.6498,
      "theoretical_loss": 3.9407565256312047,
      "tokens_seen": 503054336
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008561083249749247,
      "loss": 1.7782,
      "theoretical_loss": 3.940701690436601,
      "tokens_seen": 503119872
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008560882647943832,
      "loss": 1.7973,
      "theoretical_loss": 3.9406468643840054,
      "tokens_seen": 503185408
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008560682046138415,
      "loss": 1.7796,
      "theoretical_loss": 3.940592047470704,
      "tokens_seen": 503250944
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008560481444332999,
      "loss": 1.6904,
      "theoretical_loss": 3.940537239693983,
      "tokens_seen": 503316480
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008560280842527582,
      "loss": 1.7036,
      "theoretical_loss": 3.9404824410511305,
      "tokens_seen": 503382016
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008560080240722167,
      "loss": 1.7104,
      "theoretical_loss": 3.940427651539435,
      "tokens_seen": 503447552
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000855987963891675,
      "loss": 1.8244,
      "theoretical_loss": 3.940372871156187,
      "tokens_seen": 503513088
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008559679037111334,
      "loss": 1.7562,
      "theoretical_loss": 3.9403180998986778,
      "tokens_seen": 503578624
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008559478435305918,
      "loss": 1.6141,
      "theoretical_loss": 3.9402633377641996,
      "tokens_seen": 503644160
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008559277833500502,
      "loss": 1.7405,
      "theoretical_loss": 3.940208584750046,
      "tokens_seen": 503709696
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008559077231695085,
      "loss": 1.6818,
      "theoretical_loss": 3.940153840853512,
      "tokens_seen": 503775232
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000855887662988967,
      "loss": 1.6255,
      "theoretical_loss": 3.9400991060718935,
      "tokens_seen": 503840768
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008558676028084252,
      "loss": 1.7052,
      "theoretical_loss": 3.940044380402487,
      "tokens_seen": 503906304
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008558475426278837,
      "loss": 1.8162,
      "theoretical_loss": 3.9399896638425913,
      "tokens_seen": 503971840
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008558274824473421,
      "loss": 1.8732,
      "theoretical_loss": 3.9399349563895054,
      "tokens_seen": 504037376
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008558074222668004,
      "loss": 1.7249,
      "theoretical_loss": 3.93988025804053,
      "tokens_seen": 504102912
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008557873620862589,
      "loss": 1.7259,
      "theoretical_loss": 3.939825568792967,
      "tokens_seen": 504168448
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008557673019057171,
      "loss": 1.7787,
      "theoretical_loss": 3.9397708886441185,
      "tokens_seen": 504233984
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008557472417251756,
      "loss": 1.6362,
      "theoretical_loss": 3.939716217591289,
      "tokens_seen": 504299520
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000855727181544634,
      "loss": 1.7864,
      "theoretical_loss": 3.9396615556317838,
      "tokens_seen": 504365056
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008557071213640923,
      "loss": 1.664,
      "theoretical_loss": 3.9396069027629084,
      "tokens_seen": 504430592
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008556870611835507,
      "loss": 1.688,
      "theoretical_loss": 3.939552258981971,
      "tokens_seen": 504496128
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000855667001003009,
      "loss": 1.6894,
      "theoretical_loss": 3.93949762428628,
      "tokens_seen": 504561664
    },
    {
      "epoch": 0.15,
      "objective/train/advantage_avg": 0.4919562339782715,
      "objective/train/docs_used": 189911,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.488186001777649,
      "objective/train/original_loss": 1.4881861209869385,
      "objective/train/theoretical_loss": 3.939442998673144,
      "objective/train/tokens_used": 525087200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24271796643733978,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504294633865356,
      "objective/train/weighted_lm_loss": 1.563111424446106,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.939442998673144,
      "tokens_seen": 504627200
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008556469408224674,
      "loss": 1.6751,
      "theoretical_loss": 3.939442998673144,
      "tokens_seen": 504627200
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008556268806419259,
      "loss": 1.6742,
      "theoretical_loss": 3.9393883821398754,
      "tokens_seen": 504692736
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008556068204613841,
      "loss": 1.6823,
      "theoretical_loss": 3.939333774683785,
      "tokens_seen": 504758272
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008555867602808426,
      "loss": 1.6976,
      "theoretical_loss": 3.939279176302187,
      "tokens_seen": 504823808
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000855566700100301,
      "loss": 1.6921,
      "theoretical_loss": 3.9392245869923954,
      "tokens_seen": 504889344
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008555466399197593,
      "loss": 1.7448,
      "theoretical_loss": 3.939170006751725,
      "tokens_seen": 504954880
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008555265797392177,
      "loss": 1.6712,
      "theoretical_loss": 3.9391154355774924,
      "tokens_seen": 505020416
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000855506519558676,
      "loss": 1.752,
      "theoretical_loss": 3.9390608734670156,
      "tokens_seen": 505085952
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008554864593781344,
      "loss": 1.87,
      "theoretical_loss": 3.939006320417614,
      "tokens_seen": 505151488
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008554663991975929,
      "loss": 1.7513,
      "theoretical_loss": 3.938951776426607,
      "tokens_seen": 505217024
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008554463390170511,
      "loss": 1.6259,
      "theoretical_loss": 3.9388972414913157,
      "tokens_seen": 505282560
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008554262788365096,
      "loss": 1.7146,
      "theoretical_loss": 3.938842715609062,
      "tokens_seen": 505348096
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008554062186559679,
      "loss": 1.6968,
      "theoretical_loss": 3.9387881987771705,
      "tokens_seen": 505413632
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008553861584754263,
      "loss": 1.6711,
      "theoretical_loss": 3.938733690992964,
      "tokens_seen": 505479168
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008553660982948847,
      "loss": 1.7096,
      "theoretical_loss": 3.93867919225377,
      "tokens_seen": 505544704
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008553460381143431,
      "loss": 1.7244,
      "theoretical_loss": 3.938624702556914,
      "tokens_seen": 505610240
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008553259779338014,
      "loss": 1.7492,
      "theoretical_loss": 3.9385702218997247,
      "tokens_seen": 505675776
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008553059177532598,
      "loss": 1.803,
      "theoretical_loss": 3.9385157502795316,
      "tokens_seen": 505741312
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008552858575727181,
      "loss": 1.6991,
      "theoretical_loss": 3.9384612876936633,
      "tokens_seen": 505806848
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008552657973921766,
      "loss": 1.7034,
      "theoretical_loss": 3.9384068341394523,
      "tokens_seen": 505872384
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008552457372116349,
      "loss": 1.6652,
      "theoretical_loss": 3.9383523896142316,
      "tokens_seen": 505937920
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008552256770310933,
      "loss": 1.6977,
      "theoretical_loss": 3.9382979541153333,
      "tokens_seen": 506003456
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008552056168505517,
      "loss": 1.7646,
      "theoretical_loss": 3.9382435276400938,
      "tokens_seen": 506068992
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008551855566700101,
      "loss": 1.613,
      "theoretical_loss": 3.9381891101858475,
      "tokens_seen": 506134528
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008551654964894684,
      "loss": 1.7528,
      "theoretical_loss": 3.9381347017499326,
      "tokens_seen": 506200064
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008551454363089268,
      "loss": 1.8456,
      "theoretical_loss": 3.938080302329687,
      "tokens_seen": 506265600
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008551253761283851,
      "loss": 1.7717,
      "theoretical_loss": 3.938025911922449,
      "tokens_seen": 506331136
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008551053159478436,
      "loss": 1.5916,
      "theoretical_loss": 3.9379715305255605,
      "tokens_seen": 506396672
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008550852557673019,
      "loss": 1.6501,
      "theoretical_loss": 3.9379171581363623,
      "tokens_seen": 506462208
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008550651955867603,
      "loss": 1.7902,
      "theoretical_loss": 3.9378627947521965,
      "tokens_seen": 506527744
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008550451354062186,
      "loss": 1.595,
      "theoretical_loss": 3.937808440370408,
      "tokens_seen": 506593280
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008550250752256771,
      "loss": 1.672,
      "theoretical_loss": 3.9377540949883416,
      "tokens_seen": 506658816
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008550050150451354,
      "loss": 1.7568,
      "theoretical_loss": 3.937699758603342,
      "tokens_seen": 506724352
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008549849548645938,
      "loss": 1.6198,
      "theoretical_loss": 3.9376454312127587,
      "tokens_seen": 506789888
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008549648946840522,
      "loss": 1.797,
      "theoretical_loss": 3.9375911128139376,
      "tokens_seen": 506855424
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008549448345035106,
      "loss": 1.7229,
      "theoretical_loss": 3.9375368034042295,
      "tokens_seen": 506920960
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008549247743229689,
      "loss": 1.6817,
      "theoretical_loss": 3.937482502980985,
      "tokens_seen": 506986496
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008549047141424273,
      "loss": 1.7819,
      "theoretical_loss": 3.9374282115415546,
      "tokens_seen": 507052032
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008548846539618856,
      "loss": 1.6991,
      "theoretical_loss": 3.9373739290832925,
      "tokens_seen": 507117568
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008548645937813441,
      "loss": 1.7333,
      "theoretical_loss": 3.937319655603552,
      "tokens_seen": 507183104
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008548445336008024,
      "loss": 1.8329,
      "theoretical_loss": 3.937265391099688,
      "tokens_seen": 507248640
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008548244734202608,
      "loss": 1.5592,
      "theoretical_loss": 3.937211135569057,
      "tokens_seen": 507314176
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008548044132397192,
      "loss": 1.6507,
      "theoretical_loss": 3.9371568890090165,
      "tokens_seen": 507379712
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008547843530591775,
      "loss": 1.6943,
      "theoretical_loss": 3.937102651416924,
      "tokens_seen": 507445248
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008547642928786359,
      "loss": 1.6051,
      "theoretical_loss": 3.9370484227901397,
      "tokens_seen": 507510784
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008547442326980944,
      "loss": 1.6238,
      "theoretical_loss": 3.936994203126024,
      "tokens_seen": 507576320
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008547241725175526,
      "loss": 1.6631,
      "theoretical_loss": 3.936939992421939,
      "tokens_seen": 507641856
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008547041123370111,
      "loss": 1.6778,
      "theoretical_loss": 3.936885790675247,
      "tokens_seen": 507707392
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008546840521564693,
      "loss": 1.5877,
      "theoretical_loss": 3.9368315978833124,
      "tokens_seen": 507772928
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008546639919759278,
      "loss": 1.7614,
      "theoretical_loss": 3.9367774140435,
      "tokens_seen": 507838464
    },
    {
      "epoch": 0.15,
      "objective/train/advantage_avg": 0.49557170271873474,
      "objective/train/docs_used": 191028,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5263073444366455,
      "objective/train/original_loss": 1.5263073444366455,
      "objective/train/theoretical_loss": 3.936723239153177,
      "objective/train/tokens_used": 528364000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24599750339984894,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508077144622803,
      "objective/train/weighted_lm_loss": 1.6038414239883423,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.936723239153177,
      "tokens_seen": 507904000
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008546439317953862,
      "loss": 1.5987,
      "theoretical_loss": 3.936723239153177,
      "tokens_seen": 507904000
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008546238716148445,
      "loss": 1.6881,
      "theoretical_loss": 3.9366690732097096,
      "tokens_seen": 507969536
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008546038114343029,
      "loss": 1.5569,
      "theoretical_loss": 3.936614916210466,
      "tokens_seen": 508035072
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008545837512537614,
      "loss": 1.627,
      "theoretical_loss": 3.936560768152817,
      "tokens_seen": 508100608
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008545636910732196,
      "loss": 1.5718,
      "theoretical_loss": 3.936506629034133,
      "tokens_seen": 508166144
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008545436308926781,
      "loss": 1.6788,
      "theoretical_loss": 3.9364524988517844,
      "tokens_seen": 508231680
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008545235707121363,
      "loss": 1.6036,
      "theoretical_loss": 3.9363983776031457,
      "tokens_seen": 508297216
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008545035105315948,
      "loss": 1.6642,
      "theoretical_loss": 3.936344265285591,
      "tokens_seen": 508362752
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008544834503510532,
      "loss": 1.6405,
      "theoretical_loss": 3.936290161896494,
      "tokens_seen": 508428288
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008544633901705115,
      "loss": 1.6672,
      "theoretical_loss": 3.9362360674332315,
      "tokens_seen": 508493824
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008544433299899699,
      "loss": 1.7934,
      "theoretical_loss": 3.936181981893182,
      "tokens_seen": 508559360
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008544232698094283,
      "loss": 1.7661,
      "theoretical_loss": 3.9361279052737217,
      "tokens_seen": 508624896
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008544032096288866,
      "loss": 1.6901,
      "theoretical_loss": 3.9360738375722324,
      "tokens_seen": 508690432
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008543831494483451,
      "loss": 1.8077,
      "theoretical_loss": 3.9360197787860938,
      "tokens_seen": 508755968
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008543630892678033,
      "loss": 1.6619,
      "theoretical_loss": 3.9359657289126875,
      "tokens_seen": 508821504
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008543430290872618,
      "loss": 1.6664,
      "theoretical_loss": 3.9359116879493965,
      "tokens_seen": 508887040
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008543229689067202,
      "loss": 1.6839,
      "theoretical_loss": 3.9358576558936047,
      "tokens_seen": 508952576
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008543029087261785,
      "loss": 1.71,
      "theoretical_loss": 3.9358036327426977,
      "tokens_seen": 509018112
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000854282848545637,
      "loss": 1.6584,
      "theoretical_loss": 3.935749618494061,
      "tokens_seen": 509083648
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008542627883650953,
      "loss": 1.5792,
      "theoretical_loss": 3.935695613145082,
      "tokens_seen": 509149184
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008542427281845537,
      "loss": 1.6885,
      "theoretical_loss": 3.9356416166931503,
      "tokens_seen": 509214720
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008542226680040121,
      "loss": 1.7117,
      "theoretical_loss": 3.9355876291356533,
      "tokens_seen": 509280256
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008542026078234705,
      "loss": 1.6857,
      "theoretical_loss": 3.935533650469983,
      "tokens_seen": 509345792
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008541825476429288,
      "loss": 1.595,
      "theoretical_loss": 3.9354796806935313,
      "tokens_seen": 509411328
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008541624874623872,
      "loss": 1.6484,
      "theoretical_loss": 3.93542571980369,
      "tokens_seen": 509476864
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008541424272818455,
      "loss": 1.6633,
      "theoretical_loss": 3.9353717677978537,
      "tokens_seen": 509542400
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000854122367101304,
      "loss": 1.6041,
      "theoretical_loss": 3.935317824673417,
      "tokens_seen": 509607936
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008541023069207623,
      "loss": 1.715,
      "theoretical_loss": 3.935263890427777,
      "tokens_seen": 509673472
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008540822467402207,
      "loss": 1.5171,
      "theoretical_loss": 3.935209965058329,
      "tokens_seen": 509739008
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000854062186559679,
      "loss": 1.6945,
      "theoretical_loss": 3.9351560485624733,
      "tokens_seen": 509804544
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008540421263791375,
      "loss": 1.7677,
      "theoretical_loss": 3.935102140937608,
      "tokens_seen": 509870080
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008540220661985958,
      "loss": 1.7678,
      "theoretical_loss": 3.935048242181134,
      "tokens_seen": 509935616
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008540020060180542,
      "loss": 1.5264,
      "theoretical_loss": 3.9349943522904525,
      "tokens_seen": 510001152
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008539819458375126,
      "loss": 1.7062,
      "theoretical_loss": 3.934940471262967,
      "tokens_seen": 510066688
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000853961885656971,
      "loss": 1.5977,
      "theoretical_loss": 3.934886599096081,
      "tokens_seen": 510132224
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008539418254764293,
      "loss": 1.6127,
      "theoretical_loss": 3.9348327357871984,
      "tokens_seen": 510197760
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008539217652958877,
      "loss": 1.5628,
      "theoretical_loss": 3.9347788813337266,
      "tokens_seen": 510263296
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000853901705115346,
      "loss": 1.7656,
      "theoretical_loss": 3.934725035733072,
      "tokens_seen": 510328832
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008538816449348045,
      "loss": 1.6011,
      "theoretical_loss": 3.9346711989826426,
      "tokens_seen": 510394368
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008538615847542628,
      "loss": 1.6009,
      "theoretical_loss": 3.934617371079848,
      "tokens_seen": 510459904
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008538415245737212,
      "loss": 1.6237,
      "theoretical_loss": 3.934563552022098,
      "tokens_seen": 510525440
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008538214643931796,
      "loss": 1.736,
      "theoretical_loss": 3.9345097418068047,
      "tokens_seen": 510590976
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008538014042126379,
      "loss": 1.6643,
      "theoretical_loss": 3.93445594043138,
      "tokens_seen": 510656512
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008537813440320963,
      "loss": 1.6711,
      "theoretical_loss": 3.9344021478932376,
      "tokens_seen": 510722048
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008537612838515548,
      "loss": 1.6318,
      "theoretical_loss": 3.934348364189793,
      "tokens_seen": 510787584
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.000853741223671013,
      "loss": 1.5485,
      "theoretical_loss": 3.93429458931846,
      "tokens_seen": 510853120
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008537211634904715,
      "loss": 1.6557,
      "theoretical_loss": 3.9342408232766584,
      "tokens_seen": 510918656
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008537011033099297,
      "loss": 1.4878,
      "theoretical_loss": 3.9341870660618037,
      "tokens_seen": 510984192
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008536810431293882,
      "loss": 1.7565,
      "theoretical_loss": 3.9341333176713156,
      "tokens_seen": 511049728
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008536609829488466,
      "loss": 1.6649,
      "theoretical_loss": 3.9340795781026148,
      "tokens_seen": 511115264
    },
    {
      "epoch": 0.15,
      "objective/train/advantage_avg": 0.4954502582550049,
      "objective/train/docs_used": 192198,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6345678567886353,
      "objective/train/original_loss": 1.6345679759979248,
      "objective/train/theoretical_loss": 3.934025847353122,
      "objective/train/tokens_used": 531640800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24563312530517578,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507938861846924,
      "objective/train/weighted_lm_loss": 1.7175711393356323,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.934025847353122,
      "tokens_seen": 511180800
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008536409227683049,
      "loss": 1.755,
      "theoretical_loss": 3.934025847353122,
      "tokens_seen": 511180800
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008536208625877633,
      "loss": 1.5519,
      "theoretical_loss": 3.933972125420259,
      "tokens_seen": 511246336
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008536008024072218,
      "loss": 1.6695,
      "theoretical_loss": 3.93391841230145,
      "tokens_seen": 511311872
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00085358074222668,
      "loss": 1.6085,
      "theoretical_loss": 3.933864707994119,
      "tokens_seen": 511377408
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0008535606820461385,
      "loss": 1.6707,
      "theoretical_loss": 3.9338110124956924,
      "tokens_seen": 511442944
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008535406218655967,
      "loss": 1.65,
      "theoretical_loss": 3.933757325803596,
      "tokens_seen": 511508480
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008535205616850552,
      "loss": 1.5947,
      "theoretical_loss": 3.933703647915258,
      "tokens_seen": 511574016
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008535005015045136,
      "loss": 1.6617,
      "theoretical_loss": 3.9336499788281056,
      "tokens_seen": 511639552
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008534804413239719,
      "loss": 1.7443,
      "theoretical_loss": 3.9335963185395713,
      "tokens_seen": 511705088
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008534603811434303,
      "loss": 1.4573,
      "theoretical_loss": 3.9335426670470834,
      "tokens_seen": 511770624
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008534403209628887,
      "loss": 1.6172,
      "theoretical_loss": 3.933489024348076,
      "tokens_seen": 511836160
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000853420260782347,
      "loss": 1.7263,
      "theoretical_loss": 3.933435390439981,
      "tokens_seen": 511901696
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008534002006018055,
      "loss": 1.6474,
      "theoretical_loss": 3.933381765320233,
      "tokens_seen": 511967232
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008533801404212637,
      "loss": 1.7552,
      "theoretical_loss": 3.9333281489862673,
      "tokens_seen": 512032768
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008533600802407222,
      "loss": 1.5181,
      "theoretical_loss": 3.9332745414355204,
      "tokens_seen": 512098304
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008533400200601805,
      "loss": 1.6143,
      "theoretical_loss": 3.9332209426654288,
      "tokens_seen": 512163840
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008533199598796389,
      "loss": 1.5941,
      "theoretical_loss": 3.933167352673432,
      "tokens_seen": 512229376
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008532998996990973,
      "loss": 1.6446,
      "theoretical_loss": 3.933113771456969,
      "tokens_seen": 512294912
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008532798395185557,
      "loss": 1.8387,
      "theoretical_loss": 3.9330601990134806,
      "tokens_seen": 512360448
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000853259779338014,
      "loss": 1.5672,
      "theoretical_loss": 3.9330066353404085,
      "tokens_seen": 512425984
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008532397191574725,
      "loss": 1.602,
      "theoretical_loss": 3.9329530804351958,
      "tokens_seen": 512491520
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008532196589769308,
      "loss": 1.6394,
      "theoretical_loss": 3.932899534295286,
      "tokens_seen": 512557056
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008531995987963892,
      "loss": 1.7659,
      "theoretical_loss": 3.932845996918124,
      "tokens_seen": 512622592
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008531795386158475,
      "loss": 1.7233,
      "theoretical_loss": 3.9327924683011557,
      "tokens_seen": 512688128
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000853159478435306,
      "loss": 1.6986,
      "theoretical_loss": 3.9327389484418287,
      "tokens_seen": 512753664
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008531394182547643,
      "loss": 1.5626,
      "theoretical_loss": 3.9326854373375904,
      "tokens_seen": 512819200
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008531193580742227,
      "loss": 1.5647,
      "theoretical_loss": 3.9326319349858903,
      "tokens_seen": 512884736
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000853099297893681,
      "loss": 1.6313,
      "theoretical_loss": 3.9325784413841793,
      "tokens_seen": 512950272
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008530792377131394,
      "loss": 1.6366,
      "theoretical_loss": 3.9325249565299076,
      "tokens_seen": 513015808
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008530591775325978,
      "loss": 1.7395,
      "theoretical_loss": 3.932471480420528,
      "tokens_seen": 513081344
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008530391173520562,
      "loss": 1.7295,
      "theoretical_loss": 3.932418013053495,
      "tokens_seen": 513146880
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008530190571715145,
      "loss": 1.5878,
      "theoretical_loss": 3.932364554426262,
      "tokens_seen": 513212416
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000852998996990973,
      "loss": 1.6803,
      "theoretical_loss": 3.932311104536285,
      "tokens_seen": 513277952
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008529789368104312,
      "loss": 1.6529,
      "theoretical_loss": 3.9322576633810202,
      "tokens_seen": 513343488
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008529588766298897,
      "loss": 1.5544,
      "theoretical_loss": 3.9322042309579257,
      "tokens_seen": 513409024
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000852938816449348,
      "loss": 1.6762,
      "theoretical_loss": 3.9321508072644606,
      "tokens_seen": 513474560
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008529187562688064,
      "loss": 1.6738,
      "theoretical_loss": 3.9320973922980844,
      "tokens_seen": 513540096
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008528986960882648,
      "loss": 1.7436,
      "theoretical_loss": 3.932043986056258,
      "tokens_seen": 513605632
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008528786359077232,
      "loss": 1.7212,
      "theoretical_loss": 3.9319905885364443,
      "tokens_seen": 513671168
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008528585757271815,
      "loss": 1.5518,
      "theoretical_loss": 3.9319371997361054,
      "tokens_seen": 513736704
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00085283851554664,
      "loss": 1.7026,
      "theoretical_loss": 3.931883819652705,
      "tokens_seen": 513802240
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008528184553660983,
      "loss": 1.6241,
      "theoretical_loss": 3.9318304482837094,
      "tokens_seen": 513867776
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008527983951855567,
      "loss": 1.5239,
      "theoretical_loss": 3.931777085626584,
      "tokens_seen": 513933312
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008527783350050152,
      "loss": 1.6441,
      "theoretical_loss": 3.9317237316787965,
      "tokens_seen": 513998848
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008527582748244734,
      "loss": 1.7269,
      "theoretical_loss": 3.9316703864378155,
      "tokens_seen": 514064384
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008527382146439319,
      "loss": 1.636,
      "theoretical_loss": 3.9316170499011096,
      "tokens_seen": 514129920
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008527181544633901,
      "loss": 1.5734,
      "theoretical_loss": 3.93156372206615,
      "tokens_seen": 514195456
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008526980942828486,
      "loss": 1.6645,
      "theoretical_loss": 3.9315104029304084,
      "tokens_seen": 514260992
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000852678034102307,
      "loss": 1.641,
      "theoretical_loss": 3.9314570924913568,
      "tokens_seen": 514326528
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008526579739217653,
      "loss": 1.6756,
      "theoretical_loss": 3.9314037907464687,
      "tokens_seen": 514392064
    },
    {
      "epoch": 0.16,
      "objective/train/advantage_avg": 0.49453940987586975,
      "objective/train/docs_used": 193343,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4929040670394897,
      "objective/train/original_loss": 1.4929039478302002,
      "objective/train/theoretical_loss": 3.931350497693219,
      "objective/train/tokens_used": 534917600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24488162994384766,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506988763809204,
      "objective/train/weighted_lm_loss": 1.568619966506958,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.931350497693219,
      "tokens_seen": 514457600
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008526379137412237,
      "loss": 1.6691,
      "theoretical_loss": 3.931350497693219,
      "tokens_seen": 514457600
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008526178535606822,
      "loss": 1.681,
      "theoretical_loss": 3.9312972133290844,
      "tokens_seen": 514523136
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008525977933801404,
      "loss": 1.6421,
      "theoretical_loss": 3.9312439376515407,
      "tokens_seen": 514588672
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008525777331995989,
      "loss": 1.7608,
      "theoretical_loss": 3.931190670658066,
      "tokens_seen": 514654208
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008525576730190571,
      "loss": 1.584,
      "theoretical_loss": 3.9311374123461387,
      "tokens_seen": 514719744
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008525376128385156,
      "loss": 1.5738,
      "theoretical_loss": 3.9310841627132396,
      "tokens_seen": 514785280
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000852517552657974,
      "loss": 1.7031,
      "theoretical_loss": 3.9310309217568493,
      "tokens_seen": 514850816
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008524974924774323,
      "loss": 1.641,
      "theoretical_loss": 3.93097768947445,
      "tokens_seen": 514916352
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008524774322968907,
      "loss": 1.6874,
      "theoretical_loss": 3.9309244658635247,
      "tokens_seen": 514981888
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008524573721163491,
      "loss": 1.6679,
      "theoretical_loss": 3.930871250921558,
      "tokens_seen": 515047424
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008524373119358074,
      "loss": 1.7494,
      "theoretical_loss": 3.9308180446460343,
      "tokens_seen": 515112960
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008524172517552659,
      "loss": 1.6921,
      "theoretical_loss": 3.9307648470344407,
      "tokens_seen": 515178496
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008523971915747241,
      "loss": 1.5922,
      "theoretical_loss": 3.930711658084264,
      "tokens_seen": 515244032
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008523771313941826,
      "loss": 1.6567,
      "theoretical_loss": 3.930658477792993,
      "tokens_seen": 515309568
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008523570712136409,
      "loss": 1.7634,
      "theoretical_loss": 3.9306053061581165,
      "tokens_seen": 515375104
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008523370110330993,
      "loss": 1.6344,
      "theoretical_loss": 3.9305521431771253,
      "tokens_seen": 515440640
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008523169508525577,
      "loss": 1.6002,
      "theoretical_loss": 3.9304989888475115,
      "tokens_seen": 515506176
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008522968906720161,
      "loss": 1.7391,
      "theoretical_loss": 3.9304458431667664,
      "tokens_seen": 515571712
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008522768304914744,
      "loss": 1.6446,
      "theoretical_loss": 3.930392706132385,
      "tokens_seen": 515637248
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008522567703109329,
      "loss": 1.7342,
      "theoretical_loss": 3.9303395777418606,
      "tokens_seen": 515702784
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008522367101303912,
      "loss": 1.6347,
      "theoretical_loss": 3.93028645799269,
      "tokens_seen": 515768320
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008522166499498496,
      "loss": 1.6956,
      "theoretical_loss": 3.9302333468823694,
      "tokens_seen": 515833856
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008521965897693079,
      "loss": 1.654,
      "theoretical_loss": 3.9301802444083966,
      "tokens_seen": 515899392
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008521765295887663,
      "loss": 1.6685,
      "theoretical_loss": 3.930127150568271,
      "tokens_seen": 515964928
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008521564694082247,
      "loss": 1.7572,
      "theoretical_loss": 3.9300740653594914,
      "tokens_seen": 516030464
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008521364092276831,
      "loss": 1.6354,
      "theoretical_loss": 3.9300209887795594,
      "tokens_seen": 516096000
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008521163490471414,
      "loss": 1.6736,
      "theoretical_loss": 3.929967920825977,
      "tokens_seen": 516161536
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008520962888665998,
      "loss": 1.6179,
      "theoretical_loss": 3.929914861496247,
      "tokens_seen": 516227072
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008520762286860582,
      "loss": 1.57,
      "theoretical_loss": 3.9298618107878727,
      "tokens_seen": 516292608
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008520561685055166,
      "loss": 1.5583,
      "theoretical_loss": 3.9298087686983605,
      "tokens_seen": 516358144
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008520361083249749,
      "loss": 1.6439,
      "theoretical_loss": 3.929755735225216,
      "tokens_seen": 516423680
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008520160481444334,
      "loss": 1.6558,
      "theoretical_loss": 3.929702710365946,
      "tokens_seen": 516489216
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008519959879638916,
      "loss": 1.6313,
      "theoretical_loss": 3.9296496941180585,
      "tokens_seen": 516554752
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008519759277833501,
      "loss": 1.6295,
      "theoretical_loss": 3.929596686479064,
      "tokens_seen": 516620288
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008519558676028084,
      "loss": 1.683,
      "theoretical_loss": 3.9295436874464715,
      "tokens_seen": 516685824
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008519358074222668,
      "loss": 1.6023,
      "theoretical_loss": 3.9294906970177923,
      "tokens_seen": 516751360
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008519157472417252,
      "loss": 1.5942,
      "theoretical_loss": 3.9294377151905397,
      "tokens_seen": 516816896
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008518956870611836,
      "loss": 1.6842,
      "theoretical_loss": 3.929384741962226,
      "tokens_seen": 516882432
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008518756268806419,
      "loss": 1.7586,
      "theoretical_loss": 3.929331777330366,
      "tokens_seen": 516947968
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008518555667001004,
      "loss": 1.5151,
      "theoretical_loss": 3.929278821292476,
      "tokens_seen": 517013504
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008518355065195586,
      "loss": 1.5127,
      "theoretical_loss": 3.929225873846071,
      "tokens_seen": 517079040
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008518154463390171,
      "loss": 1.6306,
      "theoretical_loss": 3.929172934988669,
      "tokens_seen": 517144576
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008517953861584754,
      "loss": 1.6748,
      "theoretical_loss": 3.9291200047177886,
      "tokens_seen": 517210112
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008517753259779338,
      "loss": 1.574,
      "theoretical_loss": 3.9290670830309495,
      "tokens_seen": 517275648
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008517552657973922,
      "loss": 1.6791,
      "theoretical_loss": 3.929014169925672,
      "tokens_seen": 517341184
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008517352056168505,
      "loss": 1.6078,
      "theoretical_loss": 3.928961265399478,
      "tokens_seen": 517406720
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008517151454363089,
      "loss": 1.6236,
      "theoretical_loss": 3.9289083694498905,
      "tokens_seen": 517472256
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008516950852557674,
      "loss": 1.4969,
      "theoretical_loss": 3.9288554820744324,
      "tokens_seen": 517537792
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008516750250752256,
      "loss": 1.6019,
      "theoretical_loss": 3.9288026032706282,
      "tokens_seen": 517603328
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008516549648946841,
      "loss": 1.6628,
      "theoretical_loss": 3.928749733036005,
      "tokens_seen": 517668864
    },
    {
      "epoch": 0.16,
      "objective/train/advantage_avg": 0.49043938517570496,
      "objective/train/docs_used": 194581,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5922307968139648,
      "objective/train/original_loss": 1.5922307968139648,
      "objective/train/theoretical_loss": 3.9286968713680883,
      "objective/train/tokens_used": 538194400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24175012111663818,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502729415893555,
      "objective/train/weighted_lm_loss": 1.672400951385498,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9286968713680883,
      "tokens_seen": 517734400
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008516349047141423,
      "loss": 1.6263,
      "theoretical_loss": 3.9286968713680883,
      "tokens_seen": 517734400
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008516148445336008,
      "loss": 1.64,
      "theoretical_loss": 3.9286440182644062,
      "tokens_seen": 517799936
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008515947843530592,
      "loss": 1.6413,
      "theoretical_loss": 3.928591173722488,
      "tokens_seen": 517865472
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008515747241725175,
      "loss": 1.6881,
      "theoretical_loss": 3.9285383377398624,
      "tokens_seen": 517931008
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008515546639919759,
      "loss": 1.5773,
      "theoretical_loss": 3.9284855103140615,
      "tokens_seen": 517996544
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008515346038114344,
      "loss": 1.5778,
      "theoretical_loss": 3.9284326914426164,
      "tokens_seen": 518062080
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008515145436308926,
      "loss": 1.6953,
      "theoretical_loss": 3.9283798811230604,
      "tokens_seen": 518127616
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008514944834503511,
      "loss": 1.6031,
      "theoretical_loss": 3.9283270793529272,
      "tokens_seen": 518193152
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008514744232698094,
      "loss": 1.6565,
      "theoretical_loss": 3.9282742861297524,
      "tokens_seen": 518258688
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008514543630892678,
      "loss": 1.5558,
      "theoretical_loss": 3.9282215014510706,
      "tokens_seen": 518324224
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008514343029087262,
      "loss": 1.5033,
      "theoretical_loss": 3.9281687253144204,
      "tokens_seen": 518389760
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008514142427281845,
      "loss": 1.6238,
      "theoretical_loss": 3.9281159577173392,
      "tokens_seen": 518455296
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008513941825476429,
      "loss": 1.579,
      "theoretical_loss": 3.928063198657365,
      "tokens_seen": 518520832
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008513741223671013,
      "loss": 1.6135,
      "theoretical_loss": 3.9280104481320395,
      "tokens_seen": 518586368
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008513540621865596,
      "loss": 1.6128,
      "theoretical_loss": 3.927957706138903,
      "tokens_seen": 518651904
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008513340020060181,
      "loss": 1.6363,
      "theoretical_loss": 3.9279049726754973,
      "tokens_seen": 518717440
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008513139418254765,
      "loss": 1.573,
      "theoretical_loss": 3.9278522477393656,
      "tokens_seen": 518782976
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008512938816449348,
      "loss": 1.7497,
      "theoretical_loss": 3.927799531328053,
      "tokens_seen": 518848512
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008512738214643933,
      "loss": 1.5822,
      "theoretical_loss": 3.9277468234391035,
      "tokens_seen": 518914048
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008512537612838516,
      "loss": 1.565,
      "theoretical_loss": 3.9276941240700642,
      "tokens_seen": 518979584
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00085123370110331,
      "loss": 1.6436,
      "theoretical_loss": 3.9276414332184815,
      "tokens_seen": 519045120
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008512136409227683,
      "loss": 1.6952,
      "theoretical_loss": 3.9275887508819043,
      "tokens_seen": 519110656
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008511935807422267,
      "loss": 1.5753,
      "theoretical_loss": 3.927536077057881,
      "tokens_seen": 519176192
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008511735205616851,
      "loss": 1.5902,
      "theoretical_loss": 3.927483411743962,
      "tokens_seen": 519241728
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008511534603811435,
      "loss": 1.6483,
      "theoretical_loss": 3.927430754937699,
      "tokens_seen": 519307264
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008511334002006018,
      "loss": 1.5533,
      "theoretical_loss": 3.9273781066366444,
      "tokens_seen": 519372800
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008511133400200602,
      "loss": 1.6394,
      "theoretical_loss": 3.9273254668383513,
      "tokens_seen": 519438336
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008510932798395186,
      "loss": 1.5954,
      "theoretical_loss": 3.9272728355403737,
      "tokens_seen": 519503872
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000851073219658977,
      "loss": 1.5294,
      "theoretical_loss": 3.927220212740267,
      "tokens_seen": 519569408
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008510531594784353,
      "loss": 1.5487,
      "theoretical_loss": 3.927167598435587,
      "tokens_seen": 519634944
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008510330992978938,
      "loss": 1.5729,
      "theoretical_loss": 3.9271149926238924,
      "tokens_seen": 519700480
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000851013039117352,
      "loss": 1.5498,
      "theoretical_loss": 3.92706239530274,
      "tokens_seen": 519766016
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008509929789368105,
      "loss": 1.6295,
      "theoretical_loss": 3.9270098064696906,
      "tokens_seen": 519831552
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008509729187562688,
      "loss": 1.6667,
      "theoretical_loss": 3.9269572261223034,
      "tokens_seen": 519897088
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008509528585757272,
      "loss": 1.719,
      "theoretical_loss": 3.9269046542581396,
      "tokens_seen": 519962624
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008509327983951856,
      "loss": 1.5316,
      "theoretical_loss": 3.926852090874763,
      "tokens_seen": 520028160
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000850912738214644,
      "loss": 1.5447,
      "theoretical_loss": 3.9267995359697356,
      "tokens_seen": 520093696
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008508926780341023,
      "loss": 1.5462,
      "theoretical_loss": 3.9267469895406224,
      "tokens_seen": 520159232
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008508726178535608,
      "loss": 1.713,
      "theoretical_loss": 3.9266944515849884,
      "tokens_seen": 520224768
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000850852557673019,
      "loss": 1.619,
      "theoretical_loss": 3.9266419221004005,
      "tokens_seen": 520290304
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008508324974924775,
      "loss": 1.6332,
      "theoretical_loss": 3.926589401084426,
      "tokens_seen": 520355840
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008508124373119358,
      "loss": 1.5951,
      "theoretical_loss": 3.9265368885346326,
      "tokens_seen": 520421376
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008507923771313942,
      "loss": 1.5693,
      "theoretical_loss": 3.9264843844485906,
      "tokens_seen": 520486912
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008507723169508526,
      "loss": 1.674,
      "theoretical_loss": 3.9264318888238705,
      "tokens_seen": 520552448
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008507522567703109,
      "loss": 1.5264,
      "theoretical_loss": 3.9263794016580427,
      "tokens_seen": 520617984
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008507321965897693,
      "loss": 1.6321,
      "theoretical_loss": 3.9263269229486797,
      "tokens_seen": 520683520
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008507121364092278,
      "loss": 1.597,
      "theoretical_loss": 3.9262744526933564,
      "tokens_seen": 520749056
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000850692076228686,
      "loss": 1.5582,
      "theoretical_loss": 3.926221990889646,
      "tokens_seen": 520814592
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008506720160481445,
      "loss": 1.5381,
      "theoretical_loss": 3.9261695375351238,
      "tokens_seen": 520880128
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008506519558676027,
      "loss": 1.6321,
      "theoretical_loss": 3.9261170926273667,
      "tokens_seen": 520945664
    },
    {
      "epoch": 0.16,
      "objective/train/advantage_avg": 0.4877818822860718,
      "objective/train/docs_used": 195804,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6571053266525269,
      "objective/train/original_loss": 1.6571054458618164,
      "objective/train/theoretical_loss": 3.926064656163952,
      "objective/train/tokens_used": 541471200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2390846312046051,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0499935150146484,
      "objective/train/weighted_lm_loss": 1.7400037050247192,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.926064656163952,
      "tokens_seen": 521011200
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008506318956870612,
      "loss": 1.681,
      "theoretical_loss": 3.926064656163952,
      "tokens_seen": 521011200
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008506118355065196,
      "loss": 1.6877,
      "theoretical_loss": 3.926012228142458,
      "tokens_seen": 521076736
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008505917753259779,
      "loss": 1.6522,
      "theoretical_loss": 3.9259598085604646,
      "tokens_seen": 521142272
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008505717151454363,
      "loss": 1.5564,
      "theoretical_loss": 3.9259073974155516,
      "tokens_seen": 521207808
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008505516549648948,
      "loss": 1.6742,
      "theoretical_loss": 3.9258549947053005,
      "tokens_seen": 521273344
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000850531594784353,
      "loss": 1.5631,
      "theoretical_loss": 3.9258026004272946,
      "tokens_seen": 521338880
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008505115346038115,
      "loss": 1.6367,
      "theoretical_loss": 3.925750214579116,
      "tokens_seen": 521404416
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008504914744232698,
      "loss": 1.7537,
      "theoretical_loss": 3.9256978371583493,
      "tokens_seen": 521469952
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008504714142427282,
      "loss": 1.5956,
      "theoretical_loss": 3.925645468162581,
      "tokens_seen": 521535488
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008504513540621866,
      "loss": 1.8158,
      "theoretical_loss": 3.925593107589397,
      "tokens_seen": 521601024
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000850431293881645,
      "loss": 1.5445,
      "theoretical_loss": 3.9255407554363835,
      "tokens_seen": 521666560
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008504112337011033,
      "loss": 1.7284,
      "theoretical_loss": 3.9254884117011306,
      "tokens_seen": 521732096
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008503911735205617,
      "loss": 1.5485,
      "theoretical_loss": 3.925436076381227,
      "tokens_seen": 521797632
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00085037111334002,
      "loss": 1.6211,
      "theoretical_loss": 3.925383749474263,
      "tokens_seen": 521863168
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008503510531594785,
      "loss": 1.5507,
      "theoretical_loss": 3.92533143097783,
      "tokens_seen": 521928704
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008503309929789368,
      "loss": 1.5523,
      "theoretical_loss": 3.9252791208895204,
      "tokens_seen": 521994240
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008503109327983952,
      "loss": 1.5814,
      "theoretical_loss": 3.9252268192069266,
      "tokens_seen": 522059776
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008502908726178536,
      "loss": 1.6998,
      "theoretical_loss": 3.9251745259276447,
      "tokens_seen": 522125312
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000850270812437312,
      "loss": 1.7234,
      "theoretical_loss": 3.9251222410492694,
      "tokens_seen": 522190848
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008502507522567703,
      "loss": 1.7072,
      "theoretical_loss": 3.925069964569396,
      "tokens_seen": 522256384
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008502306920762287,
      "loss": 1.6071,
      "theoretical_loss": 3.9250176964856234,
      "tokens_seen": 522321920
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000850210631895687,
      "loss": 1.7023,
      "theoretical_loss": 3.9249654367955484,
      "tokens_seen": 522387456
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008501905717151455,
      "loss": 1.5816,
      "theoretical_loss": 3.924913185496772,
      "tokens_seen": 522452992
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008501705115346038,
      "loss": 1.4997,
      "theoretical_loss": 3.9248609425868923,
      "tokens_seen": 522518528
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008501504513540622,
      "loss": 1.5808,
      "theoretical_loss": 3.9248087080635123,
      "tokens_seen": 522584064
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008501303911735205,
      "loss": 1.7183,
      "theoretical_loss": 3.9247564819242333,
      "tokens_seen": 522649600
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000850110330992979,
      "loss": 1.6624,
      "theoretical_loss": 3.924704264166659,
      "tokens_seen": 522715136
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008500902708124373,
      "loss": 1.6772,
      "theoretical_loss": 3.9246520547883934,
      "tokens_seen": 522780672
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008500702106318957,
      "loss": 1.7125,
      "theoretical_loss": 3.9245998537870426,
      "tokens_seen": 522846208
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000850050150451354,
      "loss": 1.5976,
      "theoretical_loss": 3.924547661160211,
      "tokens_seen": 522911744
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008500300902708124,
      "loss": 1.4942,
      "theoretical_loss": 3.9244954769055074,
      "tokens_seen": 522977280
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008500100300902708,
      "loss": 1.6971,
      "theoretical_loss": 3.9244433010205393,
      "tokens_seen": 523042816
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008499899699097292,
      "loss": 1.6379,
      "theoretical_loss": 3.9243911335029154,
      "tokens_seen": 523108352
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008499699097291875,
      "loss": 1.6794,
      "theoretical_loss": 3.924338974350247,
      "tokens_seen": 523173888
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000849949849548646,
      "loss": 1.6712,
      "theoretical_loss": 3.924286823560144,
      "tokens_seen": 523239424
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008499297893681043,
      "loss": 1.5988,
      "theoretical_loss": 3.9242346811302187,
      "tokens_seen": 523304960
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008499097291875627,
      "loss": 1.5973,
      "theoretical_loss": 3.924182547058085,
      "tokens_seen": 523370496
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000849889669007021,
      "loss": 1.5487,
      "theoretical_loss": 3.9241304213413564,
      "tokens_seen": 523436032
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008498696088264794,
      "loss": 1.6642,
      "theoretical_loss": 3.9240783039776472,
      "tokens_seen": 523501568
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008498495486459379,
      "loss": 1.6403,
      "theoretical_loss": 3.9240261949645747,
      "tokens_seen": 523567104
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008498294884653962,
      "loss": 1.7722,
      "theoretical_loss": 3.9239740942997554,
      "tokens_seen": 523632640
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008498094282848546,
      "loss": 1.5607,
      "theoretical_loss": 3.9239220019808068,
      "tokens_seen": 523698176
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000849789368104313,
      "loss": 1.5919,
      "theoretical_loss": 3.9238699180053485,
      "tokens_seen": 523763712
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008497693079237713,
      "loss": 1.6647,
      "theoretical_loss": 3.923817842371,
      "tokens_seen": 523829248
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008497492477432297,
      "loss": 1.7784,
      "theoretical_loss": 3.923765775075382,
      "tokens_seen": 523894784
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008497291875626882,
      "loss": 1.5481,
      "theoretical_loss": 3.9237137161161173,
      "tokens_seen": 523960320
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008497091273821464,
      "loss": 1.6527,
      "theoretical_loss": 3.923661665490828,
      "tokens_seen": 524025856
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008496890672016049,
      "loss": 1.6881,
      "theoretical_loss": 3.9236096231971374,
      "tokens_seen": 524091392
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008496690070210631,
      "loss": 1.6147,
      "theoretical_loss": 3.9235575892326713,
      "tokens_seen": 524156928
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008496489468405216,
      "loss": 1.6174,
      "theoretical_loss": 3.9235055635950555,
      "tokens_seen": 524222464
    },
    {
      "epoch": 0.16,
      "objective/train/advantage_avg": 0.49302032589912415,
      "objective/train/docs_used": 196964,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4459208250045776,
      "objective/train/original_loss": 1.4459209442138672,
      "objective/train/theoretical_loss": 3.9234535462819156,
      "objective/train/tokens_used": 544748000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24345248937606812,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505396127700806,
      "objective/train/weighted_lm_loss": 1.5191477537155151,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9234535462819156,
      "tokens_seen": 524288000
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00084962888665998,
      "loss": 1.5534,
      "theoretical_loss": 3.9234535462819156,
      "tokens_seen": 524288000
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008496088264794383,
      "loss": 1.6074,
      "theoretical_loss": 3.9234015372908804,
      "tokens_seen": 524353536
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008495887662988967,
      "loss": 1.6905,
      "theoretical_loss": 3.923349536619578,
      "tokens_seen": 524419072
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008495687061183552,
      "loss": 1.6289,
      "theoretical_loss": 3.923297544265639,
      "tokens_seen": 524484608
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008495486459378134,
      "loss": 1.591,
      "theoretical_loss": 3.923245560226693,
      "tokens_seen": 524550144
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008495285857572719,
      "loss": 1.5203,
      "theoretical_loss": 3.923193584500372,
      "tokens_seen": 524615680
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008495085255767302,
      "loss": 1.6374,
      "theoretical_loss": 3.9231416170843083,
      "tokens_seen": 524681216
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008494884653961886,
      "loss": 1.6512,
      "theoretical_loss": 3.9230896579761354,
      "tokens_seen": 524746752
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000849468405215647,
      "loss": 1.4736,
      "theoretical_loss": 3.9230377071734885,
      "tokens_seen": 524812288
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008494483450351053,
      "loss": 1.7909,
      "theoretical_loss": 3.922985764674002,
      "tokens_seen": 524877824
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008494282848545637,
      "loss": 1.6283,
      "theoretical_loss": 3.922933830475314,
      "tokens_seen": 524943360
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008494082246740221,
      "loss": 1.4642,
      "theoretical_loss": 3.9228819045750596,
      "tokens_seen": 525008896
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008493881644934804,
      "loss": 1.6107,
      "theoretical_loss": 3.9228299869708794,
      "tokens_seen": 525074432
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008493681043129389,
      "loss": 1.5806,
      "theoretical_loss": 3.9227780776604115,
      "tokens_seen": 525139968
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008493480441323972,
      "loss": 1.6152,
      "theoretical_loss": 3.922726176641297,
      "tokens_seen": 525205504
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008493279839518556,
      "loss": 1.6536,
      "theoretical_loss": 3.922674283911176,
      "tokens_seen": 525271040
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000849307923771314,
      "loss": 1.6945,
      "theoretical_loss": 3.9226223994676923,
      "tokens_seen": 525336576
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008492878635907724,
      "loss": 1.6635,
      "theoretical_loss": 3.9225705233084875,
      "tokens_seen": 525402112
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008492678034102307,
      "loss": 1.7497,
      "theoretical_loss": 3.922518655431207,
      "tokens_seen": 525467648
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008492477432296891,
      "loss": 1.6683,
      "theoretical_loss": 3.922466795833495,
      "tokens_seen": 525533184
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008492276830491474,
      "loss": 1.6898,
      "theoretical_loss": 3.9224149445129983,
      "tokens_seen": 525598720
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008492076228686059,
      "loss": 1.6656,
      "theoretical_loss": 3.922363101467364,
      "tokens_seen": 525664256
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008491875626880642,
      "loss": 1.5702,
      "theoretical_loss": 3.9223112666942397,
      "tokens_seen": 525729792
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008491675025075226,
      "loss": 1.5852,
      "theoretical_loss": 3.9222594401912745,
      "tokens_seen": 525795328
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008491474423269809,
      "loss": 1.6441,
      "theoretical_loss": 3.922207621956119,
      "tokens_seen": 525860864
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008491273821464394,
      "loss": 1.5662,
      "theoretical_loss": 3.9221558119864226,
      "tokens_seen": 525926400
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008491073219658977,
      "loss": 1.64,
      "theoretical_loss": 3.922104010279839,
      "tokens_seen": 525991936
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008490872617853561,
      "loss": 1.6588,
      "theoretical_loss": 3.922052216834019,
      "tokens_seen": 526057472
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008490672016048144,
      "loss": 1.5135,
      "theoretical_loss": 3.9220004316466186,
      "tokens_seen": 526123008
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008490471414242728,
      "loss": 1.7018,
      "theoretical_loss": 3.921948654715291,
      "tokens_seen": 526188544
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008490270812437312,
      "loss": 1.6147,
      "theoretical_loss": 3.9218968860376924,
      "tokens_seen": 526254080
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008490070210631896,
      "loss": 1.6883,
      "theoretical_loss": 3.9218451256114792,
      "tokens_seen": 526319616
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008489869608826479,
      "loss": 1.6665,
      "theoretical_loss": 3.9217933734343093,
      "tokens_seen": 526385152
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008489669007021064,
      "loss": 1.6813,
      "theoretical_loss": 3.9217416295038414,
      "tokens_seen": 526450688
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008489468405215647,
      "loss": 1.5609,
      "theoretical_loss": 3.921689893817735,
      "tokens_seen": 526516224
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008489267803410231,
      "loss": 1.6527,
      "theoretical_loss": 3.9216381663736506,
      "tokens_seen": 526581760
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008489067201604815,
      "loss": 1.6266,
      "theoretical_loss": 3.9215864471692488,
      "tokens_seen": 526647296
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008488866599799398,
      "loss": 1.6327,
      "theoretical_loss": 3.921534736202193,
      "tokens_seen": 526712832
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008488665997993982,
      "loss": 1.7006,
      "theoretical_loss": 3.921483033470146,
      "tokens_seen": 526778368
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008488465396188566,
      "loss": 1.5604,
      "theoretical_loss": 3.921431338970773,
      "tokens_seen": 526843904
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008488264794383149,
      "loss": 1.5901,
      "theoretical_loss": 3.921379652701738,
      "tokens_seen": 526909440
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008488064192577734,
      "loss": 1.6418,
      "theoretical_loss": 3.9213279746607084,
      "tokens_seen": 526974976
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008487863590772316,
      "loss": 1.5446,
      "theoretical_loss": 3.9212763048453505,
      "tokens_seen": 527040512
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008487662988966901,
      "loss": 1.6844,
      "theoretical_loss": 3.921224643253332,
      "tokens_seen": 527106048
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008487462387161485,
      "loss": 1.561,
      "theoretical_loss": 3.9211729898823235,
      "tokens_seen": 527171584
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008487261785356068,
      "loss": 1.584,
      "theoretical_loss": 3.9211213447299937,
      "tokens_seen": 527237120
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008487061183550652,
      "loss": 1.6675,
      "theoretical_loss": 3.921069707794014,
      "tokens_seen": 527302656
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008486860581745235,
      "loss": 1.6663,
      "theoretical_loss": 3.9210180790720566,
      "tokens_seen": 527368192
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008486659979939819,
      "loss": 1.6013,
      "theoretical_loss": 3.920966458561794,
      "tokens_seen": 527433728
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008486459378134404,
      "loss": 1.6349,
      "theoretical_loss": 3.9209148462609003,
      "tokens_seen": 527499264
    },
    {
      "epoch": 0.16,
      "objective/train/advantage_avg": 0.4926772713661194,
      "objective/train/docs_used": 198271,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.406691074371338,
      "objective/train/original_loss": 1.4066911935806274,
      "objective/train/theoretical_loss": 3.9208632421670497,
      "objective/train/tokens_used": 548024800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2433115839958191,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505046844482422,
      "objective/train/weighted_lm_loss": 1.4775863885879517,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9208632421670497,
      "tokens_seen": 527564800
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008486258776328986,
      "loss": 1.559,
      "theoretical_loss": 3.9208632421670497,
      "tokens_seen": 527564800
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008486058174523571,
      "loss": 1.6536,
      "theoretical_loss": 3.920811646277918,
      "tokens_seen": 527630336
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008485857572718155,
      "loss": 1.6171,
      "theoretical_loss": 3.920760058591182,
      "tokens_seen": 527695872
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008485656970912738,
      "loss": 1.6706,
      "theoretical_loss": 3.92070847910452,
      "tokens_seen": 527761408
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008485456369107322,
      "loss": 1.5576,
      "theoretical_loss": 3.9206569078156104,
      "tokens_seen": 527826944
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008485255767301906,
      "loss": 1.7217,
      "theoretical_loss": 3.920605344722132,
      "tokens_seen": 527892480
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008485055165496489,
      "loss": 1.6116,
      "theoretical_loss": 3.9205537898217644,
      "tokens_seen": 527958016
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008484854563691074,
      "loss": 1.6864,
      "theoretical_loss": 3.920502243112191,
      "tokens_seen": 528023552
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008484653961885656,
      "loss": 1.6173,
      "theoretical_loss": 3.920450704591093,
      "tokens_seen": 528089088
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008484453360080241,
      "loss": 1.657,
      "theoretical_loss": 3.920399174256154,
      "tokens_seen": 528154624
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008484252758274824,
      "loss": 1.5317,
      "theoretical_loss": 3.920347652105058,
      "tokens_seen": 528220160
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008484052156469408,
      "loss": 1.6393,
      "theoretical_loss": 3.92029613813549,
      "tokens_seen": 528285696
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008483851554663992,
      "loss": 1.6641,
      "theoretical_loss": 3.9202446323451365,
      "tokens_seen": 528351232
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008483650952858576,
      "loss": 1.6252,
      "theoretical_loss": 3.920193134731684,
      "tokens_seen": 528416768
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000848345035105316,
      "loss": 1.6232,
      "theoretical_loss": 3.920141645292821,
      "tokens_seen": 528482304
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008483249749247744,
      "loss": 1.5546,
      "theoretical_loss": 3.9200901640262362,
      "tokens_seen": 528547840
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008483049147442328,
      "loss": 1.6653,
      "theoretical_loss": 3.92003869092962,
      "tokens_seen": 528613376
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008482848545636911,
      "loss": 1.6492,
      "theoretical_loss": 3.919987226000662,
      "tokens_seen": 528678912
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008482647943831495,
      "loss": 1.5318,
      "theoretical_loss": 3.919935769237055,
      "tokens_seen": 528744448
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008482447342026078,
      "loss": 1.5988,
      "theoretical_loss": 3.9198843206364913,
      "tokens_seen": 528809984
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008482246740220663,
      "loss": 1.7496,
      "theoretical_loss": 3.919832880196664,
      "tokens_seen": 528875520
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008482046138415246,
      "loss": 1.6519,
      "theoretical_loss": 3.919781447915269,
      "tokens_seen": 528941056
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000848184553660983,
      "loss": 1.6635,
      "theoretical_loss": 3.91973002379,
      "tokens_seen": 529006592
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008481644934804413,
      "loss": 1.639,
      "theoretical_loss": 3.919678607818555,
      "tokens_seen": 529072128
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008481444332998998,
      "loss": 1.5742,
      "theoretical_loss": 3.9196271999986303,
      "tokens_seen": 529137664
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008481243731193581,
      "loss": 1.5977,
      "theoretical_loss": 3.919575800327925,
      "tokens_seen": 529203200
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008481043129388165,
      "loss": 1.6649,
      "theoretical_loss": 3.919524408804137,
      "tokens_seen": 529268736
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008480842527582748,
      "loss": 1.5606,
      "theoretical_loss": 3.919473025424969,
      "tokens_seen": 529334272
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008480641925777332,
      "loss": 1.6166,
      "theoretical_loss": 3.9194216501881187,
      "tokens_seen": 529399808
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008480441323971916,
      "loss": 1.5941,
      "theoretical_loss": 3.9193702830912907,
      "tokens_seen": 529465344
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00084802407221665,
      "loss": 1.7999,
      "theoretical_loss": 3.9193189241321873,
      "tokens_seen": 529530880
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008480040120361083,
      "loss": 1.5353,
      "theoretical_loss": 3.9192675733085123,
      "tokens_seen": 529596416
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008479839518555668,
      "loss": 1.6426,
      "theoretical_loss": 3.91921623061797,
      "tokens_seen": 529661952
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008479638916750251,
      "loss": 1.6522,
      "theoretical_loss": 3.919164896058267,
      "tokens_seen": 529727488
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008479438314944835,
      "loss": 1.6843,
      "theoretical_loss": 3.9191135696271098,
      "tokens_seen": 529793024
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008479237713139419,
      "loss": 1.6878,
      "theoretical_loss": 3.9190622513222055,
      "tokens_seen": 529858560
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008479037111334002,
      "loss": 1.7081,
      "theoretical_loss": 3.9190109411412632,
      "tokens_seen": 529924096
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008478836509528586,
      "loss": 1.6211,
      "theoretical_loss": 3.918959639081992,
      "tokens_seen": 529989632
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000847863590772317,
      "loss": 1.7312,
      "theoretical_loss": 3.9189083451421025,
      "tokens_seen": 530055168
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008478435305917753,
      "loss": 1.4901,
      "theoretical_loss": 3.918857059319306,
      "tokens_seen": 530120704
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008478234704112338,
      "loss": 1.4875,
      "theoretical_loss": 3.918805781611315,
      "tokens_seen": 530186240
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000847803410230692,
      "loss": 1.583,
      "theoretical_loss": 3.918754512015842,
      "tokens_seen": 530251776
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008477833500501505,
      "loss": 1.6291,
      "theoretical_loss": 3.9187032505306023,
      "tokens_seen": 530317312
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008477632898696089,
      "loss": 1.6093,
      "theoretical_loss": 3.91865199715331,
      "tokens_seen": 530382848
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008477432296890672,
      "loss": 1.5456,
      "theoretical_loss": 3.918600751881681,
      "tokens_seen": 530448384
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008477231695085256,
      "loss": 1.6374,
      "theoretical_loss": 3.9185495147134324,
      "tokens_seen": 530513920
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000847703109327984,
      "loss": 1.6139,
      "theoretical_loss": 3.918498285646282,
      "tokens_seen": 530579456
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008476830491474423,
      "loss": 1.6597,
      "theoretical_loss": 3.9184470646779492,
      "tokens_seen": 530644992
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008476629889669008,
      "loss": 1.6118,
      "theoretical_loss": 3.918395851806153,
      "tokens_seen": 530710528
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000847642928786359,
      "loss": 1.6052,
      "theoretical_loss": 3.9183446470286136,
      "tokens_seen": 530776064
    },
    {
      "epoch": 0.16,
      "objective/train/advantage_avg": 0.4919034540653229,
      "objective/train/docs_used": 199409,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5748982429504395,
      "objective/train/original_loss": 1.5748980045318604,
      "objective/train/theoretical_loss": 3.9182934503430538,
      "objective/train/tokens_used": 551301600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2425510436296463,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050423264503479,
      "objective/train/weighted_lm_loss": 1.654366374015808,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9182934503430538,
      "tokens_seen": 530841600
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008476228686058175,
      "loss": 1.5934,
      "theoretical_loss": 3.9182934503430538,
      "tokens_seen": 530841600
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008476028084252759,
      "loss": 1.6358,
      "theoretical_loss": 3.918242261747195,
      "tokens_seen": 530907136
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008475827482447342,
      "loss": 1.6104,
      "theoretical_loss": 3.9181910812387604,
      "tokens_seen": 530972672
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008475626880641926,
      "loss": 1.5103,
      "theoretical_loss": 3.9181399088154745,
      "tokens_seen": 531038208
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000847542627883651,
      "loss": 1.5375,
      "theoretical_loss": 3.918088744475064,
      "tokens_seen": 531103744
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008475225677031093,
      "loss": 1.724,
      "theoretical_loss": 3.9180375882152525,
      "tokens_seen": 531169280
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008475025075225678,
      "loss": 1.6848,
      "theoretical_loss": 3.9179864400337685,
      "tokens_seen": 531234816
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000847482447342026,
      "loss": 1.6819,
      "theoretical_loss": 3.91793529992834,
      "tokens_seen": 531300352
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008474623871614845,
      "loss": 1.5828,
      "theoretical_loss": 3.9178841678966956,
      "tokens_seen": 531365888
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008474423269809428,
      "loss": 1.5993,
      "theoretical_loss": 3.9178330439365654,
      "tokens_seen": 531431424
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008474222668004012,
      "loss": 1.6697,
      "theoretical_loss": 3.917781928045679,
      "tokens_seen": 531496960
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008474022066198596,
      "loss": 1.7663,
      "theoretical_loss": 3.9177308202217693,
      "tokens_seen": 531562496
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000847382146439318,
      "loss": 1.6602,
      "theoretical_loss": 3.9176797204625693,
      "tokens_seen": 531628032
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008473620862587763,
      "loss": 1.6338,
      "theoretical_loss": 3.9176286287658106,
      "tokens_seen": 531693568
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008473420260782348,
      "loss": 1.716,
      "theoretical_loss": 3.9175775451292285,
      "tokens_seen": 531759104
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000847321965897693,
      "loss": 1.7191,
      "theoretical_loss": 3.917526469550559,
      "tokens_seen": 531824640
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008473019057171515,
      "loss": 1.5278,
      "theoretical_loss": 3.917475402027537,
      "tokens_seen": 531890176
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008472818455366098,
      "loss": 1.5785,
      "theoretical_loss": 3.9174243425579007,
      "tokens_seen": 531955712
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008472617853560682,
      "loss": 1.642,
      "theoretical_loss": 3.917373291139388,
      "tokens_seen": 532021248
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008472417251755266,
      "loss": 1.5372,
      "theoretical_loss": 3.9173222477697376,
      "tokens_seen": 532086784
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000847221664994985,
      "loss": 1.5666,
      "theoretical_loss": 3.917271212446689,
      "tokens_seen": 532152320
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008472016048144433,
      "loss": 1.5001,
      "theoretical_loss": 3.917220185167984,
      "tokens_seen": 532217856
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008471815446339017,
      "loss": 1.568,
      "theoretical_loss": 3.9171691659313637,
      "tokens_seen": 532283392
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00084716148445336,
      "loss": 1.8325,
      "theoretical_loss": 3.9171181547345704,
      "tokens_seen": 532348928
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008471414242728185,
      "loss": 1.5937,
      "theoretical_loss": 3.917067151575348,
      "tokens_seen": 532414464
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008471213640922768,
      "loss": 1.6996,
      "theoretical_loss": 3.917016156451441,
      "tokens_seen": 532480000
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008471013039117352,
      "loss": 1.7402,
      "theoretical_loss": 3.9169651693605942,
      "tokens_seen": 532545536
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008470812437311935,
      "loss": 1.6802,
      "theoretical_loss": 3.916914190300555,
      "tokens_seen": 532611072
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000847061183550652,
      "loss": 1.6203,
      "theoretical_loss": 3.916863219269069,
      "tokens_seen": 532676608
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008470411233701103,
      "loss": 1.524,
      "theoretical_loss": 3.916812256263886,
      "tokens_seen": 532742144
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008470210631895687,
      "loss": 1.6564,
      "theoretical_loss": 3.9167613012827536,
      "tokens_seen": 532807680
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008470010030090271,
      "loss": 1.6716,
      "theoretical_loss": 3.9167103543234223,
      "tokens_seen": 532873216
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008469809428284855,
      "loss": 1.6233,
      "theoretical_loss": 3.9166594153836427,
      "tokens_seen": 532938752
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008469608826479438,
      "loss": 1.6648,
      "theoretical_loss": 3.9166084844611664,
      "tokens_seen": 533004288
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008469408224674023,
      "loss": 1.6156,
      "theoretical_loss": 3.9165575615537467,
      "tokens_seen": 533069824
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008469207622868605,
      "loss": 1.668,
      "theoretical_loss": 3.916506646659136,
      "tokens_seen": 533135360
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000846900702106319,
      "loss": 1.6171,
      "theoretical_loss": 3.9164557397750897,
      "tokens_seen": 533200896
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008468806419257773,
      "loss": 1.6386,
      "theoretical_loss": 3.916404840899363,
      "tokens_seen": 533266432
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008468605817452357,
      "loss": 1.6073,
      "theoretical_loss": 3.9163539500297118,
      "tokens_seen": 533331968
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008468405215646942,
      "loss": 1.6268,
      "theoretical_loss": 3.916303067163893,
      "tokens_seen": 533397504
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008468204613841524,
      "loss": 1.6233,
      "theoretical_loss": 3.916252192299665,
      "tokens_seen": 533463040
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008468004012036109,
      "loss": 1.6126,
      "theoretical_loss": 3.916201325434787,
      "tokens_seen": 533528576
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008467803410230693,
      "loss": 1.5906,
      "theoretical_loss": 3.916150466567019,
      "tokens_seen": 533594112
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008467602808425276,
      "loss": 1.6757,
      "theoretical_loss": 3.9160996156941206,
      "tokens_seen": 533659648
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000846740220661986,
      "loss": 1.6084,
      "theoretical_loss": 3.9160487728138538,
      "tokens_seen": 533725184
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008467201604814443,
      "loss": 1.63,
      "theoretical_loss": 3.915997937923982,
      "tokens_seen": 533790720
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008467001003009027,
      "loss": 1.5667,
      "theoretical_loss": 3.915947111022269,
      "tokens_seen": 533856256
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008466800401203612,
      "loss": 1.7042,
      "theoretical_loss": 3.9158962921064773,
      "tokens_seen": 533921792
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008466599799398194,
      "loss": 1.6378,
      "theoretical_loss": 3.9158454811743733,
      "tokens_seen": 533987328
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008466399197592779,
      "loss": 1.6598,
      "theoretical_loss": 3.9157946782237234,
      "tokens_seen": 534052864
    },
    {
      "epoch": 0.16,
      "objective/train/advantage_avg": 0.49396708607673645,
      "objective/train/docs_used": 200516,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5956155061721802,
      "objective/train/original_loss": 1.5956156253814697,
      "objective/train/theoretical_loss": 3.9157438832522944,
      "objective/train/tokens_used": 554578400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24434596300125122,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050639033317566,
      "objective/train/weighted_lm_loss": 1.6763161420822144,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9157438832522944,
      "tokens_seen": 534118400
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008466198595787363,
      "loss": 1.6087,
      "theoretical_loss": 3.9157438832522944,
      "tokens_seen": 534118400
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008465997993981946,
      "loss": 1.7316,
      "theoretical_loss": 3.9156930962578533,
      "tokens_seen": 534183936
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000846579739217653,
      "loss": 1.7161,
      "theoretical_loss": 3.915642317238171,
      "tokens_seen": 534249472
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008465596790371114,
      "loss": 1.6406,
      "theoretical_loss": 3.915591546191015,
      "tokens_seen": 534315008
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008465396188565697,
      "loss": 1.6527,
      "theoretical_loss": 3.9155407831141575,
      "tokens_seen": 534380544
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008465195586760282,
      "loss": 1.6793,
      "theoretical_loss": 3.9154900280053697,
      "tokens_seen": 534446080
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008464994984954864,
      "loss": 1.6632,
      "theoretical_loss": 3.915439280862423,
      "tokens_seen": 534511616
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008464794383149449,
      "loss": 1.6393,
      "theoretical_loss": 3.9153885416830922,
      "tokens_seen": 534577152
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008464593781344032,
      "loss": 1.5558,
      "theoretical_loss": 3.915337810465151,
      "tokens_seen": 534642688
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008464393179538616,
      "loss": 1.6084,
      "theoretical_loss": 3.9152870872063747,
      "tokens_seen": 534708224
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00084641925777332,
      "loss": 1.6117,
      "theoretical_loss": 3.915236371904539,
      "tokens_seen": 534773760
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008463991975927784,
      "loss": 1.7532,
      "theoretical_loss": 3.915185664557421,
      "tokens_seen": 534839296
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008463791374122367,
      "loss": 1.5398,
      "theoretical_loss": 3.915134965162798,
      "tokens_seen": 534904832
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008463590772316952,
      "loss": 1.6308,
      "theoretical_loss": 3.915084273718449,
      "tokens_seen": 534970368
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008463390170511534,
      "loss": 1.6037,
      "theoretical_loss": 3.915033590222153,
      "tokens_seen": 535035904
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008463189568706119,
      "loss": 1.5928,
      "theoretical_loss": 3.9149829146716923,
      "tokens_seen": 535101440
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008462988966900702,
      "loss": 1.6595,
      "theoretical_loss": 3.914932247064847,
      "tokens_seen": 535166976
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008462788365095286,
      "loss": 1.6967,
      "theoretical_loss": 3.914881587399399,
      "tokens_seen": 535232512
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000846258776328987,
      "loss": 1.5722,
      "theoretical_loss": 3.914830935673132,
      "tokens_seen": 535298048
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008462387161484454,
      "loss": 1.5449,
      "theoretical_loss": 3.9147802918838304,
      "tokens_seen": 535363584
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008462186559679037,
      "loss": 1.5417,
      "theoretical_loss": 3.914729656029279,
      "tokens_seen": 535429120
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008461985957873621,
      "loss": 1.6225,
      "theoretical_loss": 3.9146790281072628,
      "tokens_seen": 535494656
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008461785356068205,
      "loss": 1.5895,
      "theoretical_loss": 3.914628408115569,
      "tokens_seen": 535560192
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008461584754262789,
      "loss": 1.6551,
      "theoretical_loss": 3.9145777960519856,
      "tokens_seen": 535625728
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008461384152457372,
      "loss": 1.6158,
      "theoretical_loss": 3.9145271919143005,
      "tokens_seen": 535691264
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008461183550651956,
      "loss": 1.6353,
      "theoretical_loss": 3.9144765957003034,
      "tokens_seen": 535756800
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008460982948846539,
      "loss": 1.5529,
      "theoretical_loss": 3.9144260074077843,
      "tokens_seen": 535822336
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008460782347041124,
      "loss": 1.7125,
      "theoretical_loss": 3.914375427034535,
      "tokens_seen": 535887872
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008460581745235707,
      "loss": 1.6906,
      "theoretical_loss": 3.914324854578347,
      "tokens_seen": 535953408
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008460381143430291,
      "loss": 1.6646,
      "theoretical_loss": 3.914274290037013,
      "tokens_seen": 536018944
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008460180541624875,
      "loss": 1.6138,
      "theoretical_loss": 3.9142237334083276,
      "tokens_seen": 536084480
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008459979939819459,
      "loss": 1.6063,
      "theoretical_loss": 3.914173184690085,
      "tokens_seen": 536150016
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008459779338014042,
      "loss": 1.5708,
      "theoretical_loss": 3.9141226438800802,
      "tokens_seen": 536215552
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008459578736208627,
      "loss": 1.6743,
      "theoretical_loss": 3.9140721109761105,
      "tokens_seen": 536281088
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008459378134403209,
      "loss": 1.6015,
      "theoretical_loss": 3.914021585975973,
      "tokens_seen": 536346624
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008459177532597794,
      "loss": 1.7059,
      "theoretical_loss": 3.9139710688774656,
      "tokens_seen": 536412160
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008458976930792377,
      "loss": 1.6028,
      "theoretical_loss": 3.913920559678388,
      "tokens_seen": 536477696
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008458776328986961,
      "loss": 1.5366,
      "theoretical_loss": 3.9138700583765402,
      "tokens_seen": 536543232
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008458575727181545,
      "loss": 1.5262,
      "theoretical_loss": 3.9138195649697227,
      "tokens_seen": 536608768
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008458375125376128,
      "loss": 1.6228,
      "theoretical_loss": 3.9137690794557374,
      "tokens_seen": 536674304
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008458174523570712,
      "loss": 1.7002,
      "theoretical_loss": 3.913718601832387,
      "tokens_seen": 536739840
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008457973921765297,
      "loss": 1.5999,
      "theoretical_loss": 3.9136681320974738,
      "tokens_seen": 536805376
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008457773319959879,
      "loss": 1.6347,
      "theoretical_loss": 3.9136176702488044,
      "tokens_seen": 536870912
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008457572718154464,
      "loss": 1.5275,
      "theoretical_loss": 3.913567216284182,
      "tokens_seen": 536936448
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008457372116349046,
      "loss": 1.5859,
      "theoretical_loss": 3.913516770201414,
      "tokens_seen": 537001984
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008457171514543631,
      "loss": 1.5626,
      "theoretical_loss": 3.9134663319983076,
      "tokens_seen": 537067520
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008456970912738215,
      "loss": 1.5884,
      "theoretical_loss": 3.91341590167267,
      "tokens_seen": 537133056
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008456770310932798,
      "loss": 1.5622,
      "theoretical_loss": 3.91336547922231,
      "tokens_seen": 537198592
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008456569709127382,
      "loss": 1.6479,
      "theoretical_loss": 3.9133150646450376,
      "tokens_seen": 537264128
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008456369107321967,
      "loss": 1.6251,
      "theoretical_loss": 3.913264657938663,
      "tokens_seen": 537329664
    },
    {
      "epoch": 0.16,
      "objective/train/advantage_avg": 0.4932006597518921,
      "objective/train/docs_used": 201641,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4497480392456055,
      "objective/train/original_loss": 1.449748158454895,
      "objective/train/theoretical_loss": 3.9132142591009975,
      "objective/train/tokens_used": 557855200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2437322437763214,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050559163093567,
      "objective/train/weighted_lm_loss": 1.5229429006576538,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9132142591009975,
      "tokens_seen": 537395200
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008456168505516549,
      "loss": 1.5278,
      "theoretical_loss": 3.9132142591009975,
      "tokens_seen": 537395200
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008455967903711134,
      "loss": 1.6539,
      "theoretical_loss": 3.913163868129854,
      "tokens_seen": 537460736
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008455767301905716,
      "loss": 1.6771,
      "theoretical_loss": 3.913113485023046,
      "tokens_seen": 537526272
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008455566700100301,
      "loss": 1.5341,
      "theoretical_loss": 3.9130631097783866,
      "tokens_seen": 537591808
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008455366098294885,
      "loss": 1.5754,
      "theoretical_loss": 3.9130127423936907,
      "tokens_seen": 537657344
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008455165496489468,
      "loss": 1.6758,
      "theoretical_loss": 3.912962382866774,
      "tokens_seen": 537722880
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008454964894684052,
      "loss": 1.6715,
      "theoretical_loss": 3.9129120311954537,
      "tokens_seen": 537788416
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008454764292878636,
      "loss": 1.6385,
      "theoretical_loss": 3.9128616873775472,
      "tokens_seen": 537853952
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008454563691073219,
      "loss": 1.6726,
      "theoretical_loss": 3.9128113514108733,
      "tokens_seen": 537919488
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008454363089267804,
      "loss": 1.6917,
      "theoretical_loss": 3.9127610232932506,
      "tokens_seen": 537985024
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008454162487462387,
      "loss": 1.6347,
      "theoretical_loss": 3.912710703022499,
      "tokens_seen": 538050560
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008453961885656971,
      "loss": 1.6559,
      "theoretical_loss": 3.9126603905964403,
      "tokens_seen": 538116096
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008453761283851554,
      "loss": 1.5037,
      "theoretical_loss": 3.9126100860128963,
      "tokens_seen": 538181632
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008453560682046138,
      "loss": 1.6297,
      "theoretical_loss": 3.9125597892696895,
      "tokens_seen": 538247168
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008453360080240723,
      "loss": 1.732,
      "theoretical_loss": 3.912509500364643,
      "tokens_seen": 538312704
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008453159478435306,
      "loss": 1.7173,
      "theoretical_loss": 3.9124592192955827,
      "tokens_seen": 538378240
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000845295887662989,
      "loss": 1.5541,
      "theoretical_loss": 3.9124089460603324,
      "tokens_seen": 538443776
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008452758274824474,
      "loss": 1.5886,
      "theoretical_loss": 3.9123586806567188,
      "tokens_seen": 538509312
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008452557673019058,
      "loss": 1.5959,
      "theoretical_loss": 3.9123084230825698,
      "tokens_seen": 538574848
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008452357071213641,
      "loss": 1.7745,
      "theoretical_loss": 3.9122581733357125,
      "tokens_seen": 538640384
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008452156469408225,
      "loss": 1.5802,
      "theoretical_loss": 3.9122079314139766,
      "tokens_seen": 538705920
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008451955867602809,
      "loss": 1.693,
      "theoretical_loss": 3.9121576973151906,
      "tokens_seen": 538771456
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008451755265797393,
      "loss": 1.5976,
      "theoretical_loss": 3.912107471037186,
      "tokens_seen": 538836992
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008451554663991976,
      "loss": 1.6027,
      "theoretical_loss": 3.9120572525777937,
      "tokens_seen": 538902528
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000845135406218656,
      "loss": 1.6657,
      "theoretical_loss": 3.9120070419348463,
      "tokens_seen": 538968064
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008451153460381143,
      "loss": 1.4962,
      "theoretical_loss": 3.9119568391061765,
      "tokens_seen": 539033600
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008450952858575728,
      "loss": 1.5403,
      "theoretical_loss": 3.911906644089619,
      "tokens_seen": 539099136
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008450752256770311,
      "loss": 1.5892,
      "theoretical_loss": 3.9118564568830085,
      "tokens_seen": 539164672
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008450551654964895,
      "loss": 1.6834,
      "theoretical_loss": 3.9118062774841804,
      "tokens_seen": 539230208
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008450351053159479,
      "loss": 1.7064,
      "theoretical_loss": 3.911756105890972,
      "tokens_seen": 539295744
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008450150451354063,
      "loss": 1.767,
      "theoretical_loss": 3.9117059421012197,
      "tokens_seen": 539361280
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008449949849548646,
      "loss": 1.6195,
      "theoretical_loss": 3.911655786112763,
      "tokens_seen": 539426816
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008449749247743231,
      "loss": 1.5392,
      "theoretical_loss": 3.91160563792344,
      "tokens_seen": 539492352
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008449548645937813,
      "loss": 1.6299,
      "theoretical_loss": 3.9115554975310918,
      "tokens_seen": 539557888
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008449348044132398,
      "loss": 1.6654,
      "theoretical_loss": 3.9115053649335585,
      "tokens_seen": 539623424
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008449147442326981,
      "loss": 1.6519,
      "theoretical_loss": 3.9114552401286824,
      "tokens_seen": 539688960
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008448946840521565,
      "loss": 1.612,
      "theoretical_loss": 3.911405123114305,
      "tokens_seen": 539754496
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008448746238716149,
      "loss": 1.7025,
      "theoretical_loss": 3.9113550138882722,
      "tokens_seen": 539820032
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008448545636910732,
      "loss": 1.6332,
      "theoretical_loss": 3.9113049124484256,
      "tokens_seen": 539885568
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008448345035105316,
      "loss": 1.7054,
      "theoretical_loss": 3.9112548187926124,
      "tokens_seen": 539951104
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008448144433299901,
      "loss": 1.6892,
      "theoretical_loss": 3.9112047329186783,
      "tokens_seen": 540016640
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008447943831494483,
      "loss": 1.64,
      "theoretical_loss": 3.911154654824469,
      "tokens_seen": 540082176
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008447743229689068,
      "loss": 1.5573,
      "theoretical_loss": 3.911104584507834,
      "tokens_seen": 540147712
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000844754262788365,
      "loss": 1.8344,
      "theoretical_loss": 3.91105452196662,
      "tokens_seen": 540213248
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008447342026078235,
      "loss": 1.5585,
      "theoretical_loss": 3.911004467198679,
      "tokens_seen": 540278784
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008447141424272819,
      "loss": 1.7415,
      "theoretical_loss": 3.9109544202018593,
      "tokens_seen": 540344320
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008446940822467402,
      "loss": 1.656,
      "theoretical_loss": 3.9109043809740127,
      "tokens_seen": 540409856
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008446740220661986,
      "loss": 1.5673,
      "theoretical_loss": 3.9108543495129915,
      "tokens_seen": 540475392
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008446539618856571,
      "loss": 1.7181,
      "theoretical_loss": 3.9108043258166485,
      "tokens_seen": 540540928
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008446339017051153,
      "loss": 1.5633,
      "theoretical_loss": 3.9107543098828375,
      "tokens_seen": 540606464
    },
    {
      "epoch": 0.16,
      "objective/train/advantage_avg": 0.49212127923965454,
      "objective/train/docs_used": 202818,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4933364391326904,
      "objective/train/original_loss": 1.4933364391326904,
      "objective/train/theoretical_loss": 3.910704301709413,
      "objective/train/tokens_used": 561132000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24271930754184723,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504459142684937,
      "objective/train/weighted_lm_loss": 1.5686144828796387,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.910704301709413,
      "tokens_seen": 540672000
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008446138415245738,
      "loss": 1.6177,
      "theoretical_loss": 3.910704301709413,
      "tokens_seen": 540672000
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000844593781344032,
      "loss": 1.6962,
      "theoretical_loss": 3.91065430129423,
      "tokens_seen": 540737536
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008445737211634905,
      "loss": 1.5858,
      "theoretical_loss": 3.910604308635146,
      "tokens_seen": 540803072
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008445536609829489,
      "loss": 1.5983,
      "theoretical_loss": 3.9105543237300173,
      "tokens_seen": 540868608
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008445336008024072,
      "loss": 1.609,
      "theoretical_loss": 3.910504346576702,
      "tokens_seen": 540934144
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008445135406218656,
      "loss": 1.5867,
      "theoretical_loss": 3.9104543771730595,
      "tokens_seen": 540999680
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000844493480441324,
      "loss": 1.5341,
      "theoretical_loss": 3.9104044155169495,
      "tokens_seen": 541065216
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008444734202607823,
      "loss": 1.5828,
      "theoretical_loss": 3.9103544616062322,
      "tokens_seen": 541130752
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008444533600802408,
      "loss": 1.7507,
      "theoretical_loss": 3.910304515438769,
      "tokens_seen": 541196288
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000844433299899699,
      "loss": 1.7463,
      "theoretical_loss": 3.910254577012422,
      "tokens_seen": 541261824
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008444132397191575,
      "loss": 1.6767,
      "theoretical_loss": 3.910204646325055,
      "tokens_seen": 541327360
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008443931795386158,
      "loss": 1.656,
      "theoretical_loss": 3.9101547233745313,
      "tokens_seen": 541392896
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008443731193580742,
      "loss": 1.6428,
      "theoretical_loss": 3.9101048081587164,
      "tokens_seen": 541458432
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008443530591775326,
      "loss": 1.6341,
      "theoretical_loss": 3.9100549006754752,
      "tokens_seen": 541523968
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000844332998996991,
      "loss": 1.7341,
      "theoretical_loss": 3.9100050009226752,
      "tokens_seen": 541589504
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008443129388164493,
      "loss": 1.7356,
      "theoretical_loss": 3.909955108898183,
      "tokens_seen": 541655040
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008442928786359078,
      "loss": 1.6214,
      "theoretical_loss": 3.909905224599867,
      "tokens_seen": 541720576
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008442728184553661,
      "loss": 1.6616,
      "theoretical_loss": 3.909855348025596,
      "tokens_seen": 541786112
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008442527582748245,
      "loss": 1.5306,
      "theoretical_loss": 3.9098054791732406,
      "tokens_seen": 541851648
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008442326980942828,
      "loss": 1.7722,
      "theoretical_loss": 3.9097556180406716,
      "tokens_seen": 541917184
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008442126379137413,
      "loss": 1.6713,
      "theoretical_loss": 3.909705764625759,
      "tokens_seen": 541982720
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008441925777331996,
      "loss": 1.6604,
      "theoretical_loss": 3.9096559189263775,
      "tokens_seen": 542048256
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000844172517552658,
      "loss": 1.5467,
      "theoretical_loss": 3.909606080940399,
      "tokens_seen": 542113792
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008441524573721163,
      "loss": 1.6024,
      "theoretical_loss": 3.9095562506656982,
      "tokens_seen": 542179328
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008441323971915747,
      "loss": 1.6659,
      "theoretical_loss": 3.909506428100149,
      "tokens_seen": 542244864
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008441123370110331,
      "loss": 1.6401,
      "theoretical_loss": 3.9094566132416286,
      "tokens_seen": 542310400
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008440922768304915,
      "loss": 1.6249,
      "theoretical_loss": 3.909406806088013,
      "tokens_seen": 542375936
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008440722166499498,
      "loss": 1.5873,
      "theoretical_loss": 3.9093570066371797,
      "tokens_seen": 542441472
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008440521564694083,
      "loss": 1.7236,
      "theoretical_loss": 3.9093072148870074,
      "tokens_seen": 542507008
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008440320962888666,
      "loss": 1.6741,
      "theoretical_loss": 3.9092574308353747,
      "tokens_seen": 542572544
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000844012036108325,
      "loss": 1.657,
      "theoretical_loss": 3.909207654480162,
      "tokens_seen": 542638080
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008439919759277833,
      "loss": 1.6883,
      "theoretical_loss": 3.9091578858192504,
      "tokens_seen": 542703616
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008439719157472417,
      "loss": 1.675,
      "theoretical_loss": 3.909108124850521,
      "tokens_seen": 542769152
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008439518555667001,
      "loss": 1.6478,
      "theoretical_loss": 3.9090583715718568,
      "tokens_seen": 542834688
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008439317953861585,
      "loss": 1.4922,
      "theoretical_loss": 3.9090086259811403,
      "tokens_seen": 542900224
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008439117352056168,
      "loss": 1.8396,
      "theoretical_loss": 3.908958888076257,
      "tokens_seen": 542965760
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008438916750250753,
      "loss": 1.6117,
      "theoretical_loss": 3.908909157855091,
      "tokens_seen": 543031296
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008438716148445335,
      "loss": 1.5594,
      "theoretical_loss": 3.908859435315529,
      "tokens_seen": 543096832
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000843851554663992,
      "loss": 1.5383,
      "theoretical_loss": 3.908809720455457,
      "tokens_seen": 543162368
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008438314944834505,
      "loss": 1.6345,
      "theoretical_loss": 3.908760013272763,
      "tokens_seen": 543227904
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008438114343029087,
      "loss": 1.5347,
      "theoretical_loss": 3.908710313765335,
      "tokens_seen": 543293440
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008437913741223672,
      "loss": 1.677,
      "theoretical_loss": 3.908660621931063,
      "tokens_seen": 543358976
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008437713139418254,
      "loss": 1.659,
      "theoretical_loss": 3.908610937767836,
      "tokens_seen": 543424512
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008437512537612839,
      "loss": 1.6521,
      "theoretical_loss": 3.9085612612735456,
      "tokens_seen": 543490048
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008437311935807423,
      "loss": 1.6628,
      "theoretical_loss": 3.908511592446083,
      "tokens_seen": 543555584
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008437111334002006,
      "loss": 1.6616,
      "theoretical_loss": 3.908461931283342,
      "tokens_seen": 543621120
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000843691073219659,
      "loss": 1.6106,
      "theoretical_loss": 3.9084122777832144,
      "tokens_seen": 543686656
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008436710130391175,
      "loss": 1.6654,
      "theoretical_loss": 3.9083626319435956,
      "tokens_seen": 543752192
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008436509528585757,
      "loss": 1.6849,
      "theoretical_loss": 3.9083129937623804,
      "tokens_seen": 543817728
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008436308926780342,
      "loss": 1.5901,
      "theoretical_loss": 3.908263363237464,
      "tokens_seen": 543883264
    },
    {
      "epoch": 0.16,
      "objective/train/advantage_avg": 0.4907113313674927,
      "objective/train/docs_used": 204061,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5793161392211914,
      "objective/train/original_loss": 1.579316258430481,
      "objective/train/theoretical_loss": 3.908213740366744,
      "objective/train/tokens_used": 564408800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24126523733139038,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050297498703003,
      "objective/train/weighted_lm_loss": 1.6586852073669434,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.908213740366744,
      "tokens_seen": 543948800
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008436108324974924,
      "loss": 1.6774,
      "theoretical_loss": 3.908213740366744,
      "tokens_seen": 543948800
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008435907723169509,
      "loss": 1.6269,
      "theoretical_loss": 3.9081641251481174,
      "tokens_seen": 544014336
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008435707121364093,
      "loss": 1.6298,
      "theoretical_loss": 3.908114517579483,
      "tokens_seen": 544079872
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008435506519558676,
      "loss": 1.6212,
      "theoretical_loss": 3.9080649176587396,
      "tokens_seen": 544145408
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000843530591775326,
      "loss": 1.6775,
      "theoretical_loss": 3.908015325383788,
      "tokens_seen": 544210944
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008435105315947844,
      "loss": 1.6693,
      "theoretical_loss": 3.9079657407525286,
      "tokens_seen": 544276480
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008434904714142427,
      "loss": 1.6731,
      "theoretical_loss": 3.907916163762863,
      "tokens_seen": 544342016
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008434704112337012,
      "loss": 1.4552,
      "theoretical_loss": 3.9078665944126936,
      "tokens_seen": 544407552
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0008434503510531595,
      "loss": 1.7013,
      "theoretical_loss": 3.907817032699924,
      "tokens_seen": 544473088
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008434302908726179,
      "loss": 1.6329,
      "theoretical_loss": 3.9077674786224588,
      "tokens_seen": 544538624
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008434102306920762,
      "loss": 1.5692,
      "theoretical_loss": 3.907717932178203,
      "tokens_seen": 544604160
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008433901705115346,
      "loss": 1.5835,
      "theoretical_loss": 3.9076683933650616,
      "tokens_seen": 544669696
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000843370110330993,
      "loss": 1.6368,
      "theoretical_loss": 3.9076188621809416,
      "tokens_seen": 544735232
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008433500501504514,
      "loss": 1.5922,
      "theoretical_loss": 3.907569338623751,
      "tokens_seen": 544800768
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008433299899699097,
      "loss": 1.6241,
      "theoretical_loss": 3.9075198226913974,
      "tokens_seen": 544866304
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008433099297893682,
      "loss": 1.6416,
      "theoretical_loss": 3.907470314381791,
      "tokens_seen": 544931840
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008432898696088265,
      "loss": 1.796,
      "theoretical_loss": 3.9074208136928408,
      "tokens_seen": 544997376
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008432698094282849,
      "loss": 1.6469,
      "theoretical_loss": 3.907371320622458,
      "tokens_seen": 545062912
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008432497492477432,
      "loss": 1.7268,
      "theoretical_loss": 3.9073218351685544,
      "tokens_seen": 545128448
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008432296890672017,
      "loss": 1.6564,
      "theoretical_loss": 3.9072723573290418,
      "tokens_seen": 545193984
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00084320962888666,
      "loss": 1.5664,
      "theoretical_loss": 3.907222887101834,
      "tokens_seen": 545259520
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008431895687061184,
      "loss": 1.7664,
      "theoretical_loss": 3.907173424484845,
      "tokens_seen": 545325056
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008431695085255767,
      "loss": 1.5977,
      "theoretical_loss": 3.9071239694759896,
      "tokens_seen": 545390592
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008431494483450351,
      "loss": 1.7512,
      "theoretical_loss": 3.9070745220731844,
      "tokens_seen": 545456128
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008431293881644935,
      "loss": 1.6896,
      "theoretical_loss": 3.9070250822743446,
      "tokens_seen": 545521664
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008431093279839519,
      "loss": 1.659,
      "theoretical_loss": 3.9069756500773885,
      "tokens_seen": 545587200
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008430892678034102,
      "loss": 1.7349,
      "theoretical_loss": 3.9069262254802335,
      "tokens_seen": 545652736
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008430692076228687,
      "loss": 1.6861,
      "theoretical_loss": 3.906876808480799,
      "tokens_seen": 545718272
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000843049147442327,
      "loss": 1.6246,
      "theoretical_loss": 3.906827399077006,
      "tokens_seen": 545783808
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008430290872617854,
      "loss": 1.6346,
      "theoretical_loss": 3.906777997266773,
      "tokens_seen": 545849344
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008430090270812437,
      "loss": 1.6471,
      "theoretical_loss": 3.9067286030480237,
      "tokens_seen": 545914880
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008429889669007021,
      "loss": 1.689,
      "theoretical_loss": 3.9066792164186785,
      "tokens_seen": 545980416
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008429689067201605,
      "loss": 1.6054,
      "theoretical_loss": 3.9066298373766615,
      "tokens_seen": 546045952
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008429488465396189,
      "loss": 1.6965,
      "theoretical_loss": 3.9065804659198964,
      "tokens_seen": 546111488
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008429287863590772,
      "loss": 1.6824,
      "theoretical_loss": 3.9065311020463085,
      "tokens_seen": 546177024
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008429087261785357,
      "loss": 1.6068,
      "theoretical_loss": 3.9064817457538226,
      "tokens_seen": 546242560
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008428886659979939,
      "loss": 1.6234,
      "theoretical_loss": 3.9064323970403656,
      "tokens_seen": 546308096
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008428686058174524,
      "loss": 1.7045,
      "theoretical_loss": 3.9063830559038637,
      "tokens_seen": 546373632
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008428485456369108,
      "loss": 1.6106,
      "theoretical_loss": 3.9063337223422465,
      "tokens_seen": 546439168
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008428284854563691,
      "loss": 1.6302,
      "theoretical_loss": 3.906284396353441,
      "tokens_seen": 546504704
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008428084252758275,
      "loss": 1.6731,
      "theoretical_loss": 3.9062350779353787,
      "tokens_seen": 546570240
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008427883650952858,
      "loss": 1.6182,
      "theoretical_loss": 3.906185767085989,
      "tokens_seen": 546635776
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008427683049147442,
      "loss": 1.6749,
      "theoretical_loss": 3.906136463803204,
      "tokens_seen": 546701312
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008427482447342027,
      "loss": 1.5789,
      "theoretical_loss": 3.9060871680849543,
      "tokens_seen": 546766848
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008427281845536609,
      "loss": 1.6737,
      "theoretical_loss": 3.906037879929174,
      "tokens_seen": 546832384
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008427081243731194,
      "loss": 1.6823,
      "theoretical_loss": 3.9059885993337966,
      "tokens_seen": 546897920
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008426880641925778,
      "loss": 1.6538,
      "theoretical_loss": 3.905939326296756,
      "tokens_seen": 546963456
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008426680040120361,
      "loss": 1.4244,
      "theoretical_loss": 3.9058900608159886,
      "tokens_seen": 547028992
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008426479438314945,
      "loss": 1.7044,
      "theoretical_loss": 3.90584080288943,
      "tokens_seen": 547094528
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008426278836509528,
      "loss": 1.6694,
      "theoretical_loss": 3.9057915525150166,
      "tokens_seen": 547160064
    },
    {
      "epoch": 0.17,
      "objective/train/advantage_avg": 0.49299463629722595,
      "objective/train/docs_used": 204823,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4490361213684082,
      "objective/train/original_loss": 1.4490363597869873,
      "objective/train/theoretical_loss": 3.9057423096906874,
      "objective/train/tokens_used": 567685600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.244096577167511,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505404472351074,
      "objective/train/weighted_lm_loss": 1.5221048593521118,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9057423096906874,
      "tokens_seen": 547225600
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008426078234704112,
      "loss": 1.5892,
      "theoretical_loss": 3.9057423096906874,
      "tokens_seen": 547225600
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008425877632898697,
      "loss": 1.61,
      "theoretical_loss": 3.90569307441438,
      "tokens_seen": 547291136
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008425677031093279,
      "loss": 1.7166,
      "theoretical_loss": 3.905643846684034,
      "tokens_seen": 547356672
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008425476429287864,
      "loss": 1.6075,
      "theoretical_loss": 3.9055946264975896,
      "tokens_seen": 547422208
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008425275827482447,
      "loss": 1.5245,
      "theoretical_loss": 3.9055454138529884,
      "tokens_seen": 547487744
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008425075225677031,
      "loss": 1.6103,
      "theoretical_loss": 3.905496208748171,
      "tokens_seen": 547553280
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008424874623871615,
      "loss": 1.5812,
      "theoretical_loss": 3.9054470111810815,
      "tokens_seen": 547618816
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008424674022066199,
      "loss": 1.5837,
      "theoretical_loss": 3.905397821149662,
      "tokens_seen": 547684352
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008424473420260782,
      "loss": 1.6127,
      "theoretical_loss": 3.9053486386518577,
      "tokens_seen": 547749888
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008424272818455366,
      "loss": 1.6257,
      "theoretical_loss": 3.905299463685613,
      "tokens_seen": 547815424
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008424072216649949,
      "loss": 1.6105,
      "theoretical_loss": 3.9052502962488735,
      "tokens_seen": 547880960
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008423871614844534,
      "loss": 1.6986,
      "theoretical_loss": 3.9052011363395867,
      "tokens_seen": 547946496
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008423671013039117,
      "loss": 1.5628,
      "theoretical_loss": 3.9051519839557,
      "tokens_seen": 548012032
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008423470411233701,
      "loss": 1.6445,
      "theoretical_loss": 3.905102839095161,
      "tokens_seen": 548077568
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008423269809428286,
      "loss": 1.6641,
      "theoretical_loss": 3.9050537017559197,
      "tokens_seen": 548143104
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008423069207622869,
      "loss": 1.6682,
      "theoretical_loss": 3.9050045719359248,
      "tokens_seen": 548208640
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008422868605817453,
      "loss": 1.7039,
      "theoretical_loss": 3.904955449633128,
      "tokens_seen": 548274176
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008422668004012036,
      "loss": 1.6177,
      "theoretical_loss": 3.9049063348454798,
      "tokens_seen": 548339712
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008422467402206621,
      "loss": 1.5919,
      "theoretical_loss": 3.904857227570934,
      "tokens_seen": 548405248
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008422266800401204,
      "loss": 1.6974,
      "theoretical_loss": 3.904808127807442,
      "tokens_seen": 548470784
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008422066198595788,
      "loss": 1.5215,
      "theoretical_loss": 3.9047590355529587,
      "tokens_seen": 548536320
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008421865596790371,
      "loss": 1.5496,
      "theoretical_loss": 3.9047099508054384,
      "tokens_seen": 548601856
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008421664994984955,
      "loss": 1.4681,
      "theoretical_loss": 3.904660873562837,
      "tokens_seen": 548667392
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008421464393179539,
      "loss": 1.4619,
      "theoretical_loss": 3.90461180382311,
      "tokens_seen": 548732928
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008421263791374123,
      "loss": 1.6295,
      "theoretical_loss": 3.9045627415842157,
      "tokens_seen": 548798464
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008421063189568706,
      "loss": 1.7297,
      "theoretical_loss": 3.9045136868441106,
      "tokens_seen": 548864000
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008420862587763291,
      "loss": 1.6703,
      "theoretical_loss": 3.9044646396007545,
      "tokens_seen": 548929536
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008420661985957874,
      "loss": 1.6258,
      "theoretical_loss": 3.9044155998521064,
      "tokens_seen": 548995072
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008420461384152458,
      "loss": 1.6198,
      "theoretical_loss": 3.904366567596127,
      "tokens_seen": 549060608
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008420260782347041,
      "loss": 1.6694,
      "theoretical_loss": 3.9043175428307766,
      "tokens_seen": 549126144
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008420060180541625,
      "loss": 1.6021,
      "theoretical_loss": 3.9042685255540177,
      "tokens_seen": 549191680
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008419859578736209,
      "loss": 1.655,
      "theoretical_loss": 3.9042195157638133,
      "tokens_seen": 549257216
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008419658976930793,
      "loss": 1.6636,
      "theoretical_loss": 3.904170513458126,
      "tokens_seen": 549322752
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008419458375125376,
      "loss": 1.6395,
      "theoretical_loss": 3.9041215186349203,
      "tokens_seen": 549388288
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008419257773319961,
      "loss": 1.6057,
      "theoretical_loss": 3.9040725312921616,
      "tokens_seen": 549453824
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008419057171514543,
      "loss": 1.6732,
      "theoretical_loss": 3.9040235514278163,
      "tokens_seen": 549519360
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008418856569709128,
      "loss": 1.7559,
      "theoretical_loss": 3.9039745790398497,
      "tokens_seen": 549584896
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008418655967903712,
      "loss": 1.4984,
      "theoretical_loss": 3.903925614126231,
      "tokens_seen": 549650432
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008418455366098295,
      "loss": 1.6358,
      "theoretical_loss": 3.9038766566849263,
      "tokens_seen": 549715968
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008418254764292879,
      "loss": 1.6838,
      "theoretical_loss": 3.903827706713906,
      "tokens_seen": 549781504
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008418054162487462,
      "loss": 1.6195,
      "theoretical_loss": 3.9037787642111406,
      "tokens_seen": 549847040
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008417853560682046,
      "loss": 1.5633,
      "theoretical_loss": 3.903729829174599,
      "tokens_seen": 549912576
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008417652958876631,
      "loss": 1.6828,
      "theoretical_loss": 3.903680901602254,
      "tokens_seen": 549978112
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008417452357071213,
      "loss": 1.5739,
      "theoretical_loss": 3.903631981492077,
      "tokens_seen": 550043648
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008417251755265798,
      "loss": 1.5889,
      "theoretical_loss": 3.9035830688420416,
      "tokens_seen": 550109184
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008417051153460382,
      "loss": 1.5274,
      "theoretical_loss": 3.9035341636501215,
      "tokens_seen": 550174720
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008416850551654965,
      "loss": 1.7497,
      "theoretical_loss": 3.9034852659142913,
      "tokens_seen": 550240256
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008416649949849549,
      "loss": 1.6579,
      "theoretical_loss": 3.903436375632526,
      "tokens_seen": 550305792
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008416449348044132,
      "loss": 1.4647,
      "theoretical_loss": 3.9033874928028025,
      "tokens_seen": 550371328
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008416248746238716,
      "loss": 1.6656,
      "theoretical_loss": 3.9033386174230973,
      "tokens_seen": 550436864
    },
    {
      "epoch": 0.17,
      "objective/train/advantage_avg": 0.49245110154151917,
      "objective/train/docs_used": 206347,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6513526439666748,
      "objective/train/original_loss": 1.6513524055480957,
      "objective/train/theoretical_loss": 3.9032897494913876,
      "objective/train/tokens_used": 570962400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24299608170986176,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504804849624634,
      "objective/train/weighted_lm_loss": 1.73475980758667,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9032897494913876,
      "tokens_seen": 550502400
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008416048144433301,
      "loss": 1.6543,
      "theoretical_loss": 3.9032897494913876,
      "tokens_seen": 550502400
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008415847542627883,
      "loss": 1.6547,
      "theoretical_loss": 3.903240889005653,
      "tokens_seen": 550567936
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008415646940822468,
      "loss": 1.6404,
      "theoretical_loss": 3.9031920359638725,
      "tokens_seen": 550633472
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008415446339017051,
      "loss": 1.5895,
      "theoretical_loss": 3.903143190364026,
      "tokens_seen": 550699008
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008415245737211635,
      "loss": 1.5233,
      "theoretical_loss": 3.9030943522040946,
      "tokens_seen": 550764544
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008415045135406219,
      "loss": 1.6824,
      "theoretical_loss": 3.90304552148206,
      "tokens_seen": 550830080
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008414844533600803,
      "loss": 1.6634,
      "theoretical_loss": 3.9029966981959046,
      "tokens_seen": 550895616
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008414643931795386,
      "loss": 1.6046,
      "theoretical_loss": 3.9029478823436117,
      "tokens_seen": 550961152
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000841444332998997,
      "loss": 1.781,
      "theoretical_loss": 3.902899073923166,
      "tokens_seen": 551026688
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008414242728184553,
      "loss": 1.6708,
      "theoretical_loss": 3.902850272932551,
      "tokens_seen": 551092224
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008414042126379138,
      "loss": 1.6083,
      "theoretical_loss": 3.902801479369754,
      "tokens_seen": 551157760
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008413841524573721,
      "loss": 1.6505,
      "theoretical_loss": 3.9027526932327596,
      "tokens_seen": 551223296
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008413640922768305,
      "loss": 1.6389,
      "theoretical_loss": 3.902703914519557,
      "tokens_seen": 551288832
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008413440320962889,
      "loss": 1.6305,
      "theoretical_loss": 3.9026551432281322,
      "tokens_seen": 551354368
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008413239719157473,
      "loss": 1.6152,
      "theoretical_loss": 3.9026063793564756,
      "tokens_seen": 551419904
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008413039117352056,
      "loss": 1.5958,
      "theoretical_loss": 3.9025576229025756,
      "tokens_seen": 551485440
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000841283851554664,
      "loss": 1.7422,
      "theoretical_loss": 3.9025088738644236,
      "tokens_seen": 551550976
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008412637913741223,
      "loss": 1.5482,
      "theoretical_loss": 3.90246013224001,
      "tokens_seen": 551616512
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008412437311935808,
      "loss": 1.5257,
      "theoretical_loss": 3.902411398027327,
      "tokens_seen": 551682048
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008412236710130391,
      "loss": 1.6078,
      "theoretical_loss": 3.902362671224367,
      "tokens_seen": 551747584
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008412036108324975,
      "loss": 1.6325,
      "theoretical_loss": 3.9023139518291243,
      "tokens_seen": 551813120
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008411835506519558,
      "loss": 1.6692,
      "theoretical_loss": 3.902265239839592,
      "tokens_seen": 551878656
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008411634904714143,
      "loss": 1.5291,
      "theoretical_loss": 3.902216535253766,
      "tokens_seen": 551944192
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008411434302908726,
      "loss": 1.5662,
      "theoretical_loss": 3.9021678380696416,
      "tokens_seen": 552009728
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000841123370110331,
      "loss": 1.6418,
      "theoretical_loss": 3.902119148285216,
      "tokens_seen": 552075264
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008411033099297894,
      "loss": 1.5989,
      "theoretical_loss": 3.9020704658984857,
      "tokens_seen": 552140800
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008410832497492478,
      "loss": 1.5853,
      "theoretical_loss": 3.90202179090745,
      "tokens_seen": 552206336
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008410631895687061,
      "loss": 1.7523,
      "theoretical_loss": 3.901973123310107,
      "tokens_seen": 552271872
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008410431293881645,
      "loss": 1.6091,
      "theoretical_loss": 3.9019244631044563,
      "tokens_seen": 552337408
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008410230692076228,
      "loss": 1.4845,
      "theoretical_loss": 3.901875810288499,
      "tokens_seen": 552402944
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008410030090270813,
      "loss": 1.6014,
      "theoretical_loss": 3.9018271648602365,
      "tokens_seen": 552468480
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008409829488465396,
      "loss": 1.7345,
      "theoretical_loss": 3.9017785268176697,
      "tokens_seen": 552534016
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000840962888665998,
      "loss": 1.519,
      "theoretical_loss": 3.9017298961588027,
      "tokens_seen": 552599552
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008409428284854564,
      "loss": 1.7287,
      "theoretical_loss": 3.9016812728816386,
      "tokens_seen": 552665088
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008409227683049147,
      "loss": 1.656,
      "theoretical_loss": 3.901632656984182,
      "tokens_seen": 552730624
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008409027081243731,
      "loss": 1.568,
      "theoretical_loss": 3.901584048464438,
      "tokens_seen": 552796160
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008408826479438316,
      "loss": 1.5216,
      "theoretical_loss": 3.901535447320412,
      "tokens_seen": 552861696
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008408625877632898,
      "loss": 1.4595,
      "theoretical_loss": 3.9014868535501113,
      "tokens_seen": 552927232
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008408425275827483,
      "loss": 1.6847,
      "theoretical_loss": 3.9014382671515433,
      "tokens_seen": 552992768
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008408224674022066,
      "loss": 1.6083,
      "theoretical_loss": 3.901389688122716,
      "tokens_seen": 553058304
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000840802407221665,
      "loss": 1.6974,
      "theoretical_loss": 3.901341116461639,
      "tokens_seen": 553123840
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008407823470411235,
      "loss": 1.5456,
      "theoretical_loss": 3.9012925521663213,
      "tokens_seen": 553189376
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008407622868605817,
      "loss": 1.6157,
      "theoretical_loss": 3.901243995234774,
      "tokens_seen": 553254912
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008407422266800402,
      "loss": 1.6511,
      "theoretical_loss": 3.901195445665009,
      "tokens_seen": 553320448
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008407221664994986,
      "loss": 1.5942,
      "theoretical_loss": 3.9011469034550372,
      "tokens_seen": 553385984
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008407021063189569,
      "loss": 1.6496,
      "theoretical_loss": 3.901098368602872,
      "tokens_seen": 553451520
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008406820461384153,
      "loss": 1.5681,
      "theoretical_loss": 3.901049841106528,
      "tokens_seen": 553517056
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008406619859578736,
      "loss": 1.7091,
      "theoretical_loss": 3.901001320964018,
      "tokens_seen": 553582592
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000840641925777332,
      "loss": 1.5802,
      "theoretical_loss": 3.900952808173358,
      "tokens_seen": 553648128
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008406218655967905,
      "loss": 1.7343,
      "theoretical_loss": 3.900904302732564,
      "tokens_seen": 553713664
    },
    {
      "epoch": 0.17,
      "objective/train/advantage_avg": 0.49144139885902405,
      "objective/train/docs_used": 207589,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7685081958770752,
      "objective/train/original_loss": 1.768507957458496,
      "objective/train/theoretical_loss": 3.9008558046396535,
      "objective/train/tokens_used": 574239200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24197053909301758,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503742694854736,
      "objective/train/weighted_lm_loss": 1.8574825525283813,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.9008558046396535,
      "tokens_seen": 553779200
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008406018054162487,
      "loss": 1.6778,
      "theoretical_loss": 3.9008558046396535,
      "tokens_seen": 553779200
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008405817452357072,
      "loss": 1.571,
      "theoretical_loss": 3.900807313892643,
      "tokens_seen": 553844736
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008405616850551655,
      "loss": 1.6578,
      "theoretical_loss": 3.900758830489551,
      "tokens_seen": 553910272
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008405416248746239,
      "loss": 1.7218,
      "theoretical_loss": 3.900710354428397,
      "tokens_seen": 553975808
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008405215646940823,
      "loss": 1.6324,
      "theoretical_loss": 3.9006618857072,
      "tokens_seen": 554041344
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008405015045135407,
      "loss": 1.5992,
      "theoretical_loss": 3.900613424323981,
      "tokens_seen": 554106880
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000840481444332999,
      "loss": 1.6166,
      "theoretical_loss": 3.900564970276762,
      "tokens_seen": 554172416
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008404613841524574,
      "loss": 1.5832,
      "theoretical_loss": 3.9005165235635646,
      "tokens_seen": 554237952
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008404413239719157,
      "loss": 1.5677,
      "theoretical_loss": 3.9004680841824113,
      "tokens_seen": 554303488
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008404212637913742,
      "loss": 1.7651,
      "theoretical_loss": 3.900419652131327,
      "tokens_seen": 554369024
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008404012036108325,
      "loss": 1.5291,
      "theoretical_loss": 3.9003712274083346,
      "tokens_seen": 554434560
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008403811434302909,
      "loss": 1.7042,
      "theoretical_loss": 3.9003228100114606,
      "tokens_seen": 554500096
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008403610832497493,
      "loss": 1.6628,
      "theoretical_loss": 3.9002743999387297,
      "tokens_seen": 554565632
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008403410230692077,
      "loss": 1.572,
      "theoretical_loss": 3.9002259971881696,
      "tokens_seen": 554631168
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000840320962888666,
      "loss": 1.5983,
      "theoretical_loss": 3.9001776017578074,
      "tokens_seen": 554696704
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008403009027081244,
      "loss": 1.6355,
      "theoretical_loss": 3.900129213645672,
      "tokens_seen": 554762240
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008402808425275827,
      "loss": 1.5939,
      "theoretical_loss": 3.9000808328497913,
      "tokens_seen": 554827776
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008402607823470412,
      "loss": 1.5598,
      "theoretical_loss": 3.9000324593681963,
      "tokens_seen": 554893312
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008402407221664995,
      "loss": 1.6882,
      "theoretical_loss": 3.899984093198916,
      "tokens_seen": 554958848
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008402206619859579,
      "loss": 1.5463,
      "theoretical_loss": 3.8999357343399836,
      "tokens_seen": 555024384
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008402006018054162,
      "loss": 1.6846,
      "theoretical_loss": 3.89988738278943,
      "tokens_seen": 555089920
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008401805416248747,
      "loss": 1.5995,
      "theoretical_loss": 3.8998390385452883,
      "tokens_seen": 555155456
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000840160481444333,
      "loss": 1.5115,
      "theoretical_loss": 3.899790701605592,
      "tokens_seen": 555220992
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008401404212637914,
      "loss": 1.5238,
      "theoretical_loss": 3.899742371968376,
      "tokens_seen": 555286528
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008401203610832498,
      "loss": 1.5713,
      "theoretical_loss": 3.899694049631674,
      "tokens_seen": 555352064
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008401003009027082,
      "loss": 1.6603,
      "theoretical_loss": 3.899645734593524,
      "tokens_seen": 555417600
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008400802407221665,
      "loss": 1.6839,
      "theoretical_loss": 3.899597426851961,
      "tokens_seen": 555483136
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000840060180541625,
      "loss": 1.5337,
      "theoretical_loss": 3.899549126405023,
      "tokens_seen": 555548672
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008400401203610832,
      "loss": 1.4908,
      "theoretical_loss": 3.8995008332507477,
      "tokens_seen": 555614208
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008400200601805417,
      "loss": 1.5943,
      "theoretical_loss": 3.899452547387175,
      "tokens_seen": 555679744
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00084,
      "loss": 1.5668,
      "theoretical_loss": 3.899404268812343,
      "tokens_seen": 555745280
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008399799398194584,
      "loss": 1.5763,
      "theoretical_loss": 3.899355997524294,
      "tokens_seen": 555810816
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008399598796389168,
      "loss": 1.7422,
      "theoretical_loss": 3.899307733521068,
      "tokens_seen": 555876352
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008399398194583751,
      "loss": 1.6995,
      "theoretical_loss": 3.899259476800707,
      "tokens_seen": 555941888
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008399197592778335,
      "loss": 1.3834,
      "theoretical_loss": 3.8992112273612545,
      "tokens_seen": 556007424
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000839899699097292,
      "loss": 1.6003,
      "theoretical_loss": 3.8991629852007534,
      "tokens_seen": 556072960
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008398796389167502,
      "loss": 1.7357,
      "theoretical_loss": 3.899114750317248,
      "tokens_seen": 556138496
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008398595787362087,
      "loss": 1.7076,
      "theoretical_loss": 3.899066522708783,
      "tokens_seen": 556204032
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008398395185556669,
      "loss": 1.4848,
      "theoretical_loss": 3.8990183023734044,
      "tokens_seen": 556269568
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008398194583751254,
      "loss": 1.6815,
      "theoretical_loss": 3.8989700893091586,
      "tokens_seen": 556335104
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008397993981945838,
      "loss": 1.6465,
      "theoretical_loss": 3.898921883514093,
      "tokens_seen": 556400640
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008397793380140421,
      "loss": 1.5887,
      "theoretical_loss": 3.898873684986256,
      "tokens_seen": 556466176
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008397592778335005,
      "loss": 1.5622,
      "theoretical_loss": 3.8988254937236952,
      "tokens_seen": 556531712
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000839739217652959,
      "loss": 1.5486,
      "theoretical_loss": 3.898777309724461,
      "tokens_seen": 556597248
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008397191574724172,
      "loss": 1.5783,
      "theoretical_loss": 3.8987291329866034,
      "tokens_seen": 556662784
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008396990972918757,
      "loss": 1.735,
      "theoretical_loss": 3.898680963508174,
      "tokens_seen": 556728320
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008396790371113339,
      "loss": 1.6335,
      "theoretical_loss": 3.8986328012872233,
      "tokens_seen": 556793856
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008396589769307924,
      "loss": 1.6146,
      "theoretical_loss": 3.898584646321805,
      "tokens_seen": 556859392
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008396389167502508,
      "loss": 1.5333,
      "theoretical_loss": 3.8985364986099715,
      "tokens_seen": 556924928
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008396188565697091,
      "loss": 1.5544,
      "theoretical_loss": 3.898488358149778,
      "tokens_seen": 556990464
    },
    {
      "epoch": 0.17,
      "objective/train/advantage_avg": 0.4942537546157837,
      "objective/train/docs_used": 208740,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5796339511871338,
      "objective/train/original_loss": 1.5796339511871338,
      "objective/train/theoretical_loss": 3.8984402249392778,
      "objective/train/tokens_used": 577516000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24452345073223114,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050668478012085,
      "objective/train/weighted_lm_loss": 1.6596615314483643,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8984402249392778,
      "tokens_seen": 557056000
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008395987963891675,
      "loss": 1.5798,
      "theoretical_loss": 3.8984402249392778,
      "tokens_seen": 557056000
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008395787362086259,
      "loss": 1.5101,
      "theoretical_loss": 3.8983920989765273,
      "tokens_seen": 557121536
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008395586760280842,
      "loss": 1.6682,
      "theoretical_loss": 3.8983439802595825,
      "tokens_seen": 557187072
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008395386158475427,
      "loss": 1.5552,
      "theoretical_loss": 3.8982958687865006,
      "tokens_seen": 557252608
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008395185556670009,
      "loss": 1.4999,
      "theoretical_loss": 3.8982477645553395,
      "tokens_seen": 557318144
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008394984954864594,
      "loss": 1.5518,
      "theoretical_loss": 3.8981996675641573,
      "tokens_seen": 557383680
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008394784353059177,
      "loss": 1.6179,
      "theoretical_loss": 3.8981515778110136,
      "tokens_seen": 557449216
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008394583751253761,
      "loss": 1.5914,
      "theoretical_loss": 3.8981034952939684,
      "tokens_seen": 557514752
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008394383149448345,
      "loss": 1.5776,
      "theoretical_loss": 3.898055420011082,
      "tokens_seen": 557580288
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008394182547642929,
      "loss": 1.5414,
      "theoretical_loss": 3.898007351960416,
      "tokens_seen": 557645824
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008393981945837512,
      "loss": 1.5064,
      "theoretical_loss": 3.897959291140034,
      "tokens_seen": 557711360
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008393781344032097,
      "loss": 1.5765,
      "theoretical_loss": 3.8979112375479974,
      "tokens_seen": 557776896
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000839358074222668,
      "loss": 1.5801,
      "theoretical_loss": 3.8978631911823705,
      "tokens_seen": 557842432
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008393380140421264,
      "loss": 1.5644,
      "theoretical_loss": 3.8978151520412174,
      "tokens_seen": 557907968
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008393179538615848,
      "loss": 1.5375,
      "theoretical_loss": 3.8977671201226043,
      "tokens_seen": 557973504
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008392978936810431,
      "loss": 1.566,
      "theoretical_loss": 3.8977190954245966,
      "tokens_seen": 558039040
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008392778335005016,
      "loss": 1.6901,
      "theoretical_loss": 3.8976710779452612,
      "tokens_seen": 558104576
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008392577733199599,
      "loss": 1.5803,
      "theoretical_loss": 3.897623067682665,
      "tokens_seen": 558170112
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008392377131394183,
      "loss": 1.6792,
      "theoretical_loss": 3.8975750646348772,
      "tokens_seen": 558235648
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008392176529588766,
      "loss": 1.6074,
      "theoretical_loss": 3.897527068799966,
      "tokens_seen": 558301184
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008391975927783351,
      "loss": 1.6105,
      "theoretical_loss": 3.8974790801760015,
      "tokens_seen": 558366720
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008391775325977934,
      "loss": 1.7481,
      "theoretical_loss": 3.8974310987610536,
      "tokens_seen": 558432256
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008391574724172518,
      "loss": 1.6748,
      "theoretical_loss": 3.8973831245531945,
      "tokens_seen": 558497792
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008391374122367102,
      "loss": 1.6032,
      "theoretical_loss": 3.8973351575504953,
      "tokens_seen": 558563328
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008391173520561686,
      "loss": 1.6181,
      "theoretical_loss": 3.897287197751029,
      "tokens_seen": 558628864
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008390972918756269,
      "loss": 1.6316,
      "theoretical_loss": 3.8972392451528686,
      "tokens_seen": 558694400
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008390772316950853,
      "loss": 1.6653,
      "theoretical_loss": 3.8971912997540885,
      "tokens_seen": 558759936
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008390571715145436,
      "loss": 1.6679,
      "theoretical_loss": 3.8971433615527644,
      "tokens_seen": 558825472
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008390371113340021,
      "loss": 1.629,
      "theoretical_loss": 3.897095430546971,
      "tokens_seen": 558891008
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008390170511534604,
      "loss": 1.6606,
      "theoretical_loss": 3.8970475067347854,
      "tokens_seen": 558956544
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008389969909729188,
      "loss": 1.6832,
      "theoretical_loss": 3.8969995901142838,
      "tokens_seen": 559022080
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008389769307923772,
      "loss": 1.5703,
      "theoretical_loss": 3.8969516806835443,
      "tokens_seen": 559087616
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008389568706118355,
      "loss": 1.6939,
      "theoretical_loss": 3.896903778440646,
      "tokens_seen": 559153152
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008389368104312939,
      "loss": 1.6771,
      "theoretical_loss": 3.8968558833836684,
      "tokens_seen": 559218688
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008389167502507524,
      "loss": 1.6217,
      "theoretical_loss": 3.8968079955106907,
      "tokens_seen": 559284224
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008388966900702106,
      "loss": 1.5746,
      "theoretical_loss": 3.8967601148197946,
      "tokens_seen": 559349760
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008388766298896691,
      "loss": 1.6534,
      "theoretical_loss": 3.896712241309061,
      "tokens_seen": 559415296
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008388565697091273,
      "loss": 1.6365,
      "theoretical_loss": 3.8966643749765724,
      "tokens_seen": 559480832
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008388365095285858,
      "loss": 1.6652,
      "theoretical_loss": 3.896616515820412,
      "tokens_seen": 559546368
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008388164493480442,
      "loss": 1.6126,
      "theoretical_loss": 3.8965686638386634,
      "tokens_seen": 559611904
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008387963891675025,
      "loss": 1.6858,
      "theoretical_loss": 3.896520819029411,
      "tokens_seen": 559677440
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008387763289869609,
      "loss": 1.6394,
      "theoretical_loss": 3.896472981390741,
      "tokens_seen": 559742976
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008387562688064194,
      "loss": 1.6711,
      "theoretical_loss": 3.896425150920738,
      "tokens_seen": 559808512
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008387362086258776,
      "loss": 1.6029,
      "theoretical_loss": 3.896377327617489,
      "tokens_seen": 559874048
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008387161484453361,
      "loss": 1.597,
      "theoretical_loss": 3.896329511479082,
      "tokens_seen": 559939584
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008386960882647943,
      "loss": 1.5347,
      "theoretical_loss": 3.8962817025036047,
      "tokens_seen": 560005120
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008386760280842528,
      "loss": 1.5564,
      "theoretical_loss": 3.8962339006891464,
      "tokens_seen": 560070656
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008386559679037112,
      "loss": 1.5286,
      "theoretical_loss": 3.896186106033796,
      "tokens_seen": 560136192
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008386359077231695,
      "loss": 1.6609,
      "theoretical_loss": 3.8961383185356455,
      "tokens_seen": 560201728
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008386158475426279,
      "loss": 1.5907,
      "theoretical_loss": 3.8960905381927837,
      "tokens_seen": 560267264
    },
    {
      "epoch": 0.17,
      "objective/train/advantage_avg": 0.49280041456222534,
      "objective/train/docs_used": 209977,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4882906675338745,
      "objective/train/original_loss": 1.4882906675338745,
      "objective/train/theoretical_loss": 3.8960427650033047,
      "objective/train/tokens_used": 580792800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24333533644676208,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505170822143555,
      "objective/train/weighted_lm_loss": 1.5633784532546997,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8960427650033047,
      "tokens_seen": 560332800
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008385957873620863,
      "loss": 1.5401,
      "theoretical_loss": 3.8960427650033047,
      "tokens_seen": 560332800
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008385757271815446,
      "loss": 1.5833,
      "theoretical_loss": 3.8959949989652998,
      "tokens_seen": 560398336
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008385556670010031,
      "loss": 1.5688,
      "theoretical_loss": 3.895947240076862,
      "tokens_seen": 560463872
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008385356068204613,
      "loss": 1.5936,
      "theoretical_loss": 3.8958994883360862,
      "tokens_seen": 560529408
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008385155466399198,
      "loss": 1.4294,
      "theoretical_loss": 3.8958517437410674,
      "tokens_seen": 560594944
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008384954864593781,
      "loss": 1.554,
      "theoretical_loss": 3.8958040062899,
      "tokens_seen": 560660480
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008384754262788365,
      "loss": 1.6376,
      "theoretical_loss": 3.895756275980681,
      "tokens_seen": 560726016
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008384553660982949,
      "loss": 1.55,
      "theoretical_loss": 3.895708552811507,
      "tokens_seen": 560791552
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008384353059177533,
      "loss": 1.6216,
      "theoretical_loss": 3.895660836780476,
      "tokens_seen": 560857088
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008384152457372116,
      "loss": 1.6825,
      "theoretical_loss": 3.895613127885686,
      "tokens_seen": 560922624
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008383951855566701,
      "loss": 1.6211,
      "theoretical_loss": 3.895565426125237,
      "tokens_seen": 560988160
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008383751253761284,
      "loss": 1.7213,
      "theoretical_loss": 3.8955177314972276,
      "tokens_seen": 561053696
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008383550651955868,
      "loss": 1.5804,
      "theoretical_loss": 3.895470043999759,
      "tokens_seen": 561119232
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008383350050150451,
      "loss": 1.5192,
      "theoretical_loss": 3.8954223636309333,
      "tokens_seen": 561184768
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008383149448345035,
      "loss": 1.6263,
      "theoretical_loss": 3.8953746903888513,
      "tokens_seen": 561250304
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008382948846539619,
      "loss": 1.5549,
      "theoretical_loss": 3.8953270242716167,
      "tokens_seen": 561315840
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008382748244734203,
      "loss": 1.4904,
      "theoretical_loss": 3.8952793652773323,
      "tokens_seen": 561381376
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008382547642928786,
      "loss": 1.6546,
      "theoretical_loss": 3.895231713404103,
      "tokens_seen": 561446912
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000838234704112337,
      "loss": 1.5228,
      "theoretical_loss": 3.895184068650033,
      "tokens_seen": 561512448
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008382146439317954,
      "loss": 1.6943,
      "theoretical_loss": 3.895136431013229,
      "tokens_seen": 561577984
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008381945837512538,
      "loss": 1.5857,
      "theoretical_loss": 3.8950888004917967,
      "tokens_seen": 561643520
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008381745235707121,
      "loss": 1.6202,
      "theoretical_loss": 3.8950411770838427,
      "tokens_seen": 561709056
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008381544633901706,
      "loss": 1.5768,
      "theoretical_loss": 3.8949935607874764,
      "tokens_seen": 561774592
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008381344032096288,
      "loss": 1.6026,
      "theoretical_loss": 3.894945951600805,
      "tokens_seen": 561840128
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008381143430290873,
      "loss": 1.488,
      "theoretical_loss": 3.894898349521939,
      "tokens_seen": 561905664
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008380942828485456,
      "loss": 1.6545,
      "theoretical_loss": 3.8948507545489877,
      "tokens_seen": 561971200
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000838074222668004,
      "loss": 1.6421,
      "theoretical_loss": 3.8948031666800613,
      "tokens_seen": 562036736
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008380541624874624,
      "loss": 1.5853,
      "theoretical_loss": 3.894755585913272,
      "tokens_seen": 562102272
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008380341023069208,
      "loss": 1.6885,
      "theoretical_loss": 3.894708012246732,
      "tokens_seen": 562167808
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008380140421263791,
      "loss": 1.6247,
      "theoretical_loss": 3.8946604456785545,
      "tokens_seen": 562233344
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008379939819458376,
      "loss": 1.756,
      "theoretical_loss": 3.8946128862068528,
      "tokens_seen": 562298880
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008379739217652958,
      "loss": 1.6148,
      "theoretical_loss": 3.894565333829741,
      "tokens_seen": 562364416
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008379538615847543,
      "loss": 1.537,
      "theoretical_loss": 3.8945177885453344,
      "tokens_seen": 562429952
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008379338014042126,
      "loss": 1.5882,
      "theoretical_loss": 3.8944702503517483,
      "tokens_seen": 562495488
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000837913741223671,
      "loss": 1.6343,
      "theoretical_loss": 3.8944227192471006,
      "tokens_seen": 562561024
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008378936810431294,
      "loss": 1.5708,
      "theoretical_loss": 3.894375195229507,
      "tokens_seen": 562626560
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008378736208625877,
      "loss": 1.6487,
      "theoretical_loss": 3.8943276782970866,
      "tokens_seen": 562692096
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008378535606820461,
      "loss": 1.6422,
      "theoretical_loss": 3.8942801684479575,
      "tokens_seen": 562757632
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008378335005015046,
      "loss": 1.6344,
      "theoretical_loss": 3.8942326656802395,
      "tokens_seen": 562823168
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008378134403209629,
      "loss": 1.6309,
      "theoretical_loss": 3.8941851699920518,
      "tokens_seen": 562888704
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008377933801404213,
      "loss": 1.7459,
      "theoretical_loss": 3.8941376813815167,
      "tokens_seen": 562954240
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008377733199598798,
      "loss": 1.6037,
      "theoretical_loss": 3.8940901998467545,
      "tokens_seen": 563019776
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000837753259779338,
      "loss": 1.5997,
      "theoretical_loss": 3.894042725385888,
      "tokens_seen": 563085312
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008377331995987965,
      "loss": 1.7549,
      "theoretical_loss": 3.89399525799704,
      "tokens_seen": 563150848
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008377131394182547,
      "loss": 1.6288,
      "theoretical_loss": 3.8939477976783348,
      "tokens_seen": 563216384
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008376930792377132,
      "loss": 1.496,
      "theoretical_loss": 3.893900344427896,
      "tokens_seen": 563281920
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008376730190571716,
      "loss": 1.7518,
      "theoretical_loss": 3.893852898243849,
      "tokens_seen": 563347456
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008376529588766299,
      "loss": 1.6264,
      "theoretical_loss": 3.8938054591243194,
      "tokens_seen": 563412992
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008376328986960883,
      "loss": 1.5521,
      "theoretical_loss": 3.893758027067435,
      "tokens_seen": 563478528
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008376128385155467,
      "loss": 1.5585,
      "theoretical_loss": 3.8937106020713212,
      "tokens_seen": 563544064
    },
    {
      "epoch": 0.17,
      "objective/train/advantage_avg": 0.49166640639305115,
      "objective/train/docs_used": 211161,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.651707410812378,
      "objective/train/original_loss": 1.651707410812378,
      "objective/train/theoretical_loss": 3.8936631841341076,
      "objective/train/tokens_used": 584069600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24235595762729645,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503987073898315,
      "objective/train/weighted_lm_loss": 1.734801173210144,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8936631841341076,
      "tokens_seen": 563609600
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000837592778335005,
      "loss": 1.6154,
      "theoretical_loss": 3.8936631841341076,
      "tokens_seen": 563609600
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008375727181544635,
      "loss": 1.6302,
      "theoretical_loss": 3.8936157732539223,
      "tokens_seen": 563675136
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008375526579739217,
      "loss": 1.6018,
      "theoretical_loss": 3.8935683694288943,
      "tokens_seen": 563740672
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008375325977933802,
      "loss": 1.5432,
      "theoretical_loss": 3.8935209726571545,
      "tokens_seen": 563806208
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008375125376128385,
      "loss": 1.687,
      "theoretical_loss": 3.893473582936833,
      "tokens_seen": 563871744
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008374924774322969,
      "loss": 1.6945,
      "theoretical_loss": 3.8934262002660622,
      "tokens_seen": 563937280
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008374724172517553,
      "loss": 1.6496,
      "theoretical_loss": 3.8933788246429737,
      "tokens_seen": 564002816
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008374523570712137,
      "loss": 1.5673,
      "theoretical_loss": 3.893331456065701,
      "tokens_seen": 564068352
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000837432296890672,
      "loss": 1.6722,
      "theoretical_loss": 3.8932840945323774,
      "tokens_seen": 564133888
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008374122367101305,
      "loss": 1.6569,
      "theoretical_loss": 3.8932367400411376,
      "tokens_seen": 564199424
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008373921765295888,
      "loss": 1.5352,
      "theoretical_loss": 3.8931893925901155,
      "tokens_seen": 564264960
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008373721163490472,
      "loss": 1.6271,
      "theoretical_loss": 3.893142052177449,
      "tokens_seen": 564330496
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008373520561685055,
      "loss": 1.6128,
      "theoretical_loss": 3.8930947188012737,
      "tokens_seen": 564396032
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000837331995987964,
      "loss": 1.5328,
      "theoretical_loss": 3.893047392459726,
      "tokens_seen": 564461568
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008373119358074223,
      "loss": 1.5962,
      "theoretical_loss": 3.8930000731509455,
      "tokens_seen": 564527104
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008372918756268807,
      "loss": 1.5037,
      "theoretical_loss": 3.8929527608730696,
      "tokens_seen": 564592640
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000837271815446339,
      "loss": 1.6041,
      "theoretical_loss": 3.8929054556242377,
      "tokens_seen": 564658176
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008372517552657974,
      "loss": 1.6258,
      "theoretical_loss": 3.8928581574025904,
      "tokens_seen": 564723712
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008372316950852558,
      "loss": 1.5506,
      "theoretical_loss": 3.8928108662062684,
      "tokens_seen": 564789248
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008372116349047142,
      "loss": 1.5471,
      "theoretical_loss": 3.8927635820334134,
      "tokens_seen": 564854784
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008371915747241725,
      "loss": 1.551,
      "theoretical_loss": 3.892716304882167,
      "tokens_seen": 564920320
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000837171514543631,
      "loss": 1.6422,
      "theoretical_loss": 3.8926690347506727,
      "tokens_seen": 564985856
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008371514543630892,
      "loss": 1.6627,
      "theoretical_loss": 3.892621771637074,
      "tokens_seen": 565051392
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008371313941825477,
      "loss": 1.5411,
      "theoretical_loss": 3.892574515539515,
      "tokens_seen": 565116928
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000837111334002006,
      "loss": 1.6011,
      "theoretical_loss": 3.892527266456141,
      "tokens_seen": 565182464
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008370912738214644,
      "loss": 1.7549,
      "theoretical_loss": 3.892480024385097,
      "tokens_seen": 565248000
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008370712136409228,
      "loss": 1.5504,
      "theoretical_loss": 3.8924327893245305,
      "tokens_seen": 565313536
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008370511534603812,
      "loss": 1.6712,
      "theoretical_loss": 3.892385561272588,
      "tokens_seen": 565379072
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008370310932798395,
      "loss": 1.6375,
      "theoretical_loss": 3.8923383402274174,
      "tokens_seen": 565444608
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000837011033099298,
      "loss": 1.6303,
      "theoretical_loss": 3.8922911261871675,
      "tokens_seen": 565510144
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008369909729187562,
      "loss": 1.6977,
      "theoretical_loss": 3.892243919149988,
      "tokens_seen": 565575680
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008369709127382147,
      "loss": 1.5744,
      "theoretical_loss": 3.8921967191140276,
      "tokens_seen": 565641216
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000836950852557673,
      "loss": 1.5975,
      "theoretical_loss": 3.8921495260774375,
      "tokens_seen": 565706752
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008369307923771314,
      "loss": 1.5785,
      "theoretical_loss": 3.89210234003837,
      "tokens_seen": 565772288
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008369107321965898,
      "loss": 1.5981,
      "theoretical_loss": 3.892055160994976,
      "tokens_seen": 565837824
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008368906720160481,
      "loss": 1.7243,
      "theoretical_loss": 3.8920079889454087,
      "tokens_seen": 565903360
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008368706118355065,
      "loss": 1.558,
      "theoretical_loss": 3.8919608238878216,
      "tokens_seen": 565968896
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000836850551654965,
      "loss": 1.5975,
      "theoretical_loss": 3.8919136658203684,
      "tokens_seen": 566034432
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008368304914744232,
      "loss": 1.5676,
      "theoretical_loss": 3.8918665147412055,
      "tokens_seen": 566099968
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008368104312938817,
      "loss": 1.3604,
      "theoretical_loss": 3.8918193706484865,
      "tokens_seen": 566165504
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008367903711133399,
      "loss": 1.6092,
      "theoretical_loss": 3.891772233540369,
      "tokens_seen": 566231040
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008367703109327984,
      "loss": 1.6393,
      "theoretical_loss": 3.891725103415009,
      "tokens_seen": 566296576
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008367502507522568,
      "loss": 1.6181,
      "theoretical_loss": 3.891677980270565,
      "tokens_seen": 566362112
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008367301905717151,
      "loss": 1.5997,
      "theoretical_loss": 3.8916308641051955,
      "tokens_seen": 566427648
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008367101303911735,
      "loss": 1.4944,
      "theoretical_loss": 3.8915837549170584,
      "tokens_seen": 566493184
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000836690070210632,
      "loss": 1.59,
      "theoretical_loss": 3.891536652704315,
      "tokens_seen": 566558720
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008366700100300902,
      "loss": 1.5597,
      "theoretical_loss": 3.8914895574651247,
      "tokens_seen": 566624256
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008366499498495487,
      "loss": 1.5383,
      "theoretical_loss": 3.891442469197649,
      "tokens_seen": 566689792
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000836629889669007,
      "loss": 1.6051,
      "theoretical_loss": 3.89139538790005,
      "tokens_seen": 566755328
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008366098294884654,
      "loss": 1.662,
      "theoretical_loss": 3.8913483135704894,
      "tokens_seen": 566820864
    },
    {
      "epoch": 0.17,
      "objective/train/advantage_avg": 0.4925703704357147,
      "objective/train/docs_used": 212374,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.529457688331604,
      "objective/train/original_loss": 1.529457688331604,
      "objective/train/theoretical_loss": 3.8913012462071315,
      "objective/train/tokens_used": 587346400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24295853078365326,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504921674728394,
      "objective/train/weighted_lm_loss": 1.606675386428833,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.8913012462071315,
      "tokens_seen": 566886400
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008365897693079238,
      "loss": 1.6686,
      "theoretical_loss": 3.8913012462071315,
      "tokens_seen": 566886400
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008365697091273821,
      "loss": 1.6344,
      "theoretical_loss": 3.89125418580814,
      "tokens_seen": 566951936
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008365496489468405,
      "loss": 1.6334,
      "theoretical_loss": 3.8912071323716795,
      "tokens_seen": 567017472
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008365295887662989,
      "loss": 1.5775,
      "theoretical_loss": 3.8911600858959146,
      "tokens_seen": 567083008
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008365095285857572,
      "loss": 1.6099,
      "theoretical_loss": 3.8911130463790125,
      "tokens_seen": 567148544
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008364894684052157,
      "loss": 1.5115,
      "theoretical_loss": 3.891066013819139,
      "tokens_seen": 567214080
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000836469408224674,
      "loss": 1.7663,
      "theoretical_loss": 3.8910189882144626,
      "tokens_seen": 567279616
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008364493480441324,
      "loss": 1.6417,
      "theoretical_loss": 3.89097196956315,
      "tokens_seen": 567345152
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008364292878635908,
      "loss": 1.5897,
      "theoretical_loss": 3.890924957863372,
      "tokens_seen": 567410688
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008364092276830492,
      "loss": 1.7878,
      "theoretical_loss": 3.890877953113296,
      "tokens_seen": 567476224
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008363891675025075,
      "loss": 1.6121,
      "theoretical_loss": 3.8908309553110936,
      "tokens_seen": 567541760
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008363691073219659,
      "loss": 1.5549,
      "theoretical_loss": 3.890783964454935,
      "tokens_seen": 567607296
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008363490471414242,
      "loss": 1.5688,
      "theoretical_loss": 3.890736980542992,
      "tokens_seen": 567672832
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008363289869608827,
      "loss": 1.6342,
      "theoretical_loss": 3.890690003573437,
      "tokens_seen": 567738368
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008363089267803411,
      "loss": 1.6381,
      "theoretical_loss": 3.8906430335444426,
      "tokens_seen": 567803904
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008362888665997994,
      "loss": 1.4584,
      "theoretical_loss": 3.8905960704541833,
      "tokens_seen": 567869440
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008362688064192578,
      "loss": 1.6969,
      "theoretical_loss": 3.890549114300833,
      "tokens_seen": 567934976
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008362487462387162,
      "loss": 1.6092,
      "theoretical_loss": 3.890502165082566,
      "tokens_seen": 568000512
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008362286860581746,
      "loss": 1.7035,
      "theoretical_loss": 3.8904552227975593,
      "tokens_seen": 568066048
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008362086258776329,
      "loss": 1.6279,
      "theoretical_loss": 3.8904082874439885,
      "tokens_seen": 568131584
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008361885656970914,
      "loss": 1.6839,
      "theoretical_loss": 3.890361359020031,
      "tokens_seen": 568197120
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008361685055165496,
      "loss": 1.5912,
      "theoretical_loss": 3.890314437523865,
      "tokens_seen": 568262656
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008361484453360081,
      "loss": 1.6122,
      "theoretical_loss": 3.8902675229536685,
      "tokens_seen": 568328192
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008361283851554664,
      "loss": 1.6228,
      "theoretical_loss": 3.8902206153076206,
      "tokens_seen": 568393728
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008361083249749248,
      "loss": 1.6964,
      "theoretical_loss": 3.8901737145839017,
      "tokens_seen": 568459264
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008360882647943832,
      "loss": 1.6192,
      "theoretical_loss": 3.890126820780692,
      "tokens_seen": 568524800
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008360682046138416,
      "loss": 1.6188,
      "theoretical_loss": 3.8900799338961725,
      "tokens_seen": 568590336
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008360481444332999,
      "loss": 1.6078,
      "theoretical_loss": 3.8900330539285255,
      "tokens_seen": 568655872
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008360280842527584,
      "loss": 1.7185,
      "theoretical_loss": 3.8899861808759337,
      "tokens_seen": 568721408
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008360080240722166,
      "loss": 1.6761,
      "theoretical_loss": 3.889939314736581,
      "tokens_seen": 568786944
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008359879638916751,
      "loss": 1.6227,
      "theoretical_loss": 3.8898924555086496,
      "tokens_seen": 568852480
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008359679037111334,
      "loss": 1.6534,
      "theoretical_loss": 3.8898456031903255,
      "tokens_seen": 568918016
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008359478435305918,
      "loss": 1.5411,
      "theoretical_loss": 3.8897987577797943,
      "tokens_seen": 568983552
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008359277833500502,
      "loss": 1.5184,
      "theoretical_loss": 3.889751919275242,
      "tokens_seen": 569049088
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008359077231695085,
      "loss": 1.7104,
      "theoretical_loss": 3.8897050876748542,
      "tokens_seen": 569114624
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008358876629889669,
      "loss": 1.5735,
      "theoretical_loss": 3.8896582629768197,
      "tokens_seen": 569180160
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008358676028084254,
      "loss": 1.6771,
      "theoretical_loss": 3.8896114451793258,
      "tokens_seen": 569245696
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008358475426278836,
      "loss": 1.7303,
      "theoretical_loss": 3.8895646342805614,
      "tokens_seen": 569311232
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008358274824473421,
      "loss": 1.5807,
      "theoretical_loss": 3.8895178302787166,
      "tokens_seen": 569376768
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008358074222668003,
      "loss": 1.6122,
      "theoretical_loss": 3.8894710331719806,
      "tokens_seen": 569442304
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008357873620862588,
      "loss": 1.7613,
      "theoretical_loss": 3.889424242958545,
      "tokens_seen": 569507840
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008357673019057172,
      "loss": 1.5845,
      "theoretical_loss": 3.889377459636601,
      "tokens_seen": 569573376
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008357472417251755,
      "loss": 1.5114,
      "theoretical_loss": 3.8893306832043404,
      "tokens_seen": 569638912
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008357271815446339,
      "loss": 1.6762,
      "theoretical_loss": 3.8892839136599573,
      "tokens_seen": 569704448
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008357071213640924,
      "loss": 1.744,
      "theoretical_loss": 3.8892371510016446,
      "tokens_seen": 569769984
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008356870611835506,
      "loss": 1.6138,
      "theoretical_loss": 3.8891903952275966,
      "tokens_seen": 569835520
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008356670010030091,
      "loss": 1.6083,
      "theoretical_loss": 3.8891436463360076,
      "tokens_seen": 569901056
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008356469408224674,
      "loss": 1.5887,
      "theoretical_loss": 3.8890969043250743,
      "tokens_seen": 569966592
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008356268806419258,
      "loss": 1.7441,
      "theoretical_loss": 3.889050169192992,
      "tokens_seen": 570032128
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008356068204613842,
      "loss": 1.5472,
      "theoretical_loss": 3.889003440937959,
      "tokens_seen": 570097664
    },
    {
      "epoch": 0.17,
      "objective/train/advantage_avg": 0.4926186800003052,
      "objective/train/docs_used": 213119,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4580154418945312,
      "objective/train/original_loss": 1.4580153226852417,
      "objective/train/theoretical_loss": 3.8889567195581716,
      "objective/train/tokens_used": 590623200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2429921180009842,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504971742630005,
      "objective/train/weighted_lm_loss": 1.5317919254302979,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0100501775741577,
      "theoretical_loss": 3.8889567195581716,
      "tokens_seen": 570163200
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008355867602808425,
      "loss": 1.6331,
      "theoretical_loss": 3.8889567195581716,
      "tokens_seen": 570163200
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008355667001003009,
      "loss": 1.5682,
      "theoretical_loss": 3.8889100050518284,
      "tokens_seen": 570228736
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008355466399197593,
      "loss": 1.544,
      "theoretical_loss": 3.8888632974171298,
      "tokens_seen": 570294272
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008355265797392176,
      "loss": 1.5229,
      "theoretical_loss": 3.8888165966522736,
      "tokens_seen": 570359808
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008355065195586761,
      "loss": 1.6094,
      "theoretical_loss": 3.8887699027554614,
      "tokens_seen": 570425344
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008354864593781344,
      "loss": 1.5273,
      "theoretical_loss": 3.8887232157248937,
      "tokens_seen": 570490880
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008354663991975928,
      "loss": 1.5283,
      "theoretical_loss": 3.8886765355587722,
      "tokens_seen": 570556416
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008354463390170512,
      "loss": 1.6902,
      "theoretical_loss": 3.8886298622553,
      "tokens_seen": 570621952
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008354262788365096,
      "loss": 1.6482,
      "theoretical_loss": 3.8885831958126786,
      "tokens_seen": 570687488
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008354062186559679,
      "loss": 1.6233,
      "theoretical_loss": 3.888536536229114,
      "tokens_seen": 570753024
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008353861584754263,
      "loss": 1.5259,
      "theoretical_loss": 3.8884898835028094,
      "tokens_seen": 570818560
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008353660982948846,
      "loss": 1.6376,
      "theoretical_loss": 3.8884432376319698,
      "tokens_seen": 570884096
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008353460381143431,
      "loss": 1.7704,
      "theoretical_loss": 3.8883965986148015,
      "tokens_seen": 570949632
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008353259779338014,
      "loss": 1.5043,
      "theoretical_loss": 3.8883499664495105,
      "tokens_seen": 571015168
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008353059177532598,
      "loss": 1.6525,
      "theoretical_loss": 3.8883033411343044,
      "tokens_seen": 571080704
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008352858575727181,
      "loss": 1.5998,
      "theoretical_loss": 3.88825672266739,
      "tokens_seen": 571146240
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008352657973921766,
      "loss": 1.5162,
      "theoretical_loss": 3.888210111046978,
      "tokens_seen": 571211776
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008352457372116349,
      "loss": 1.6696,
      "theoretical_loss": 3.888163506271275,
      "tokens_seen": 571277312
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008352256770310933,
      "loss": 1.7017,
      "theoretical_loss": 3.888116908338492,
      "tokens_seen": 571342848
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008352056168505516,
      "loss": 1.6289,
      "theoretical_loss": 3.88807031724684,
      "tokens_seen": 571408384
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00083518555667001,
      "loss": 1.5372,
      "theoretical_loss": 3.8880237329945295,
      "tokens_seen": 571473920
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008351654964894684,
      "loss": 1.7852,
      "theoretical_loss": 3.8879771555797724,
      "tokens_seen": 571539456
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008351454363089268,
      "loss": 1.5828,
      "theoretical_loss": 3.8879305850007815,
      "tokens_seen": 571604992
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008351253761283851,
      "loss": 1.6125,
      "theoretical_loss": 3.88788402125577,
      "tokens_seen": 571670528
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008351053159478436,
      "loss": 1.6661,
      "theoretical_loss": 3.887837464342952,
      "tokens_seen": 571736064
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008350852557673019,
      "loss": 1.7204,
      "theoretical_loss": 3.887790914260541,
      "tokens_seen": 571801600
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008350651955867603,
      "loss": 1.5715,
      "theoretical_loss": 3.887744371006753,
      "tokens_seen": 571867136
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008350451354062187,
      "loss": 1.763,
      "theoretical_loss": 3.8876978345798046,
      "tokens_seen": 571932672
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000835025075225677,
      "loss": 1.5181,
      "theoretical_loss": 3.8876513049779113,
      "tokens_seen": 571998208
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008350050150451354,
      "loss": 1.6728,
      "theoretical_loss": 3.8876047821992903,
      "tokens_seen": 572063744
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008349849548645938,
      "loss": 1.6322,
      "theoretical_loss": 3.88755826624216,
      "tokens_seen": 572129280
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008349648946840521,
      "loss": 1.6106,
      "theoretical_loss": 3.887511757104739,
      "tokens_seen": 572194816
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008349448345035106,
      "loss": 1.6854,
      "theoretical_loss": 3.887465254785246,
      "tokens_seen": 572260352
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008349247743229688,
      "loss": 1.7422,
      "theoretical_loss": 3.8874187592819016,
      "tokens_seen": 572325888
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008349047141424273,
      "loss": 1.5888,
      "theoretical_loss": 3.887372270592926,
      "tokens_seen": 572391424
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008348846539618857,
      "loss": 1.7566,
      "theoretical_loss": 3.8873257887165407,
      "tokens_seen": 572456960
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000834864593781344,
      "loss": 1.5776,
      "theoretical_loss": 3.887279313650967,
      "tokens_seen": 572522496
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008348445336008024,
      "loss": 1.5842,
      "theoretical_loss": 3.8872328453944283,
      "tokens_seen": 572588032
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008348244734202607,
      "loss": 1.7323,
      "theoretical_loss": 3.887186383945147,
      "tokens_seen": 572653568
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008348044132397192,
      "loss": 1.6703,
      "theoretical_loss": 3.8871399293013473,
      "tokens_seen": 572719104
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008347843530591776,
      "loss": 1.6304,
      "theoretical_loss": 3.8870934814612546,
      "tokens_seen": 572784640
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008347642928786359,
      "loss": 1.5136,
      "theoretical_loss": 3.8870470404230932,
      "tokens_seen": 572850176
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008347442326980943,
      "loss": 1.4513,
      "theoretical_loss": 3.8870006061850892,
      "tokens_seen": 572915712
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008347241725175528,
      "loss": 1.694,
      "theoretical_loss": 3.88695417874547,
      "tokens_seen": 572981248
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000834704112337011,
      "loss": 1.5685,
      "theoretical_loss": 3.886907758102461,
      "tokens_seen": 573046784
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008346840521564695,
      "loss": 1.7508,
      "theoretical_loss": 3.8868613442542923,
      "tokens_seen": 573112320
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008346639919759278,
      "loss": 1.704,
      "theoretical_loss": 3.886814937199191,
      "tokens_seen": 573177856
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008346439317953862,
      "loss": 1.6086,
      "theoretical_loss": 3.886768536935387,
      "tokens_seen": 573243392
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008346238716148446,
      "loss": 1.5357,
      "theoretical_loss": 3.8867221434611094,
      "tokens_seen": 573308928
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000834603811434303,
      "loss": 1.7271,
      "theoretical_loss": 3.88667575677459,
      "tokens_seen": 573374464
    },
    {
      "epoch": 0.17,
      "objective/train/advantage_avg": 0.490614116191864,
      "objective/train/docs_used": 214228,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5626332759857178,
      "objective/train/original_loss": 1.5626335144042969,
      "objective/train/theoretical_loss": 3.8866293768740587,
      "objective/train/tokens_used": 593900000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24128560721874237,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502879619598389,
      "objective/train/weighted_lm_loss": 1.641161561012268,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.8866293768740587,
      "tokens_seen": 573440000
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008345837512537613,
      "loss": 1.5757,
      "theoretical_loss": 3.8866293768740587,
      "tokens_seen": 573440000
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008345636910732197,
      "loss": 1.5246,
      "theoretical_loss": 3.8865830037577487,
      "tokens_seen": 573505536
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000834543630892678,
      "loss": 1.6753,
      "theoretical_loss": 3.8865366374238914,
      "tokens_seen": 573571072
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008345235707121365,
      "loss": 1.7239,
      "theoretical_loss": 3.8864902778707213,
      "tokens_seen": 573636608
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008345035105315948,
      "loss": 1.5884,
      "theoretical_loss": 3.8864439250964704,
      "tokens_seen": 573702144
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008344834503510532,
      "loss": 1.6644,
      "theoretical_loss": 3.886397579099375,
      "tokens_seen": 573767680
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008344633901705116,
      "loss": 1.5428,
      "theoretical_loss": 3.88635123987767,
      "tokens_seen": 573833216
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00083444332998997,
      "loss": 1.5276,
      "theoretical_loss": 3.8863049074295906,
      "tokens_seen": 573898752
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008344232698094283,
      "loss": 1.6814,
      "theoretical_loss": 3.886258581753374,
      "tokens_seen": 573964288
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008344032096288867,
      "loss": 1.7092,
      "theoretical_loss": 3.8862122628472564,
      "tokens_seen": 574029824
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000834383149448345,
      "loss": 1.6209,
      "theoretical_loss": 3.8861659507094766,
      "tokens_seen": 574095360
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008343630892678035,
      "loss": 1.5473,
      "theoretical_loss": 3.8861196453382734,
      "tokens_seen": 574160896
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008343430290872618,
      "loss": 1.629,
      "theoretical_loss": 3.8860733467318846,
      "tokens_seen": 574226432
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008343229689067202,
      "loss": 1.7029,
      "theoretical_loss": 3.886027054888551,
      "tokens_seen": 574291968
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008343029087261785,
      "loss": 1.5779,
      "theoretical_loss": 3.885980769806513,
      "tokens_seen": 574357504
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000834282848545637,
      "loss": 1.6816,
      "theoretical_loss": 3.8859344914840115,
      "tokens_seen": 574423040
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008342627883650953,
      "loss": 1.6854,
      "theoretical_loss": 3.8858882199192886,
      "tokens_seen": 574488576
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008342427281845537,
      "loss": 1.6028,
      "theoretical_loss": 3.8858419551105863,
      "tokens_seen": 574554112
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000834222668004012,
      "loss": 1.7471,
      "theoretical_loss": 3.8857956970561487,
      "tokens_seen": 574619648
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008342026078234704,
      "loss": 1.6562,
      "theoretical_loss": 3.885749445754218,
      "tokens_seen": 574685184
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008341825476429288,
      "loss": 1.5984,
      "theoretical_loss": 3.8857032012030404,
      "tokens_seen": 574750720
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008341624874623872,
      "loss": 1.4557,
      "theoretical_loss": 3.8856569634008595,
      "tokens_seen": 574816256
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008341424272818455,
      "loss": 1.541,
      "theoretical_loss": 3.8856107323459215,
      "tokens_seen": 574881792
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000834122367101304,
      "loss": 1.7277,
      "theoretical_loss": 3.885564508036473,
      "tokens_seen": 574947328
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008341023069207623,
      "loss": 1.6655,
      "theoretical_loss": 3.8855182904707615,
      "tokens_seen": 575012864
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008340822467402207,
      "loss": 1.6032,
      "theoretical_loss": 3.885472079647034,
      "tokens_seen": 575078400
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000834062186559679,
      "loss": 1.5355,
      "theoretical_loss": 3.8854258755635387,
      "tokens_seen": 575143936
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008340421263791374,
      "loss": 1.6302,
      "theoretical_loss": 3.885379678218525,
      "tokens_seen": 575209472
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008340220661985958,
      "loss": 1.6603,
      "theoretical_loss": 3.885333487610243,
      "tokens_seen": 575275008
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008340020060180542,
      "loss": 1.6416,
      "theoretical_loss": 3.885287303736942,
      "tokens_seen": 575340544
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008339819458375125,
      "loss": 1.6654,
      "theoretical_loss": 3.885241126596874,
      "tokens_seen": 575406080
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000833961885656971,
      "loss": 1.5494,
      "theoretical_loss": 3.8851949561882897,
      "tokens_seen": 575471616
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008339418254764292,
      "loss": 1.4818,
      "theoretical_loss": 3.885148792509442,
      "tokens_seen": 575537152
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008339217652958877,
      "loss": 1.6263,
      "theoretical_loss": 3.8851026355585843,
      "tokens_seen": 575602688
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008339017051153461,
      "loss": 1.7115,
      "theoretical_loss": 3.885056485333969,
      "tokens_seen": 575668224
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008338816449348044,
      "loss": 1.6162,
      "theoretical_loss": 3.8850103418338513,
      "tokens_seen": 575733760
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008338615847542628,
      "loss": 1.7863,
      "theoretical_loss": 3.884964205056486,
      "tokens_seen": 575799296
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008338415245737211,
      "loss": 1.5634,
      "theoretical_loss": 3.884918075000128,
      "tokens_seen": 575864832
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008338214643931795,
      "loss": 1.6212,
      "theoretical_loss": 3.884871951663034,
      "tokens_seen": 575930368
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000833801404212638,
      "loss": 1.5676,
      "theoretical_loss": 3.8848258350434612,
      "tokens_seen": 575995904
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008337813440320962,
      "loss": 1.4362,
      "theoretical_loss": 3.884779725139666,
      "tokens_seen": 576061440
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008337612838515547,
      "loss": 1.5206,
      "theoretical_loss": 3.8847336219499073,
      "tokens_seen": 576126976
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008337412236710131,
      "loss": 1.651,
      "theoretical_loss": 3.8846875254724442,
      "tokens_seen": 576192512
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008337211634904714,
      "loss": 1.6546,
      "theoretical_loss": 3.884641435705535,
      "tokens_seen": 576258048
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008337011033099298,
      "loss": 1.6489,
      "theoretical_loss": 3.884595352647441,
      "tokens_seen": 576323584
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008336810431293882,
      "loss": 1.5317,
      "theoretical_loss": 3.884549276296423,
      "tokens_seen": 576389120
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008336609829488465,
      "loss": 1.5102,
      "theoretical_loss": 3.8845032066507414,
      "tokens_seen": 576454656
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000833640922768305,
      "loss": 1.5993,
      "theoretical_loss": 3.8844571437086586,
      "tokens_seen": 576520192
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008336208625877632,
      "loss": 1.4862,
      "theoretical_loss": 3.884411087468438,
      "tokens_seen": 576585728
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008336008024072217,
      "loss": 1.583,
      "theoretical_loss": 3.8843650379283416,
      "tokens_seen": 576651264
    },
    {
      "epoch": 0.17,
      "objective/train/advantage_avg": 0.49178701639175415,
      "objective/train/docs_used": 215374,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.578829288482666,
      "objective/train/original_loss": 1.5788294076919556,
      "objective/train/theoretical_loss": 3.884318995086635,
      "objective/train/tokens_used": 597176800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24241264164447784,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504109859466553,
      "objective/train/weighted_lm_loss": 1.6582890748977661,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.884318995086635,
      "tokens_seen": 576716800
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00083358074222668,
      "loss": 1.6054,
      "theoretical_loss": 3.884318995086635,
      "tokens_seen": 576716800
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008335606820461384,
      "loss": 1.6457,
      "theoretical_loss": 3.884272958941581,
      "tokens_seen": 576782336
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008335406218655968,
      "loss": 1.5649,
      "theoretical_loss": 3.8842269294914464,
      "tokens_seen": 576847872
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008335205616850552,
      "loss": 1.6548,
      "theoretical_loss": 3.8841809067344966,
      "tokens_seen": 576913408
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008335005015045135,
      "loss": 1.6309,
      "theoretical_loss": 3.8841348906689985,
      "tokens_seen": 576978944
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000833480441323972,
      "loss": 1.6167,
      "theoretical_loss": 3.8840888812932186,
      "tokens_seen": 577044480
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008334603811434302,
      "loss": 1.648,
      "theoretical_loss": 3.8840428786054253,
      "tokens_seen": 577110016
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008334403209628887,
      "loss": 1.667,
      "theoretical_loss": 3.8839968826038866,
      "tokens_seen": 577175552
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000833420260782347,
      "loss": 1.6751,
      "theoretical_loss": 3.8839508932868725,
      "tokens_seen": 577241088
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008334002006018054,
      "loss": 1.6029,
      "theoretical_loss": 3.883904910652652,
      "tokens_seen": 577306624
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008333801404212638,
      "loss": 1.7712,
      "theoretical_loss": 3.8838589346994965,
      "tokens_seen": 577372160
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0008333600802407222,
      "loss": 1.543,
      "theoretical_loss": 3.8838129654256757,
      "tokens_seen": 577437696
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008333400200601805,
      "loss": 1.4709,
      "theoretical_loss": 3.8837670028294626,
      "tokens_seen": 577503232
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008333199598796389,
      "loss": 1.5841,
      "theoretical_loss": 3.8837210469091286,
      "tokens_seen": 577568768
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008332998996990974,
      "loss": 1.625,
      "theoretical_loss": 3.8836750976629473,
      "tokens_seen": 577634304
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008332798395185557,
      "loss": 1.6486,
      "theoretical_loss": 3.883629155089192,
      "tokens_seen": 577699840
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008332597793380141,
      "loss": 1.6697,
      "theoretical_loss": 3.883583219186138,
      "tokens_seen": 577765376
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008332397191574724,
      "loss": 1.6512,
      "theoretical_loss": 3.8835372899520593,
      "tokens_seen": 577830912
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008332196589769308,
      "loss": 1.5178,
      "theoretical_loss": 3.883491367385231,
      "tokens_seen": 577896448
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008331995987963892,
      "loss": 1.6438,
      "theoretical_loss": 3.8834454514839303,
      "tokens_seen": 577961984
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008331795386158476,
      "loss": 1.7216,
      "theoretical_loss": 3.8833995422464342,
      "tokens_seen": 578027520
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008331594784353059,
      "loss": 1.5895,
      "theoretical_loss": 3.8833536396710198,
      "tokens_seen": 578093056
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008331394182547644,
      "loss": 1.6638,
      "theoretical_loss": 3.8833077437559647,
      "tokens_seen": 578158592
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008331193580742227,
      "loss": 1.6073,
      "theoretical_loss": 3.8832618544995485,
      "tokens_seen": 578224128
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008330992978936811,
      "loss": 1.6409,
      "theoretical_loss": 3.88321597190005,
      "tokens_seen": 578289664
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008330792377131395,
      "loss": 1.6481,
      "theoretical_loss": 3.88317009595575,
      "tokens_seen": 578355200
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008330591775325978,
      "loss": 1.7624,
      "theoretical_loss": 3.8831242266649286,
      "tokens_seen": 578420736
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008330391173520562,
      "loss": 1.5285,
      "theoretical_loss": 3.8830783640258675,
      "tokens_seen": 578486272
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008330190571715146,
      "loss": 1.6166,
      "theoretical_loss": 3.883032508036848,
      "tokens_seen": 578551808
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008329989969909729,
      "loss": 1.5131,
      "theoretical_loss": 3.8829866586961543,
      "tokens_seen": 578617344
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008329789368104314,
      "loss": 1.6345,
      "theoretical_loss": 3.882940816002068,
      "tokens_seen": 578682880
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008329588766298896,
      "loss": 1.6407,
      "theoretical_loss": 3.882894979952874,
      "tokens_seen": 578748416
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008329388164493481,
      "loss": 1.5498,
      "theoretical_loss": 3.882849150546856,
      "tokens_seen": 578813952
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008329187562688065,
      "loss": 1.681,
      "theoretical_loss": 3.8828033277823,
      "tokens_seen": 578879488
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008328986960882648,
      "loss": 1.707,
      "theoretical_loss": 3.8827575116574913,
      "tokens_seen": 578945024
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008328786359077232,
      "loss": 1.6368,
      "theoretical_loss": 3.8827117021707167,
      "tokens_seen": 579010560
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008328585757271815,
      "loss": 1.5782,
      "theoretical_loss": 3.8826658993202625,
      "tokens_seen": 579076096
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008328385155466399,
      "loss": 1.6314,
      "theoretical_loss": 3.8826201031044176,
      "tokens_seen": 579141632
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008328184553660984,
      "loss": 1.6671,
      "theoretical_loss": 3.8825743135214696,
      "tokens_seen": 579207168
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008327983951855566,
      "loss": 1.5509,
      "theoretical_loss": 3.882528530569708,
      "tokens_seen": 579272704
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008327783350050151,
      "loss": 1.6627,
      "theoretical_loss": 3.8824827542474214,
      "tokens_seen": 579338240
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008327582748244735,
      "loss": 1.601,
      "theoretical_loss": 3.8824369845529008,
      "tokens_seen": 579403776
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008327382146439318,
      "loss": 1.577,
      "theoretical_loss": 3.8823912214844376,
      "tokens_seen": 579469312
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008327181544633902,
      "loss": 1.744,
      "theoretical_loss": 3.882345465040322,
      "tokens_seen": 579534848
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008326980942828486,
      "loss": 1.6169,
      "theoretical_loss": 3.882299715218848,
      "tokens_seen": 579600384
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008326780341023069,
      "loss": 1.6419,
      "theoretical_loss": 3.8822539720183062,
      "tokens_seen": 579665920
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008326579739217654,
      "loss": 1.6029,
      "theoretical_loss": 3.882208235436992,
      "tokens_seen": 579731456
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008326379137412236,
      "loss": 1.6054,
      "theoretical_loss": 3.8821625054731976,
      "tokens_seen": 579796992
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008326178535606821,
      "loss": 1.669,
      "theoretical_loss": 3.8821167821252196,
      "tokens_seen": 579862528
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008325977933801404,
      "loss": 1.5815,
      "theoretical_loss": 3.8820710653913526,
      "tokens_seen": 579928064
    },
    {
      "epoch": 0.18,
      "objective/train/advantage_avg": 0.4898017644882202,
      "objective/train/docs_used": 216668,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6693953275680542,
      "objective/train/original_loss": 1.6693953275680542,
      "objective/train/theoretical_loss": 3.8820253552698913,
      "objective/train/tokens_used": 600453600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24118079245090485,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502060651779175,
      "objective/train/weighted_lm_loss": 1.7532109022140503,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8820253552698913,
      "tokens_seen": 579993600
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008325777331995988,
      "loss": 1.6935,
      "theoretical_loss": 3.8820253552698913,
      "tokens_seen": 579993600
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008325576730190572,
      "loss": 1.7155,
      "theoretical_loss": 3.881979651759134,
      "tokens_seen": 580059136
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008325376128385156,
      "loss": 1.6103,
      "theoretical_loss": 3.8819339548573772,
      "tokens_seen": 580124672
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008325175526579739,
      "loss": 1.5621,
      "theoretical_loss": 3.881888264562919,
      "tokens_seen": 580190208
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008324974924774324,
      "loss": 1.6421,
      "theoretical_loss": 3.8818425808740575,
      "tokens_seen": 580255744
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008324774322968906,
      "loss": 1.6178,
      "theoretical_loss": 3.881796903789092,
      "tokens_seen": 580321280
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008324573721163491,
      "loss": 1.676,
      "theoretical_loss": 3.881751233306322,
      "tokens_seen": 580386816
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008324373119358074,
      "loss": 1.6241,
      "theoretical_loss": 3.881705569424049,
      "tokens_seen": 580452352
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008324172517552658,
      "loss": 1.6982,
      "theoretical_loss": 3.8816599121405724,
      "tokens_seen": 580517888
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008323971915747242,
      "loss": 1.6278,
      "theoretical_loss": 3.881614261454195,
      "tokens_seen": 580583424
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008323771313941826,
      "loss": 1.5379,
      "theoretical_loss": 3.881568617363218,
      "tokens_seen": 580648960
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008323570712136409,
      "loss": 1.6818,
      "theoretical_loss": 3.881522979865945,
      "tokens_seen": 580714496
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008323370110330993,
      "loss": 1.69,
      "theoretical_loss": 3.88147734896068,
      "tokens_seen": 580780032
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008323169508525577,
      "loss": 1.5399,
      "theoretical_loss": 3.8814317246457266,
      "tokens_seen": 580845568
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008322968906720161,
      "loss": 1.6669,
      "theoretical_loss": 3.881386106919389,
      "tokens_seen": 580911104
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008322768304914744,
      "loss": 1.6094,
      "theoretical_loss": 3.8813404957799733,
      "tokens_seen": 580976640
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008322567703109328,
      "loss": 1.6382,
      "theoretical_loss": 3.881294891225785,
      "tokens_seen": 581042176
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008322367101303911,
      "loss": 1.5975,
      "theoretical_loss": 3.8812492932551317,
      "tokens_seen": 581107712
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008322166499498496,
      "loss": 1.7833,
      "theoretical_loss": 3.88120370186632,
      "tokens_seen": 581173248
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008321965897693079,
      "loss": 1.6112,
      "theoretical_loss": 3.8811581170576575,
      "tokens_seen": 581238784
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008321765295887663,
      "loss": 1.6593,
      "theoretical_loss": 3.881112538827453,
      "tokens_seen": 581304320
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008321564694082247,
      "loss": 1.5667,
      "theoretical_loss": 3.881066967174016,
      "tokens_seen": 581369856
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008321364092276831,
      "loss": 1.5722,
      "theoretical_loss": 3.881021402095657,
      "tokens_seen": 581435392
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008321163490471414,
      "loss": 1.6631,
      "theoretical_loss": 3.8809758435906843,
      "tokens_seen": 581500928
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008320962888665999,
      "loss": 1.5772,
      "theoretical_loss": 3.88093029165741,
      "tokens_seen": 581566464
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008320762286860581,
      "loss": 1.6153,
      "theoretical_loss": 3.8808847462941465,
      "tokens_seen": 581632000
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008320561685055166,
      "loss": 1.5789,
      "theoretical_loss": 3.880839207499205,
      "tokens_seen": 581697536
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008320361083249749,
      "loss": 1.5706,
      "theoretical_loss": 3.8807936752708985,
      "tokens_seen": 581763072
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008320160481444333,
      "loss": 1.697,
      "theoretical_loss": 3.880748149607541,
      "tokens_seen": 581828608
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008319959879638917,
      "loss": 1.6527,
      "theoretical_loss": 3.8807026305074475,
      "tokens_seen": 581894144
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00083197592778335,
      "loss": 1.6531,
      "theoretical_loss": 3.880657117968931,
      "tokens_seen": 581959680
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008319558676028084,
      "loss": 1.5888,
      "theoretical_loss": 3.8806116119903074,
      "tokens_seen": 582025216
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008319358074222669,
      "loss": 1.6148,
      "theoretical_loss": 3.880566112569893,
      "tokens_seen": 582090752
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008319157472417251,
      "loss": 1.7515,
      "theoretical_loss": 3.8805206197060045,
      "tokens_seen": 582156288
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008318956870611836,
      "loss": 1.7871,
      "theoretical_loss": 3.880475133396959,
      "tokens_seen": 582221824
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008318756268806418,
      "loss": 1.5739,
      "theoretical_loss": 3.880429653641075,
      "tokens_seen": 582287360
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008318555667001003,
      "loss": 1.6752,
      "theoretical_loss": 3.8803841804366694,
      "tokens_seen": 582352896
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008318355065195587,
      "loss": 1.5701,
      "theoretical_loss": 3.880338713782063,
      "tokens_seen": 582418432
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000831815446339017,
      "loss": 1.6733,
      "theoretical_loss": 3.8802932536755748,
      "tokens_seen": 582483968
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008317953861584755,
      "loss": 1.5975,
      "theoretical_loss": 3.8802478001155247,
      "tokens_seen": 582549504
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008317753259779339,
      "loss": 1.8073,
      "theoretical_loss": 3.880202353100235,
      "tokens_seen": 582615040
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008317552657973922,
      "loss": 1.6272,
      "theoretical_loss": 3.880156912628026,
      "tokens_seen": 582680576
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008317352056168506,
      "loss": 1.5908,
      "theoretical_loss": 3.880111478697221,
      "tokens_seen": 582746112
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000831715145436309,
      "loss": 1.7125,
      "theoretical_loss": 3.880066051306142,
      "tokens_seen": 582811648
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008316950852557673,
      "loss": 1.7426,
      "theoretical_loss": 3.880020630453113,
      "tokens_seen": 582877184
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008316750250752258,
      "loss": 1.6432,
      "theoretical_loss": 3.879975216136457,
      "tokens_seen": 582942720
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000831654964894684,
      "loss": 1.6708,
      "theoretical_loss": 3.8799298083545004,
      "tokens_seen": 583008256
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008316349047141425,
      "loss": 1.4938,
      "theoretical_loss": 3.8798844071055667,
      "tokens_seen": 583073792
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008316148445336008,
      "loss": 1.6715,
      "theoretical_loss": 3.8798390123879836,
      "tokens_seen": 583139328
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008315947843530592,
      "loss": 1.668,
      "theoretical_loss": 3.879793624200077,
      "tokens_seen": 583204864
    },
    {
      "epoch": 0.18,
      "objective/train/advantage_avg": 0.48885348439216614,
      "objective/train/docs_used": 217803,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.614006519317627,
      "objective/train/original_loss": 1.614006519317627,
      "objective/train/theoretical_loss": 3.879748242540173,
      "objective/train/tokens_used": 603730400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23961599171161652,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501033067703247,
      "objective/train/weighted_lm_loss": 1.6949542760849,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.879748242540173,
      "tokens_seen": 583270400
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008315747241725176,
      "loss": 1.6647,
      "theoretical_loss": 3.879748242540173,
      "tokens_seen": 583270400
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000831554663991976,
      "loss": 1.6274,
      "theoretical_loss": 3.8797028674066008,
      "tokens_seen": 583335936
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008315346038114343,
      "loss": 1.6726,
      "theoretical_loss": 3.8796574987976875,
      "tokens_seen": 583401472
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008315145436308928,
      "loss": 1.6305,
      "theoretical_loss": 3.8796121367117635,
      "tokens_seen": 583467008
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000831494483450351,
      "loss": 1.7564,
      "theoretical_loss": 3.8795667811471573,
      "tokens_seen": 583532544
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008314744232698095,
      "loss": 1.6404,
      "theoretical_loss": 3.8795214321022,
      "tokens_seen": 583598080
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008314543630892678,
      "loss": 1.6265,
      "theoretical_loss": 3.8794760895752223,
      "tokens_seen": 583663616
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008314343029087262,
      "loss": 1.7085,
      "theoretical_loss": 3.879430753564555,
      "tokens_seen": 583729152
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008314142427281846,
      "loss": 1.5586,
      "theoretical_loss": 3.8793854240685306,
      "tokens_seen": 583794688
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000831394182547643,
      "loss": 1.6759,
      "theoretical_loss": 3.8793401010854813,
      "tokens_seen": 583860224
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008313741223671013,
      "loss": 1.6092,
      "theoretical_loss": 3.8792947846137418,
      "tokens_seen": 583925760
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008313540621865597,
      "loss": 1.5635,
      "theoretical_loss": 3.879249474651645,
      "tokens_seen": 583991296
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000831334002006018,
      "loss": 1.6645,
      "theoretical_loss": 3.879204171197525,
      "tokens_seen": 584056832
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008313139418254765,
      "loss": 1.5576,
      "theoretical_loss": 3.8791588742497183,
      "tokens_seen": 584122368
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008312938816449348,
      "loss": 1.5826,
      "theoretical_loss": 3.8791135838065594,
      "tokens_seen": 584187904
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008312738214643932,
      "loss": 1.5733,
      "theoretical_loss": 3.8790682998663852,
      "tokens_seen": 584253440
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008312537612838515,
      "loss": 1.7476,
      "theoretical_loss": 3.879023022427533,
      "tokens_seen": 584318976
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00083123370110331,
      "loss": 1.6421,
      "theoretical_loss": 3.87897775148834,
      "tokens_seen": 584384512
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008312136409227683,
      "loss": 1.65,
      "theoretical_loss": 3.878932487047144,
      "tokens_seen": 584450048
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008311935807422267,
      "loss": 1.6201,
      "theoretical_loss": 3.8788872291022845,
      "tokens_seen": 584515584
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008311735205616851,
      "loss": 1.5899,
      "theoretical_loss": 3.878841977652101,
      "tokens_seen": 584581120
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008311534603811435,
      "loss": 1.5468,
      "theoretical_loss": 3.8787967326949335,
      "tokens_seen": 584646656
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008311334002006018,
      "loss": 1.6185,
      "theoretical_loss": 3.878751494229122,
      "tokens_seen": 584712192
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008311133400200603,
      "loss": 1.7026,
      "theoretical_loss": 3.8787062622530084,
      "tokens_seen": 584777728
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008310932798395185,
      "loss": 1.6082,
      "theoretical_loss": 3.8786610367649343,
      "tokens_seen": 584843264
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000831073219658977,
      "loss": 1.6696,
      "theoretical_loss": 3.878615817763242,
      "tokens_seen": 584908800
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008310531594784353,
      "loss": 1.6589,
      "theoretical_loss": 3.8785706052462756,
      "tokens_seen": 584974336
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008310330992978937,
      "loss": 1.5331,
      "theoretical_loss": 3.8785253992123776,
      "tokens_seen": 585039872
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008310130391173521,
      "loss": 1.7219,
      "theoretical_loss": 3.8784801996598928,
      "tokens_seen": 585105408
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008309929789368104,
      "loss": 1.5674,
      "theoretical_loss": 3.878435006587166,
      "tokens_seen": 585170944
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008309729187562688,
      "loss": 1.6549,
      "theoretical_loss": 3.8783898199925426,
      "tokens_seen": 585236480
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008309528585757273,
      "loss": 1.6957,
      "theoretical_loss": 3.8783446398743693,
      "tokens_seen": 585302016
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008309327983951855,
      "loss": 1.6219,
      "theoretical_loss": 3.878299466230992,
      "tokens_seen": 585367552
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000830912738214644,
      "loss": 1.5805,
      "theoretical_loss": 3.8782542990607594,
      "tokens_seen": 585433088
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008308926780341022,
      "loss": 1.5979,
      "theoretical_loss": 3.8782091383620183,
      "tokens_seen": 585498624
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008308726178535607,
      "loss": 1.6933,
      "theoretical_loss": 3.878163984133117,
      "tokens_seen": 585564160
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008308525576730191,
      "loss": 1.5984,
      "theoretical_loss": 3.8781188363724057,
      "tokens_seen": 585629696
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008308324974924774,
      "loss": 1.6588,
      "theoretical_loss": 3.8780736950782337,
      "tokens_seen": 585695232
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008308124373119358,
      "loss": 1.547,
      "theoretical_loss": 3.8780285602489513,
      "tokens_seen": 585760768
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008307923771313943,
      "loss": 1.6371,
      "theoretical_loss": 3.87798343188291,
      "tokens_seen": 585826304
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008307723169508525,
      "loss": 1.5934,
      "theoretical_loss": 3.87793830997846,
      "tokens_seen": 585891840
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000830752256770311,
      "loss": 1.5106,
      "theoretical_loss": 3.8778931945339554,
      "tokens_seen": 585957376
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008307321965897692,
      "loss": 1.5949,
      "theoretical_loss": 3.8778480855477477,
      "tokens_seen": 586022912
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008307121364092277,
      "loss": 1.7127,
      "theoretical_loss": 3.8778029830181904,
      "tokens_seen": 586088448
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008306920762286861,
      "loss": 1.5959,
      "theoretical_loss": 3.8777578869436384,
      "tokens_seen": 586153984
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008306720160481444,
      "loss": 1.7218,
      "theoretical_loss": 3.8777127973224452,
      "tokens_seen": 586219520
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008306519558676028,
      "loss": 1.6409,
      "theoretical_loss": 3.8776677141529667,
      "tokens_seen": 586285056
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008306318956870612,
      "loss": 1.5809,
      "theoretical_loss": 3.877622637433558,
      "tokens_seen": 586350592
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008306118355065195,
      "loss": 1.5752,
      "theoretical_loss": 3.8775775671625765,
      "tokens_seen": 586416128
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000830591775325978,
      "loss": 1.6401,
      "theoretical_loss": 3.8775325033383785,
      "tokens_seen": 586481664
    },
    {
      "epoch": 0.18,
      "objective/train/advantage_avg": 0.4921160936355591,
      "objective/train/docs_used": 218927,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7031680345535278,
      "objective/train/original_loss": 1.7031681537628174,
      "objective/train/theoretical_loss": 3.8774874459593223,
      "objective/train/tokens_used": 607007200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.242936372756958,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050446629524231,
      "objective/train/weighted_lm_loss": 1.7889472246170044,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8774874459593223,
      "tokens_seen": 586547200
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008305717151454363,
      "loss": 1.7481,
      "theoretical_loss": 3.8774874459593223,
      "tokens_seen": 586547200
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008305516549648947,
      "loss": 1.5825,
      "theoretical_loss": 3.877442395023765,
      "tokens_seen": 586612736
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000830531594784353,
      "loss": 1.5842,
      "theoretical_loss": 3.8773973505300674,
      "tokens_seen": 586678272
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008305115346038114,
      "loss": 1.5653,
      "theoretical_loss": 3.8773523124765865,
      "tokens_seen": 586743808
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008304914744232698,
      "loss": 1.653,
      "theoretical_loss": 3.8773072808616837,
      "tokens_seen": 586809344
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008304714142427282,
      "loss": 1.5497,
      "theoretical_loss": 3.8772622556837195,
      "tokens_seen": 586874880
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008304513540621865,
      "loss": 1.6068,
      "theoretical_loss": 3.877217236941055,
      "tokens_seen": 586940416
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000830431293881645,
      "loss": 1.7406,
      "theoretical_loss": 3.8771722246320524,
      "tokens_seen": 587005952
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008304112337011033,
      "loss": 1.632,
      "theoretical_loss": 3.8771272187550734,
      "tokens_seen": 587071488
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008303911735205617,
      "loss": 1.6805,
      "theoretical_loss": 3.8770822193084817,
      "tokens_seen": 587137024
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00083037111334002,
      "loss": 1.6109,
      "theoretical_loss": 3.877037226290641,
      "tokens_seen": 587202560
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008303510531594785,
      "loss": 1.6222,
      "theoretical_loss": 3.8769922396999146,
      "tokens_seen": 587268096
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008303309929789368,
      "loss": 1.5739,
      "theoretical_loss": 3.8769472595346683,
      "tokens_seen": 587333632
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008303109327983952,
      "loss": 1.5446,
      "theoretical_loss": 3.876902285793267,
      "tokens_seen": 587399168
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008302908726178536,
      "loss": 1.6361,
      "theoretical_loss": 3.8768573184740767,
      "tokens_seen": 587464704
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008302708124373119,
      "loss": 1.5991,
      "theoretical_loss": 3.8768123575754645,
      "tokens_seen": 587530240
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008302507522567704,
      "loss": 1.6164,
      "theoretical_loss": 3.8767674030957977,
      "tokens_seen": 587595776
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008302306920762287,
      "loss": 1.6073,
      "theoretical_loss": 3.876722455033443,
      "tokens_seen": 587661312
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008302106318956871,
      "loss": 1.5783,
      "theoretical_loss": 3.87667751338677,
      "tokens_seen": 587726848
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008301905717151455,
      "loss": 1.5493,
      "theoretical_loss": 3.8766325781541475,
      "tokens_seen": 587792384
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008301705115346039,
      "loss": 1.6647,
      "theoretical_loss": 3.876587649333944,
      "tokens_seen": 587857920
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008301504513540622,
      "loss": 1.6051,
      "theoretical_loss": 3.8765427269245314,
      "tokens_seen": 587923456
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008301303911735207,
      "loss": 1.5461,
      "theoretical_loss": 3.8764978109242794,
      "tokens_seen": 587988992
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008301103309929789,
      "loss": 1.5925,
      "theoretical_loss": 3.8764529013315596,
      "tokens_seen": 588054528
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008300902708124374,
      "loss": 1.6489,
      "theoretical_loss": 3.876407998144744,
      "tokens_seen": 588120064
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008300702106318957,
      "loss": 1.6229,
      "theoretical_loss": 3.8763631013622053,
      "tokens_seen": 588185600
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008300501504513541,
      "loss": 1.6815,
      "theoretical_loss": 3.8763182109823173,
      "tokens_seen": 588251136
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008300300902708125,
      "loss": 1.5642,
      "theoretical_loss": 3.876273327003452,
      "tokens_seen": 588316672
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008300100300902708,
      "loss": 1.4461,
      "theoretical_loss": 3.876228449423986,
      "tokens_seen": 588382208
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008299899699097292,
      "loss": 1.5003,
      "theoretical_loss": 3.8761835782422924,
      "tokens_seen": 588447744
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008299699097291877,
      "loss": 1.5458,
      "theoretical_loss": 3.8761387134567475,
      "tokens_seen": 588513280
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008299498495486459,
      "loss": 1.5513,
      "theoretical_loss": 3.8760938550657276,
      "tokens_seen": 588578816
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008299297893681044,
      "loss": 1.6935,
      "theoretical_loss": 3.8760490030676094,
      "tokens_seen": 588644352
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008299097291875626,
      "loss": 1.6617,
      "theoretical_loss": 3.8760041574607698,
      "tokens_seen": 588709888
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008298896690070211,
      "loss": 1.5717,
      "theoretical_loss": 3.8759593182435874,
      "tokens_seen": 588775424
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008298696088264795,
      "loss": 1.7028,
      "theoretical_loss": 3.87591448541444,
      "tokens_seen": 588840960
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008298495486459378,
      "loss": 1.6496,
      "theoretical_loss": 3.875869658971707,
      "tokens_seen": 588906496
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008298294884653962,
      "loss": 1.7078,
      "theoretical_loss": 3.875824838913769,
      "tokens_seen": 588972032
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008298094282848547,
      "loss": 1.6872,
      "theoretical_loss": 3.875780025239005,
      "tokens_seen": 589037568
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008297893681043129,
      "loss": 1.6225,
      "theoretical_loss": 3.8757352179457962,
      "tokens_seen": 589103104
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008297693079237714,
      "loss": 1.5302,
      "theoretical_loss": 3.8756904170325246,
      "tokens_seen": 589168640
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008297492477432296,
      "loss": 1.5347,
      "theoretical_loss": 3.8756456224975713,
      "tokens_seen": 589234176
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008297291875626881,
      "loss": 1.6227,
      "theoretical_loss": 3.8756008343393202,
      "tokens_seen": 589299712
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008297091273821465,
      "loss": 1.6826,
      "theoretical_loss": 3.8755560525561537,
      "tokens_seen": 589365248
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008296890672016048,
      "loss": 1.5985,
      "theoretical_loss": 3.8755112771464555,
      "tokens_seen": 589430784
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008296690070210632,
      "loss": 1.7784,
      "theoretical_loss": 3.8754665081086106,
      "tokens_seen": 589496320
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008296489468405216,
      "loss": 1.5733,
      "theoretical_loss": 3.8754217454410043,
      "tokens_seen": 589561856
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008296288866599799,
      "loss": 1.5693,
      "theoretical_loss": 3.875376989142021,
      "tokens_seen": 589627392
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008296088264794384,
      "loss": 1.6317,
      "theoretical_loss": 3.8753322392100475,
      "tokens_seen": 589692928
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008295887662988967,
      "loss": 1.6052,
      "theoretical_loss": 3.8752874956434713,
      "tokens_seen": 589758464
    },
    {
      "epoch": 0.18,
      "objective/train/advantage_avg": 0.4909812808036804,
      "objective/train/docs_used": 220165,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6117116212844849,
      "objective/train/original_loss": 1.6117116212844849,
      "objective/train/theoretical_loss": 3.875242758440679,
      "objective/train/tokens_used": 610284000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24177321791648865,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503270626068115,
      "objective/train/weighted_lm_loss": 1.6927611827850342,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.875242758440679,
      "tokens_seen": 589824000
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008295687061183551,
      "loss": 1.7046,
      "theoretical_loss": 3.875242758440679,
      "tokens_seen": 589824000
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008295486459378134,
      "loss": 1.5758,
      "theoretical_loss": 3.8751980276000584,
      "tokens_seen": 589889536
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008295285857572718,
      "loss": 1.6406,
      "theoretical_loss": 3.8751533031199985,
      "tokens_seen": 589955072
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008295085255767302,
      "loss": 1.5666,
      "theoretical_loss": 3.875108584998888,
      "tokens_seen": 590020608
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008294884653961886,
      "loss": 1.6276,
      "theoretical_loss": 3.875063873235117,
      "tokens_seen": 590086144
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008294684052156469,
      "loss": 1.6714,
      "theoretical_loss": 3.875019167827076,
      "tokens_seen": 590151680
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008294483450351054,
      "loss": 1.8013,
      "theoretical_loss": 3.8749744687731553,
      "tokens_seen": 590217216
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008294282848545637,
      "loss": 1.632,
      "theoretical_loss": 3.8749297760717463,
      "tokens_seen": 590282752
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008294082246740221,
      "loss": 1.6734,
      "theoretical_loss": 3.874885089721242,
      "tokens_seen": 590348288
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008293881644934804,
      "loss": 1.6035,
      "theoretical_loss": 3.8748404097200346,
      "tokens_seen": 590413824
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008293681043129389,
      "loss": 1.6007,
      "theoretical_loss": 3.874795736066517,
      "tokens_seen": 590479360
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008293480441323972,
      "loss": 1.5973,
      "theoretical_loss": 3.8747510687590827,
      "tokens_seen": 590544896
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008293279839518556,
      "loss": 1.4097,
      "theoretical_loss": 3.8747064077961264,
      "tokens_seen": 590610432
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008293079237713139,
      "loss": 1.6399,
      "theoretical_loss": 3.874661753176044,
      "tokens_seen": 590675968
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008292878635907723,
      "loss": 1.4542,
      "theoretical_loss": 3.87461710489723,
      "tokens_seen": 590741504
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008292678034102307,
      "loss": 1.6091,
      "theoretical_loss": 3.8745724629580813,
      "tokens_seen": 590807040
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008292477432296891,
      "loss": 1.5679,
      "theoretical_loss": 3.874527827356994,
      "tokens_seen": 590872576
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008292276830491474,
      "loss": 1.6006,
      "theoretical_loss": 3.874483198092366,
      "tokens_seen": 590938112
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008292076228686059,
      "loss": 1.5959,
      "theoretical_loss": 3.8744385751625945,
      "tokens_seen": 591003648
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008291875626880642,
      "loss": 1.6411,
      "theoretical_loss": 3.8743939585660785,
      "tokens_seen": 591069184
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008291675025075226,
      "loss": 1.7404,
      "theoretical_loss": 3.8743493483012172,
      "tokens_seen": 591134720
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000829147442326981,
      "loss": 1.5292,
      "theoretical_loss": 3.87430474436641,
      "tokens_seen": 591200256
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008291273821464393,
      "loss": 1.6631,
      "theoretical_loss": 3.8742601467600566,
      "tokens_seen": 591265792
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008291073219658977,
      "loss": 1.5965,
      "theoretical_loss": 3.874215555480559,
      "tokens_seen": 591331328
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008290872617853561,
      "loss": 1.6225,
      "theoretical_loss": 3.874170970526317,
      "tokens_seen": 591396864
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008290672016048144,
      "loss": 1.5969,
      "theoretical_loss": 3.874126391895734,
      "tokens_seen": 591462400
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008290471414242729,
      "loss": 1.656,
      "theoretical_loss": 3.874081819587212,
      "tokens_seen": 591527936
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008290270812437311,
      "loss": 1.5633,
      "theoretical_loss": 3.8740372535991545,
      "tokens_seen": 591593472
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008290070210631896,
      "loss": 1.7388,
      "theoretical_loss": 3.873992693929965,
      "tokens_seen": 591659008
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000828986960882648,
      "loss": 1.4588,
      "theoretical_loss": 3.8739481405780474,
      "tokens_seen": 591724544
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008289669007021063,
      "loss": 1.599,
      "theoretical_loss": 3.873903593541807,
      "tokens_seen": 591790080
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008289468405215647,
      "loss": 1.6276,
      "theoretical_loss": 3.8738590528196486,
      "tokens_seen": 591855616
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000828926780341023,
      "loss": 1.5245,
      "theoretical_loss": 3.8738145184099797,
      "tokens_seen": 591921152
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008289067201604814,
      "loss": 1.6088,
      "theoretical_loss": 3.8737699903112057,
      "tokens_seen": 591986688
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008288866599799399,
      "loss": 1.5537,
      "theoretical_loss": 3.8737254685217337,
      "tokens_seen": 592052224
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008288665997993981,
      "loss": 1.5496,
      "theoretical_loss": 3.8736809530399725,
      "tokens_seen": 592117760
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008288465396188566,
      "loss": 1.6273,
      "theoretical_loss": 3.8736364438643296,
      "tokens_seen": 592183296
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008288264794383151,
      "loss": 1.5786,
      "theoretical_loss": 3.8735919409932134,
      "tokens_seen": 592248832
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008288064192577733,
      "loss": 1.6814,
      "theoretical_loss": 3.873547444425035,
      "tokens_seen": 592314368
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008287863590772318,
      "loss": 1.6656,
      "theoretical_loss": 3.8735029541582033,
      "tokens_seen": 592379904
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00082876629889669,
      "loss": 1.6918,
      "theoretical_loss": 3.87345847019113,
      "tokens_seen": 592445440
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008287462387161485,
      "loss": 1.5025,
      "theoretical_loss": 3.8734139925222246,
      "tokens_seen": 592510976
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008287261785356069,
      "loss": 1.6032,
      "theoretical_loss": 3.8733695211499004,
      "tokens_seen": 592576512
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008287061183550652,
      "loss": 1.6024,
      "theoretical_loss": 3.8733250560725696,
      "tokens_seen": 592642048
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008286860581745236,
      "loss": 1.5592,
      "theoretical_loss": 3.8732805972886446,
      "tokens_seen": 592707584
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000828665997993982,
      "loss": 1.6073,
      "theoretical_loss": 3.873236144796539,
      "tokens_seen": 592773120
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008286459378134403,
      "loss": 1.6096,
      "theoretical_loss": 3.8731916985946677,
      "tokens_seen": 592838656
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008286258776328988,
      "loss": 1.5658,
      "theoretical_loss": 3.8731472586814446,
      "tokens_seen": 592904192
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000828605817452357,
      "loss": 1.5567,
      "theoretical_loss": 3.873102825055285,
      "tokens_seen": 592969728
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008285857572718155,
      "loss": 1.6636,
      "theoretical_loss": 3.8730583977146047,
      "tokens_seen": 593035264
    },
    {
      "epoch": 0.18,
      "objective/train/advantage_avg": 0.4914178252220154,
      "objective/train/docs_used": 221319,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5754297971725464,
      "objective/train/original_loss": 1.5754297971725464,
      "objective/train/theoretical_loss": 3.873013976657821,
      "objective/train/tokens_used": 613560800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24230973422527313,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503736734390259,
      "objective/train/weighted_lm_loss": 1.6547985076904297,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.873013976657821,
      "tokens_seen": 593100800
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008285656970912738,
      "loss": 1.6206,
      "theoretical_loss": 3.873013976657821,
      "tokens_seen": 593100800
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008285456369107322,
      "loss": 1.6143,
      "theoretical_loss": 3.87296956188335,
      "tokens_seen": 593166336
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008285255767301906,
      "loss": 1.4283,
      "theoretical_loss": 3.87292515338961,
      "tokens_seen": 593231872
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000828505516549649,
      "loss": 1.6161,
      "theoretical_loss": 3.8728807511750176,
      "tokens_seen": 593297408
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008284854563691073,
      "loss": 1.6033,
      "theoretical_loss": 3.872836355237993,
      "tokens_seen": 593362944
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008284653961885658,
      "loss": 1.6209,
      "theoretical_loss": 3.8727919655769547,
      "tokens_seen": 593428480
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008284453360080241,
      "loss": 1.5363,
      "theoretical_loss": 3.872747582190324,
      "tokens_seen": 593494016
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008284252758274825,
      "loss": 1.6599,
      "theoretical_loss": 3.872703205076519,
      "tokens_seen": 593559552
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008284052156469408,
      "loss": 1.6903,
      "theoretical_loss": 3.8726588342339623,
      "tokens_seen": 593625088
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008283851554663993,
      "loss": 1.6047,
      "theoretical_loss": 3.872614469661075,
      "tokens_seen": 593690624
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008283650952858576,
      "loss": 1.538,
      "theoretical_loss": 3.8725701113562794,
      "tokens_seen": 593756160
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000828345035105316,
      "loss": 1.5646,
      "theoretical_loss": 3.8725257593179983,
      "tokens_seen": 593821696
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008283249749247743,
      "loss": 1.5392,
      "theoretical_loss": 3.872481413544654,
      "tokens_seen": 593887232
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008283049147442327,
      "loss": 1.6154,
      "theoretical_loss": 3.8724370740346723,
      "tokens_seen": 593952768
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008282848545636911,
      "loss": 1.6049,
      "theoretical_loss": 3.8723927407864758,
      "tokens_seen": 594018304
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008282647943831495,
      "loss": 1.6822,
      "theoretical_loss": 3.8723484137984903,
      "tokens_seen": 594083840
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008282447342026078,
      "loss": 1.4928,
      "theoretical_loss": 3.872304093069141,
      "tokens_seen": 594149376
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008282246740220663,
      "loss": 1.6281,
      "theoretical_loss": 3.8722597785968547,
      "tokens_seen": 594214912
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008282046138415246,
      "loss": 1.6902,
      "theoretical_loss": 3.8722154703800573,
      "tokens_seen": 594280448
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000828184553660983,
      "loss": 1.5848,
      "theoretical_loss": 3.8721711684171765,
      "tokens_seen": 594345984
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008281644934804413,
      "loss": 1.5481,
      "theoretical_loss": 3.87212687270664,
      "tokens_seen": 594411520
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008281444332998997,
      "loss": 1.7675,
      "theoretical_loss": 3.8720825832468764,
      "tokens_seen": 594477056
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008281243731193581,
      "loss": 1.6445,
      "theoretical_loss": 3.8720383000363148,
      "tokens_seen": 594542592
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008281043129388165,
      "loss": 1.6253,
      "theoretical_loss": 3.871994023073384,
      "tokens_seen": 594608128
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008280842527582748,
      "loss": 1.5153,
      "theoretical_loss": 3.8719497523565147,
      "tokens_seen": 594673664
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008280641925777333,
      "loss": 1.6036,
      "theoretical_loss": 3.871905487884138,
      "tokens_seen": 594739200
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008280441323971915,
      "loss": 1.7277,
      "theoretical_loss": 3.871861229654684,
      "tokens_seen": 594804736
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00082802407221665,
      "loss": 1.4775,
      "theoretical_loss": 3.871816977666586,
      "tokens_seen": 594870272
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008280040120361084,
      "loss": 1.494,
      "theoretical_loss": 3.8717727319182744,
      "tokens_seen": 594935808
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008279839518555667,
      "loss": 1.6224,
      "theoretical_loss": 3.8717284924081836,
      "tokens_seen": 595001344
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008279638916750251,
      "loss": 1.5985,
      "theoretical_loss": 3.8716842591347476,
      "tokens_seen": 595066880
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008279438314944834,
      "loss": 1.5947,
      "theoretical_loss": 3.8716400320963986,
      "tokens_seen": 595132416
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008279237713139418,
      "loss": 1.6328,
      "theoretical_loss": 3.8715958112915727,
      "tokens_seen": 595197952
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008279037111334003,
      "loss": 1.592,
      "theoretical_loss": 3.871551596718705,
      "tokens_seen": 595263488
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008278836509528585,
      "loss": 1.7456,
      "theoretical_loss": 3.871507388376231,
      "tokens_seen": 595329024
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000827863590772317,
      "loss": 1.6105,
      "theoretical_loss": 3.8714631862625866,
      "tokens_seen": 595394560
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008278435305917754,
      "loss": 1.5579,
      "theoretical_loss": 3.8714189903762093,
      "tokens_seen": 595460096
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008278234704112337,
      "loss": 1.5667,
      "theoretical_loss": 3.8713748007155364,
      "tokens_seen": 595525632
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008278034102306921,
      "loss": 1.5983,
      "theoretical_loss": 3.871330617279006,
      "tokens_seen": 595591168
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008277833500501504,
      "loss": 1.6201,
      "theoretical_loss": 3.871286440065057,
      "tokens_seen": 595656704
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008277632898696088,
      "loss": 1.6441,
      "theoretical_loss": 3.8712422690721278,
      "tokens_seen": 595722240
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008277432296890673,
      "loss": 1.4712,
      "theoretical_loss": 3.8711981042986583,
      "tokens_seen": 595787776
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008277231695085255,
      "loss": 1.7621,
      "theoretical_loss": 3.8711539457430897,
      "tokens_seen": 595853312
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000827703109327984,
      "loss": 1.599,
      "theoretical_loss": 3.8711097934038614,
      "tokens_seen": 595918848
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008276830491474423,
      "loss": 1.5107,
      "theoretical_loss": 3.8710656472794156,
      "tokens_seen": 595984384
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008276629889669007,
      "loss": 1.6662,
      "theoretical_loss": 3.8710215073681944,
      "tokens_seen": 596049920
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008276429287863591,
      "loss": 1.5608,
      "theoretical_loss": 3.87097737366864,
      "tokens_seen": 596115456
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008276228686058175,
      "loss": 1.7127,
      "theoretical_loss": 3.870933246179196,
      "tokens_seen": 596180992
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008276028084252758,
      "loss": 1.5064,
      "theoretical_loss": 3.870889124898306,
      "tokens_seen": 596246528
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008275827482447342,
      "loss": 1.579,
      "theoretical_loss": 3.8708450098244134,
      "tokens_seen": 596312064
    },
    {
      "epoch": 0.18,
      "objective/train/advantage_avg": 0.4931483566761017,
      "objective/train/docs_used": 222574,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4788110256195068,
      "objective/train/original_loss": 1.4788110256195068,
      "objective/train/theoretical_loss": 3.870800900955963,
      "objective/train/tokens_used": 616837600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2440672218799591,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505554676055908,
      "objective/train/weighted_lm_loss": 1.5536764860153198,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.870800900955963,
      "tokens_seen": 596377600
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008275626880641925,
      "loss": 1.5266,
      "theoretical_loss": 3.870800900955963,
      "tokens_seen": 596377600
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000827542627883651,
      "loss": 1.5448,
      "theoretical_loss": 3.8707567982914015,
      "tokens_seen": 596443136
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008275225677031093,
      "loss": 1.7643,
      "theoretical_loss": 3.8707127018291736,
      "tokens_seen": 596508672
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008275025075225677,
      "loss": 1.7255,
      "theoretical_loss": 3.8706686115677265,
      "tokens_seen": 596574208
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008274824473420261,
      "loss": 1.4608,
      "theoretical_loss": 3.8706245275055062,
      "tokens_seen": 596639744
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008274623871614845,
      "loss": 1.6486,
      "theoretical_loss": 3.8705804496409613,
      "tokens_seen": 596705280
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008274423269809428,
      "loss": 1.5271,
      "theoretical_loss": 3.87053637797254,
      "tokens_seen": 596770816
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008274222668004012,
      "loss": 1.5131,
      "theoretical_loss": 3.8704923124986896,
      "tokens_seen": 596836352
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008274022066198595,
      "loss": 1.6762,
      "theoretical_loss": 3.8704482532178606,
      "tokens_seen": 596901888
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000827382146439318,
      "loss": 1.6253,
      "theoretical_loss": 3.870404200128503,
      "tokens_seen": 596967424
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008273620862587763,
      "loss": 1.7444,
      "theoretical_loss": 3.870360153229066,
      "tokens_seen": 597032960
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008273420260782347,
      "loss": 1.6915,
      "theoretical_loss": 3.8703161125180015,
      "tokens_seen": 597098496
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008273219658976931,
      "loss": 1.6255,
      "theoretical_loss": 3.8702720779937607,
      "tokens_seen": 597164032
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008273019057171515,
      "loss": 1.6267,
      "theoretical_loss": 3.8702280496547954,
      "tokens_seen": 597229568
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008272818455366099,
      "loss": 1.6136,
      "theoretical_loss": 3.870184027499559,
      "tokens_seen": 597295104
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008272617853560682,
      "loss": 1.482,
      "theoretical_loss": 3.8701400115265034,
      "tokens_seen": 597360640
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008272417251755267,
      "loss": 1.6012,
      "theoretical_loss": 3.8700960017340833,
      "tokens_seen": 597426176
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000827221664994985,
      "loss": 1.5897,
      "theoretical_loss": 3.870051998120753,
      "tokens_seen": 597491712
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008272016048144434,
      "loss": 1.4931,
      "theoretical_loss": 3.8700080006849666,
      "tokens_seen": 597557248
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008271815446339017,
      "loss": 1.6767,
      "theoretical_loss": 3.86996400942518,
      "tokens_seen": 597622784
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008271614844533601,
      "loss": 1.6117,
      "theoretical_loss": 3.8699200243398493,
      "tokens_seen": 597688320
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008271414242728185,
      "loss": 1.7407,
      "theoretical_loss": 3.8698760454274304,
      "tokens_seen": 597753856
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008271213640922769,
      "loss": 1.6222,
      "theoretical_loss": 3.869832072686381,
      "tokens_seen": 597819392
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008271013039117352,
      "loss": 1.6342,
      "theoretical_loss": 3.869788106115158,
      "tokens_seen": 597884928
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008270812437311937,
      "loss": 1.7497,
      "theoretical_loss": 3.8697441457122204,
      "tokens_seen": 597950464
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008270611835506519,
      "loss": 1.527,
      "theoretical_loss": 3.869700191476026,
      "tokens_seen": 598016000
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008270411233701104,
      "loss": 1.6388,
      "theoretical_loss": 3.869656243405035,
      "tokens_seen": 598081536
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008270210631895688,
      "loss": 1.6888,
      "theoretical_loss": 3.869612301497707,
      "tokens_seen": 598147072
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008270010030090271,
      "loss": 1.6937,
      "theoretical_loss": 3.8695683657525013,
      "tokens_seen": 598212608
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008269809428284855,
      "loss": 1.7128,
      "theoretical_loss": 3.8695244361678798,
      "tokens_seen": 598278144
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008269608826479438,
      "loss": 1.5467,
      "theoretical_loss": 3.869480512742304,
      "tokens_seen": 598343680
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008269408224674022,
      "loss": 1.6887,
      "theoretical_loss": 3.8694365954742356,
      "tokens_seen": 598409216
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008269207622868607,
      "loss": 1.6296,
      "theoretical_loss": 3.8693926843621376,
      "tokens_seen": 598474752
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008269007021063189,
      "loss": 1.6146,
      "theoretical_loss": 3.869348779404473,
      "tokens_seen": 598540288
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008268806419257774,
      "loss": 1.5223,
      "theoretical_loss": 3.869304880599705,
      "tokens_seen": 598605824
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008268605817452358,
      "loss": 1.65,
      "theoretical_loss": 3.8692609879462987,
      "tokens_seen": 598671360
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008268405215646941,
      "loss": 1.635,
      "theoretical_loss": 3.8692171014427177,
      "tokens_seen": 598736896
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008268204613841525,
      "loss": 1.5668,
      "theoretical_loss": 3.8691732210874283,
      "tokens_seen": 598802432
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008268004012036108,
      "loss": 1.5304,
      "theoretical_loss": 3.8691293468788963,
      "tokens_seen": 598867968
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008267803410230692,
      "loss": 1.5948,
      "theoretical_loss": 3.869085478815588,
      "tokens_seen": 598933504
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008267602808425277,
      "loss": 1.6285,
      "theoretical_loss": 3.86904161689597,
      "tokens_seen": 598999040
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008267402206619859,
      "loss": 1.6064,
      "theoretical_loss": 3.8689977611185107,
      "tokens_seen": 599064576
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008267201604814444,
      "loss": 1.6543,
      "theoretical_loss": 3.868953911481678,
      "tokens_seen": 599130112
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008267001003009027,
      "loss": 1.6609,
      "theoretical_loss": 3.86891006798394,
      "tokens_seen": 599195648
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008266800401203611,
      "loss": 1.694,
      "theoretical_loss": 3.868866230623766,
      "tokens_seen": 599261184
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008266599799398195,
      "loss": 1.7556,
      "theoretical_loss": 3.868822399399626,
      "tokens_seen": 599326720
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008266399197592779,
      "loss": 1.5449,
      "theoretical_loss": 3.8687785743099896,
      "tokens_seen": 599392256
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008266198595787362,
      "loss": 1.5468,
      "theoretical_loss": 3.868734755353329,
      "tokens_seen": 599457792
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008265997993981946,
      "loss": 1.5997,
      "theoretical_loss": 3.8686909425281146,
      "tokens_seen": 599523328
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008265797392176529,
      "loss": 1.55,
      "theoretical_loss": 3.8686471358328185,
      "tokens_seen": 599588864
    },
    {
      "epoch": 0.18,
      "objective/train/advantage_avg": 0.4909364879131317,
      "objective/train/docs_used": 223680,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6145431995391846,
      "objective/train/original_loss": 1.6145431995391846,
      "objective/train/theoretical_loss": 3.868603335265913,
      "objective/train/tokens_used": 620114400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24157437682151794,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503215789794922,
      "objective/train/weighted_lm_loss": 1.695879578590393,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.868603335265913,
      "tokens_seen": 599654400
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008265596790371114,
      "loss": 1.594,
      "theoretical_loss": 3.868603335265913,
      "tokens_seen": 599654400
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008265396188565697,
      "loss": 1.6526,
      "theoretical_loss": 3.8685595408258724,
      "tokens_seen": 599719936
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008265195586760281,
      "loss": 1.6839,
      "theoretical_loss": 3.8685157525111684,
      "tokens_seen": 599785472
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008264994984954865,
      "loss": 1.5806,
      "theoretical_loss": 3.868471970320276,
      "tokens_seen": 599851008
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008264794383149449,
      "loss": 1.6813,
      "theoretical_loss": 3.86842819425167,
      "tokens_seen": 599916544
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008264593781344032,
      "loss": 1.6634,
      "theoretical_loss": 3.8683844243038257,
      "tokens_seen": 599982080
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008264393179538616,
      "loss": 1.5757,
      "theoretical_loss": 3.8683406604752184,
      "tokens_seen": 600047616
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008264192577733199,
      "loss": 1.5867,
      "theoretical_loss": 3.8682969027643246,
      "tokens_seen": 600113152
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008263991975927784,
      "loss": 1.6753,
      "theoretical_loss": 3.8682531511696214,
      "tokens_seen": 600178688
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008263791374122367,
      "loss": 1.5887,
      "theoretical_loss": 3.868209405689586,
      "tokens_seen": 600244224
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008263590772316951,
      "loss": 1.7754,
      "theoretical_loss": 3.868165666322696,
      "tokens_seen": 600309760
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008263390170511534,
      "loss": 1.6739,
      "theoretical_loss": 3.868121933067431,
      "tokens_seen": 600375296
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008263189568706119,
      "loss": 1.589,
      "theoretical_loss": 3.8680782059222687,
      "tokens_seen": 600440832
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008262988966900702,
      "loss": 1.6292,
      "theoretical_loss": 3.8680344848856896,
      "tokens_seen": 600506368
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008262788365095286,
      "loss": 1.6832,
      "theoretical_loss": 3.8679907699561733,
      "tokens_seen": 600571904
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000826258776328987,
      "loss": 1.6069,
      "theoretical_loss": 3.8679470611322007,
      "tokens_seen": 600637440
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008262387161484454,
      "loss": 1.6778,
      "theoretical_loss": 3.8679033584122533,
      "tokens_seen": 600702976
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008262186559679037,
      "loss": 1.5571,
      "theoretical_loss": 3.867859661794812,
      "tokens_seen": 600768512
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008261985957873621,
      "loss": 1.7327,
      "theoretical_loss": 3.86781597127836,
      "tokens_seen": 600834048
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008261785356068204,
      "loss": 1.6719,
      "theoretical_loss": 3.8677722868613804,
      "tokens_seen": 600899584
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008261584754262789,
      "loss": 1.7021,
      "theoretical_loss": 3.867728608542355,
      "tokens_seen": 600965120
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008261384152457372,
      "loss": 1.5693,
      "theoretical_loss": 3.867684936319769,
      "tokens_seen": 601030656
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008261183550651956,
      "loss": 1.6478,
      "theoretical_loss": 3.867641270192107,
      "tokens_seen": 601096192
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000826098294884654,
      "loss": 1.6648,
      "theoretical_loss": 3.867597610157853,
      "tokens_seen": 601161728
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008260782347041123,
      "loss": 1.6999,
      "theoretical_loss": 3.8675539562154935,
      "tokens_seen": 601227264
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008260581745235707,
      "loss": 1.6054,
      "theoretical_loss": 3.8675103083635145,
      "tokens_seen": 601292800
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008260381143430292,
      "loss": 1.6909,
      "theoretical_loss": 3.867466666600402,
      "tokens_seen": 601358336
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008260180541624874,
      "loss": 1.7423,
      "theoretical_loss": 3.867423030924644,
      "tokens_seen": 601423872
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008259979939819459,
      "loss": 1.6517,
      "theoretical_loss": 3.8673794013347274,
      "tokens_seen": 601489408
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008259779338014041,
      "loss": 1.5173,
      "theoretical_loss": 3.8673357778291404,
      "tokens_seen": 601554944
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008259578736208626,
      "loss": 1.594,
      "theoretical_loss": 3.867292160406373,
      "tokens_seen": 601620480
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000825937813440321,
      "loss": 1.5296,
      "theoretical_loss": 3.8672485490649136,
      "tokens_seen": 601686016
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008259177532597793,
      "loss": 1.6407,
      "theoretical_loss": 3.8672049438032525,
      "tokens_seen": 601751552
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008258976930792377,
      "loss": 1.515,
      "theoretical_loss": 3.8671613446198796,
      "tokens_seen": 601817088
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008258776328986962,
      "loss": 1.694,
      "theoretical_loss": 3.8671177515132857,
      "tokens_seen": 601882624
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008258575727181544,
      "loss": 1.7133,
      "theoretical_loss": 3.8670741644819637,
      "tokens_seen": 601948160
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008258375125376129,
      "loss": 1.5727,
      "theoretical_loss": 3.867030583524404,
      "tokens_seen": 602013696
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008258174523570712,
      "loss": 1.6105,
      "theoretical_loss": 3.8669870086391,
      "tokens_seen": 602079232
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008257973921765296,
      "loss": 1.6828,
      "theoretical_loss": 3.866943439824545,
      "tokens_seen": 602144768
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008257773319959881,
      "loss": 1.4657,
      "theoretical_loss": 3.866899877079232,
      "tokens_seen": 602210304
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008257572718154463,
      "loss": 1.6577,
      "theoretical_loss": 3.8668563204016557,
      "tokens_seen": 602275840
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008257372116349048,
      "loss": 1.5897,
      "theoretical_loss": 3.86681276979031,
      "tokens_seen": 602341376
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008257171514543631,
      "loss": 1.6591,
      "theoretical_loss": 3.8667692252436914,
      "tokens_seen": 602406912
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008256970912738215,
      "loss": 1.7114,
      "theoretical_loss": 3.866725686760295,
      "tokens_seen": 602472448
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008256770310932799,
      "loss": 1.5512,
      "theoretical_loss": 3.8666821543386174,
      "tokens_seen": 602537984
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008256569709127383,
      "loss": 1.579,
      "theoretical_loss": 3.8666386279771547,
      "tokens_seen": 602603520
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008256369107321966,
      "loss": 1.6909,
      "theoretical_loss": 3.8665951076744056,
      "tokens_seen": 602669056
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000825616850551655,
      "loss": 1.5716,
      "theoretical_loss": 3.8665515934288672,
      "tokens_seen": 602734592
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008255967903711133,
      "loss": 1.5794,
      "theoretical_loss": 3.8665080852390377,
      "tokens_seen": 602800128
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008255767301905718,
      "loss": 1.6037,
      "theoretical_loss": 3.866464583103417,
      "tokens_seen": 602865664
    },
    {
      "epoch": 0.18,
      "objective/train/advantage_avg": 0.49399736523628235,
      "objective/train/docs_used": 224736,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.500554084777832,
      "objective/train/original_loss": 1.5005537271499634,
      "objective/train/theoretical_loss": 3.866421087020504,
      "objective/train/tokens_used": 623391200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24457556009292603,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506432056427002,
      "objective/train/weighted_lm_loss": 1.5765460729599,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.866421087020504,
      "tokens_seen": 602931200
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008255566700100301,
      "loss": 1.5648,
      "theoretical_loss": 3.866421087020504,
      "tokens_seen": 602931200
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008255366098294885,
      "loss": 1.619,
      "theoretical_loss": 3.866377596988799,
      "tokens_seen": 602996736
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008255165496489469,
      "loss": 1.5137,
      "theoretical_loss": 3.8663341130068023,
      "tokens_seen": 603062272
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008254964894684053,
      "loss": 1.5574,
      "theoretical_loss": 3.8662906350730157,
      "tokens_seen": 603127808
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008254764292878636,
      "loss": 1.4186,
      "theoretical_loss": 3.8662471631859407,
      "tokens_seen": 603193344
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000825456369107322,
      "loss": 1.701,
      "theoretical_loss": 3.8662036973440794,
      "tokens_seen": 603258880
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008254363089267803,
      "loss": 1.6054,
      "theoretical_loss": 3.866160237545934,
      "tokens_seen": 603324416
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008254162487462388,
      "loss": 1.6217,
      "theoretical_loss": 3.866116783790009,
      "tokens_seen": 603389952
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008253961885656971,
      "loss": 1.612,
      "theoretical_loss": 3.866073336074807,
      "tokens_seen": 603455488
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008253761283851555,
      "loss": 1.5529,
      "theoretical_loss": 3.8660298943988334,
      "tokens_seen": 603521024
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008253560682046138,
      "loss": 1.5292,
      "theoretical_loss": 3.8659864587605917,
      "tokens_seen": 603586560
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008253360080240723,
      "loss": 1.6586,
      "theoretical_loss": 3.865943029158589,
      "tokens_seen": 603652096
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008253159478435306,
      "loss": 1.5184,
      "theoretical_loss": 3.8658996055913297,
      "tokens_seen": 603717632
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000825295887662989,
      "loss": 1.6554,
      "theoretical_loss": 3.8658561880573217,
      "tokens_seen": 603783168
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008252758274824474,
      "loss": 1.6835,
      "theoretical_loss": 3.865812776555071,
      "tokens_seen": 603848704
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008252557673019058,
      "loss": 1.6259,
      "theoretical_loss": 3.8657693710830863,
      "tokens_seen": 603914240
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008252357071213641,
      "loss": 1.6972,
      "theoretical_loss": 3.8657259716398737,
      "tokens_seen": 603979776
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008252156469408225,
      "loss": 1.5604,
      "theoretical_loss": 3.865682578223944,
      "tokens_seen": 604045312
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008251955867602808,
      "loss": 1.6027,
      "theoretical_loss": 3.8656391908338046,
      "tokens_seen": 604110848
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008251755265797393,
      "loss": 1.6397,
      "theoretical_loss": 3.8655958094679663,
      "tokens_seen": 604176384
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008251554663991976,
      "loss": 1.5769,
      "theoretical_loss": 3.8655524341249388,
      "tokens_seen": 604241920
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000825135406218656,
      "loss": 1.5215,
      "theoretical_loss": 3.8655090648032333,
      "tokens_seen": 604307456
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008251153460381144,
      "loss": 1.7109,
      "theoretical_loss": 3.8654657015013605,
      "tokens_seen": 604372992
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008250952858575727,
      "loss": 1.7039,
      "theoretical_loss": 3.8654223442178317,
      "tokens_seen": 604438528
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008250752256770311,
      "loss": 1.6434,
      "theoretical_loss": 3.865378992951161,
      "tokens_seen": 604504064
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008250551654964896,
      "loss": 1.6853,
      "theoretical_loss": 3.865335647699859,
      "tokens_seen": 604569600
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008250351053159478,
      "loss": 1.6419,
      "theoretical_loss": 3.8652923084624415,
      "tokens_seen": 604635136
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008250150451354063,
      "loss": 1.577,
      "theoretical_loss": 3.86524897523742,
      "tokens_seen": 604700672
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008249949849548645,
      "loss": 1.5877,
      "theoretical_loss": 3.865205648023311,
      "tokens_seen": 604766208
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000824974924774323,
      "loss": 1.6295,
      "theoretical_loss": 3.865162326818628,
      "tokens_seen": 604831744
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008249548645937814,
      "loss": 1.6224,
      "theoretical_loss": 3.865119011621887,
      "tokens_seen": 604897280
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008249348044132397,
      "loss": 1.588,
      "theoretical_loss": 3.8650757024316045,
      "tokens_seen": 604962816
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008249147442326981,
      "loss": 1.6898,
      "theoretical_loss": 3.8650323992462963,
      "tokens_seen": 605028352
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008248946840521566,
      "loss": 1.4756,
      "theoretical_loss": 3.8649891020644795,
      "tokens_seen": 605093888
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008248746238716148,
      "loss": 1.6003,
      "theoretical_loss": 3.8649458108846724,
      "tokens_seen": 605159424
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008248545636910733,
      "loss": 1.7556,
      "theoretical_loss": 3.864902525705393,
      "tokens_seen": 605224960
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008248345035105315,
      "loss": 1.6128,
      "theoretical_loss": 3.8648592465251586,
      "tokens_seen": 605290496
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00082481444332999,
      "loss": 1.6823,
      "theoretical_loss": 3.8648159733424903,
      "tokens_seen": 605356032
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008247943831494484,
      "loss": 1.6702,
      "theoretical_loss": 3.8647727061559065,
      "tokens_seen": 605421568
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008247743229689067,
      "loss": 1.586,
      "theoretical_loss": 3.8647294449639276,
      "tokens_seen": 605487104
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008247542627883651,
      "loss": 1.5767,
      "theoretical_loss": 3.864686189765075,
      "tokens_seen": 605552640
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008247342026078235,
      "loss": 1.7416,
      "theoretical_loss": 3.8646429405578697,
      "tokens_seen": 605618176
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008247141424272818,
      "loss": 1.6241,
      "theoretical_loss": 3.864599697340833,
      "tokens_seen": 605683712
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008246940822467403,
      "loss": 1.5612,
      "theoretical_loss": 3.864556460112488,
      "tokens_seen": 605749248
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008246740220661985,
      "loss": 1.6083,
      "theoretical_loss": 3.864513228871357,
      "tokens_seen": 605814784
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000824653961885657,
      "loss": 1.6448,
      "theoretical_loss": 3.8644700036159634,
      "tokens_seen": 605880320
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008246339017051153,
      "loss": 1.5935,
      "theoretical_loss": 3.8644267843448317,
      "tokens_seen": 605945856
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008246138415245737,
      "loss": 1.5513,
      "theoretical_loss": 3.8643835710564858,
      "tokens_seen": 606011392
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008245937813440321,
      "loss": 1.6612,
      "theoretical_loss": 3.8643403637494504,
      "tokens_seen": 606076928
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008245737211634905,
      "loss": 1.5263,
      "theoretical_loss": 3.8642971624222513,
      "tokens_seen": 606142464
    },
    {
      "epoch": 0.18,
      "objective/train/advantage_avg": 0.49447712302207947,
      "objective/train/docs_used": 225863,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5234355926513672,
      "objective/train/original_loss": 1.5234355926513672,
      "objective/train/theoretical_loss": 3.8642539670734144,
      "objective/train/tokens_used": 626668000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2446836531162262,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506914854049683,
      "objective/train/weighted_lm_loss": 1.6006100177764893,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0293704271316528,
      "theoretical_loss": 3.8642539670734144,
      "tokens_seen": 606208000
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008245536609829488,
      "loss": 1.5679,
      "theoretical_loss": 3.8642539670734144,
      "tokens_seen": 606208000
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008245336008024073,
      "loss": 1.6517,
      "theoretical_loss": 3.8642107777014667,
      "tokens_seen": 606273536
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008245135406218656,
      "loss": 1.6136,
      "theoretical_loss": 3.8641675943049343,
      "tokens_seen": 606339072
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000824493480441324,
      "loss": 1.6884,
      "theoretical_loss": 3.8641244168823454,
      "tokens_seen": 606404608
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008244734202607823,
      "loss": 1.633,
      "theoretical_loss": 3.8640812454322284,
      "tokens_seen": 606470144
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008244533600802407,
      "loss": 1.6995,
      "theoretical_loss": 3.864038079953111,
      "tokens_seen": 606535680
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008244332998996991,
      "loss": 1.6062,
      "theoretical_loss": 3.863994920443523,
      "tokens_seen": 606601216
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008244132397191575,
      "loss": 1.6901,
      "theoretical_loss": 3.8639517669019945,
      "tokens_seen": 606666752
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008243931795386158,
      "loss": 1.5642,
      "theoretical_loss": 3.863908619327054,
      "tokens_seen": 606732288
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008243731193580742,
      "loss": 1.6432,
      "theoretical_loss": 3.863865477717234,
      "tokens_seen": 606797824
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008243530591775326,
      "loss": 1.6101,
      "theoretical_loss": 3.8638223420710647,
      "tokens_seen": 606863360
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000824332998996991,
      "loss": 1.5649,
      "theoretical_loss": 3.8637792123870778,
      "tokens_seen": 606928896
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008243129388164494,
      "loss": 1.4683,
      "theoretical_loss": 3.863736088663806,
      "tokens_seen": 606994432
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008242928786359078,
      "loss": 1.6181,
      "theoretical_loss": 3.8636929708997814,
      "tokens_seen": 607059968
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008242728184553662,
      "loss": 1.6326,
      "theoretical_loss": 3.863649859093538,
      "tokens_seen": 607125504
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008242527582748245,
      "loss": 1.5195,
      "theoretical_loss": 3.8636067532436105,
      "tokens_seen": 607191040
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000824232698094283,
      "loss": 1.5985,
      "theoretical_loss": 3.863563653348531,
      "tokens_seen": 607256576
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008242126379137412,
      "loss": 1.6593,
      "theoretical_loss": 3.8635205594068354,
      "tokens_seen": 607322112
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008241925777331997,
      "loss": 1.564,
      "theoretical_loss": 3.863477471417059,
      "tokens_seen": 607387648
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000824172517552658,
      "loss": 1.5371,
      "theoretical_loss": 3.8634343893777383,
      "tokens_seen": 607453184
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008241524573721164,
      "loss": 1.6972,
      "theoretical_loss": 3.8633913132874094,
      "tokens_seen": 607518720
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008241323971915748,
      "loss": 1.6091,
      "theoretical_loss": 3.8633482431446082,
      "tokens_seen": 607584256
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008241123370110331,
      "loss": 1.7134,
      "theoretical_loss": 3.8633051789478734,
      "tokens_seen": 607649792
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008240922768304915,
      "loss": 1.6356,
      "theoretical_loss": 3.8632621206957425,
      "tokens_seen": 607715328
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00082407221664995,
      "loss": 1.7405,
      "theoretical_loss": 3.863219068386754,
      "tokens_seen": 607780864
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008240521564694082,
      "loss": 1.6688,
      "theoretical_loss": 3.863176022019447,
      "tokens_seen": 607846400
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008240320962888667,
      "loss": 1.6639,
      "theoretical_loss": 3.8631329815923605,
      "tokens_seen": 607911936
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008240120361083249,
      "loss": 1.5745,
      "theoretical_loss": 3.8630899471040347,
      "tokens_seen": 607977472
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008239919759277834,
      "loss": 1.6226,
      "theoretical_loss": 3.863046918553011,
      "tokens_seen": 608043008
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008239719157472418,
      "loss": 1.483,
      "theoretical_loss": 3.8630038959378292,
      "tokens_seen": 608108544
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008239518555667001,
      "loss": 1.6049,
      "theoretical_loss": 3.862960879257032,
      "tokens_seen": 608174080
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008239317953861585,
      "loss": 1.5995,
      "theoretical_loss": 3.86291786850916,
      "tokens_seen": 608239616
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000823911735205617,
      "loss": 1.7919,
      "theoretical_loss": 3.862874863692758,
      "tokens_seen": 608305152
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008238916750250752,
      "loss": 1.629,
      "theoretical_loss": 3.862831864806367,
      "tokens_seen": 608370688
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008238716148445337,
      "loss": 1.6048,
      "theoretical_loss": 3.8627888718485313,
      "tokens_seen": 608436224
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008238515546639919,
      "loss": 1.4945,
      "theoretical_loss": 3.862745884817796,
      "tokens_seen": 608501760
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008238314944834504,
      "loss": 1.7123,
      "theoretical_loss": 3.8627029037127043,
      "tokens_seen": 608567296
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008238114343029088,
      "loss": 1.5196,
      "theoretical_loss": 3.862659928531802,
      "tokens_seen": 608632832
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008237913741223671,
      "loss": 1.6041,
      "theoretical_loss": 3.862616959273635,
      "tokens_seen": 608698368
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008237713139418255,
      "loss": 1.5851,
      "theoretical_loss": 3.8625739959367493,
      "tokens_seen": 608763904
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008237512537612839,
      "loss": 1.6814,
      "theoretical_loss": 3.862531038519692,
      "tokens_seen": 608829440
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008237311935807422,
      "loss": 1.5637,
      "theoretical_loss": 3.8624880870210094,
      "tokens_seen": 608894976
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008237111334002007,
      "loss": 1.5645,
      "theoretical_loss": 3.8624451414392498,
      "tokens_seen": 608960512
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008236910732196589,
      "loss": 1.6494,
      "theoretical_loss": 3.8624022017729622,
      "tokens_seen": 609026048
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008236710130391174,
      "loss": 1.6423,
      "theoretical_loss": 3.862359268020694,
      "tokens_seen": 609091584
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008236509528585757,
      "loss": 1.5312,
      "theoretical_loss": 3.8623163401809952,
      "tokens_seen": 609157120
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008236308926780341,
      "loss": 1.4567,
      "theoretical_loss": 3.8622734182524154,
      "tokens_seen": 609222656
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008236108324974925,
      "loss": 1.6388,
      "theoretical_loss": 3.862230502233505,
      "tokens_seen": 609288192
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008235907723169509,
      "loss": 1.5897,
      "theoretical_loss": 3.8621875921228153,
      "tokens_seen": 609353728
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008235707121364092,
      "loss": 1.6345,
      "theoretical_loss": 3.8621446879188963,
      "tokens_seen": 609419264
    },
    {
      "epoch": 0.18,
      "objective/train/advantage_avg": 0.49629247188568115,
      "objective/train/docs_used": 227108,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4768539667129517,
      "objective/train/original_loss": 1.4768540859222412,
      "objective/train/theoretical_loss": 3.8621017896203007,
      "objective/train/tokens_used": 629944800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24641580879688263,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05088210105896,
      "objective/train/weighted_lm_loss": 1.5519256591796875,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.8621017896203007,
      "tokens_seen": 609484800
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008235506519558677,
      "loss": 1.5544,
      "theoretical_loss": 3.8621017896203007,
      "tokens_seen": 609484800
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000823530591775326,
      "loss": 1.6336,
      "theoretical_loss": 3.862058897225581,
      "tokens_seen": 609550336
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008235105315947844,
      "loss": 1.5458,
      "theoretical_loss": 3.8620160107332904,
      "tokens_seen": 609615872
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008234904714142427,
      "loss": 1.5553,
      "theoretical_loss": 3.861973130141981,
      "tokens_seen": 609681408
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008234704112337011,
      "loss": 1.5939,
      "theoretical_loss": 3.8619302554502077,
      "tokens_seen": 609746944
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008234503510531595,
      "loss": 1.6292,
      "theoretical_loss": 3.861887386656524,
      "tokens_seen": 609812480
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008234302908726179,
      "loss": 1.4184,
      "theoretical_loss": 3.861844523759486,
      "tokens_seen": 609878016
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008234102306920762,
      "loss": 1.6175,
      "theoretical_loss": 3.861801666757648,
      "tokens_seen": 609943552
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008233901705115346,
      "loss": 1.6142,
      "theoretical_loss": 3.8617588156495666,
      "tokens_seen": 610009088
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000823370110330993,
      "loss": 1.7761,
      "theoretical_loss": 3.8617159704337984,
      "tokens_seen": 610074624
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008233500501504514,
      "loss": 1.5401,
      "theoretical_loss": 3.8616731311088994,
      "tokens_seen": 610140160
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008233299899699097,
      "loss": 1.5921,
      "theoretical_loss": 3.8616302976734276,
      "tokens_seen": 610205696
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008233099297893682,
      "loss": 1.7097,
      "theoretical_loss": 3.861587470125941,
      "tokens_seen": 610271232
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008232898696088264,
      "loss": 1.5588,
      "theoretical_loss": 3.8615446484649976,
      "tokens_seen": 610336768
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008232698094282849,
      "loss": 1.6388,
      "theoretical_loss": 3.861501832689157,
      "tokens_seen": 610402304
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0008232497492477432,
      "loss": 1.5997,
      "theoretical_loss": 3.861459022796978,
      "tokens_seen": 610467840
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008232296890672016,
      "loss": 1.6778,
      "theoretical_loss": 3.8614162187870216,
      "tokens_seen": 610533376
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00082320962888666,
      "loss": 1.4831,
      "theoretical_loss": 3.861373420657847,
      "tokens_seen": 610598912
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008231895687061184,
      "loss": 1.7374,
      "theoretical_loss": 3.861330628408016,
      "tokens_seen": 610664448
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008231695085255767,
      "loss": 1.6045,
      "theoretical_loss": 3.8612878420360905,
      "tokens_seen": 610729984
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008231494483450352,
      "loss": 1.602,
      "theoretical_loss": 3.861245061540631,
      "tokens_seen": 610795520
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008231293881644934,
      "loss": 1.6448,
      "theoretical_loss": 3.861202286920202,
      "tokens_seen": 610861056
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008231093279839519,
      "loss": 1.6841,
      "theoretical_loss": 3.8611595181733644,
      "tokens_seen": 610926592
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008230892678034102,
      "loss": 1.5038,
      "theoretical_loss": 3.861116755298683,
      "tokens_seen": 610992128
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008230692076228686,
      "loss": 1.5688,
      "theoretical_loss": 3.8610739982947218,
      "tokens_seen": 611057664
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000823049147442327,
      "loss": 1.5498,
      "theoretical_loss": 3.861031247160045,
      "tokens_seen": 611123200
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008230290872617853,
      "loss": 1.5346,
      "theoretical_loss": 3.8609885018932175,
      "tokens_seen": 611188736
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008230090270812437,
      "loss": 1.5068,
      "theoretical_loss": 3.860945762492806,
      "tokens_seen": 611254272
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008229889669007022,
      "loss": 1.5697,
      "theoretical_loss": 3.8609030289573747,
      "tokens_seen": 611319808
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008229689067201604,
      "loss": 1.6114,
      "theoretical_loss": 3.8608603012854914,
      "tokens_seen": 611385344
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008229488465396189,
      "loss": 1.6014,
      "theoretical_loss": 3.860817579475723,
      "tokens_seen": 611450880
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008229287863590771,
      "loss": 1.6972,
      "theoretical_loss": 3.860774863526637,
      "tokens_seen": 611516416
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008229087261785356,
      "loss": 1.6748,
      "theoretical_loss": 3.8607321534368007,
      "tokens_seen": 611581952
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000822888665997994,
      "loss": 1.6369,
      "theoretical_loss": 3.860689449204784,
      "tokens_seen": 611647488
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008228686058174523,
      "loss": 1.5479,
      "theoretical_loss": 3.860646750829155,
      "tokens_seen": 611713024
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008228485456369107,
      "loss": 1.4678,
      "theoretical_loss": 3.8606040583084833,
      "tokens_seen": 611778560
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008228284854563692,
      "loss": 1.5103,
      "theoretical_loss": 3.8605613716413396,
      "tokens_seen": 611844096
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008228084252758275,
      "loss": 1.5572,
      "theoretical_loss": 3.860518690826294,
      "tokens_seen": 611909632
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008227883650952859,
      "loss": 1.7006,
      "theoretical_loss": 3.8604760158619174,
      "tokens_seen": 611975168
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008227683049147443,
      "loss": 1.5919,
      "theoretical_loss": 3.860433346746782,
      "tokens_seen": 612040704
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008227482447342026,
      "loss": 1.7025,
      "theoretical_loss": 3.860390683479459,
      "tokens_seen": 612106240
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008227281845536611,
      "loss": 1.4717,
      "theoretical_loss": 3.860348026058522,
      "tokens_seen": 612171776
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008227081243731193,
      "loss": 1.5342,
      "theoretical_loss": 3.8603053744825435,
      "tokens_seen": 612237312
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008226880641925778,
      "loss": 1.5627,
      "theoretical_loss": 3.8602627287500972,
      "tokens_seen": 612302848
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008226680040120361,
      "loss": 1.6263,
      "theoretical_loss": 3.860220088859757,
      "tokens_seen": 612368384
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008226479438314945,
      "loss": 1.6842,
      "theoretical_loss": 3.8601774548100973,
      "tokens_seen": 612433920
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008226278836509529,
      "loss": 1.3856,
      "theoretical_loss": 3.860134826599694,
      "tokens_seen": 612499456
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008226078234704113,
      "loss": 1.5841,
      "theoretical_loss": 3.860092204227122,
      "tokens_seen": 612564992
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008225877632898696,
      "loss": 1.5106,
      "theoretical_loss": 3.860049587690958,
      "tokens_seen": 612630528
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008225677031093281,
      "loss": 1.6423,
      "theoretical_loss": 3.860006976989778,
      "tokens_seen": 612696064
    },
    {
      "epoch": 0.19,
      "objective/train/advantage_avg": 0.4955649971961975,
      "objective/train/docs_used": 228263,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6900759935379028,
      "objective/train/original_loss": 1.6900761127471924,
      "objective/train/theoretical_loss": 3.8599643721221586,
      "objective/train/tokens_used": 633221600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24580927193164825,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508062839508057,
      "objective/train/weighted_lm_loss": 1.7758455276489258,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0281672477722168,
      "theoretical_loss": 3.8599643721221586,
      "tokens_seen": 612761600
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008225476429287864,
      "loss": 1.6875,
      "theoretical_loss": 3.8599643721221586,
      "tokens_seen": 612761600
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008225275827482448,
      "loss": 1.4918,
      "theoretical_loss": 3.859921773086679,
      "tokens_seen": 612827136
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008225075225677031,
      "loss": 1.5538,
      "theoretical_loss": 3.8598791798819154,
      "tokens_seen": 612892672
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008224874623871615,
      "loss": 1.4891,
      "theoretical_loss": 3.859836592506448,
      "tokens_seen": 612958208
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008224674022066199,
      "loss": 1.5989,
      "theoretical_loss": 3.8597940109588547,
      "tokens_seen": 613023744
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008224473420260783,
      "loss": 1.6383,
      "theoretical_loss": 3.859751435237716,
      "tokens_seen": 613089280
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008224272818455366,
      "loss": 1.4104,
      "theoretical_loss": 3.859708865341611,
      "tokens_seen": 613154816
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000822407221664995,
      "loss": 1.6049,
      "theoretical_loss": 3.8596663012691206,
      "tokens_seen": 613220352
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008223871614844534,
      "loss": 1.6655,
      "theoretical_loss": 3.859623743018827,
      "tokens_seen": 613285888
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008223671013039118,
      "loss": 1.6336,
      "theoretical_loss": 3.85958119058931,
      "tokens_seen": 613351424
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008223470411233701,
      "loss": 1.6416,
      "theoretical_loss": 3.8595386439791532,
      "tokens_seen": 613416960
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008223269809428286,
      "loss": 1.5586,
      "theoretical_loss": 3.859496103186938,
      "tokens_seen": 613482496
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008223069207622868,
      "loss": 1.6274,
      "theoretical_loss": 3.859453568211248,
      "tokens_seen": 613548032
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008222868605817453,
      "loss": 1.6393,
      "theoretical_loss": 3.8594110390506664,
      "tokens_seen": 613613568
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008222668004012036,
      "loss": 1.5215,
      "theoretical_loss": 3.859368515703778,
      "tokens_seen": 613679104
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000822246740220662,
      "loss": 1.4824,
      "theoretical_loss": 3.8593259981691665,
      "tokens_seen": 613744640
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008222266800401204,
      "loss": 1.5944,
      "theoretical_loss": 3.8592834864454177,
      "tokens_seen": 613810176
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008222066198595788,
      "loss": 1.5089,
      "theoretical_loss": 3.8592409805311165,
      "tokens_seen": 613875712
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008221865596790371,
      "loss": 1.6413,
      "theoretical_loss": 3.859198480424849,
      "tokens_seen": 613941248
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008221664994984956,
      "loss": 1.5523,
      "theoretical_loss": 3.859155986125202,
      "tokens_seen": 614006784
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008221464393179538,
      "loss": 1.6317,
      "theoretical_loss": 3.859113497630762,
      "tokens_seen": 614072320
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008221263791374123,
      "loss": 1.6131,
      "theoretical_loss": 3.8590710149401177,
      "tokens_seen": 614137856
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008221063189568706,
      "loss": 1.4801,
      "theoretical_loss": 3.859028538051856,
      "tokens_seen": 614203392
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000822086258776329,
      "loss": 1.7205,
      "theoretical_loss": 3.8589860669645653,
      "tokens_seen": 614268928
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008220661985957874,
      "loss": 1.6196,
      "theoretical_loss": 3.8589436016768355,
      "tokens_seen": 614334464
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008220461384152457,
      "loss": 1.5722,
      "theoretical_loss": 3.858901142187256,
      "tokens_seen": 614400000
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008220260782347041,
      "loss": 1.4802,
      "theoretical_loss": 3.858858688494416,
      "tokens_seen": 614465536
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008220060180541626,
      "loss": 1.6181,
      "theoretical_loss": 3.8588162405969064,
      "tokens_seen": 614531072
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008219859578736208,
      "loss": 1.473,
      "theoretical_loss": 3.8587737984933184,
      "tokens_seen": 614596608
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008219658976930793,
      "loss": 1.6362,
      "theoretical_loss": 3.8587313621822434,
      "tokens_seen": 614662144
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008219458375125375,
      "loss": 1.5643,
      "theoretical_loss": 3.8586889316622726,
      "tokens_seen": 614727680
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000821925777331996,
      "loss": 1.5142,
      "theoretical_loss": 3.8586465069319997,
      "tokens_seen": 614793216
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008219057171514544,
      "loss": 1.4805,
      "theoretical_loss": 3.8586040879900168,
      "tokens_seen": 614858752
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008218856569709127,
      "loss": 1.4964,
      "theoretical_loss": 3.8585616748349167,
      "tokens_seen": 614924288
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008218655967903711,
      "loss": 1.5681,
      "theoretical_loss": 3.8585192674652955,
      "tokens_seen": 614989824
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008218455366098296,
      "loss": 1.4802,
      "theoretical_loss": 3.8584768658797453,
      "tokens_seen": 615055360
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008218254764292878,
      "loss": 1.6116,
      "theoretical_loss": 3.8584344700768622,
      "tokens_seen": 615120896
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008218054162487463,
      "loss": 1.5727,
      "theoretical_loss": 3.858392080055242,
      "tokens_seen": 615186432
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008217853560682046,
      "loss": 1.6323,
      "theoretical_loss": 3.8583496958134793,
      "tokens_seen": 615251968
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000821765295887663,
      "loss": 1.5314,
      "theoretical_loss": 3.858307317350171,
      "tokens_seen": 615317504
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008217452357071214,
      "loss": 1.4508,
      "theoretical_loss": 3.8582649446639143,
      "tokens_seen": 615383040
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008217251755265797,
      "loss": 1.541,
      "theoretical_loss": 3.858222577753307,
      "tokens_seen": 615448576
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008217051153460381,
      "loss": 1.6024,
      "theoretical_loss": 3.8581802166169457,
      "tokens_seen": 615514112
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008216850551654965,
      "loss": 1.6622,
      "theoretical_loss": 3.8581378612534296,
      "tokens_seen": 615579648
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008216649949849548,
      "loss": 1.5519,
      "theoretical_loss": 3.8580955116613573,
      "tokens_seen": 615645184
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008216449348044133,
      "loss": 1.5248,
      "theoretical_loss": 3.858053167839328,
      "tokens_seen": 615710720
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008216248746238716,
      "loss": 1.6547,
      "theoretical_loss": 3.8580108297859415,
      "tokens_seen": 615776256
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00082160481444333,
      "loss": 1.5183,
      "theoretical_loss": 3.8579684974997983,
      "tokens_seen": 615841792
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008215847542627884,
      "loss": 1.4812,
      "theoretical_loss": 3.8579261709794994,
      "tokens_seen": 615907328
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008215646940822468,
      "loss": 1.5726,
      "theoretical_loss": 3.8578838502236454,
      "tokens_seen": 615972864
    },
    {
      "epoch": 0.19,
      "objective/train/advantage_avg": 0.492830753326416,
      "objective/train/docs_used": 229528,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.466249704360962,
      "objective/train/original_loss": 1.466249704360962,
      "objective/train/theoretical_loss": 3.857841535230839,
      "objective/train/tokens_used": 636498400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24335187673568726,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505203008651733,
      "objective/train/weighted_lm_loss": 1.5402671098709106,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.857841535230839,
      "tokens_seen": 616038400
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008215446339017051,
      "loss": 1.608,
      "theoretical_loss": 3.857841535230839,
      "tokens_seen": 616038400
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008215245737211635,
      "loss": 1.5664,
      "theoretical_loss": 3.8577992259996816,
      "tokens_seen": 616103936
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008215045135406218,
      "loss": 1.5859,
      "theoretical_loss": 3.8577569225287762,
      "tokens_seen": 616169472
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008214844533600803,
      "loss": 1.558,
      "theoretical_loss": 3.857714624816726,
      "tokens_seen": 616235008
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008214643931795386,
      "loss": 1.6013,
      "theoretical_loss": 3.8576723328621347,
      "tokens_seen": 616300544
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000821444332998997,
      "loss": 1.6573,
      "theoretical_loss": 3.8576300466636066,
      "tokens_seen": 616366080
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008214242728184553,
      "loss": 1.5305,
      "theoretical_loss": 3.857587766219747,
      "tokens_seen": 616431616
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008214042126379138,
      "loss": 1.5478,
      "theoretical_loss": 3.8575454915291596,
      "tokens_seen": 616497152
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008213841524573721,
      "loss": 1.6022,
      "theoretical_loss": 3.8575032225904513,
      "tokens_seen": 616562688
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008213640922768305,
      "loss": 1.5505,
      "theoretical_loss": 3.8574609594022276,
      "tokens_seen": 616628224
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008213440320962888,
      "loss": 1.7057,
      "theoretical_loss": 3.8574187019630957,
      "tokens_seen": 616693760
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008213239719157472,
      "loss": 1.4051,
      "theoretical_loss": 3.857376450271662,
      "tokens_seen": 616759296
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008213039117352057,
      "loss": 1.5181,
      "theoretical_loss": 3.8573342043265346,
      "tokens_seen": 616824832
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000821283851554664,
      "loss": 1.6967,
      "theoretical_loss": 3.857291964126322,
      "tokens_seen": 616890368
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008212637913741224,
      "loss": 1.7542,
      "theoretical_loss": 3.8572497296696313,
      "tokens_seen": 616955904
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008212437311935808,
      "loss": 1.5992,
      "theoretical_loss": 3.857207500955073,
      "tokens_seen": 617021440
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008212236710130392,
      "loss": 1.6053,
      "theoretical_loss": 3.857165277981256,
      "tokens_seen": 617086976
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008212036108324975,
      "loss": 1.6847,
      "theoretical_loss": 3.8571230607467912,
      "tokens_seen": 617152512
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000821183550651956,
      "loss": 1.5465,
      "theoretical_loss": 3.8570808492502877,
      "tokens_seen": 617218048
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008211634904714142,
      "loss": 1.6344,
      "theoretical_loss": 3.8570386434903567,
      "tokens_seen": 617283584
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008211434302908727,
      "loss": 1.5999,
      "theoretical_loss": 3.8569964434656105,
      "tokens_seen": 617349120
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000821123370110331,
      "loss": 1.6031,
      "theoretical_loss": 3.8569542491746613,
      "tokens_seen": 617414656
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008211033099297894,
      "loss": 1.5452,
      "theoretical_loss": 3.8569120606161205,
      "tokens_seen": 617480192
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008210832497492478,
      "loss": 1.6047,
      "theoretical_loss": 3.8568698777886015,
      "tokens_seen": 617545728
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008210631895687061,
      "loss": 1.7153,
      "theoretical_loss": 3.856827700690718,
      "tokens_seen": 617611264
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008210431293881645,
      "loss": 1.574,
      "theoretical_loss": 3.856785529321083,
      "tokens_seen": 617676800
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000821023069207623,
      "loss": 1.6403,
      "theoretical_loss": 3.8567433636783113,
      "tokens_seen": 617742336
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008210030090270812,
      "loss": 1.637,
      "theoretical_loss": 3.8567012037610184,
      "tokens_seen": 617807872
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008209829488465397,
      "loss": 1.493,
      "theoretical_loss": 3.8566590495678192,
      "tokens_seen": 617873408
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008209628886659979,
      "loss": 1.6298,
      "theoretical_loss": 3.8566169010973295,
      "tokens_seen": 617938944
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008209428284854564,
      "loss": 1.6373,
      "theoretical_loss": 3.856574758348165,
      "tokens_seen": 618004480
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008209227683049148,
      "loss": 1.6059,
      "theoretical_loss": 3.8565326213189435,
      "tokens_seen": 618070016
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008209027081243731,
      "loss": 1.6655,
      "theoretical_loss": 3.856490490008282,
      "tokens_seen": 618135552
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008208826479438315,
      "loss": 1.5592,
      "theoretical_loss": 3.8564483644147978,
      "tokens_seen": 618201088
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00082086258776329,
      "loss": 1.7345,
      "theoretical_loss": 3.8564062445371095,
      "tokens_seen": 618266624
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008208425275827482,
      "loss": 1.6147,
      "theoretical_loss": 3.8563641303738354,
      "tokens_seen": 618332160
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008208224674022067,
      "loss": 1.5514,
      "theoretical_loss": 3.856322021923595,
      "tokens_seen": 618397696
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000820802407221665,
      "loss": 1.6845,
      "theoretical_loss": 3.856279919185008,
      "tokens_seen": 618463232
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008207823470411234,
      "loss": 1.5381,
      "theoretical_loss": 3.856237822156695,
      "tokens_seen": 618528768
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008207622868605818,
      "loss": 1.5259,
      "theoretical_loss": 3.8561957308372756,
      "tokens_seen": 618594304
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008207422266800401,
      "loss": 1.6203,
      "theoretical_loss": 3.8561536452253713,
      "tokens_seen": 618659840
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008207221664994985,
      "loss": 1.5652,
      "theoretical_loss": 3.856111565319604,
      "tokens_seen": 618725376
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008207021063189569,
      "loss": 1.6238,
      "theoretical_loss": 3.8560694911185953,
      "tokens_seen": 618790912
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008206820461384152,
      "loss": 1.4398,
      "theoretical_loss": 3.8560274226209685,
      "tokens_seen": 618856448
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008206619859578737,
      "loss": 1.5396,
      "theoretical_loss": 3.855985359825346,
      "tokens_seen": 618921984
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000820641925777332,
      "loss": 1.697,
      "theoretical_loss": 3.8559433027303514,
      "tokens_seen": 618987520
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008206218655967904,
      "loss": 1.6485,
      "theoretical_loss": 3.8559012513346085,
      "tokens_seen": 619053056
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008206018054162488,
      "loss": 1.6622,
      "theoretical_loss": 3.855859205636742,
      "tokens_seen": 619118592
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008205817452357072,
      "loss": 1.6001,
      "theoretical_loss": 3.855817165635377,
      "tokens_seen": 619184128
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008205616850551655,
      "loss": 1.532,
      "theoretical_loss": 3.855775131329138,
      "tokens_seen": 619249664
    },
    {
      "epoch": 0.19,
      "objective/train/advantage_avg": 0.4949982166290283,
      "objective/train/docs_used": 230719,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6773806810379028,
      "objective/train/original_loss": 1.6773808002471924,
      "objective/train/theoretical_loss": 3.855733102716653,
      "objective/train/tokens_used": 639775200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24520695209503174,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507465600967407,
      "objective/train/weighted_lm_loss": 1.7623634338378906,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.855733102716653,
      "tokens_seen": 619315200
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008205416248746239,
      "loss": 1.7109,
      "theoretical_loss": 3.855733102716653,
      "tokens_seen": 619315200
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008205215646940822,
      "loss": 1.5925,
      "theoretical_loss": 3.8556910797965456,
      "tokens_seen": 619380736
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008205015045135407,
      "loss": 1.7159,
      "theoretical_loss": 3.8556490625674447,
      "tokens_seen": 619446272
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000820481444332999,
      "loss": 1.7182,
      "theoretical_loss": 3.8556070510279765,
      "tokens_seen": 619511808
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008204613841524574,
      "loss": 1.6026,
      "theoretical_loss": 3.8555650451767693,
      "tokens_seen": 619577344
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008204413239719157,
      "loss": 1.68,
      "theoretical_loss": 3.8555230450124514,
      "tokens_seen": 619642880
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008204212637913742,
      "loss": 1.5025,
      "theoretical_loss": 3.855481050533651,
      "tokens_seen": 619708416
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008204012036108325,
      "loss": 1.5698,
      "theoretical_loss": 3.855439061738998,
      "tokens_seen": 619773952
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008203811434302909,
      "loss": 1.6344,
      "theoretical_loss": 3.8553970786271217,
      "tokens_seen": 619839488
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008203610832497492,
      "loss": 1.6571,
      "theoretical_loss": 3.8553551011966523,
      "tokens_seen": 619905024
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008203410230692076,
      "loss": 1.6364,
      "theoretical_loss": 3.8553131294462206,
      "tokens_seen": 619970560
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000820320962888666,
      "loss": 1.6088,
      "theoretical_loss": 3.855271163374458,
      "tokens_seen": 620036096
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008203009027081244,
      "loss": 1.6577,
      "theoretical_loss": 3.855229202979995,
      "tokens_seen": 620101632
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008202808425275827,
      "loss": 1.6194,
      "theoretical_loss": 3.8551872482614655,
      "tokens_seen": 620167168
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008202607823470412,
      "loss": 1.6074,
      "theoretical_loss": 3.8551452992175,
      "tokens_seen": 620232704
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008202407221664995,
      "loss": 1.6324,
      "theoretical_loss": 3.8551033558467327,
      "tokens_seen": 620298240
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008202206619859579,
      "loss": 1.6571,
      "theoretical_loss": 3.8550614181477965,
      "tokens_seen": 620363776
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008202006018054163,
      "loss": 1.6409,
      "theoretical_loss": 3.8550194861193265,
      "tokens_seen": 620429312
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008201805416248746,
      "loss": 1.5578,
      "theoretical_loss": 3.8549775597599556,
      "tokens_seen": 620494848
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000820160481444333,
      "loss": 1.601,
      "theoretical_loss": 3.8549356390683203,
      "tokens_seen": 620560384
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008201404212637914,
      "loss": 1.5341,
      "theoretical_loss": 3.8548937240430545,
      "tokens_seen": 620625920
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008201203610832497,
      "loss": 1.613,
      "theoretical_loss": 3.854851814682795,
      "tokens_seen": 620691456
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008201003009027082,
      "loss": 1.5435,
      "theoretical_loss": 3.8548099109861775,
      "tokens_seen": 620756992
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008200802407221664,
      "loss": 1.5476,
      "theoretical_loss": 3.8547680129518396,
      "tokens_seen": 620822528
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008200601805416249,
      "loss": 1.6422,
      "theoretical_loss": 3.854726120578418,
      "tokens_seen": 620888064
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008200401203610833,
      "loss": 1.7115,
      "theoretical_loss": 3.8546842338645497,
      "tokens_seen": 620953600
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008200200601805416,
      "loss": 1.5755,
      "theoretical_loss": 3.8546423528088747,
      "tokens_seen": 621019136
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00082,
      "loss": 1.7127,
      "theoretical_loss": 3.85460047741003,
      "tokens_seen": 621084672
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008199799398194583,
      "loss": 1.5826,
      "theoretical_loss": 3.8545586076666565,
      "tokens_seen": 621150208
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008199598796389167,
      "loss": 1.5554,
      "theoretical_loss": 3.854516743577392,
      "tokens_seen": 621215744
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008199398194583752,
      "loss": 1.5614,
      "theoretical_loss": 3.8544748851408777,
      "tokens_seen": 621281280
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008199197592778334,
      "loss": 1.6355,
      "theoretical_loss": 3.854433032355754,
      "tokens_seen": 621346816
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008198996990972919,
      "loss": 1.4199,
      "theoretical_loss": 3.8543911852206616,
      "tokens_seen": 621412352
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008198796389167503,
      "loss": 1.6398,
      "theoretical_loss": 3.8543493437342424,
      "tokens_seen": 621477888
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008198595787362086,
      "loss": 1.5345,
      "theoretical_loss": 3.8543075078951388,
      "tokens_seen": 621543424
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000819839518555667,
      "loss": 1.5795,
      "theoretical_loss": 3.8542656777019917,
      "tokens_seen": 621608960
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008198194583751254,
      "loss": 1.5518,
      "theoretical_loss": 3.8542238531534454,
      "tokens_seen": 621674496
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008197993981945838,
      "loss": 1.6004,
      "theoretical_loss": 3.8541820342481428,
      "tokens_seen": 621740032
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008197793380140422,
      "loss": 1.5146,
      "theoretical_loss": 3.8541402209847284,
      "tokens_seen": 621805568
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008197592778335005,
      "loss": 1.5916,
      "theoretical_loss": 3.8540984133618448,
      "tokens_seen": 621871104
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008197392176529589,
      "loss": 1.7015,
      "theoretical_loss": 3.854056611378139,
      "tokens_seen": 621936640
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008197191574724173,
      "loss": 1.5986,
      "theoretical_loss": 3.854014815032255,
      "tokens_seen": 622002176
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008196990972918756,
      "loss": 1.6376,
      "theoretical_loss": 3.8539730243228387,
      "tokens_seen": 622067712
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008196790371113341,
      "loss": 1.5778,
      "theoretical_loss": 3.853931239248536,
      "tokens_seen": 622133248
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008196589769307924,
      "loss": 1.6422,
      "theoretical_loss": 3.8538894598079945,
      "tokens_seen": 622198784
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008196389167502508,
      "loss": 1.5142,
      "theoretical_loss": 3.8538476859998605,
      "tokens_seen": 622264320
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008196188565697092,
      "loss": 1.4573,
      "theoretical_loss": 3.8538059178227817,
      "tokens_seen": 622329856
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008195987963891676,
      "loss": 1.7258,
      "theoretical_loss": 3.8537641552754067,
      "tokens_seen": 622395392
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008195787362086259,
      "loss": 1.7015,
      "theoretical_loss": 3.853722398356383,
      "tokens_seen": 622460928
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008195586760280843,
      "loss": 1.6101,
      "theoretical_loss": 3.853680647064361,
      "tokens_seen": 622526464
    },
    {
      "epoch": 0.19,
      "objective/train/advantage_avg": 0.4887832999229431,
      "objective/train/docs_used": 231894,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6082185506820679,
      "objective/train/original_loss": 1.6082186698913574,
      "objective/train/theoretical_loss": 3.8536389013979893,
      "objective/train/tokens_used": 643052000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23972667753696442,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500969886779785,
      "objective/train/weighted_lm_loss": 1.6888190507888794,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8536389013979893,
      "tokens_seen": 622592000
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008195386158475426,
      "loss": 1.6239,
      "theoretical_loss": 3.8536389013979893,
      "tokens_seen": 622592000
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008195185556670011,
      "loss": 1.6087,
      "theoretical_loss": 3.8535971613559177,
      "tokens_seen": 622657536
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008194984954864594,
      "loss": 1.6609,
      "theoretical_loss": 3.8535554269367966,
      "tokens_seen": 622723072
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008194784353059178,
      "loss": 1.4851,
      "theoretical_loss": 3.8535136981392775,
      "tokens_seen": 622788608
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008194583751253761,
      "loss": 1.5719,
      "theoretical_loss": 3.8534719749620114,
      "tokens_seen": 622854144
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008194383149448346,
      "loss": 1.6139,
      "theoretical_loss": 3.85343025740365,
      "tokens_seen": 622919680
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008194182547642929,
      "loss": 1.4939,
      "theoretical_loss": 3.8533885454628454,
      "tokens_seen": 622985216
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008193981945837513,
      "loss": 1.5971,
      "theoretical_loss": 3.8533468391382515,
      "tokens_seen": 623050752
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008193781344032096,
      "loss": 1.5851,
      "theoretical_loss": 3.8533051384285195,
      "tokens_seen": 623116288
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000819358074222668,
      "loss": 1.554,
      "theoretical_loss": 3.8532634433323043,
      "tokens_seen": 623181824
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008193380140421264,
      "loss": 1.8111,
      "theoretical_loss": 3.85322175384826,
      "tokens_seen": 623247360
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008193179538615848,
      "loss": 1.5707,
      "theoretical_loss": 3.853180069975041,
      "tokens_seen": 623312896
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008192978936810431,
      "loss": 1.5713,
      "theoretical_loss": 3.853138391711303,
      "tokens_seen": 623378432
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008192778335005016,
      "loss": 1.564,
      "theoretical_loss": 3.8530967190556997,
      "tokens_seen": 623443968
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008192577733199599,
      "loss": 1.5382,
      "theoretical_loss": 3.853055052006889,
      "tokens_seen": 623509504
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008192377131394183,
      "loss": 1.5818,
      "theoretical_loss": 3.853013390563526,
      "tokens_seen": 623575040
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008192176529588767,
      "loss": 1.6479,
      "theoretical_loss": 3.852971734724269,
      "tokens_seen": 623640576
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000819197592778335,
      "loss": 1.5934,
      "theoretical_loss": 3.852930084487774,
      "tokens_seen": 623706112
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008191775325977934,
      "loss": 1.5378,
      "theoretical_loss": 3.8528884398526992,
      "tokens_seen": 623771648
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008191574724172518,
      "loss": 1.5815,
      "theoretical_loss": 3.852846800817704,
      "tokens_seen": 623837184
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008191374122367101,
      "loss": 1.5563,
      "theoretical_loss": 3.8528051673814456,
      "tokens_seen": 623902720
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008191173520561686,
      "loss": 1.5153,
      "theoretical_loss": 3.852763539542584,
      "tokens_seen": 623968256
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008190972918756268,
      "loss": 1.6816,
      "theoretical_loss": 3.8527219172997786,
      "tokens_seen": 624033792
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008190772316950853,
      "loss": 1.6855,
      "theoretical_loss": 3.8526803006516896,
      "tokens_seen": 624099328
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008190571715145437,
      "loss": 1.5917,
      "theoretical_loss": 3.8526386895969775,
      "tokens_seen": 624164864
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000819037111334002,
      "loss": 1.6849,
      "theoretical_loss": 3.8525970841343042,
      "tokens_seen": 624230400
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008190170511534604,
      "loss": 1.5308,
      "theoretical_loss": 3.85255548426233,
      "tokens_seen": 624295936
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008189969909729187,
      "loss": 1.7186,
      "theoretical_loss": 3.8525138899797176,
      "tokens_seen": 624361472
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008189769307923771,
      "loss": 1.6299,
      "theoretical_loss": 3.8524723012851294,
      "tokens_seen": 624427008
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008189568706118356,
      "loss": 1.7516,
      "theoretical_loss": 3.8524307181772284,
      "tokens_seen": 624492544
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008189368104312938,
      "loss": 1.5709,
      "theoretical_loss": 3.8523891406546773,
      "tokens_seen": 624558080
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008189167502507523,
      "loss": 1.5958,
      "theoretical_loss": 3.8523475687161404,
      "tokens_seen": 624623616
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008188966900702107,
      "loss": 1.545,
      "theoretical_loss": 3.852306002360282,
      "tokens_seen": 624689152
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000818876629889669,
      "loss": 1.6013,
      "theoretical_loss": 3.852264441585767,
      "tokens_seen": 624754688
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008188565697091274,
      "loss": 1.6479,
      "theoretical_loss": 3.85222288639126,
      "tokens_seen": 624820224
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008188365095285858,
      "loss": 1.624,
      "theoretical_loss": 3.852181336775428,
      "tokens_seen": 624885760
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008188164493480441,
      "loss": 1.6374,
      "theoretical_loss": 3.852139792736936,
      "tokens_seen": 624951296
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008187963891675026,
      "loss": 1.4283,
      "theoretical_loss": 3.85209825427445,
      "tokens_seen": 625016832
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008187763289869608,
      "loss": 1.513,
      "theoretical_loss": 3.8520567213866386,
      "tokens_seen": 625082368
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008187562688064193,
      "loss": 1.7715,
      "theoretical_loss": 3.852015194072168,
      "tokens_seen": 625147904
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008187362086258776,
      "loss": 1.5732,
      "theoretical_loss": 3.8519736723297067,
      "tokens_seen": 625213440
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000818716148445336,
      "loss": 1.6342,
      "theoretical_loss": 3.8519321561579236,
      "tokens_seen": 625278976
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008186960882647944,
      "loss": 1.555,
      "theoretical_loss": 3.8518906455554864,
      "tokens_seen": 625344512
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008186760280842528,
      "loss": 1.54,
      "theoretical_loss": 3.8518491405210655,
      "tokens_seen": 625410048
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008186559679037111,
      "loss": 1.6034,
      "theoretical_loss": 3.8518076410533304,
      "tokens_seen": 625475584
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008186359077231696,
      "loss": 1.5722,
      "theoretical_loss": 3.8517661471509506,
      "tokens_seen": 625541120
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008186158475426278,
      "loss": 1.6391,
      "theoretical_loss": 3.851724658812598,
      "tokens_seen": 625606656
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008185957873620863,
      "loss": 1.5547,
      "theoretical_loss": 3.851683176036943,
      "tokens_seen": 625672192
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008185757271815446,
      "loss": 1.5753,
      "theoretical_loss": 3.8516416988226574,
      "tokens_seen": 625737728
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000818555667001003,
      "loss": 1.6727,
      "theoretical_loss": 3.8516002271684133,
      "tokens_seen": 625803264
    },
    {
      "epoch": 0.19,
      "objective/train/advantage_avg": 0.4916975498199463,
      "objective/train/docs_used": 233149,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3925726413726807,
      "objective/train/original_loss": 1.3925726413726807,
      "objective/train/theoretical_loss": 3.8515587610728836,
      "objective/train/tokens_used": 646328800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2422853410243988,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504014492034912,
      "objective/train/weighted_lm_loss": 1.4627898931503296,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8515587610728836,
      "tokens_seen": 625868800
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008185356068204614,
      "loss": 1.5648,
      "theoretical_loss": 3.8515587610728836,
      "tokens_seen": 625868800
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008185155466399198,
      "loss": 1.5074,
      "theoretical_loss": 3.8515173005347405,
      "tokens_seen": 625934336
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008184954864593781,
      "loss": 1.6098,
      "theoretical_loss": 3.851475845552658,
      "tokens_seen": 625999872
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008184754262788365,
      "loss": 1.6126,
      "theoretical_loss": 3.8514343961253092,
      "tokens_seen": 626065408
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008184553660982949,
      "loss": 1.6273,
      "theoretical_loss": 3.851392952251369,
      "tokens_seen": 626130944
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008184353059177533,
      "loss": 1.6988,
      "theoretical_loss": 3.8513515139295134,
      "tokens_seen": 626196480
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008184152457372116,
      "loss": 1.6243,
      "theoretical_loss": 3.8513100811584158,
      "tokens_seen": 626262016
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00081839518555667,
      "loss": 1.5949,
      "theoretical_loss": 3.851268653936753,
      "tokens_seen": 626327552
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008183751253761283,
      "loss": 1.6994,
      "theoretical_loss": 3.8512272322631995,
      "tokens_seen": 626393088
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008183550651955868,
      "loss": 1.5009,
      "theoretical_loss": 3.8511858161364345,
      "tokens_seen": 626458624
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008183350050150451,
      "loss": 1.775,
      "theoretical_loss": 3.851144405555134,
      "tokens_seen": 626524160
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008183149448345035,
      "loss": 1.5517,
      "theoretical_loss": 3.8511030005179743,
      "tokens_seen": 626589696
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000818294884653962,
      "loss": 1.7334,
      "theoretical_loss": 3.851061601023635,
      "tokens_seen": 626655232
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008182748244734203,
      "loss": 1.5323,
      "theoretical_loss": 3.851020207070794,
      "tokens_seen": 626720768
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008182547642928787,
      "loss": 1.531,
      "theoretical_loss": 3.85097881865813,
      "tokens_seen": 626786304
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000818234704112337,
      "loss": 1.5273,
      "theoretical_loss": 3.8509374357843225,
      "tokens_seen": 626851840
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008182146439317954,
      "loss": 1.6216,
      "theoretical_loss": 3.8508960584480514,
      "tokens_seen": 626917376
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008181945837512538,
      "loss": 1.5957,
      "theoretical_loss": 3.8508546866479962,
      "tokens_seen": 626982912
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008181745235707122,
      "loss": 1.7507,
      "theoretical_loss": 3.850813320382839,
      "tokens_seen": 627048448
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008181544633901705,
      "loss": 1.5464,
      "theoretical_loss": 3.85077195965126,
      "tokens_seen": 627113984
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000818134403209629,
      "loss": 1.6804,
      "theoretical_loss": 3.850730604451941,
      "tokens_seen": 627179520
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008181143430290872,
      "loss": 1.704,
      "theoretical_loss": 3.850689254783564,
      "tokens_seen": 627245056
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008180942828485457,
      "loss": 1.5882,
      "theoretical_loss": 3.8506479106448115,
      "tokens_seen": 627310592
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008180742226680041,
      "loss": 1.6426,
      "theoretical_loss": 3.8506065720343665,
      "tokens_seen": 627376128
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008180541624874624,
      "loss": 1.6072,
      "theoretical_loss": 3.850565238950913,
      "tokens_seen": 627441664
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008180341023069208,
      "loss": 1.734,
      "theoretical_loss": 3.8505239113931333,
      "tokens_seen": 627507200
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008180140421263791,
      "loss": 1.6028,
      "theoretical_loss": 3.8504825893597134,
      "tokens_seen": 627572736
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008179939819458375,
      "loss": 1.7099,
      "theoretical_loss": 3.850441272849337,
      "tokens_seen": 627638272
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000817973921765296,
      "loss": 1.4988,
      "theoretical_loss": 3.85039996186069,
      "tokens_seen": 627703808
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008179538615847542,
      "loss": 1.549,
      "theoretical_loss": 3.8503586563924577,
      "tokens_seen": 627769344
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008179338014042127,
      "loss": 1.6557,
      "theoretical_loss": 3.850317356443326,
      "tokens_seen": 627834880
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008179137412236711,
      "loss": 1.5299,
      "theoretical_loss": 3.8502760620119822,
      "tokens_seen": 627900416
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008178936810431294,
      "loss": 1.6339,
      "theoretical_loss": 3.850234773097112,
      "tokens_seen": 627965952
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008178736208625878,
      "loss": 1.5002,
      "theoretical_loss": 3.8501934896974044,
      "tokens_seen": 628031488
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008178535606820462,
      "loss": 1.6185,
      "theoretical_loss": 3.8501522118115465,
      "tokens_seen": 628097024
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008178335005015045,
      "loss": 1.7303,
      "theoretical_loss": 3.8501109394382267,
      "tokens_seen": 628162560
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000817813440320963,
      "loss": 1.6692,
      "theoretical_loss": 3.850069672576134,
      "tokens_seen": 628228096
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008177933801404212,
      "loss": 1.6016,
      "theoretical_loss": 3.850028411223958,
      "tokens_seen": 628293632
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008177733199598797,
      "loss": 1.6357,
      "theoretical_loss": 3.8499871553803873,
      "tokens_seen": 628359168
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000817753259779338,
      "loss": 1.516,
      "theoretical_loss": 3.8499459050441125,
      "tokens_seen": 628424704
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008177331995987964,
      "loss": 1.626,
      "theoretical_loss": 3.849904660213825,
      "tokens_seen": 628490240
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008177131394182548,
      "loss": 1.6112,
      "theoretical_loss": 3.849863420888215,
      "tokens_seen": 628555776
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008176930792377132,
      "loss": 1.6741,
      "theoretical_loss": 3.8498221870659743,
      "tokens_seen": 628621312
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008176730190571715,
      "loss": 1.5285,
      "theoretical_loss": 3.8497809587457947,
      "tokens_seen": 628686848
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00081765295887663,
      "loss": 1.4918,
      "theoretical_loss": 3.849739735926369,
      "tokens_seen": 628752384
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008176328986960882,
      "loss": 1.5062,
      "theoretical_loss": 3.84969851860639,
      "tokens_seen": 628817920
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008176128385155467,
      "loss": 1.5491,
      "theoretical_loss": 3.8496573067845503,
      "tokens_seen": 628883456
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000817592778335005,
      "loss": 1.6805,
      "theoretical_loss": 3.849616100459544,
      "tokens_seen": 628948992
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008175727181544634,
      "loss": 1.6466,
      "theoretical_loss": 3.8495748996300656,
      "tokens_seen": 629014528
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008175526579739218,
      "loss": 1.6244,
      "theoretical_loss": 3.8495337042948092,
      "tokens_seen": 629080064
    },
    {
      "epoch": 0.19,
      "objective/train/advantage_avg": 0.4909607768058777,
      "objective/train/docs_used": 234355,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4236242771148682,
      "objective/train/original_loss": 1.4236242771148682,
      "objective/train/theoretical_loss": 3.8494925144524705,
      "objective/train/tokens_used": 649605600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24220523238182068,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503273010253906,
      "objective/train/weighted_lm_loss": 1.4951283931732178,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8494925144524705,
      "tokens_seen": 629145600
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008175325977933802,
      "loss": 1.5261,
      "theoretical_loss": 3.8494925144524705,
      "tokens_seen": 629145600
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008175125376128385,
      "loss": 1.6114,
      "theoretical_loss": 3.8494513301017443,
      "tokens_seen": 629211136
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008174924774322969,
      "loss": 1.6353,
      "theoretical_loss": 3.8494101512413277,
      "tokens_seen": 629276672
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008174724172517553,
      "loss": 1.6749,
      "theoretical_loss": 3.849368977869916,
      "tokens_seen": 629342208
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008174523570712137,
      "loss": 1.5678,
      "theoretical_loss": 3.849327809986206,
      "tokens_seen": 629407744
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000817432296890672,
      "loss": 1.6866,
      "theoretical_loss": 3.8492866475888956,
      "tokens_seen": 629473280
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008174122367101304,
      "loss": 1.5259,
      "theoretical_loss": 3.849245490676682,
      "tokens_seen": 629538816
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008173921765295887,
      "loss": 1.4854,
      "theoretical_loss": 3.8492043392482644,
      "tokens_seen": 629604352
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008173721163490472,
      "loss": 1.4975,
      "theoretical_loss": 3.8491631933023407,
      "tokens_seen": 629669888
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008173520561685055,
      "loss": 1.6544,
      "theoretical_loss": 3.8491220528376098,
      "tokens_seen": 629735424
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008173319959879639,
      "loss": 1.6691,
      "theoretical_loss": 3.8490809178527714,
      "tokens_seen": 629800960
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008173119358074223,
      "loss": 1.7029,
      "theoretical_loss": 3.849039788346526,
      "tokens_seen": 629866496
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008172918756268807,
      "loss": 1.5522,
      "theoretical_loss": 3.8489986643175733,
      "tokens_seen": 629932032
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000817271815446339,
      "loss": 1.5926,
      "theoretical_loss": 3.8489575457646144,
      "tokens_seen": 629997568
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008172517552657975,
      "loss": 1.5624,
      "theoretical_loss": 3.8489164326863508,
      "tokens_seen": 630063104
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008172316950852557,
      "loss": 1.6351,
      "theoretical_loss": 3.848875325081484,
      "tokens_seen": 630128640
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008172116349047142,
      "loss": 1.6827,
      "theoretical_loss": 3.8488342229487156,
      "tokens_seen": 630194176
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008171915747241725,
      "loss": 1.6491,
      "theoretical_loss": 3.8487931262867496,
      "tokens_seen": 630259712
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008171715145436309,
      "loss": 1.6055,
      "theoretical_loss": 3.8487520350942876,
      "tokens_seen": 630325248
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008171514543630893,
      "loss": 1.686,
      "theoretical_loss": 3.8487109493700347,
      "tokens_seen": 630390784
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008171313941825476,
      "loss": 1.691,
      "theoretical_loss": 3.8486698691126935,
      "tokens_seen": 630456320
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000817111334002006,
      "loss": 1.5864,
      "theoretical_loss": 3.8486287943209687,
      "tokens_seen": 630521856
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008170912738214645,
      "loss": 1.6892,
      "theoretical_loss": 3.848587724993566,
      "tokens_seen": 630587392
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008170712136409227,
      "loss": 1.5977,
      "theoretical_loss": 3.8485466611291894,
      "tokens_seen": 630652928
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008170511534603812,
      "loss": 1.6303,
      "theoretical_loss": 3.8485056027265454,
      "tokens_seen": 630718464
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008170310932798394,
      "loss": 1.5844,
      "theoretical_loss": 3.8484645497843397,
      "tokens_seen": 630784000
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008170110330992979,
      "loss": 1.5034,
      "theoretical_loss": 3.848423502301279,
      "tokens_seen": 630849536
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008169909729187563,
      "loss": 1.7574,
      "theoretical_loss": 3.8483824602760714,
      "tokens_seen": 630915072
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008169709127382146,
      "loss": 1.5968,
      "theoretical_loss": 3.848341423707423,
      "tokens_seen": 630980608
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000816950852557673,
      "loss": 1.489,
      "theoretical_loss": 3.848300392594042,
      "tokens_seen": 631046144
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008169307923771315,
      "loss": 1.6075,
      "theoretical_loss": 3.8482593669346374,
      "tokens_seen": 631111680
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008169107321965897,
      "loss": 1.661,
      "theoretical_loss": 3.8482183467279176,
      "tokens_seen": 631177216
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008168906720160482,
      "loss": 1.5972,
      "theoretical_loss": 3.8481773319725914,
      "tokens_seen": 631242752
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008168706118355064,
      "loss": 1.6962,
      "theoretical_loss": 3.848136322667369,
      "tokens_seen": 631308288
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008168505516549649,
      "loss": 1.6175,
      "theoretical_loss": 3.8480953188109606,
      "tokens_seen": 631373824
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008168304914744233,
      "loss": 1.6909,
      "theoretical_loss": 3.8480543204020763,
      "tokens_seen": 631439360
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008168104312938816,
      "loss": 1.6689,
      "theoretical_loss": 3.8480133274394275,
      "tokens_seen": 631504896
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008167903711133401,
      "loss": 1.5611,
      "theoretical_loss": 3.847972339921726,
      "tokens_seen": 631570432
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008167703109327984,
      "loss": 1.4894,
      "theoretical_loss": 3.8479313578476826,
      "tokens_seen": 631635968
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008167502507522568,
      "loss": 1.5413,
      "theoretical_loss": 3.8478903812160103,
      "tokens_seen": 631701504
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008167301905717152,
      "loss": 1.5309,
      "theoretical_loss": 3.8478494100254217,
      "tokens_seen": 631767040
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008167101303911736,
      "loss": 1.6383,
      "theoretical_loss": 3.84780844427463,
      "tokens_seen": 631832576
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008166900702106319,
      "loss": 1.7528,
      "theoretical_loss": 3.8477674839623495,
      "tokens_seen": 631898112
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008166700100300904,
      "loss": 1.5873,
      "theoretical_loss": 3.847726529087293,
      "tokens_seen": 631963648
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008166499498495486,
      "loss": 1.6397,
      "theoretical_loss": 3.847685579648176,
      "tokens_seen": 632029184
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008166298896690071,
      "loss": 1.6154,
      "theoretical_loss": 3.8476446356437126,
      "tokens_seen": 632094720
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008166098294884654,
      "loss": 1.6617,
      "theoretical_loss": 3.8476036970726186,
      "tokens_seen": 632160256
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008165897693079238,
      "loss": 1.6953,
      "theoretical_loss": 3.8475627639336105,
      "tokens_seen": 632225792
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008165697091273822,
      "loss": 1.6312,
      "theoretical_loss": 3.847521836225404,
      "tokens_seen": 632291328
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008165496489468406,
      "loss": 1.5592,
      "theoretical_loss": 3.847480913946715,
      "tokens_seen": 632356864
    },
    {
      "epoch": 0.19,
      "objective/train/advantage_avg": 0.49222224950790405,
      "objective/train/docs_used": 235545,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.8225845098495483,
      "objective/train/original_loss": 1.8225843906402588,
      "objective/train/theoretical_loss": 3.847439997096262,
      "objective/train/tokens_used": 652882400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2427038848400116,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050455927848816,
      "objective/train/weighted_lm_loss": 1.914535403251648,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.847439997096262,
      "tokens_seen": 632422400
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008165295887662989,
      "loss": 1.7776,
      "theoretical_loss": 3.847439997096262,
      "tokens_seen": 632422400
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008165095285857573,
      "loss": 1.6502,
      "theoretical_loss": 3.8473990856727616,
      "tokens_seen": 632487936
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008164894684052157,
      "loss": 1.6379,
      "theoretical_loss": 3.8473581796749317,
      "tokens_seen": 632553472
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008164694082246741,
      "loss": 1.5226,
      "theoretical_loss": 3.8473172791014916,
      "tokens_seen": 632619008
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008164493480441324,
      "loss": 1.5734,
      "theoretical_loss": 3.8472763839511597,
      "tokens_seen": 632684544
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008164292878635908,
      "loss": 1.5846,
      "theoretical_loss": 3.847235494222655,
      "tokens_seen": 632750080
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008164092276830491,
      "loss": 1.5095,
      "theoretical_loss": 3.8471946099146983,
      "tokens_seen": 632815616
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008163891675025076,
      "loss": 1.6385,
      "theoretical_loss": 3.847153731026008,
      "tokens_seen": 632881152
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008163691073219659,
      "loss": 1.5517,
      "theoretical_loss": 3.847112857555306,
      "tokens_seen": 632946688
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008163490471414243,
      "loss": 1.6175,
      "theoretical_loss": 3.8470719895013135,
      "tokens_seen": 633012224
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008163289869608827,
      "loss": 1.5434,
      "theoretical_loss": 3.847031126862751,
      "tokens_seen": 633077760
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008163089267803411,
      "loss": 1.6316,
      "theoretical_loss": 3.846990269638341,
      "tokens_seen": 633143296
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008162888665997994,
      "loss": 1.5763,
      "theoretical_loss": 3.8469494178268056,
      "tokens_seen": 633208832
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008162688064192579,
      "loss": 1.542,
      "theoretical_loss": 3.8469085714268676,
      "tokens_seen": 633274368
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008162487462387161,
      "loss": 1.5828,
      "theoretical_loss": 3.8468677304372507,
      "tokens_seen": 633339904
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008162286860581746,
      "loss": 1.5444,
      "theoretical_loss": 3.8468268948566786,
      "tokens_seen": 633405440
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008162086258776329,
      "loss": 1.5582,
      "theoretical_loss": 3.846786064683874,
      "tokens_seen": 633470976
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008161885656970913,
      "loss": 1.5706,
      "theoretical_loss": 3.8467452399175635,
      "tokens_seen": 633536512
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008161685055165497,
      "loss": 1.5734,
      "theoretical_loss": 3.8467044205564704,
      "tokens_seen": 633602048
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000816148445336008,
      "loss": 1.5057,
      "theoretical_loss": 3.8466636065993205,
      "tokens_seen": 633667584
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008161283851554664,
      "loss": 1.6447,
      "theoretical_loss": 3.8466227980448395,
      "tokens_seen": 633733120
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008161083249749249,
      "loss": 1.6633,
      "theoretical_loss": 3.8465819948917543,
      "tokens_seen": 633798656
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008160882647943831,
      "loss": 1.6112,
      "theoretical_loss": 3.846541197138791,
      "tokens_seen": 633864192
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008160682046138416,
      "loss": 1.625,
      "theoretical_loss": 3.8465004047846767,
      "tokens_seen": 633929728
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008160481444332998,
      "loss": 1.7111,
      "theoretical_loss": 3.846459617828139,
      "tokens_seen": 633995264
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008160280842527583,
      "loss": 1.6754,
      "theoretical_loss": 3.8464188362679055,
      "tokens_seen": 634060800
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008160080240722167,
      "loss": 1.6947,
      "theoretical_loss": 3.8463780601027056,
      "tokens_seen": 634126336
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000815987963891675,
      "loss": 1.5935,
      "theoretical_loss": 3.8463372893312675,
      "tokens_seen": 634191872
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008159679037111334,
      "loss": 1.5846,
      "theoretical_loss": 3.8462965239523204,
      "tokens_seen": 634257408
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008159478435305919,
      "loss": 1.5438,
      "theoretical_loss": 3.846255763964594,
      "tokens_seen": 634322944
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008159277833500501,
      "loss": 1.5796,
      "theoretical_loss": 3.846215009366819,
      "tokens_seen": 634388480
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008159077231695086,
      "loss": 1.6794,
      "theoretical_loss": 3.8461742601577247,
      "tokens_seen": 634454016
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008158876629889668,
      "loss": 1.5451,
      "theoretical_loss": 3.8461335163360433,
      "tokens_seen": 634519552
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008158676028084253,
      "loss": 1.6057,
      "theoretical_loss": 3.8460927779005054,
      "tokens_seen": 634585088
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008158475426278837,
      "loss": 1.664,
      "theoretical_loss": 3.846052044849843,
      "tokens_seen": 634650624
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000815827482447342,
      "loss": 1.6116,
      "theoretical_loss": 3.846011317182789,
      "tokens_seen": 634716160
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008158074222668004,
      "loss": 1.5671,
      "theoretical_loss": 3.845970594898076,
      "tokens_seen": 634781696
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008157873620862588,
      "loss": 1.6006,
      "theoretical_loss": 3.845929877994436,
      "tokens_seen": 634847232
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008157673019057171,
      "loss": 1.5838,
      "theoretical_loss": 3.845889166470604,
      "tokens_seen": 634912768
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008157472417251756,
      "loss": 1.5436,
      "theoretical_loss": 3.845848460325313,
      "tokens_seen": 634978304
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008157271815446339,
      "loss": 1.7046,
      "theoretical_loss": 3.8458077595572977,
      "tokens_seen": 635043840
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008157071213640923,
      "loss": 1.5955,
      "theoretical_loss": 3.8457670641652935,
      "tokens_seen": 635109376
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008156870611835506,
      "loss": 1.5783,
      "theoretical_loss": 3.8457263741480343,
      "tokens_seen": 635174912
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000815667001003009,
      "loss": 1.573,
      "theoretical_loss": 3.8456856895042573,
      "tokens_seen": 635240448
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008156469408224674,
      "loss": 1.5688,
      "theoretical_loss": 3.8456450102326976,
      "tokens_seen": 635305984
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008156268806419258,
      "loss": 1.6171,
      "theoretical_loss": 3.8456043363320926,
      "tokens_seen": 635371520
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008156068204613841,
      "loss": 1.5597,
      "theoretical_loss": 3.845563667801178,
      "tokens_seen": 635437056
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008155867602808426,
      "loss": 1.6134,
      "theoretical_loss": 3.8455230046386926,
      "tokens_seen": 635502592
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008155667001003009,
      "loss": 1.6806,
      "theoretical_loss": 3.845482346843373,
      "tokens_seen": 635568128
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008155466399197593,
      "loss": 1.5767,
      "theoretical_loss": 3.845441694413959,
      "tokens_seen": 635633664
    },
    {
      "epoch": 0.19,
      "objective/train/advantage_avg": 0.49294331669807434,
      "objective/train/docs_used": 236526,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5541789531707764,
      "objective/train/original_loss": 1.5541788339614868,
      "objective/train/theoretical_loss": 3.8454010473491884,
      "objective/train/tokens_used": 656159200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24326615035533905,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505311489105225,
      "objective/train/weighted_lm_loss": 1.6325575113296509,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 3.8454010473491884,
      "tokens_seen": 635699200
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008155265797392176,
      "loss": 1.5994,
      "theoretical_loss": 3.8454010473491884,
      "tokens_seen": 635699200
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000815506519558676,
      "loss": 1.6319,
      "theoretical_loss": 3.8453604056477997,
      "tokens_seen": 635764736
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008154864593781344,
      "loss": 1.579,
      "theoretical_loss": 3.8453197693085333,
      "tokens_seen": 635830272
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008154663991975928,
      "loss": 1.6322,
      "theoretical_loss": 3.845279138330129,
      "tokens_seen": 635895808
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008154463390170511,
      "loss": 1.5124,
      "theoretical_loss": 3.845238512711327,
      "tokens_seen": 635961344
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008154262788365095,
      "loss": 1.5947,
      "theoretical_loss": 3.845197892450868,
      "tokens_seen": 636026880
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008154062186559679,
      "loss": 1.6175,
      "theoretical_loss": 3.845157277547494,
      "tokens_seen": 636092416
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008153861584754263,
      "loss": 1.6552,
      "theoretical_loss": 3.845116667999946,
      "tokens_seen": 636157952
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008153660982948846,
      "loss": 1.6637,
      "theoretical_loss": 3.845076063806966,
      "tokens_seen": 636223488
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008153460381143431,
      "loss": 1.6709,
      "theoretical_loss": 3.845035464967297,
      "tokens_seen": 636289024
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008153259779338014,
      "loss": 1.6537,
      "theoretical_loss": 3.8449948714796816,
      "tokens_seen": 636354560
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008153059177532598,
      "loss": 1.623,
      "theoretical_loss": 3.8449542833428634,
      "tokens_seen": 636420096
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008152858575727183,
      "loss": 1.714,
      "theoretical_loss": 3.844913700555586,
      "tokens_seen": 636485632
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008152657973921765,
      "loss": 1.5764,
      "theoretical_loss": 3.844873123116594,
      "tokens_seen": 636551168
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000815245737211635,
      "loss": 1.5439,
      "theoretical_loss": 3.844832551024631,
      "tokens_seen": 636616704
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008152256770310933,
      "loss": 1.5508,
      "theoretical_loss": 3.8447919842784435,
      "tokens_seen": 636682240
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008152056168505517,
      "loss": 1.579,
      "theoretical_loss": 3.8447514228767763,
      "tokens_seen": 636747776
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008151855566700101,
      "loss": 1.5836,
      "theoretical_loss": 3.844710866818375,
      "tokens_seen": 636813312
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008151654964894684,
      "loss": 1.862,
      "theoretical_loss": 3.844670316101986,
      "tokens_seen": 636878848
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008151454363089268,
      "loss": 1.5483,
      "theoretical_loss": 3.844629770726357,
      "tokens_seen": 636944384
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008151253761283853,
      "loss": 1.6434,
      "theoretical_loss": 3.844589230690234,
      "tokens_seen": 637009920
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008151053159478435,
      "loss": 1.5362,
      "theoretical_loss": 3.844548695992365,
      "tokens_seen": 637075456
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000815085255767302,
      "loss": 1.6246,
      "theoretical_loss": 3.8445081666314986,
      "tokens_seen": 637140992
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008150651955867602,
      "loss": 1.5281,
      "theoretical_loss": 3.8444676426063826,
      "tokens_seen": 637206528
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008150451354062187,
      "loss": 1.6882,
      "theoretical_loss": 3.844427123915766,
      "tokens_seen": 637272064
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008150250752256771,
      "loss": 1.5928,
      "theoretical_loss": 3.844386610558398,
      "tokens_seen": 637337600
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008150050150451354,
      "loss": 1.568,
      "theoretical_loss": 3.844346102533029,
      "tokens_seen": 637403136
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008149849548645938,
      "loss": 1.5826,
      "theoretical_loss": 3.8443055998384077,
      "tokens_seen": 637468672
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008149648946840523,
      "loss": 1.516,
      "theoretical_loss": 3.8442651024732863,
      "tokens_seen": 637534208
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008149448345035105,
      "loss": 1.6055,
      "theoretical_loss": 3.844224610436415,
      "tokens_seen": 637599744
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000814924774322969,
      "loss": 1.455,
      "theoretical_loss": 3.844184123726545,
      "tokens_seen": 637665280
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008149047141424272,
      "loss": 1.5898,
      "theoretical_loss": 3.844143642342428,
      "tokens_seen": 637730816
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008148846539618857,
      "loss": 1.605,
      "theoretical_loss": 3.8441031662828173,
      "tokens_seen": 637796352
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008148645937813441,
      "loss": 1.6104,
      "theoretical_loss": 3.8440626955464645,
      "tokens_seen": 637861888
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008148445336008024,
      "loss": 1.5913,
      "theoretical_loss": 3.844022230132123,
      "tokens_seen": 637927424
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008148244734202608,
      "loss": 1.5559,
      "theoretical_loss": 3.8439817700385466,
      "tokens_seen": 637992960
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008148044132397192,
      "loss": 1.6045,
      "theoretical_loss": 3.843941315264489,
      "tokens_seen": 638058496
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008147843530591775,
      "loss": 1.7066,
      "theoretical_loss": 3.8439008658087044,
      "tokens_seen": 638124032
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000814764292878636,
      "loss": 1.5802,
      "theoretical_loss": 3.8438604216699477,
      "tokens_seen": 638189568
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008147442326980943,
      "loss": 1.5175,
      "theoretical_loss": 3.843819982846974,
      "tokens_seen": 638255104
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008147241725175527,
      "loss": 1.4899,
      "theoretical_loss": 3.8437795493385387,
      "tokens_seen": 638320640
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000814704112337011,
      "loss": 1.5917,
      "theoretical_loss": 3.8437391211433987,
      "tokens_seen": 638386176
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008146840521564694,
      "loss": 1.6376,
      "theoretical_loss": 3.84369869826031,
      "tokens_seen": 638451712
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008146639919759278,
      "loss": 1.5972,
      "theoretical_loss": 3.843658280688029,
      "tokens_seen": 638517248
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008146439317953862,
      "loss": 1.6802,
      "theoretical_loss": 3.8436178684253126,
      "tokens_seen": 638582784
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008146238716148445,
      "loss": 1.6295,
      "theoretical_loss": 3.84357746147092,
      "tokens_seen": 638648320
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000814603811434303,
      "loss": 1.7287,
      "theoretical_loss": 3.843537059823608,
      "tokens_seen": 638713856
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008145837512537613,
      "loss": 1.6448,
      "theoretical_loss": 3.843496663482136,
      "tokens_seen": 638779392
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008145636910732197,
      "loss": 1.5549,
      "theoretical_loss": 3.843456272445262,
      "tokens_seen": 638844928
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000814543630892678,
      "loss": 1.5987,
      "theoretical_loss": 3.8434158867117465,
      "tokens_seen": 638910464
    },
    {
      "epoch": 0.19,
      "objective/train/advantage_avg": 0.49233630299568176,
      "objective/train/docs_used": 237858,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3402748107910156,
      "objective/train/original_loss": 1.3402745723724365,
      "objective/train/theoretical_loss": 3.8433755062803483,
      "objective/train/tokens_used": 659436000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2428392469882965,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05046808719635,
      "objective/train/weighted_lm_loss": 1.407797932624817,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8433755062803483,
      "tokens_seen": 638976000
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008145235707121365,
      "loss": 1.3713,
      "theoretical_loss": 3.8433755062803483,
      "tokens_seen": 638976000
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008145035105315948,
      "loss": 1.6112,
      "theoretical_loss": 3.8433351311498285,
      "tokens_seen": 639041536
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008144834503510532,
      "loss": 1.6616,
      "theoretical_loss": 3.8432947613189468,
      "tokens_seen": 639107072
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008144633901705115,
      "loss": 1.5815,
      "theoretical_loss": 3.843254396786465,
      "tokens_seen": 639172608
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008144433299899699,
      "loss": 1.5886,
      "theoretical_loss": 3.843214037551144,
      "tokens_seen": 639238144
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008144232698094283,
      "loss": 1.6515,
      "theoretical_loss": 3.843173683611746,
      "tokens_seen": 639303680
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008144032096288867,
      "loss": 1.7762,
      "theoretical_loss": 3.8431333349670336,
      "tokens_seen": 639369216
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000814383149448345,
      "loss": 1.5744,
      "theoretical_loss": 3.8430929916157686,
      "tokens_seen": 639434752
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008143630892678035,
      "loss": 1.4854,
      "theoretical_loss": 3.843052653556714,
      "tokens_seen": 639500288
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008143430290872618,
      "loss": 1.6078,
      "theoretical_loss": 3.8430123207886346,
      "tokens_seen": 639565824
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008143229689067202,
      "loss": 1.6266,
      "theoretical_loss": 3.842971993310294,
      "tokens_seen": 639631360
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008143029087261785,
      "loss": 1.5957,
      "theoretical_loss": 3.8429316711204557,
      "tokens_seen": 639696896
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008142828485456369,
      "loss": 1.5743,
      "theoretical_loss": 3.842891354217885,
      "tokens_seen": 639762432
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008142627883650953,
      "loss": 1.6018,
      "theoretical_loss": 3.842851042601348,
      "tokens_seen": 639827968
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008142427281845537,
      "loss": 1.5752,
      "theoretical_loss": 3.8428107362696085,
      "tokens_seen": 639893504
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000814222668004012,
      "loss": 1.5521,
      "theoretical_loss": 3.842770435221434,
      "tokens_seen": 639959040
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008142026078234705,
      "loss": 1.5399,
      "theoretical_loss": 3.8427301394555897,
      "tokens_seen": 640024576
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008141825476429287,
      "loss": 1.526,
      "theoretical_loss": 3.8426898489708434,
      "tokens_seen": 640090112
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008141624874623872,
      "loss": 1.6133,
      "theoretical_loss": 3.842649563765962,
      "tokens_seen": 640155648
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008141424272818456,
      "loss": 1.6805,
      "theoretical_loss": 3.8426092838397135,
      "tokens_seen": 640221184
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008141223671013039,
      "loss": 1.7152,
      "theoretical_loss": 3.842569009190866,
      "tokens_seen": 640286720
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008141023069207623,
      "loss": 1.5326,
      "theoretical_loss": 3.8425287398181873,
      "tokens_seen": 640352256
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008140822467402206,
      "loss": 1.6237,
      "theoretical_loss": 3.8424884757204474,
      "tokens_seen": 640417792
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000814062186559679,
      "loss": 1.6939,
      "theoretical_loss": 3.8424482168964142,
      "tokens_seen": 640483328
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008140421263791375,
      "loss": 1.5402,
      "theoretical_loss": 3.8424079633448587,
      "tokens_seen": 640548864
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008140220661985957,
      "loss": 1.5025,
      "theoretical_loss": 3.8423677150645505,
      "tokens_seen": 640614400
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008140020060180542,
      "loss": 1.6547,
      "theoretical_loss": 3.8423274720542606,
      "tokens_seen": 640679936
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008139819458375126,
      "loss": 1.5912,
      "theoretical_loss": 3.8422872343127596,
      "tokens_seen": 640745472
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008139618856569709,
      "loss": 1.5802,
      "theoretical_loss": 3.842247001838819,
      "tokens_seen": 640811008
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008139418254764293,
      "loss": 1.4553,
      "theoretical_loss": 3.8422067746312107,
      "tokens_seen": 640876544
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008139217652958876,
      "loss": 1.6781,
      "theoretical_loss": 3.842166552688706,
      "tokens_seen": 640942080
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000813901705115346,
      "loss": 1.5691,
      "theoretical_loss": 3.8421263360100797,
      "tokens_seen": 641007616
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008138816449348045,
      "loss": 1.6562,
      "theoretical_loss": 3.8420861245941027,
      "tokens_seen": 641073152
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008138615847542627,
      "loss": 1.6452,
      "theoretical_loss": 3.842045918439549,
      "tokens_seen": 641138688
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008138415245737212,
      "loss": 1.5871,
      "theoretical_loss": 3.8420057175451934,
      "tokens_seen": 641204224
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008138214643931795,
      "loss": 1.6131,
      "theoretical_loss": 3.8419655219098097,
      "tokens_seen": 641269760
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008138014042126379,
      "loss": 1.6284,
      "theoretical_loss": 3.8419253315321713,
      "tokens_seen": 641335296
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008137813440320964,
      "loss": 1.6019,
      "theoretical_loss": 3.8418851464110553,
      "tokens_seen": 641400832
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008137612838515547,
      "loss": 1.6425,
      "theoretical_loss": 3.841844966545236,
      "tokens_seen": 641466368
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008137412236710131,
      "loss": 1.5938,
      "theoretical_loss": 3.8418047919334892,
      "tokens_seen": 641531904
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008137211634904714,
      "loss": 1.5786,
      "theoretical_loss": 3.8417646225745923,
      "tokens_seen": 641597440
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008137011033099298,
      "loss": 1.5369,
      "theoretical_loss": 3.841724458467321,
      "tokens_seen": 641662976
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008136810431293882,
      "loss": 1.6833,
      "theoretical_loss": 3.841684299610453,
      "tokens_seen": 641728512
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008136609829488466,
      "loss": 1.5156,
      "theoretical_loss": 3.841644146002766,
      "tokens_seen": 641794048
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008136409227683049,
      "loss": 1.5341,
      "theoretical_loss": 3.841603997643037,
      "tokens_seen": 641859584
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008136208625877634,
      "loss": 1.687,
      "theoretical_loss": 3.8415638545300457,
      "tokens_seen": 641925120
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008136008024072217,
      "loss": 1.5458,
      "theoretical_loss": 3.8415237166625698,
      "tokens_seen": 641990656
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008135807422266801,
      "loss": 1.5651,
      "theoretical_loss": 3.8414835840393886,
      "tokens_seen": 642056192
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008135606820461384,
      "loss": 1.4712,
      "theoretical_loss": 3.8414434566592828,
      "tokens_seen": 642121728
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008135406218655969,
      "loss": 1.5997,
      "theoretical_loss": 3.8414033345210314,
      "tokens_seen": 642187264
    },
    {
      "epoch": 0.19,
      "objective/train/advantage_avg": 0.49311363697052,
      "objective/train/docs_used": 239059,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5280405282974243,
      "objective/train/original_loss": 1.5280404090881348,
      "objective/train/theoretical_loss": 3.8413632176234147,
      "objective/train/tokens_used": 662712800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24341195821762085,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050548791885376,
      "objective/train/weighted_lm_loss": 1.6052095890045166,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.8413632176234147,
      "tokens_seen": 642252800
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008135205616850552,
      "loss": 1.7193,
      "theoretical_loss": 3.8413632176234147,
      "tokens_seen": 642252800
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008135005015045136,
      "loss": 1.5913,
      "theoretical_loss": 3.8413231059652144,
      "tokens_seen": 642318336
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008134804413239719,
      "loss": 1.6132,
      "theoretical_loss": 3.8412829995452107,
      "tokens_seen": 642383872
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008134603811434303,
      "loss": 1.734,
      "theoretical_loss": 3.8412428983621867,
      "tokens_seen": 642449408
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008134403209628887,
      "loss": 1.7358,
      "theoretical_loss": 3.8412028024149225,
      "tokens_seen": 642514944
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008134202607823471,
      "loss": 1.5482,
      "theoretical_loss": 3.8411627117022022,
      "tokens_seen": 642580480
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008134002006018054,
      "loss": 1.5465,
      "theoretical_loss": 3.8411226262228078,
      "tokens_seen": 642646016
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008133801404212639,
      "loss": 1.6358,
      "theoretical_loss": 3.841082545975523,
      "tokens_seen": 642711552
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008133600802407222,
      "loss": 1.5882,
      "theoretical_loss": 3.841042470959131,
      "tokens_seen": 642777088
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008133400200601806,
      "loss": 1.5375,
      "theoretical_loss": 3.8410024011724166,
      "tokens_seen": 642842624
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008133199598796389,
      "loss": 1.5786,
      "theoretical_loss": 3.840962336614164,
      "tokens_seen": 642908160
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008132998996990973,
      "loss": 1.6254,
      "theoretical_loss": 3.8409222772831577,
      "tokens_seen": 642973696
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008132798395185557,
      "loss": 1.5837,
      "theoretical_loss": 3.840882223178183,
      "tokens_seen": 643039232
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008132597793380141,
      "loss": 1.6823,
      "theoretical_loss": 3.8408421742980265,
      "tokens_seen": 643104768
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008132397191574724,
      "loss": 1.5178,
      "theoretical_loss": 3.8408021306414737,
      "tokens_seen": 643170304
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008132196589769309,
      "loss": 1.6362,
      "theoretical_loss": 3.840762092207311,
      "tokens_seen": 643235840
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008131995987963891,
      "loss": 1.6533,
      "theoretical_loss": 3.8407220589943254,
      "tokens_seen": 643301376
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008131795386158476,
      "loss": 1.6076,
      "theoretical_loss": 3.840682031001305,
      "tokens_seen": 643366912
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000813159478435306,
      "loss": 1.5621,
      "theoretical_loss": 3.840642008227036,
      "tokens_seen": 643432448
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0008131394182547643,
      "loss": 1.5643,
      "theoretical_loss": 3.840601990670308,
      "tokens_seen": 643497984
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008131193580742227,
      "loss": 1.4832,
      "theoretical_loss": 3.8405619783299083,
      "tokens_seen": 643563520
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000813099297893681,
      "loss": 1.4709,
      "theoretical_loss": 3.8405219712046272,
      "tokens_seen": 643629056
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008130792377131394,
      "loss": 1.5877,
      "theoretical_loss": 3.8404819692932533,
      "tokens_seen": 643694592
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008130591775325979,
      "loss": 1.5957,
      "theoretical_loss": 3.840441972594576,
      "tokens_seen": 643760128
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008130391173520561,
      "loss": 1.6079,
      "theoretical_loss": 3.8404019811073864,
      "tokens_seen": 643825664
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008130190571715146,
      "loss": 1.5431,
      "theoretical_loss": 3.8403619948304746,
      "tokens_seen": 643891200
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000812998996990973,
      "loss": 1.5753,
      "theoretical_loss": 3.8403220137626315,
      "tokens_seen": 643956736
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008129789368104313,
      "loss": 1.5984,
      "theoretical_loss": 3.8402820379026483,
      "tokens_seen": 644022272
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008129588766298897,
      "loss": 1.577,
      "theoretical_loss": 3.840242067249317,
      "tokens_seen": 644087808
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000812938816449348,
      "loss": 1.5498,
      "theoretical_loss": 3.84020210180143,
      "tokens_seen": 644153344
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008129187562688064,
      "loss": 1.5375,
      "theoretical_loss": 3.84016214155778,
      "tokens_seen": 644218880
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008128986960882649,
      "loss": 1.6124,
      "theoretical_loss": 3.8401221865171595,
      "tokens_seen": 644284416
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008128786359077231,
      "loss": 1.5779,
      "theoretical_loss": 3.840082236678362,
      "tokens_seen": 644349952
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008128585757271816,
      "loss": 1.534,
      "theoretical_loss": 3.8400422920401818,
      "tokens_seen": 644415488
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008128385155466399,
      "loss": 1.5585,
      "theoretical_loss": 3.840002352601412,
      "tokens_seen": 644481024
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008128184553660983,
      "loss": 1.6898,
      "theoretical_loss": 3.8399624183608485,
      "tokens_seen": 644546560
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008127983951855567,
      "loss": 1.6361,
      "theoretical_loss": 3.8399224893172854,
      "tokens_seen": 644612096
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000812778335005015,
      "loss": 1.6244,
      "theoretical_loss": 3.8398825654695186,
      "tokens_seen": 644677632
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008127582748244734,
      "loss": 1.6029,
      "theoretical_loss": 3.8398426468163436,
      "tokens_seen": 644743168
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008127382146439318,
      "loss": 1.5932,
      "theoretical_loss": 3.8398027333565565,
      "tokens_seen": 644808704
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008127181544633901,
      "loss": 1.5407,
      "theoretical_loss": 3.839762825088955,
      "tokens_seen": 644874240
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008126980942828486,
      "loss": 1.6714,
      "theoretical_loss": 3.839722922012334,
      "tokens_seen": 644939776
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008126780341023069,
      "loss": 1.681,
      "theoretical_loss": 3.8396830241254927,
      "tokens_seen": 645005312
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008126579739217653,
      "loss": 1.7181,
      "theoretical_loss": 3.839643131427229,
      "tokens_seen": 645070848
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008126379137412237,
      "loss": 1.6032,
      "theoretical_loss": 3.8396032439163394,
      "tokens_seen": 645136384
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008126178535606821,
      "loss": 1.5822,
      "theoretical_loss": 3.8395633615916243,
      "tokens_seen": 645201920
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008125977933801404,
      "loss": 1.6175,
      "theoretical_loss": 3.839523484451882,
      "tokens_seen": 645267456
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008125777331995988,
      "loss": 1.5985,
      "theoretical_loss": 3.8394836124959113,
      "tokens_seen": 645332992
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008125576730190571,
      "loss": 1.6482,
      "theoretical_loss": 3.8394437457225132,
      "tokens_seen": 645398528
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008125376128385156,
      "loss": 1.6193,
      "theoretical_loss": 3.8394038841304874,
      "tokens_seen": 645464064
    },
    {
      "epoch": 0.2,
      "objective/train/advantage_avg": 0.4884245991706848,
      "objective/train/docs_used": 240316,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5354140996932983,
      "objective/train/original_loss": 1.5354139804840088,
      "objective/train/theoretical_loss": 3.8393640277186343,
      "objective/train/tokens_used": 665989600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23942209780216217,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500595569610596,
      "objective/train/weighted_lm_loss": 1.611963152885437,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8393640277186343,
      "tokens_seen": 645529600
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008125175526579739,
      "loss": 1.566,
      "theoretical_loss": 3.8393640277186343,
      "tokens_seen": 645529600
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008124974924774323,
      "loss": 1.6982,
      "theoretical_loss": 3.839324176485755,
      "tokens_seen": 645595136
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008124774322968906,
      "loss": 1.7881,
      "theoretical_loss": 3.839284330430651,
      "tokens_seen": 645660672
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008124573721163491,
      "loss": 1.5892,
      "theoretical_loss": 3.839244489552124,
      "tokens_seen": 645726208
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008124373119358074,
      "loss": 1.6584,
      "theoretical_loss": 3.8392046538489764,
      "tokens_seen": 645791744
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008124172517552658,
      "loss": 1.6323,
      "theoretical_loss": 3.839164823320011,
      "tokens_seen": 645857280
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008123971915747242,
      "loss": 1.6147,
      "theoretical_loss": 3.8391249979640305,
      "tokens_seen": 645922816
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008123771313941825,
      "loss": 1.6646,
      "theoretical_loss": 3.839085177779838,
      "tokens_seen": 645988352
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008123570712136409,
      "loss": 1.6631,
      "theoretical_loss": 3.839045362766238,
      "tokens_seen": 646053888
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008123370110330993,
      "loss": 1.4806,
      "theoretical_loss": 3.8390055529220337,
      "tokens_seen": 646119424
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008123169508525576,
      "loss": 1.5989,
      "theoretical_loss": 3.8389657482460313,
      "tokens_seen": 646184960
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008122968906720161,
      "loss": 1.5877,
      "theoretical_loss": 3.838925948737034,
      "tokens_seen": 646250496
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008122768304914745,
      "loss": 1.5171,
      "theoretical_loss": 3.8388861543938484,
      "tokens_seen": 646316032
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008122567703109328,
      "loss": 1.5722,
      "theoretical_loss": 3.8388463652152796,
      "tokens_seen": 646381568
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008122367101303913,
      "loss": 1.5325,
      "theoretical_loss": 3.838806581200134,
      "tokens_seen": 646447104
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008122166499498495,
      "loss": 1.5838,
      "theoretical_loss": 3.8387668023472195,
      "tokens_seen": 646512640
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000812196589769308,
      "loss": 1.6213,
      "theoretical_loss": 3.8387270286553408,
      "tokens_seen": 646578176
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008121765295887664,
      "loss": 1.699,
      "theoretical_loss": 3.838687260123306,
      "tokens_seen": 646643712
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008121564694082247,
      "loss": 1.5451,
      "theoretical_loss": 3.838647496749924,
      "tokens_seen": 646709248
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008121364092276831,
      "loss": 1.6904,
      "theoretical_loss": 3.838607738534002,
      "tokens_seen": 646774784
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008121163490471414,
      "loss": 1.7643,
      "theoretical_loss": 3.8385679854743486,
      "tokens_seen": 646840320
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008120962888665998,
      "loss": 1.6247,
      "theoretical_loss": 3.838528237569773,
      "tokens_seen": 646905856
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008120762286860583,
      "loss": 1.4815,
      "theoretical_loss": 3.8384884948190847,
      "tokens_seen": 646971392
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008120561685055165,
      "loss": 1.5986,
      "theoretical_loss": 3.8384487572210935,
      "tokens_seen": 647036928
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000812036108324975,
      "loss": 1.6638,
      "theoretical_loss": 3.8384090247746085,
      "tokens_seen": 647102464
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008120160481444334,
      "loss": 1.6351,
      "theoretical_loss": 3.8383692974784416,
      "tokens_seen": 647168000
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008119959879638917,
      "loss": 1.569,
      "theoretical_loss": 3.838329575331403,
      "tokens_seen": 647233536
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008119759277833501,
      "loss": 1.5627,
      "theoretical_loss": 3.8382898583323044,
      "tokens_seen": 647299072
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008119558676028084,
      "loss": 1.568,
      "theoretical_loss": 3.8382501464799565,
      "tokens_seen": 647364608
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008119358074222668,
      "loss": 1.5953,
      "theoretical_loss": 3.8382104397731736,
      "tokens_seen": 647430144
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008119157472417253,
      "loss": 1.6566,
      "theoretical_loss": 3.8381707382107657,
      "tokens_seen": 647495680
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008118956870611835,
      "loss": 1.649,
      "theoretical_loss": 3.8381310417915477,
      "tokens_seen": 647561216
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000811875626880642,
      "loss": 1.7117,
      "theoretical_loss": 3.8380913505143317,
      "tokens_seen": 647626752
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008118555667001003,
      "loss": 1.6518,
      "theoretical_loss": 3.8380516643779323,
      "tokens_seen": 647692288
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008118355065195587,
      "loss": 1.6882,
      "theoretical_loss": 3.838011983381162,
      "tokens_seen": 647757824
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008118154463390171,
      "loss": 1.6258,
      "theoretical_loss": 3.837972307522837,
      "tokens_seen": 647823360
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008117953861584755,
      "loss": 1.5789,
      "theoretical_loss": 3.837932636801772,
      "tokens_seen": 647888896
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008117753259779338,
      "loss": 1.5619,
      "theoretical_loss": 3.8378929712167817,
      "tokens_seen": 647954432
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008117552657973922,
      "loss": 1.5697,
      "theoretical_loss": 3.8378533107666817,
      "tokens_seen": 648019968
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008117352056168505,
      "loss": 1.612,
      "theoretical_loss": 3.8378136554502884,
      "tokens_seen": 648085504
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000811715145436309,
      "loss": 1.7064,
      "theoretical_loss": 3.837774005266418,
      "tokens_seen": 648151040
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008116950852557673,
      "loss": 1.6931,
      "theoretical_loss": 3.8377343602138874,
      "tokens_seen": 648216576
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008116750250752257,
      "loss": 1.6846,
      "theoretical_loss": 3.8376947202915144,
      "tokens_seen": 648282112
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008116549648946841,
      "loss": 1.612,
      "theoretical_loss": 3.8376550854981155,
      "tokens_seen": 648347648
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008116349047141425,
      "loss": 1.6302,
      "theoretical_loss": 3.83761545583251,
      "tokens_seen": 648413184
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008116148445336008,
      "loss": 1.5701,
      "theoretical_loss": 3.837575831293515,
      "tokens_seen": 648478720
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008115947843530592,
      "loss": 1.627,
      "theoretical_loss": 3.8375362118799505,
      "tokens_seen": 648544256
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008115747241725175,
      "loss": 1.5537,
      "theoretical_loss": 3.837496597590635,
      "tokens_seen": 648609792
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000811554663991976,
      "loss": 1.6683,
      "theoretical_loss": 3.837456988424388,
      "tokens_seen": 648675328
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008115346038114343,
      "loss": 1.5657,
      "theoretical_loss": 3.8374173843800303,
      "tokens_seen": 648740864
    },
    {
      "epoch": 0.2,
      "objective/train/advantage_avg": 0.4949766993522644,
      "objective/train/docs_used": 241544,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4393831491470337,
      "objective/train/original_loss": 1.4393832683563232,
      "objective/train/theoretical_loss": 3.8373777854563813,
      "objective/train/tokens_used": 669266400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2453908920288086,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507452487945557,
      "objective/train/weighted_lm_loss": 1.512376308441162,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8373777854563813,
      "tokens_seen": 648806400
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008115145436308927,
      "loss": 1.6015,
      "theoretical_loss": 3.8373777854563813,
      "tokens_seen": 648806400
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000811494483450351,
      "loss": 1.5641,
      "theoretical_loss": 3.8373381916522624,
      "tokens_seen": 648871936
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008114744232698095,
      "loss": 1.6364,
      "theoretical_loss": 3.8372986029664946,
      "tokens_seen": 648937472
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008114543630892678,
      "loss": 1.6498,
      "theoretical_loss": 3.8372590193978993,
      "tokens_seen": 649003008
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008114343029087262,
      "loss": 1.6915,
      "theoretical_loss": 3.837219440945298,
      "tokens_seen": 649068544
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008114142427281846,
      "loss": 1.6399,
      "theoretical_loss": 3.8371798676075146,
      "tokens_seen": 649134080
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008113941825476429,
      "loss": 1.5462,
      "theoretical_loss": 3.83714029938337,
      "tokens_seen": 649199616
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008113741223671013,
      "loss": 1.6549,
      "theoretical_loss": 3.837100736271688,
      "tokens_seen": 649265152
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008113540621865597,
      "loss": 1.5815,
      "theoretical_loss": 3.8370611782712922,
      "tokens_seen": 649330688
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000811334002006018,
      "loss": 1.6956,
      "theoretical_loss": 3.8370216253810066,
      "tokens_seen": 649396224
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008113139418254765,
      "loss": 1.7814,
      "theoretical_loss": 3.8369820775996555,
      "tokens_seen": 649461760
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008112938816449348,
      "loss": 1.5998,
      "theoretical_loss": 3.836942534926063,
      "tokens_seen": 649527296
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008112738214643932,
      "loss": 1.5409,
      "theoretical_loss": 3.8369029973590543,
      "tokens_seen": 649592832
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008112537612838516,
      "loss": 1.5699,
      "theoretical_loss": 3.8368634648974558,
      "tokens_seen": 649658368
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008112337011033099,
      "loss": 1.6917,
      "theoretical_loss": 3.8368239375400917,
      "tokens_seen": 649723904
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008112136409227683,
      "loss": 1.6013,
      "theoretical_loss": 3.8367844152857895,
      "tokens_seen": 649789440
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008111935807422268,
      "loss": 1.5228,
      "theoretical_loss": 3.836744898133376,
      "tokens_seen": 649854976
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000811173520561685,
      "loss": 1.6308,
      "theoretical_loss": 3.836705386081676,
      "tokens_seen": 649920512
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008111534603811435,
      "loss": 1.6035,
      "theoretical_loss": 3.83666587912952,
      "tokens_seen": 649986048
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008111334002006017,
      "loss": 1.5487,
      "theoretical_loss": 3.8366263772757336,
      "tokens_seen": 650051584
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008111133400200602,
      "loss": 1.5378,
      "theoretical_loss": 3.8365868805191456,
      "tokens_seen": 650117120
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008110932798395186,
      "loss": 1.5322,
      "theoretical_loss": 3.836547388858585,
      "tokens_seen": 650182656
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008110732196589769,
      "loss": 1.5584,
      "theoretical_loss": 3.8365079022928796,
      "tokens_seen": 650248192
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008110531594784353,
      "loss": 1.7369,
      "theoretical_loss": 3.836468420820859,
      "tokens_seen": 650313728
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008110330992978938,
      "loss": 1.5655,
      "theoretical_loss": 3.836428944441354,
      "tokens_seen": 650379264
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000811013039117352,
      "loss": 1.6268,
      "theoretical_loss": 3.8363894731531936,
      "tokens_seen": 650444800
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008109929789368105,
      "loss": 1.6064,
      "theoretical_loss": 3.8363500069552092,
      "tokens_seen": 650510336
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008109729187562687,
      "loss": 1.6259,
      "theoretical_loss": 3.8363105458462305,
      "tokens_seen": 650575872
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008109528585757272,
      "loss": 1.5815,
      "theoretical_loss": 3.8362710898250896,
      "tokens_seen": 650641408
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008109327983951856,
      "loss": 1.6482,
      "theoretical_loss": 3.8362316388906184,
      "tokens_seen": 650706944
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008109127382146439,
      "loss": 1.5952,
      "theoretical_loss": 3.836192193041647,
      "tokens_seen": 650772480
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008108926780341023,
      "loss": 1.6838,
      "theoretical_loss": 3.8361527522770107,
      "tokens_seen": 650838016
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008108726178535607,
      "loss": 1.5973,
      "theoretical_loss": 3.83611331659554,
      "tokens_seen": 650903552
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000810852557673019,
      "loss": 1.575,
      "theoretical_loss": 3.8360738859960692,
      "tokens_seen": 650969088
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008108324974924775,
      "loss": 1.5135,
      "theoretical_loss": 3.836034460477431,
      "tokens_seen": 651034624
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008108124373119357,
      "loss": 1.4913,
      "theoretical_loss": 3.8359950400384606,
      "tokens_seen": 651100160
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008107923771313942,
      "loss": 1.5501,
      "theoretical_loss": 3.8359556246779913,
      "tokens_seen": 651165696
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008107723169508526,
      "loss": 1.6536,
      "theoretical_loss": 3.8359162143948584,
      "tokens_seen": 651231232
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008107522567703109,
      "loss": 1.7162,
      "theoretical_loss": 3.835876809187896,
      "tokens_seen": 651296768
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008107321965897694,
      "loss": 1.622,
      "theoretical_loss": 3.8358374090559413,
      "tokens_seen": 651362304
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008107121364092277,
      "loss": 1.6346,
      "theoretical_loss": 3.8357980139978283,
      "tokens_seen": 651427840
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008106920762286861,
      "loss": 1.64,
      "theoretical_loss": 3.835758624012395,
      "tokens_seen": 651493376
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008106720160481445,
      "loss": 1.7091,
      "theoretical_loss": 3.835719239098477,
      "tokens_seen": 651558912
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008106519558676029,
      "loss": 1.6619,
      "theoretical_loss": 3.835679859254911,
      "tokens_seen": 651624448
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008106318956870612,
      "loss": 1.6016,
      "theoretical_loss": 3.8356404844805354,
      "tokens_seen": 651689984
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008106118355065196,
      "loss": 1.6643,
      "theoretical_loss": 3.835601114774188,
      "tokens_seen": 651755520
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008105917753259779,
      "loss": 1.6405,
      "theoretical_loss": 3.8355617501347057,
      "tokens_seen": 651821056
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008105717151454364,
      "loss": 1.6398,
      "theoretical_loss": 3.8355223905609277,
      "tokens_seen": 651886592
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008105516549648947,
      "loss": 1.5885,
      "theoretical_loss": 3.835483036051694,
      "tokens_seen": 651952128
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008105315947843531,
      "loss": 1.6238,
      "theoretical_loss": 3.835443686605842,
      "tokens_seen": 652017664
    },
    {
      "epoch": 0.2,
      "objective/train/advantage_avg": 0.49520987272262573,
      "objective/train/docs_used": 242832,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5519492626190186,
      "objective/train/original_loss": 1.5519492626190186,
      "objective/train/theoretical_loss": 3.8354043422222133,
      "objective/train/tokens_used": 672543200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24547232687473297,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507689714431763,
      "objective/train/weighted_lm_loss": 1.6308265924453735,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8354043422222133,
      "tokens_seen": 652083200
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008105115346038114,
      "loss": 1.5766,
      "theoretical_loss": 3.8354043422222133,
      "tokens_seen": 652083200
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008104914744232699,
      "loss": 1.6648,
      "theoretical_loss": 3.835365002899647,
      "tokens_seen": 652148736
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008104714142427282,
      "loss": 1.5517,
      "theoretical_loss": 3.835325668636983,
      "tokens_seen": 652214272
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008104513540621866,
      "loss": 1.5991,
      "theoretical_loss": 3.835286339433064,
      "tokens_seen": 652279808
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000810431293881645,
      "loss": 1.7126,
      "theoretical_loss": 3.835247015286729,
      "tokens_seen": 652345344
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008104112337011033,
      "loss": 1.6634,
      "theoretical_loss": 3.8352076961968207,
      "tokens_seen": 652410880
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008103911735205617,
      "loss": 1.5613,
      "theoretical_loss": 3.8351683821621814,
      "tokens_seen": 652476416
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008103711133400201,
      "loss": 1.5562,
      "theoretical_loss": 3.8351290731816534,
      "tokens_seen": 652541952
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008103510531594784,
      "loss": 1.6435,
      "theoretical_loss": 3.835089769254079,
      "tokens_seen": 652607488
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008103309929789369,
      "loss": 1.7236,
      "theoretical_loss": 3.8350504703783015,
      "tokens_seen": 652673024
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008103109327983952,
      "loss": 1.6332,
      "theoretical_loss": 3.8350111765531647,
      "tokens_seen": 652738560
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008102908726178536,
      "loss": 1.6058,
      "theoretical_loss": 3.8349718877775123,
      "tokens_seen": 652804096
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000810270812437312,
      "loss": 1.5609,
      "theoretical_loss": 3.8349326040501888,
      "tokens_seen": 652869632
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008102507522567703,
      "loss": 1.6216,
      "theoretical_loss": 3.834893325370038,
      "tokens_seen": 652935168
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008102306920762287,
      "loss": 1.6131,
      "theoretical_loss": 3.834854051735906,
      "tokens_seen": 653000704
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008102106318956872,
      "loss": 1.678,
      "theoretical_loss": 3.834814783146638,
      "tokens_seen": 653066240
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008101905717151454,
      "loss": 1.5165,
      "theoretical_loss": 3.834775519601079,
      "tokens_seen": 653131776
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008101705115346039,
      "loss": 1.7688,
      "theoretical_loss": 3.8347362610980764,
      "tokens_seen": 653197312
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008101504513540621,
      "loss": 1.604,
      "theoretical_loss": 3.8346970076364757,
      "tokens_seen": 653262848
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008101303911735206,
      "loss": 1.5796,
      "theoretical_loss": 3.8346577592151245,
      "tokens_seen": 653328384
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000810110330992979,
      "loss": 1.5307,
      "theoretical_loss": 3.83461851583287,
      "tokens_seen": 653393920
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008100902708124373,
      "loss": 1.6857,
      "theoretical_loss": 3.83457927748856,
      "tokens_seen": 653459456
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008100702106318957,
      "loss": 1.6034,
      "theoretical_loss": 3.8345400441810424,
      "tokens_seen": 653524992
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008100501504513542,
      "loss": 1.619,
      "theoretical_loss": 3.8345008159091654,
      "tokens_seen": 653590528
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008100300902708124,
      "loss": 1.6036,
      "theoretical_loss": 3.8344615926717784,
      "tokens_seen": 653656064
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008100100300902709,
      "loss": 1.6723,
      "theoretical_loss": 3.8344223744677297,
      "tokens_seen": 653721600
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008099899699097291,
      "loss": 1.6536,
      "theoretical_loss": 3.8343831612958703,
      "tokens_seen": 653787136
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008099699097291876,
      "loss": 1.6551,
      "theoretical_loss": 3.834343953155049,
      "tokens_seen": 653852672
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000809949849548646,
      "loss": 1.5267,
      "theoretical_loss": 3.834304750044117,
      "tokens_seen": 653918208
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008099297893681043,
      "loss": 1.7122,
      "theoretical_loss": 3.8342655519619244,
      "tokens_seen": 653983744
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008099097291875627,
      "loss": 1.6517,
      "theoretical_loss": 3.834226358907322,
      "tokens_seen": 654049280
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008098896690070211,
      "loss": 1.6659,
      "theoretical_loss": 3.8341871708791624,
      "tokens_seen": 654114816
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008098696088264794,
      "loss": 1.5852,
      "theoretical_loss": 3.8341479878762965,
      "tokens_seen": 654180352
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008098495486459379,
      "loss": 1.5873,
      "theoretical_loss": 3.834108809897577,
      "tokens_seen": 654245888
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008098294884653961,
      "loss": 1.6979,
      "theoretical_loss": 3.8340696369418565,
      "tokens_seen": 654311424
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008098094282848546,
      "loss": 1.583,
      "theoretical_loss": 3.834030469007988,
      "tokens_seen": 654376960
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008097893681043129,
      "loss": 1.5965,
      "theoretical_loss": 3.8339913060948243,
      "tokens_seen": 654442496
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008097693079237713,
      "loss": 1.5352,
      "theoretical_loss": 3.83395214820122,
      "tokens_seen": 654508032
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008097492477432297,
      "loss": 1.5438,
      "theoretical_loss": 3.833912995326029,
      "tokens_seen": 654573568
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008097291875626881,
      "loss": 1.6002,
      "theoretical_loss": 3.833873847468105,
      "tokens_seen": 654639104
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008097091273821464,
      "loss": 1.6113,
      "theoretical_loss": 3.8338347046263044,
      "tokens_seen": 654704640
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008096890672016049,
      "loss": 1.492,
      "theoretical_loss": 3.833795566799481,
      "tokens_seen": 654770176
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008096690070210631,
      "loss": 1.6333,
      "theoretical_loss": 3.8337564339864914,
      "tokens_seen": 654835712
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008096489468405216,
      "loss": 1.7008,
      "theoretical_loss": 3.83371730618619,
      "tokens_seen": 654901248
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008096288866599799,
      "loss": 1.6477,
      "theoretical_loss": 3.833678183397436,
      "tokens_seen": 654966784
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008096088264794383,
      "loss": 1.4679,
      "theoretical_loss": 3.833639065619084,
      "tokens_seen": 655032320
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008095887662988967,
      "loss": 1.5969,
      "theoretical_loss": 3.8335999528499913,
      "tokens_seen": 655097856
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008095687061183551,
      "loss": 1.7129,
      "theoretical_loss": 3.833560845089016,
      "tokens_seen": 655163392
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008095486459378134,
      "loss": 1.6872,
      "theoretical_loss": 3.833521742335016,
      "tokens_seen": 655228928
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008095285857572718,
      "loss": 1.5545,
      "theoretical_loss": 3.833482644586849,
      "tokens_seen": 655294464
    },
    {
      "debugging/Compilability": 0.9230769230769231,
      "debugging/distinct-1-grams": 0.7560193640250044,
      "debugging/entropy-1-grams": 5.2534808493234895,
      "debugging/length": 422.2307692307692,
      "debugging/num_segments": 13,
      "debugging/raw_token_scores_avg": 0.00857908371835947,
      "debugging/raw_token_scores_std": 0.02154596894979477,
      "debugging/score": 0.01638830111902192,
      "debugging/score_std": 0.019821935735086043,
      "epoch": 0.2,
      "objective/train/advantage_avg": 0.49142131209373474,
      "objective/train/docs_used": 243976,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.465399980545044,
      "objective/train/original_loss": 1.465399980545044,
      "objective/train/theoretical_loss": 3.833443551843374,
      "objective/train/tokens_used": 675820000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24195925891399384,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503722429275513,
      "objective/train/weighted_lm_loss": 1.538944125175476,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.833443551843374,
      "tokens_seen": 655360000
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008095085255767302,
      "loss": 1.6607,
      "theoretical_loss": 3.833443551843374,
      "tokens_seen": 655360000
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008094884653961886,
      "loss": 1.6405,
      "theoretical_loss": 3.83340446410345,
      "tokens_seen": 655425536
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008094684052156469,
      "loss": 1.5339,
      "theoretical_loss": 3.8333653813659363,
      "tokens_seen": 655491072
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008094483450351054,
      "loss": 1.6343,
      "theoretical_loss": 3.8333263036296925,
      "tokens_seen": 655556608
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008094282848545636,
      "loss": 1.604,
      "theoretical_loss": 3.8332872308935793,
      "tokens_seen": 655622144
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008094082246740221,
      "loss": 1.579,
      "theoretical_loss": 3.8332481631564566,
      "tokens_seen": 655687680
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008093881644934804,
      "loss": 1.6304,
      "theoretical_loss": 3.8332091004171853,
      "tokens_seen": 655753216
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008093681043129388,
      "loss": 1.6733,
      "theoretical_loss": 3.8331700426746274,
      "tokens_seen": 655818752
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008093480441323972,
      "loss": 1.5976,
      "theoretical_loss": 3.833130989927643,
      "tokens_seen": 655884288
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008093279839518556,
      "loss": 1.5416,
      "theoretical_loss": 3.8330919421750957,
      "tokens_seen": 655949824
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000809307923771314,
      "loss": 1.4586,
      "theoretical_loss": 3.833052899415847,
      "tokens_seen": 656015360
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008092878635907724,
      "loss": 1.6714,
      "theoretical_loss": 3.8330138616487597,
      "tokens_seen": 656080896
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008092678034102307,
      "loss": 1.6958,
      "theoretical_loss": 3.8329748288726972,
      "tokens_seen": 656146432
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008092477432296891,
      "loss": 1.6458,
      "theoretical_loss": 3.832935801086523,
      "tokens_seen": 656211968
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008092276830491476,
      "loss": 1.629,
      "theoretical_loss": 3.8328967782891006,
      "tokens_seen": 656277504
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008092076228686058,
      "loss": 1.8014,
      "theoretical_loss": 3.8328577604792944,
      "tokens_seen": 656343040
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008091875626880643,
      "loss": 1.6359,
      "theoretical_loss": 3.8328187476559687,
      "tokens_seen": 656408576
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008091675025075225,
      "loss": 1.5431,
      "theoretical_loss": 3.8327797398179895,
      "tokens_seen": 656474112
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000809147442326981,
      "loss": 1.6703,
      "theoretical_loss": 3.8327407369642206,
      "tokens_seen": 656539648
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008091273821464394,
      "loss": 1.7301,
      "theoretical_loss": 3.832701739093529,
      "tokens_seen": 656605184
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008091073219658977,
      "loss": 1.6201,
      "theoretical_loss": 3.83266274620478,
      "tokens_seen": 656670720
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008090872617853561,
      "loss": 1.6111,
      "theoretical_loss": 3.832623758296841,
      "tokens_seen": 656736256
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008090672016048146,
      "loss": 1.5389,
      "theoretical_loss": 3.832584775368578,
      "tokens_seen": 656801792
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008090471414242728,
      "loss": 1.5878,
      "theoretical_loss": 3.832545797418858,
      "tokens_seen": 656867328
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008090270812437313,
      "loss": 1.5454,
      "theoretical_loss": 3.832506824446549,
      "tokens_seen": 656932864
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008090070210631895,
      "loss": 1.5918,
      "theoretical_loss": 3.8324678564505192,
      "tokens_seen": 656998400
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000808986960882648,
      "loss": 1.4737,
      "theoretical_loss": 3.8324288934296367,
      "tokens_seen": 657063936
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008089669007021064,
      "loss": 1.599,
      "theoretical_loss": 3.83238993538277,
      "tokens_seen": 657129472
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008089468405215647,
      "loss": 1.6811,
      "theoretical_loss": 3.832350982308788,
      "tokens_seen": 657195008
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008089267803410231,
      "loss": 1.6363,
      "theoretical_loss": 3.8323120342065606,
      "tokens_seen": 657260544
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008089067201604815,
      "loss": 1.4928,
      "theoretical_loss": 3.832273091074957,
      "tokens_seen": 657326080
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008088866599799398,
      "loss": 1.5677,
      "theoretical_loss": 3.8322341529128483,
      "tokens_seen": 657391616
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008088665997993983,
      "loss": 1.5585,
      "theoretical_loss": 3.8321952197191043,
      "tokens_seen": 657457152
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008088465396188565,
      "loss": 1.5735,
      "theoretical_loss": 3.8321562914925957,
      "tokens_seen": 657522688
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000808826479438315,
      "loss": 1.7222,
      "theoretical_loss": 3.8321173682321943,
      "tokens_seen": 657588224
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008088064192577733,
      "loss": 1.652,
      "theoretical_loss": 3.832078449936772,
      "tokens_seen": 657653760
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008087863590772317,
      "loss": 1.5807,
      "theoretical_loss": 3.8320395366052,
      "tokens_seen": 657719296
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008087662988966901,
      "loss": 1.5722,
      "theoretical_loss": 3.832000628236351,
      "tokens_seen": 657784832
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008087462387161485,
      "loss": 1.5228,
      "theoretical_loss": 3.831961724829098,
      "tokens_seen": 657850368
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008087261785356068,
      "loss": 1.6183,
      "theoretical_loss": 3.8319228263823133,
      "tokens_seen": 657915904
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008087061183550653,
      "loss": 1.6565,
      "theoretical_loss": 3.8318839328948715,
      "tokens_seen": 657981440
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008086860581745236,
      "loss": 1.5765,
      "theoretical_loss": 3.8318450443656458,
      "tokens_seen": 658046976
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000808665997993982,
      "loss": 1.6736,
      "theoretical_loss": 3.8318061607935103,
      "tokens_seen": 658112512
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008086459378134403,
      "loss": 1.6562,
      "theoretical_loss": 3.83176728217734,
      "tokens_seen": 658178048
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008086258776328987,
      "loss": 1.5601,
      "theoretical_loss": 3.8317284085160095,
      "tokens_seen": 658243584
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008086058174523571,
      "loss": 1.6646,
      "theoretical_loss": 3.8316895398083943,
      "tokens_seen": 658309120
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008085857572718155,
      "loss": 1.5948,
      "theoretical_loss": 3.8316506760533704,
      "tokens_seen": 658374656
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008085656970912738,
      "loss": 1.6021,
      "theoretical_loss": 3.8316118172498133,
      "tokens_seen": 658440192
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008085456369107322,
      "loss": 1.6017,
      "theoretical_loss": 3.8315729633965994,
      "tokens_seen": 658505728
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008085255767301906,
      "loss": 1.5475,
      "theoretical_loss": 3.831534114492606,
      "tokens_seen": 658571264
    },
    {
      "epoch": 0.2,
      "objective/train/advantage_avg": 0.4889923930168152,
      "objective/train/docs_used": 245288,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4206156730651855,
      "objective/train/original_loss": 1.4206156730651855,
      "objective/train/theoretical_loss": 3.83149527053671,
      "objective/train/tokens_used": 679096800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23963674902915955,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501173734664917,
      "objective/train/weighted_lm_loss": 1.4917279481887817,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.83149527053671,
      "tokens_seen": 658636800
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000808505516549649,
      "loss": 1.5777,
      "theoretical_loss": 3.83149527053671,
      "tokens_seen": 658636800
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008084854563691073,
      "loss": 1.5938,
      "theoretical_loss": 3.831456431527789,
      "tokens_seen": 658702336
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008084653961885658,
      "loss": 1.5555,
      "theoretical_loss": 3.8314175974647195,
      "tokens_seen": 658767872
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000808445336008024,
      "loss": 1.6148,
      "theoretical_loss": 3.831378768346382,
      "tokens_seen": 658833408
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008084252758274825,
      "loss": 1.6196,
      "theoretical_loss": 3.8313399441716545,
      "tokens_seen": 658898944
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008084052156469408,
      "loss": 1.6469,
      "theoretical_loss": 3.831301124939415,
      "tokens_seen": 658964480
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008083851554663992,
      "loss": 1.5654,
      "theoretical_loss": 3.831262310648544,
      "tokens_seen": 659030016
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008083650952858576,
      "loss": 1.5986,
      "theoretical_loss": 3.83122350129792,
      "tokens_seen": 659095552
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000808345035105316,
      "loss": 1.6141,
      "theoretical_loss": 3.8311846968864245,
      "tokens_seen": 659161088
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008083249749247743,
      "loss": 1.6455,
      "theoretical_loss": 3.831145897412937,
      "tokens_seen": 659226624
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008083049147442328,
      "loss": 1.6642,
      "theoretical_loss": 3.831107102876338,
      "tokens_seen": 659292160
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000808284854563691,
      "loss": 1.6648,
      "theoretical_loss": 3.8310683132755097,
      "tokens_seen": 659357696
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008082647943831495,
      "loss": 1.666,
      "theoretical_loss": 3.831029528609333,
      "tokens_seen": 659423232
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008082447342026078,
      "loss": 1.6229,
      "theoretical_loss": 3.83099074887669,
      "tokens_seen": 659488768
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008082246740220662,
      "loss": 1.6607,
      "theoretical_loss": 3.830951974076463,
      "tokens_seen": 659554304
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008082046138415246,
      "loss": 1.6777,
      "theoretical_loss": 3.830913204207534,
      "tokens_seen": 659619840
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008081845536609829,
      "loss": 1.715,
      "theoretical_loss": 3.830874439268787,
      "tokens_seen": 659685376
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008081644934804413,
      "loss": 1.6152,
      "theoretical_loss": 3.8308356792591045,
      "tokens_seen": 659750912
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008081444332998998,
      "loss": 1.6225,
      "theoretical_loss": 3.830796924177371,
      "tokens_seen": 659816448
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000808124373119358,
      "loss": 1.5804,
      "theoretical_loss": 3.8307581740224697,
      "tokens_seen": 659881984
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008081043129388165,
      "loss": 1.6829,
      "theoretical_loss": 3.8307194287932855,
      "tokens_seen": 659947520
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008080842527582747,
      "loss": 1.5084,
      "theoretical_loss": 3.8306806884887035,
      "tokens_seen": 660013056
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008080641925777332,
      "loss": 1.5822,
      "theoretical_loss": 3.830641953107609,
      "tokens_seen": 660078592
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008080441323971916,
      "loss": 1.6421,
      "theoretical_loss": 3.830603222648887,
      "tokens_seen": 660144128
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008080240722166499,
      "loss": 1.6527,
      "theoretical_loss": 3.8305644971114234,
      "tokens_seen": 660209664
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008080040120361083,
      "loss": 1.6338,
      "theoretical_loss": 3.8305257764941048,
      "tokens_seen": 660275200
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008079839518555668,
      "loss": 1.5111,
      "theoretical_loss": 3.8304870607958175,
      "tokens_seen": 660340736
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000807963891675025,
      "loss": 1.5984,
      "theoretical_loss": 3.8304483500154487,
      "tokens_seen": 660406272
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008079438314944835,
      "loss": 1.6517,
      "theoretical_loss": 3.830409644151886,
      "tokens_seen": 660471808
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008079237713139417,
      "loss": 1.6072,
      "theoretical_loss": 3.830370943204016,
      "tokens_seen": 660537344
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008079037111334002,
      "loss": 1.5927,
      "theoretical_loss": 3.8303322471707286,
      "tokens_seen": 660602880
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008078836509528586,
      "loss": 1.6451,
      "theoretical_loss": 3.8302935560509104,
      "tokens_seen": 660668416
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008078635907723169,
      "loss": 1.5211,
      "theoretical_loss": 3.8302548698434515,
      "tokens_seen": 660733952
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008078435305917753,
      "loss": 1.5369,
      "theoretical_loss": 3.830216188547241,
      "tokens_seen": 660799488
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008078234704112337,
      "loss": 1.5686,
      "theoretical_loss": 3.830177512161167,
      "tokens_seen": 660865024
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008078034102306921,
      "loss": 1.7609,
      "theoretical_loss": 3.8301388406841212,
      "tokens_seen": 660930560
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008077833500501505,
      "loss": 1.6087,
      "theoretical_loss": 3.8301001741149934,
      "tokens_seen": 660996096
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008077632898696089,
      "loss": 1.6037,
      "theoretical_loss": 3.830061512452673,
      "tokens_seen": 661061632
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008077432296890672,
      "loss": 1.576,
      "theoretical_loss": 3.8300228556960523,
      "tokens_seen": 661127168
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008077231695085257,
      "loss": 1.481,
      "theoretical_loss": 3.829984203844022,
      "tokens_seen": 661192704
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000807703109327984,
      "loss": 1.5844,
      "theoretical_loss": 3.829945556895474,
      "tokens_seen": 661258240
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008076830491474424,
      "loss": 1.5123,
      "theoretical_loss": 3.8299069148493,
      "tokens_seen": 661323776
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008076629889669007,
      "loss": 1.6352,
      "theoretical_loss": 3.829868277704393,
      "tokens_seen": 661389312
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008076429287863591,
      "loss": 1.665,
      "theoretical_loss": 3.829829645459646,
      "tokens_seen": 661454848
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008076228686058175,
      "loss": 1.5309,
      "theoretical_loss": 3.8297910181139505,
      "tokens_seen": 661520384
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008076028084252759,
      "loss": 1.655,
      "theoretical_loss": 3.829752395666202,
      "tokens_seen": 661585920
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008075827482447342,
      "loss": 1.629,
      "theoretical_loss": 3.829713778115293,
      "tokens_seen": 661651456
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008075626880641926,
      "loss": 1.6145,
      "theoretical_loss": 3.8296751654601184,
      "tokens_seen": 661716992
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000807542627883651,
      "loss": 1.5272,
      "theoretical_loss": 3.829636557699572,
      "tokens_seen": 661782528
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008075225677031094,
      "loss": 1.6803,
      "theoretical_loss": 3.8295979548325496,
      "tokens_seen": 661848064
    },
    {
      "epoch": 0.2,
      "objective/train/advantage_avg": 0.48947271704673767,
      "objective/train/docs_used": 246530,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5787867307662964,
      "objective/train/original_loss": 1.578786849975586,
      "objective/train/theoretical_loss": 3.8295593568579465,
      "objective/train/tokens_used": 682373600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24023346602916718,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501683950424194,
      "objective/train/weighted_lm_loss": 1.657979130744934,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8295593568579465,
      "tokens_seen": 661913600
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008075025075225677,
      "loss": 1.7248,
      "theoretical_loss": 3.8295593568579465,
      "tokens_seen": 661913600
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008074824473420262,
      "loss": 1.5461,
      "theoretical_loss": 3.8295207637746573,
      "tokens_seen": 661979136
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008074623871614844,
      "loss": 1.5983,
      "theoretical_loss": 3.829482175581579,
      "tokens_seen": 662044672
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008074423269809429,
      "loss": 1.6615,
      "theoretical_loss": 3.8294435922776078,
      "tokens_seen": 662110208
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008074222668004012,
      "loss": 1.7078,
      "theoretical_loss": 3.8294050138616402,
      "tokens_seen": 662175744
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008074022066198596,
      "loss": 1.7188,
      "theoretical_loss": 3.829366440332573,
      "tokens_seen": 662241280
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000807382146439318,
      "loss": 1.6146,
      "theoretical_loss": 3.829327871689304,
      "tokens_seen": 662306816
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008073620862587764,
      "loss": 1.6156,
      "theoretical_loss": 3.829289307930731,
      "tokens_seen": 662372352
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008073420260782347,
      "loss": 1.5238,
      "theoretical_loss": 3.8292507490557526,
      "tokens_seen": 662437888
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008073219658976932,
      "loss": 1.5607,
      "theoretical_loss": 3.829212195063266,
      "tokens_seen": 662503424
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008073019057171514,
      "loss": 1.5285,
      "theoretical_loss": 3.8291736459521712,
      "tokens_seen": 662568960
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008072818455366099,
      "loss": 1.4719,
      "theoretical_loss": 3.8291351017213673,
      "tokens_seen": 662634496
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008072617853560682,
      "loss": 1.5779,
      "theoretical_loss": 3.8290965623697537,
      "tokens_seen": 662700032
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008072417251755266,
      "loss": 1.698,
      "theoretical_loss": 3.82905802789623,
      "tokens_seen": 662765568
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000807221664994985,
      "loss": 1.5219,
      "theoretical_loss": 3.8290194982996972,
      "tokens_seen": 662831104
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008072016048144433,
      "loss": 1.4465,
      "theoretical_loss": 3.8289809735790548,
      "tokens_seen": 662896640
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008071815446339017,
      "loss": 1.6047,
      "theoretical_loss": 3.8289424537332053,
      "tokens_seen": 662962176
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008071614844533602,
      "loss": 1.6498,
      "theoretical_loss": 3.828903938761049,
      "tokens_seen": 663027712
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008071414242728184,
      "loss": 1.6072,
      "theoretical_loss": 3.828865428661488,
      "tokens_seen": 663093248
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008071213640922769,
      "loss": 1.7724,
      "theoretical_loss": 3.8288269234334242,
      "tokens_seen": 663158784
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008071013039117351,
      "loss": 1.6244,
      "theoretical_loss": 3.82878842307576,
      "tokens_seen": 663224320
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008070812437311936,
      "loss": 1.5777,
      "theoretical_loss": 3.828749927587398,
      "tokens_seen": 663289856
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000807061183550652,
      "loss": 1.5811,
      "theoretical_loss": 3.828711436967243,
      "tokens_seen": 663355392
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008070411233701103,
      "loss": 1.5661,
      "theoretical_loss": 3.8286729512141964,
      "tokens_seen": 663420928
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008070210631895687,
      "loss": 1.646,
      "theoretical_loss": 3.828634470327162,
      "tokens_seen": 663486464
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008070010030090272,
      "loss": 1.6593,
      "theoretical_loss": 3.8285959943050463,
      "tokens_seen": 663552000
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008069809428284854,
      "loss": 1.7145,
      "theoretical_loss": 3.8285575231467517,
      "tokens_seen": 663617536
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008069608826479439,
      "loss": 1.5623,
      "theoretical_loss": 3.8285190568511833,
      "tokens_seen": 663683072
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008069408224674021,
      "loss": 1.5455,
      "theoretical_loss": 3.8284805954172474,
      "tokens_seen": 663748608
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008069207622868606,
      "loss": 1.5862,
      "theoretical_loss": 3.828442138843849,
      "tokens_seen": 663814144
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000806900702106319,
      "loss": 1.7224,
      "theoretical_loss": 3.828403687129894,
      "tokens_seen": 663879680
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008068806419257773,
      "loss": 1.6059,
      "theoretical_loss": 3.828365240274289,
      "tokens_seen": 663945216
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008068605817452357,
      "loss": 1.6,
      "theoretical_loss": 3.828326798275941,
      "tokens_seen": 664010752
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008068405215646941,
      "loss": 1.5576,
      "theoretical_loss": 3.8282883611337564,
      "tokens_seen": 664076288
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008068204613841524,
      "loss": 1.5852,
      "theoretical_loss": 3.8282499288466427,
      "tokens_seen": 664141824
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008068004012036109,
      "loss": 1.6618,
      "theoretical_loss": 3.8282115014135076,
      "tokens_seen": 664207360
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008067803410230692,
      "loss": 1.574,
      "theoretical_loss": 3.82817307883326,
      "tokens_seen": 664272896
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008067602808425276,
      "loss": 1.7069,
      "theoretical_loss": 3.8281346611048073,
      "tokens_seen": 664338432
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000806740220661986,
      "loss": 1.6034,
      "theoretical_loss": 3.828096248227059,
      "tokens_seen": 664403968
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008067201604814444,
      "loss": 1.4631,
      "theoretical_loss": 3.8280578401989236,
      "tokens_seen": 664469504
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008067001003009027,
      "loss": 1.6192,
      "theoretical_loss": 3.8280194370193112,
      "tokens_seen": 664535040
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008066800401203611,
      "loss": 1.6552,
      "theoretical_loss": 3.827981038687132,
      "tokens_seen": 664600576
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008066599799398194,
      "loss": 1.5685,
      "theoretical_loss": 3.8279426452012952,
      "tokens_seen": 664666112
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008066399197592779,
      "loss": 1.6436,
      "theoretical_loss": 3.827904256560712,
      "tokens_seen": 664731648
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008066198595787362,
      "loss": 1.5219,
      "theoretical_loss": 3.827865872764293,
      "tokens_seen": 664797184
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008065997993981946,
      "loss": 1.6272,
      "theoretical_loss": 3.82782749381095,
      "tokens_seen": 664862720
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008065797392176529,
      "loss": 1.5302,
      "theoretical_loss": 3.8277891196995943,
      "tokens_seen": 664928256
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008065596790371114,
      "loss": 1.6373,
      "theoretical_loss": 3.8277507504291384,
      "tokens_seen": 664993792
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008065396188565697,
      "loss": 1.6188,
      "theoretical_loss": 3.8277123859984936,
      "tokens_seen": 665059328
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008065195586760281,
      "loss": 1.7614,
      "theoretical_loss": 3.827674026406573,
      "tokens_seen": 665124864
    },
    {
      "epoch": 0.2,
      "objective/train/advantage_avg": 0.4918051064014435,
      "objective/train/docs_used": 247583,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.641771674156189,
      "objective/train/original_loss": 1.6417717933654785,
      "objective/train/theoretical_loss": 3.82763567165229,
      "objective/train/tokens_used": 685650400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24246832728385925,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504131317138672,
      "objective/train/weighted_lm_loss": 1.7244248390197754,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.82763567165229,
      "tokens_seen": 665190400
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008064994984954864,
      "loss": 1.6398,
      "theoretical_loss": 3.82763567165229,
      "tokens_seen": 665190400
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008064794383149448,
      "loss": 1.5133,
      "theoretical_loss": 3.827597321734557,
      "tokens_seen": 665255936
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008064593781344032,
      "loss": 1.6254,
      "theoretical_loss": 3.8275589766522895,
      "tokens_seen": 665321472
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008064393179538616,
      "loss": 1.703,
      "theoretical_loss": 3.8275206364044,
      "tokens_seen": 665387008
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008064192577733199,
      "loss": 1.5623,
      "theoretical_loss": 3.8274823009898045,
      "tokens_seen": 665452544
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008063991975927784,
      "loss": 1.5296,
      "theoretical_loss": 3.8274439704074155,
      "tokens_seen": 665518080
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008063791374122367,
      "loss": 1.5814,
      "theoretical_loss": 3.8274056446561504,
      "tokens_seen": 665583616
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008063590772316951,
      "loss": 1.5752,
      "theoretical_loss": 3.8273673237349235,
      "tokens_seen": 665649152
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008063390170511535,
      "loss": 1.4642,
      "theoretical_loss": 3.8273290076426507,
      "tokens_seen": 665714688
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008063189568706118,
      "loss": 1.5254,
      "theoretical_loss": 3.8272906963782485,
      "tokens_seen": 665780224
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008062988966900703,
      "loss": 1.6066,
      "theoretical_loss": 3.827252389940633,
      "tokens_seen": 665845760
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008062788365095286,
      "loss": 1.5965,
      "theoretical_loss": 3.827214088328722,
      "tokens_seen": 665911296
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000806258776328987,
      "loss": 1.5741,
      "theoretical_loss": 3.827175791541432,
      "tokens_seen": 665976832
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008062387161484454,
      "loss": 1.5974,
      "theoretical_loss": 3.8271374995776806,
      "tokens_seen": 666042368
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008062186559679037,
      "loss": 1.5623,
      "theoretical_loss": 3.827099212436386,
      "tokens_seen": 666107904
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008061985957873621,
      "loss": 1.6402,
      "theoretical_loss": 3.827060930116466,
      "tokens_seen": 666173440
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008061785356068206,
      "loss": 1.6192,
      "theoretical_loss": 3.82702265261684,
      "tokens_seen": 666238976
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008061584754262788,
      "loss": 1.615,
      "theoretical_loss": 3.8269843799364267,
      "tokens_seen": 666304512
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008061384152457373,
      "loss": 1.5594,
      "theoretical_loss": 3.826946112074145,
      "tokens_seen": 666370048
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008061183550651955,
      "loss": 1.5905,
      "theoretical_loss": 3.826907849028915,
      "tokens_seen": 666435584
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000806098294884654,
      "loss": 1.6237,
      "theoretical_loss": 3.8268695907996566,
      "tokens_seen": 666501120
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008060782347041124,
      "loss": 1.4994,
      "theoretical_loss": 3.8268313373852902,
      "tokens_seen": 666566656
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008060581745235707,
      "loss": 1.6246,
      "theoretical_loss": 3.826793088784737,
      "tokens_seen": 666632192
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008060381143430291,
      "loss": 1.69,
      "theoretical_loss": 3.826754844996917,
      "tokens_seen": 666697728
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008060180541624876,
      "loss": 1.5547,
      "theoretical_loss": 3.826716606020753,
      "tokens_seen": 666763264
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008059979939819458,
      "loss": 1.6581,
      "theoretical_loss": 3.826678371855165,
      "tokens_seen": 666828800
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008059779338014043,
      "loss": 1.6429,
      "theoretical_loss": 3.826640142499077,
      "tokens_seen": 666894336
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008059578736208625,
      "loss": 1.5683,
      "theoretical_loss": 3.8266019179514106,
      "tokens_seen": 666959872
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000805937813440321,
      "loss": 1.5195,
      "theoretical_loss": 3.826563698211088,
      "tokens_seen": 667025408
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008059177532597794,
      "loss": 1.7288,
      "theoretical_loss": 3.8265254832770337,
      "tokens_seen": 667090944
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008058976930792377,
      "loss": 1.4852,
      "theoretical_loss": 3.8264872731481705,
      "tokens_seen": 667156480
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008058776328986961,
      "loss": 1.5198,
      "theoretical_loss": 3.8264490678234218,
      "tokens_seen": 667222016
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008058575727181545,
      "loss": 1.6224,
      "theoretical_loss": 3.8264108673017128,
      "tokens_seen": 667287552
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008058375125376128,
      "loss": 1.5776,
      "theoretical_loss": 3.826372671581967,
      "tokens_seen": 667353088
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008058174523570713,
      "loss": 1.5839,
      "theoretical_loss": 3.8263344806631103,
      "tokens_seen": 667418624
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008057973921765296,
      "loss": 1.6449,
      "theoretical_loss": 3.826296294544067,
      "tokens_seen": 667484160
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000805777331995988,
      "loss": 1.6329,
      "theoretical_loss": 3.8262581132237634,
      "tokens_seen": 667549696
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008057572718154464,
      "loss": 1.5845,
      "theoretical_loss": 3.8262199367011256,
      "tokens_seen": 667615232
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008057372116349048,
      "loss": 1.5933,
      "theoretical_loss": 3.8261817649750784,
      "tokens_seen": 667680768
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008057171514543631,
      "loss": 1.5366,
      "theoretical_loss": 3.8261435980445504,
      "tokens_seen": 667746304
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008056970912738215,
      "loss": 1.4927,
      "theoretical_loss": 3.8261054359084667,
      "tokens_seen": 667811840
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008056770310932798,
      "loss": 1.6556,
      "theoretical_loss": 3.826067278565757,
      "tokens_seen": 667877376
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008056569709127383,
      "loss": 1.5709,
      "theoretical_loss": 3.8260291260153463,
      "tokens_seen": 667942912
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008056369107321966,
      "loss": 1.6031,
      "theoretical_loss": 3.825990978256164,
      "tokens_seen": 668008448
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000805616850551655,
      "loss": 1.5309,
      "theoretical_loss": 3.825952835287139,
      "tokens_seen": 668073984
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008055967903711133,
      "loss": 1.577,
      "theoretical_loss": 3.8259146971071987,
      "tokens_seen": 668139520
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008055767301905718,
      "loss": 1.6768,
      "theoretical_loss": 3.8258765637152727,
      "tokens_seen": 668205056
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008055566700100301,
      "loss": 1.5768,
      "theoretical_loss": 3.825838435110291,
      "tokens_seen": 668270592
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008055366098294885,
      "loss": 1.6087,
      "theoretical_loss": 3.8258003112911823,
      "tokens_seen": 668336128
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008055165496489468,
      "loss": 1.5799,
      "theoretical_loss": 3.825762192256877,
      "tokens_seen": 668401664
    },
    {
      "epoch": 0.2,
      "objective/train/advantage_avg": 0.49283918738365173,
      "objective/train/docs_used": 248744,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5623775720596313,
      "objective/train/original_loss": 1.5623774528503418,
      "objective/train/theoretical_loss": 3.8257240780063055,
      "objective/train/tokens_used": 688927200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24347048997879028,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050521731376648,
      "objective/train/weighted_lm_loss": 1.6415270566940308,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8257240780063055,
      "tokens_seen": 668467200
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008054964894684052,
      "loss": 1.6941,
      "theoretical_loss": 3.8257240780063055,
      "tokens_seen": 668467200
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008054764292878636,
      "loss": 1.5395,
      "theoretical_loss": 3.825685968538399,
      "tokens_seen": 668532736
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000805456369107322,
      "loss": 1.6504,
      "theoretical_loss": 3.8256478638520885,
      "tokens_seen": 668598272
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000805456369107322,
      "loss": 1.5479,
      "theoretical_loss": 3.825609763946305,
      "tokens_seen": 668663808
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008054363089267803,
      "loss": 1.5018,
      "theoretical_loss": 3.8255716688199803,
      "tokens_seen": 668729344
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008054162487462388,
      "loss": 1.6585,
      "theoretical_loss": 3.8255335784720472,
      "tokens_seen": 668794880
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008053961885656971,
      "loss": 1.5092,
      "theoretical_loss": 3.825495492901438,
      "tokens_seen": 668860416
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008053761283851555,
      "loss": 1.6239,
      "theoretical_loss": 3.8254574121070846,
      "tokens_seen": 668925952
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008053560682046139,
      "loss": 1.5325,
      "theoretical_loss": 3.825419336087921,
      "tokens_seen": 668991488
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008053360080240722,
      "loss": 1.5311,
      "theoretical_loss": 3.8253812648428807,
      "tokens_seen": 669057024
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008053159478435306,
      "loss": 1.5849,
      "theoretical_loss": 3.825343198370897,
      "tokens_seen": 669122560
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000805295887662989,
      "loss": 1.6427,
      "theoretical_loss": 3.825305136670905,
      "tokens_seen": 669188096
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008052758274824473,
      "loss": 1.5594,
      "theoretical_loss": 3.825267079741839,
      "tokens_seen": 669253632
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008052557673019058,
      "loss": 1.6214,
      "theoretical_loss": 3.8252290275826333,
      "tokens_seen": 669319168
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000805235707121364,
      "loss": 1.5823,
      "theoretical_loss": 3.8251909801922235,
      "tokens_seen": 669384704
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008052156469408225,
      "loss": 1.5861,
      "theoretical_loss": 3.8251529375695448,
      "tokens_seen": 669450240
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008051955867602809,
      "loss": 1.5721,
      "theoretical_loss": 3.825114899713533,
      "tokens_seen": 669515776
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008051755265797392,
      "loss": 1.6808,
      "theoretical_loss": 3.8250768666231254,
      "tokens_seen": 669581312
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008051554663991976,
      "loss": 1.5711,
      "theoretical_loss": 3.825038838297258,
      "tokens_seen": 669646848
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008051354062186559,
      "loss": 1.58,
      "theoretical_loss": 3.8250008147348673,
      "tokens_seen": 669712384
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008051153460381143,
      "loss": 1.6596,
      "theoretical_loss": 3.8249627959348915,
      "tokens_seen": 669777920
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008050952858575728,
      "loss": 1.5548,
      "theoretical_loss": 3.824924781896267,
      "tokens_seen": 669843456
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000805075225677031,
      "loss": 1.5535,
      "theoretical_loss": 3.8248867726179325,
      "tokens_seen": 669908992
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008050551654964895,
      "loss": 1.6706,
      "theoretical_loss": 3.8248487680988266,
      "tokens_seen": 669974528
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008050351053159479,
      "loss": 1.5618,
      "theoretical_loss": 3.824810768337887,
      "tokens_seen": 670040064
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008050150451354062,
      "loss": 1.5837,
      "theoretical_loss": 3.824772773334053,
      "tokens_seen": 670105600
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008049949849548646,
      "loss": 1.5237,
      "theoretical_loss": 3.8247347830862646,
      "tokens_seen": 670171136
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000804974924774323,
      "loss": 1.6789,
      "theoretical_loss": 3.824696797593461,
      "tokens_seen": 670236672
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008049548645937813,
      "loss": 1.6224,
      "theoretical_loss": 3.8246588168545816,
      "tokens_seen": 670302208
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008049348044132398,
      "loss": 1.5468,
      "theoretical_loss": 3.824620840868568,
      "tokens_seen": 670367744
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000804914744232698,
      "loss": 1.3762,
      "theoretical_loss": 3.82458286963436,
      "tokens_seen": 670433280
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008048946840521565,
      "loss": 1.4916,
      "theoretical_loss": 3.8245449031508985,
      "tokens_seen": 670498816
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008048746238716148,
      "loss": 1.5651,
      "theoretical_loss": 3.824506941417125,
      "tokens_seen": 670564352
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008048545636910732,
      "loss": 1.6503,
      "theoretical_loss": 3.8244689844319817,
      "tokens_seen": 670629888
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008048345035105316,
      "loss": 1.5854,
      "theoretical_loss": 3.8244310321944095,
      "tokens_seen": 670695424
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00080481444332999,
      "loss": 1.6439,
      "theoretical_loss": 3.8243930847033525,
      "tokens_seen": 670760960
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008047943831494484,
      "loss": 1.6139,
      "theoretical_loss": 3.824355141957752,
      "tokens_seen": 670826496
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008047743229689068,
      "loss": 1.6497,
      "theoretical_loss": 3.824317203956551,
      "tokens_seen": 670892032
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008047542627883652,
      "loss": 1.5332,
      "theoretical_loss": 3.8242792706986934,
      "tokens_seen": 670957568
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008047342026078235,
      "loss": 1.564,
      "theoretical_loss": 3.8242413421831234,
      "tokens_seen": 671023104
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008047141424272819,
      "loss": 1.6738,
      "theoretical_loss": 3.8242034184087847,
      "tokens_seen": 671088640
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008046940822467402,
      "loss": 1.5866,
      "theoretical_loss": 3.824165499374621,
      "tokens_seen": 671154176
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008046740220661987,
      "loss": 1.5079,
      "theoretical_loss": 3.8241275850795775,
      "tokens_seen": 671219712
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000804653961885657,
      "loss": 1.546,
      "theoretical_loss": 3.8240896755226,
      "tokens_seen": 671285248
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008046339017051154,
      "loss": 1.5593,
      "theoretical_loss": 3.824051770702633,
      "tokens_seen": 671350784
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008046138415245737,
      "loss": 1.6308,
      "theoretical_loss": 3.8240138706186224,
      "tokens_seen": 671416320
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008045937813440322,
      "loss": 1.6196,
      "theoretical_loss": 3.823975975269514,
      "tokens_seen": 671481856
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008045737211634905,
      "loss": 1.6442,
      "theoretical_loss": 3.823938084654255,
      "tokens_seen": 671547392
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008045536609829489,
      "loss": 1.6753,
      "theoretical_loss": 3.823900198771792,
      "tokens_seen": 671612928
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008045336008024072,
      "loss": 1.6353,
      "theoretical_loss": 3.8238623176210718,
      "tokens_seen": 671678464
    },
    {
      "epoch": 0.2,
      "objective/train/advantage_avg": 0.4925973117351532,
      "objective/train/docs_used": 249880,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4061464071273804,
      "objective/train/original_loss": 1.40614652633667,
      "objective/train/theoretical_loss": 3.823824441201042,
      "objective/train/tokens_used": 692204000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24312211573123932,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504958629608154,
      "objective/train/weighted_lm_loss": 1.4770593643188477,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.823824441201042,
      "tokens_seen": 671744000
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008045135406218656,
      "loss": 1.5135,
      "theoretical_loss": 3.823824441201042,
      "tokens_seen": 671744000
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000804493480441324,
      "loss": 1.4581,
      "theoretical_loss": 3.8237865695106503,
      "tokens_seen": 671809536
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008044734202607824,
      "loss": 1.5251,
      "theoretical_loss": 3.823748702548845,
      "tokens_seen": 671875072
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008044533600802407,
      "loss": 1.5031,
      "theoretical_loss": 3.8237108403145736,
      "tokens_seen": 671940608
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008044332998996992,
      "loss": 1.6428,
      "theoretical_loss": 3.8236729828067864,
      "tokens_seen": 672006144
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008044132397191575,
      "loss": 1.5619,
      "theoretical_loss": 3.8236351300244316,
      "tokens_seen": 672071680
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008043931795386159,
      "loss": 1.4918,
      "theoretical_loss": 3.823597281966459,
      "tokens_seen": 672137216
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008043731193580743,
      "loss": 1.5527,
      "theoretical_loss": 3.823559438631818,
      "tokens_seen": 672202752
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008043530591775326,
      "loss": 1.6208,
      "theoretical_loss": 3.823521600019459,
      "tokens_seen": 672268288
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000804332998996991,
      "loss": 1.6099,
      "theoretical_loss": 3.8234837661283327,
      "tokens_seen": 672333824
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008043129388164494,
      "loss": 1.5344,
      "theoretical_loss": 3.8234459369573894,
      "tokens_seen": 672399360
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008042928786359077,
      "loss": 1.6276,
      "theoretical_loss": 3.8234081125055805,
      "tokens_seen": 672464896
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008042728184553662,
      "loss": 1.4961,
      "theoretical_loss": 3.823370292771857,
      "tokens_seen": 672530432
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008042527582748244,
      "loss": 1.6389,
      "theoretical_loss": 3.823332477755172,
      "tokens_seen": 672595968
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008042326980942829,
      "loss": 1.6051,
      "theoretical_loss": 3.8232946674544763,
      "tokens_seen": 672661504
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008042126379137413,
      "loss": 1.6242,
      "theoretical_loss": 3.823256861868723,
      "tokens_seen": 672727040
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008041925777331996,
      "loss": 1.6606,
      "theoretical_loss": 3.8232190609968644,
      "tokens_seen": 672792576
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000804172517552658,
      "loss": 1.5083,
      "theoretical_loss": 3.8231812648378547,
      "tokens_seen": 672858112
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008041524573721163,
      "loss": 1.6659,
      "theoretical_loss": 3.8231434733906458,
      "tokens_seen": 672923648
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008041323971915747,
      "loss": 1.5494,
      "theoretical_loss": 3.823105686654193,
      "tokens_seen": 672989184
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008041123370110332,
      "loss": 1.6282,
      "theoretical_loss": 3.8230679046274503,
      "tokens_seen": 673054720
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008040922768304914,
      "loss": 1.5388,
      "theoretical_loss": 3.823030127309371,
      "tokens_seen": 673120256
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008040722166499499,
      "loss": 1.7218,
      "theoretical_loss": 3.822992354698911,
      "tokens_seen": 673185792
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008040521564694083,
      "loss": 1.6344,
      "theoretical_loss": 3.8229545867950248,
      "tokens_seen": 673251328
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008040320962888666,
      "loss": 1.4885,
      "theoretical_loss": 3.8229168235966684,
      "tokens_seen": 673316864
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000804012036108325,
      "loss": 1.5777,
      "theoretical_loss": 3.8228790651027973,
      "tokens_seen": 673382400
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008039919759277834,
      "loss": 1.5871,
      "theoretical_loss": 3.822841311312368,
      "tokens_seen": 673447936
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008039719157472417,
      "loss": 1.6214,
      "theoretical_loss": 3.822803562224337,
      "tokens_seen": 673513472
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008039518555667002,
      "loss": 1.6379,
      "theoretical_loss": 3.8227658178376602,
      "tokens_seen": 673579008
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008039317953861584,
      "loss": 1.6012,
      "theoretical_loss": 3.822728078151296,
      "tokens_seen": 673644544
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008039117352056169,
      "loss": 1.4917,
      "theoretical_loss": 3.8226903431642008,
      "tokens_seen": 673710080
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008038916750250752,
      "loss": 1.5202,
      "theoretical_loss": 3.822652612875333,
      "tokens_seen": 673775616
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008038716148445336,
      "loss": 1.5413,
      "theoretical_loss": 3.8226148872836507,
      "tokens_seen": 673841152
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000803851554663992,
      "loss": 1.6453,
      "theoretical_loss": 3.822577166388112,
      "tokens_seen": 673906688
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008038314944834504,
      "loss": 1.6092,
      "theoretical_loss": 3.8225394501876764,
      "tokens_seen": 673972224
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008038114343029087,
      "loss": 1.558,
      "theoretical_loss": 3.8225017386813027,
      "tokens_seen": 674037760
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008037913741223672,
      "loss": 1.6233,
      "theoretical_loss": 3.82246403186795,
      "tokens_seen": 674103296
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008037713139418254,
      "loss": 1.6469,
      "theoretical_loss": 3.822426329746579,
      "tokens_seen": 674168832
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008037512537612839,
      "loss": 1.591,
      "theoretical_loss": 3.822388632316149,
      "tokens_seen": 674234368
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008037311935807422,
      "loss": 1.5824,
      "theoretical_loss": 3.822350939575621,
      "tokens_seen": 674299904
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008037111334002006,
      "loss": 1.4637,
      "theoretical_loss": 3.8223132515239553,
      "tokens_seen": 674365440
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000803691073219659,
      "loss": 1.6411,
      "theoretical_loss": 3.822275568160114,
      "tokens_seen": 674430976
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008036710130391174,
      "loss": 1.5376,
      "theoretical_loss": 3.822237889483057,
      "tokens_seen": 674496512
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008036509528585757,
      "loss": 1.5464,
      "theoretical_loss": 3.822200215491747,
      "tokens_seen": 674562048
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008036308926780341,
      "loss": 1.5831,
      "theoretical_loss": 3.8221625461851465,
      "tokens_seen": 674627584
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008036108324974924,
      "loss": 1.5933,
      "theoretical_loss": 3.8221248815622175,
      "tokens_seen": 674693120
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008035907723169509,
      "loss": 1.648,
      "theoretical_loss": 3.822087221621923,
      "tokens_seen": 674758656
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008035707121364092,
      "loss": 1.4938,
      "theoretical_loss": 3.8220495663632255,
      "tokens_seen": 674824192
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008035506519558676,
      "loss": 1.504,
      "theoretical_loss": 3.8220119157850894,
      "tokens_seen": 674889728
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008035305917753259,
      "loss": 1.4771,
      "theoretical_loss": 3.8219742698864776,
      "tokens_seen": 674955264
    },
    {
      "epoch": 0.2,
      "objective/train/advantage_avg": 0.492687463760376,
      "objective/train/docs_used": 251058,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5048130750656128,
      "objective/train/original_loss": 1.5048130750656128,
      "objective/train/theoretical_loss": 3.821936628666355,
      "objective/train/tokens_used": 695480800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2432936728000641,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050505518913269,
      "objective/train/weighted_lm_loss": 1.5806468725204468,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.821936628666355,
      "tokens_seen": 675020800
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008035105315947844,
      "loss": 1.556,
      "theoretical_loss": 3.821936628666355,
      "tokens_seen": 675020800
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008034904714142427,
      "loss": 1.5471,
      "theoretical_loss": 3.8218989921236854,
      "tokens_seen": 675086336
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008034704112337011,
      "loss": 1.587,
      "theoretical_loss": 3.8218613602574334,
      "tokens_seen": 675151872
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008034503510531595,
      "loss": 1.58,
      "theoretical_loss": 3.821823733066565,
      "tokens_seen": 675217408
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008034302908726179,
      "loss": 1.5252,
      "theoretical_loss": 3.821786110550045,
      "tokens_seen": 675282944
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008034102306920762,
      "loss": 1.5797,
      "theoretical_loss": 3.8217484927068393,
      "tokens_seen": 675348480
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008033901705115347,
      "loss": 1.5183,
      "theoretical_loss": 3.821710879535914,
      "tokens_seen": 675414016
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008033701103309929,
      "loss": 1.5176,
      "theoretical_loss": 3.8216732710362353,
      "tokens_seen": 675479552
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008033500501504514,
      "loss": 1.54,
      "theoretical_loss": 3.8216356672067704,
      "tokens_seen": 675545088
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008033299899699097,
      "loss": 1.5901,
      "theoretical_loss": 3.8215980680464856,
      "tokens_seen": 675610624
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008033099297893681,
      "loss": 1.6545,
      "theoretical_loss": 3.8215604735543494,
      "tokens_seen": 675676160
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008032898696088266,
      "loss": 1.593,
      "theoretical_loss": 3.8215228837293282,
      "tokens_seen": 675741696
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008032698094282848,
      "loss": 1.5136,
      "theoretical_loss": 3.821485298570391,
      "tokens_seen": 675807232
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008032497492477433,
      "loss": 1.5327,
      "theoretical_loss": 3.8214477180765063,
      "tokens_seen": 675872768
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008032296890672017,
      "loss": 1.5051,
      "theoretical_loss": 3.821410142246642,
      "tokens_seen": 675938304
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00080320962888666,
      "loss": 1.615,
      "theoretical_loss": 3.821372571079768,
      "tokens_seen": 676003840
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008031895687061184,
      "loss": 1.5315,
      "theoretical_loss": 3.8213350045748533,
      "tokens_seen": 676069376
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008031695085255767,
      "loss": 1.5044,
      "theoretical_loss": 3.8212974427308675,
      "tokens_seen": 676134912
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008031494483450351,
      "loss": 1.6371,
      "theoretical_loss": 3.8212598855467803,
      "tokens_seen": 676200448
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008031293881644936,
      "loss": 1.5681,
      "theoretical_loss": 3.821222333021563,
      "tokens_seen": 676265984
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008031093279839518,
      "loss": 1.583,
      "theoretical_loss": 3.821184785154186,
      "tokens_seen": 676331520
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008030892678034103,
      "loss": 1.5636,
      "theoretical_loss": 3.8211472419436197,
      "tokens_seen": 676397056
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0008030692076228687,
      "loss": 1.6033,
      "theoretical_loss": 3.821109703388835,
      "tokens_seen": 676462592
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000803049147442327,
      "loss": 1.6203,
      "theoretical_loss": 3.8210721694888052,
      "tokens_seen": 676528128
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008030290872617854,
      "loss": 1.5691,
      "theoretical_loss": 3.8210346402425013,
      "tokens_seen": 676593664
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008030090270812438,
      "loss": 1.6195,
      "theoretical_loss": 3.820997115648896,
      "tokens_seen": 676659200
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008029889669007021,
      "loss": 1.6045,
      "theoretical_loss": 3.8209595957069613,
      "tokens_seen": 676724736
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008029689067201606,
      "loss": 1.4939,
      "theoretical_loss": 3.8209220804156705,
      "tokens_seen": 676790272
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008029488465396188,
      "loss": 1.6088,
      "theoretical_loss": 3.820884569773997,
      "tokens_seen": 676855808
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008029287863590773,
      "loss": 1.6113,
      "theoretical_loss": 3.820847063780914,
      "tokens_seen": 676921344
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008029087261785356,
      "loss": 1.4363,
      "theoretical_loss": 3.8208095624353966,
      "tokens_seen": 676986880
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000802888665997994,
      "loss": 1.5372,
      "theoretical_loss": 3.8207720657364175,
      "tokens_seen": 677052416
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008028686058174524,
      "loss": 1.5815,
      "theoretical_loss": 3.8207345736829526,
      "tokens_seen": 677117952
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008028485456369108,
      "loss": 1.437,
      "theoretical_loss": 3.820697086273976,
      "tokens_seen": 677183488
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008028284854563691,
      "loss": 1.6432,
      "theoretical_loss": 3.8206596035084632,
      "tokens_seen": 677249024
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008028084252758276,
      "loss": 1.5943,
      "theoretical_loss": 3.82062212538539,
      "tokens_seen": 677314560
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008027883650952858,
      "loss": 1.5809,
      "theoretical_loss": 3.820584651903732,
      "tokens_seen": 677380096
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008027683049147443,
      "loss": 1.5974,
      "theoretical_loss": 3.820547183062465,
      "tokens_seen": 677445632
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008027482447342026,
      "loss": 1.5033,
      "theoretical_loss": 3.820509718860567,
      "tokens_seen": 677511168
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000802728184553661,
      "loss": 1.5847,
      "theoretical_loss": 3.8204722592970137,
      "tokens_seen": 677576704
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008027081243731194,
      "loss": 1.5899,
      "theoretical_loss": 3.820434804370782,
      "tokens_seen": 677642240
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008026880641925778,
      "loss": 1.6461,
      "theoretical_loss": 3.8203973540808507,
      "tokens_seen": 677707776
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008026680040120361,
      "loss": 1.5392,
      "theoretical_loss": 3.8203599084261963,
      "tokens_seen": 677773312
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008026479438314945,
      "loss": 1.586,
      "theoretical_loss": 3.8203224674057976,
      "tokens_seen": 677838848
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008026278836509529,
      "loss": 1.4787,
      "theoretical_loss": 3.820285031018633,
      "tokens_seen": 677904384
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008026078234704113,
      "loss": 1.551,
      "theoretical_loss": 3.820247599263682,
      "tokens_seen": 677969920
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008025877632898696,
      "loss": 1.5822,
      "theoretical_loss": 3.8202101721399226,
      "tokens_seen": 678035456
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000802567703109328,
      "loss": 1.5526,
      "theoretical_loss": 3.8201727496463347,
      "tokens_seen": 678100992
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008025476429287863,
      "loss": 1.6435,
      "theoretical_loss": 3.8201353317818985,
      "tokens_seen": 678166528
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008025275827482448,
      "loss": 1.6288,
      "theoretical_loss": 3.820097918545594,
      "tokens_seen": 678232064
    },
    {
      "epoch": 0.21,
      "objective/train/advantage_avg": 0.49411365389823914,
      "objective/train/docs_used": 252282,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.430901288986206,
      "objective/train/original_loss": 1.4309011697769165,
      "objective/train/theoretical_loss": 3.8200605099364005,
      "objective/train/tokens_used": 698757600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24448274075984955,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506542921066284,
      "objective/train/weighted_lm_loss": 1.5033304691314697,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8200605099364005,
      "tokens_seen": 678297600
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008025075225677031,
      "loss": 1.4869,
      "theoretical_loss": 3.8200605099364005,
      "tokens_seen": 678297600
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008024874623871615,
      "loss": 1.57,
      "theoretical_loss": 3.8200231059533003,
      "tokens_seen": 678363136
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008024674022066199,
      "loss": 1.4631,
      "theoretical_loss": 3.819985706595274,
      "tokens_seen": 678428672
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008024473420260783,
      "loss": 1.6472,
      "theoretical_loss": 3.8199483118613022,
      "tokens_seen": 678494208
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008024272818455366,
      "loss": 1.5785,
      "theoretical_loss": 3.8199109217503677,
      "tokens_seen": 678559744
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000802407221664995,
      "loss": 1.6879,
      "theoretical_loss": 3.8198735362614524,
      "tokens_seen": 678625280
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008023871614844533,
      "loss": 1.631,
      "theoretical_loss": 3.8198361553935385,
      "tokens_seen": 678690816
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008023671013039118,
      "loss": 1.6277,
      "theoretical_loss": 3.819798779145608,
      "tokens_seen": 678756352
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008023470411233701,
      "loss": 1.6684,
      "theoretical_loss": 3.8197614075166446,
      "tokens_seen": 678821888
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008023269809428285,
      "loss": 1.6442,
      "theoretical_loss": 3.8197240405056316,
      "tokens_seen": 678887424
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008023069207622869,
      "loss": 1.5968,
      "theoretical_loss": 3.8196866781115526,
      "tokens_seen": 678952960
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008022868605817452,
      "loss": 1.6437,
      "theoretical_loss": 3.819649320333392,
      "tokens_seen": 679018496
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008022668004012036,
      "loss": 1.66,
      "theoretical_loss": 3.819611967170133,
      "tokens_seen": 679084032
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008022467402206621,
      "loss": 1.6421,
      "theoretical_loss": 3.819574618620761,
      "tokens_seen": 679149568
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008022266800401203,
      "loss": 1.7343,
      "theoretical_loss": 3.8195372746842615,
      "tokens_seen": 679215104
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008022066198595788,
      "loss": 1.4686,
      "theoretical_loss": 3.8194999353596186,
      "tokens_seen": 679280640
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000802186559679037,
      "loss": 1.5529,
      "theoretical_loss": 3.819462600645818,
      "tokens_seen": 679346176
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008021664994984955,
      "loss": 1.5401,
      "theoretical_loss": 3.819425270541847,
      "tokens_seen": 679411712
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008021464393179539,
      "loss": 1.5726,
      "theoretical_loss": 3.8193879450466905,
      "tokens_seen": 679477248
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008021263791374122,
      "loss": 1.7235,
      "theoretical_loss": 3.819350624159335,
      "tokens_seen": 679542784
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008021063189568706,
      "loss": 1.5611,
      "theoretical_loss": 3.819313307878768,
      "tokens_seen": 679608320
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008020862587763291,
      "loss": 1.5859,
      "theoretical_loss": 3.8192759962039764,
      "tokens_seen": 679673856
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008020661985957873,
      "loss": 1.5364,
      "theoretical_loss": 3.819238689133948,
      "tokens_seen": 679739392
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008020461384152458,
      "loss": 1.6159,
      "theoretical_loss": 3.8192013866676704,
      "tokens_seen": 679804928
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000802026078234704,
      "loss": 1.5549,
      "theoretical_loss": 3.819164088804132,
      "tokens_seen": 679870464
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008020060180541625,
      "loss": 1.62,
      "theoretical_loss": 3.8191267955423203,
      "tokens_seen": 679936000
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008019859578736209,
      "loss": 1.5928,
      "theoretical_loss": 3.819089506881225,
      "tokens_seen": 680001536
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008019658976930792,
      "loss": 1.5809,
      "theoretical_loss": 3.8190522228198356,
      "tokens_seen": 680067072
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008019458375125376,
      "loss": 1.5064,
      "theoretical_loss": 3.81901494335714,
      "tokens_seen": 680132608
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000801925777331996,
      "loss": 1.5809,
      "theoretical_loss": 3.8189776684921295,
      "tokens_seen": 680198144
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008019057171514543,
      "loss": 1.6652,
      "theoretical_loss": 3.8189403982237935,
      "tokens_seen": 680263680
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008018856569709128,
      "loss": 1.4886,
      "theoretical_loss": 3.8189031325511227,
      "tokens_seen": 680329216
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000801865596790371,
      "loss": 1.5602,
      "theoretical_loss": 3.8188658714731076,
      "tokens_seen": 680394752
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008018455366098295,
      "loss": 1.638,
      "theoretical_loss": 3.818828614988739,
      "tokens_seen": 680460288
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008018254764292878,
      "loss": 1.5084,
      "theoretical_loss": 3.818791363097008,
      "tokens_seen": 680525824
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008018054162487462,
      "loss": 1.6107,
      "theoretical_loss": 3.818754115796907,
      "tokens_seen": 680591360
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008017853560682047,
      "loss": 1.6343,
      "theoretical_loss": 3.8187168730874275,
      "tokens_seen": 680656896
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000801765295887663,
      "loss": 1.5922,
      "theoretical_loss": 3.8186796349675625,
      "tokens_seen": 680722432
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008017452357071214,
      "loss": 1.5014,
      "theoretical_loss": 3.818642401436304,
      "tokens_seen": 680787968
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008017251755265798,
      "loss": 1.6048,
      "theoretical_loss": 3.818605172492645,
      "tokens_seen": 680853504
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008017051153460382,
      "loss": 1.6152,
      "theoretical_loss": 3.8185679481355788,
      "tokens_seen": 680919040
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008016850551654965,
      "loss": 1.4628,
      "theoretical_loss": 3.8185307283640983,
      "tokens_seen": 680984576
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008016649949849549,
      "loss": 1.6703,
      "theoretical_loss": 3.8184935131771987,
      "tokens_seen": 681050112
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008016449348044133,
      "loss": 1.6035,
      "theoretical_loss": 3.8184563025738734,
      "tokens_seen": 681115648
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008016248746238717,
      "loss": 1.4916,
      "theoretical_loss": 3.818419096553117,
      "tokens_seen": 681181184
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00080160481444333,
      "loss": 1.6209,
      "theoretical_loss": 3.818381895113925,
      "tokens_seen": 681246720
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008015847542627884,
      "loss": 1.5641,
      "theoretical_loss": 3.8183446982552915,
      "tokens_seen": 681312256
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008015646940822467,
      "loss": 1.6276,
      "theoretical_loss": 3.8183075059762124,
      "tokens_seen": 681377792
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008015446339017052,
      "loss": 1.6178,
      "theoretical_loss": 3.8182703182756836,
      "tokens_seen": 681443328
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008015245737211635,
      "loss": 1.6189,
      "theoretical_loss": 3.8182331351527017,
      "tokens_seen": 681508864
    },
    {
      "epoch": 0.21,
      "objective/train/advantage_avg": 0.4955112338066101,
      "objective/train/docs_used": 253498,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5169893503189087,
      "objective/train/original_loss": 1.5169892311096191,
      "objective/train/theoretical_loss": 3.818195956606262,
      "objective/train/tokens_used": 702034400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24576736986637115,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508005619049072,
      "objective/train/weighted_lm_loss": 1.5940165519714355,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.818195956606262,
      "tokens_seen": 681574400
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008015045135406219,
      "loss": 1.621,
      "theoretical_loss": 3.818195956606262,
      "tokens_seen": 681574400
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008014844533600803,
      "loss": 1.5552,
      "theoretical_loss": 3.818158782635362,
      "tokens_seen": 681639936
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008014643931795387,
      "loss": 1.6968,
      "theoretical_loss": 3.8181216132389983,
      "tokens_seen": 681705472
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000801444332998997,
      "loss": 1.6022,
      "theoretical_loss": 3.818084448416169,
      "tokens_seen": 681771008
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008014242728184555,
      "loss": 1.5787,
      "theoretical_loss": 3.8180472881658707,
      "tokens_seen": 681836544
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008014042126379137,
      "loss": 1.6462,
      "theoretical_loss": 3.818010132487103,
      "tokens_seen": 681902080
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008013841524573722,
      "loss": 1.5661,
      "theoretical_loss": 3.8179729813788628,
      "tokens_seen": 681967616
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008013640922768305,
      "loss": 1.6059,
      "theoretical_loss": 3.8179358348401493,
      "tokens_seen": 682033152
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008013440320962889,
      "loss": 1.6553,
      "theoretical_loss": 3.817898692869961,
      "tokens_seen": 682098688
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008013239719157473,
      "loss": 1.5404,
      "theoretical_loss": 3.817861555467298,
      "tokens_seen": 682164224
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008013039117352056,
      "loss": 1.6554,
      "theoretical_loss": 3.8178244226311597,
      "tokens_seen": 682229760
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000801283851554664,
      "loss": 1.6585,
      "theoretical_loss": 3.817787294360545,
      "tokens_seen": 682295296
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008012637913741225,
      "loss": 1.6564,
      "theoretical_loss": 3.817750170654455,
      "tokens_seen": 682360832
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008012437311935807,
      "loss": 1.6004,
      "theoretical_loss": 3.817713051511891,
      "tokens_seen": 682426368
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008012236710130392,
      "loss": 1.6653,
      "theoretical_loss": 3.8176759369318525,
      "tokens_seen": 682491904
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008012036108324974,
      "loss": 1.6103,
      "theoretical_loss": 3.8176388269133406,
      "tokens_seen": 682557440
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008011835506519559,
      "loss": 1.5103,
      "theoretical_loss": 3.8176017214553575,
      "tokens_seen": 682622976
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008011634904714143,
      "loss": 1.6258,
      "theoretical_loss": 3.8175646205569045,
      "tokens_seen": 682688512
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008011434302908726,
      "loss": 1.6678,
      "theoretical_loss": 3.8175275242169846,
      "tokens_seen": 682754048
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000801123370110331,
      "loss": 1.6727,
      "theoretical_loss": 3.817490432434599,
      "tokens_seen": 682819584
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008011033099297895,
      "loss": 1.4748,
      "theoretical_loss": 3.8174533452087513,
      "tokens_seen": 682885120
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008010832497492477,
      "loss": 1.5809,
      "theoretical_loss": 3.817416262538444,
      "tokens_seen": 682950656
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008010631895687062,
      "loss": 1.5509,
      "theoretical_loss": 3.817379184422681,
      "tokens_seen": 683016192
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008010431293881644,
      "loss": 1.5299,
      "theoretical_loss": 3.8173421108604657,
      "tokens_seen": 683081728
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008010230692076229,
      "loss": 1.5632,
      "theoretical_loss": 3.817305041850802,
      "tokens_seen": 683147264
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008010030090270813,
      "loss": 1.6129,
      "theoretical_loss": 3.817267977392694,
      "tokens_seen": 683212800
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008009829488465396,
      "loss": 1.5801,
      "theoretical_loss": 3.817230917485147,
      "tokens_seen": 683278336
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000800962888665998,
      "loss": 1.6283,
      "theoretical_loss": 3.817193862127165,
      "tokens_seen": 683343872
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008009428284854564,
      "loss": 1.6558,
      "theoretical_loss": 3.817156811317754,
      "tokens_seen": 683409408
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008009227683049147,
      "loss": 1.5399,
      "theoretical_loss": 3.81711976505592,
      "tokens_seen": 683474944
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008009027081243732,
      "loss": 1.7733,
      "theoretical_loss": 3.8170827233406674,
      "tokens_seen": 683540480
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008008826479438314,
      "loss": 1.6144,
      "theoretical_loss": 3.817045686171003,
      "tokens_seen": 683606016
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008008625877632899,
      "loss": 1.5258,
      "theoretical_loss": 3.8170086535459333,
      "tokens_seen": 683671552
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008008425275827482,
      "loss": 1.5926,
      "theoretical_loss": 3.816971625464466,
      "tokens_seen": 683737088
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008008224674022066,
      "loss": 1.5895,
      "theoretical_loss": 3.816934601925607,
      "tokens_seen": 683802624
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000800802407221665,
      "loss": 1.6567,
      "theoretical_loss": 3.8168975829283642,
      "tokens_seen": 683868160
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008007823470411234,
      "loss": 1.5822,
      "theoretical_loss": 3.8168605684717454,
      "tokens_seen": 683933696
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008007622868605817,
      "loss": 1.5838,
      "theoretical_loss": 3.8168235585547587,
      "tokens_seen": 683999232
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008007422266800402,
      "loss": 1.7083,
      "theoretical_loss": 3.816786553176412,
      "tokens_seen": 684064768
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008007221664994985,
      "loss": 1.5816,
      "theoretical_loss": 3.8167495523357147,
      "tokens_seen": 684130304
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008007021063189569,
      "loss": 1.5373,
      "theoretical_loss": 3.816712556031675,
      "tokens_seen": 684195840
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008006820461384152,
      "loss": 1.7265,
      "theoretical_loss": 3.816675564263303,
      "tokens_seen": 684261376
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008006619859578737,
      "loss": 1.4646,
      "theoretical_loss": 3.8166385770296074,
      "tokens_seen": 684326912
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000800641925777332,
      "loss": 1.5245,
      "theoretical_loss": 3.816601594329599,
      "tokens_seen": 684392448
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008006218655967904,
      "loss": 1.6153,
      "theoretical_loss": 3.816564616162287,
      "tokens_seen": 684457984
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008006018054162487,
      "loss": 1.5494,
      "theoretical_loss": 3.816527642526683,
      "tokens_seen": 684523520
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008005817452357071,
      "loss": 1.587,
      "theoretical_loss": 3.8164906734217974,
      "tokens_seen": 684589056
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008005616850551655,
      "loss": 1.5202,
      "theoretical_loss": 3.816453708846642,
      "tokens_seen": 684654592
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008005416248746239,
      "loss": 1.5808,
      "theoretical_loss": 3.8164167488002265,
      "tokens_seen": 684720128
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008005215646940822,
      "loss": 1.5932,
      "theoretical_loss": 3.8163797932815644,
      "tokens_seen": 684785664
    },
    {
      "epoch": 0.21,
      "objective/train/advantage_avg": 0.49446743726730347,
      "objective/train/docs_used": 254674,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4938709735870361,
      "objective/train/original_loss": 1.4938709735870361,
      "objective/train/theoretical_loss": 3.816342842289668,
      "objective/train/tokens_used": 705311200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2447027564048767,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050690770149231,
      "objective/train/weighted_lm_loss": 1.5695686340332031,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.816342842289668,
      "tokens_seen": 684851200
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008005015045135407,
      "loss": 1.5732,
      "theoretical_loss": 3.816342842289668,
      "tokens_seen": 684851200
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008004814443329989,
      "loss": 1.6135,
      "theoretical_loss": 3.8163058958235476,
      "tokens_seen": 684916736
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008004613841524574,
      "loss": 1.6286,
      "theoretical_loss": 3.816268953882218,
      "tokens_seen": 684982272
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008004413239719157,
      "loss": 1.5237,
      "theoretical_loss": 3.8162320164646912,
      "tokens_seen": 685047808
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008004212637913741,
      "loss": 1.5382,
      "theoretical_loss": 3.816195083569981,
      "tokens_seen": 685113344
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008004012036108325,
      "loss": 1.5408,
      "theoretical_loss": 3.816158155197101,
      "tokens_seen": 685178880
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008003811434302909,
      "loss": 1.6047,
      "theoretical_loss": 3.8161212313450648,
      "tokens_seen": 685244416
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008003610832497492,
      "loss": 1.6447,
      "theoretical_loss": 3.8160843120128867,
      "tokens_seen": 685309952
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008003410230692077,
      "loss": 1.5582,
      "theoretical_loss": 3.816047397199582,
      "tokens_seen": 685375488
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008003209628886659,
      "loss": 1.6029,
      "theoretical_loss": 3.8160104869041644,
      "tokens_seen": 685441024
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008003009027081244,
      "loss": 1.5618,
      "theoretical_loss": 3.81597358112565,
      "tokens_seen": 685506560
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008002808425275829,
      "loss": 1.7055,
      "theoretical_loss": 3.8159366798630536,
      "tokens_seen": 685572096
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008002607823470411,
      "loss": 1.5197,
      "theoretical_loss": 3.8158997831153916,
      "tokens_seen": 685637632
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008002407221664996,
      "loss": 1.5811,
      "theoretical_loss": 3.81586289088168,
      "tokens_seen": 685703168
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008002206619859578,
      "loss": 1.6259,
      "theoretical_loss": 3.815826003160935,
      "tokens_seen": 685768704
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008002006018054163,
      "loss": 1.5439,
      "theoretical_loss": 3.8157891199521736,
      "tokens_seen": 685834240
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008001805416248747,
      "loss": 1.5799,
      "theoretical_loss": 3.8157522412544127,
      "tokens_seen": 685899776
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000800160481444333,
      "loss": 1.7461,
      "theoretical_loss": 3.8157153670666695,
      "tokens_seen": 685965312
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008001404212637914,
      "loss": 1.6544,
      "theoretical_loss": 3.815678497387962,
      "tokens_seen": 686030848
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008001203610832499,
      "loss": 1.4647,
      "theoretical_loss": 3.815641632217308,
      "tokens_seen": 686096384
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008001003009027081,
      "loss": 1.5925,
      "theoretical_loss": 3.8156047715537253,
      "tokens_seen": 686161920
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008000802407221666,
      "loss": 1.5962,
      "theoretical_loss": 3.8155679153962336,
      "tokens_seen": 686227456
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008000601805416248,
      "loss": 1.6627,
      "theoretical_loss": 3.8155310637438506,
      "tokens_seen": 686292992
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008000401203610833,
      "loss": 1.6375,
      "theoretical_loss": 3.815494216595596,
      "tokens_seen": 686358528
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008000200601805417,
      "loss": 1.4875,
      "theoretical_loss": 3.8154573739504896,
      "tokens_seen": 686424064
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0008,
      "loss": 1.6557,
      "theoretical_loss": 3.815420535807551,
      "tokens_seen": 686489600
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007999799398194584,
      "loss": 1.5671,
      "theoretical_loss": 3.8153837021657995,
      "tokens_seen": 686555136
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007999598796389168,
      "loss": 1.5943,
      "theoretical_loss": 3.815346873024257,
      "tokens_seen": 686620672
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007999398194583751,
      "loss": 1.5247,
      "theoretical_loss": 3.8153100483819427,
      "tokens_seen": 686686208
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007999197592778336,
      "loss": 1.533,
      "theoretical_loss": 3.8152732282378787,
      "tokens_seen": 686751744
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007998996990972918,
      "loss": 1.5665,
      "theoretical_loss": 3.8152364125910863,
      "tokens_seen": 686817280
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007998796389167503,
      "loss": 1.5765,
      "theoretical_loss": 3.815199601440587,
      "tokens_seen": 686882816
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007998595787362086,
      "loss": 1.6301,
      "theoretical_loss": 3.815162794785402,
      "tokens_seen": 686948352
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000799839518555667,
      "loss": 1.5349,
      "theoretical_loss": 3.815125992624554,
      "tokens_seen": 687013888
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007998194583751254,
      "loss": 1.5315,
      "theoretical_loss": 3.8150891949570664,
      "tokens_seen": 687079424
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007997993981945838,
      "loss": 1.5514,
      "theoretical_loss": 3.815052401781961,
      "tokens_seen": 687144960
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007997793380140421,
      "loss": 1.6791,
      "theoretical_loss": 3.8150156130982618,
      "tokens_seen": 687210496
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007997592778335006,
      "loss": 1.5273,
      "theoretical_loss": 3.8149788289049917,
      "tokens_seen": 687276032
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007997392176529589,
      "loss": 1.5077,
      "theoretical_loss": 3.814942049201175,
      "tokens_seen": 687341568
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007997191574724173,
      "loss": 1.6135,
      "theoretical_loss": 3.8149052739858345,
      "tokens_seen": 687407104
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007996990972918756,
      "loss": 1.7439,
      "theoretical_loss": 3.8148685032579963,
      "tokens_seen": 687472640
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000799679037111334,
      "loss": 1.6633,
      "theoretical_loss": 3.814831737016684,
      "tokens_seen": 687538176
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007996589769307924,
      "loss": 1.5305,
      "theoretical_loss": 3.8147949752609236,
      "tokens_seen": 687603712
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007996389167502508,
      "loss": 1.7444,
      "theoretical_loss": 3.814758217989739,
      "tokens_seen": 687669248
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007996188565697091,
      "loss": 1.593,
      "theoretical_loss": 3.814721465202157,
      "tokens_seen": 687734784
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007995987963891675,
      "loss": 1.5504,
      "theoretical_loss": 3.8146847168972027,
      "tokens_seen": 687800320
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007995787362086259,
      "loss": 1.453,
      "theoretical_loss": 3.814647973073903,
      "tokens_seen": 687865856
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007995586760280843,
      "loss": 1.5786,
      "theoretical_loss": 3.8146112337312843,
      "tokens_seen": 687931392
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007995386158475426,
      "loss": 1.5101,
      "theoretical_loss": 3.8145744988683727,
      "tokens_seen": 687996928
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007995185556670011,
      "loss": 1.6444,
      "theoretical_loss": 3.814537768484197,
      "tokens_seen": 688062464
    },
    {
      "epoch": 0.21,
      "objective/train/advantage_avg": 0.4902955889701843,
      "objective/train/docs_used": 255827,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4737927913665771,
      "objective/train/original_loss": 1.4737927913665771,
      "objective/train/theoretical_loss": 3.8145010425777826,
      "objective/train/tokens_used": 708588000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24105313420295715,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502549409866333,
      "objective/train/weighted_lm_loss": 1.548177719116211,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.8145010425777826,
      "tokens_seen": 688128000
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007994984954864593,
      "loss": 1.5881,
      "theoretical_loss": 3.8145010425777826,
      "tokens_seen": 688128000
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007994784353059178,
      "loss": 1.5737,
      "theoretical_loss": 3.8144643211481584,
      "tokens_seen": 688193536
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007994583751253761,
      "loss": 1.6205,
      "theoretical_loss": 3.8144276041943526,
      "tokens_seen": 688259072
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007994383149448345,
      "loss": 1.6015,
      "theoretical_loss": 3.8143908917153926,
      "tokens_seen": 688324608
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007994182547642929,
      "loss": 1.5934,
      "theoretical_loss": 3.814354183710308,
      "tokens_seen": 688390144
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007993981945837513,
      "loss": 1.5689,
      "theoretical_loss": 3.814317480178128,
      "tokens_seen": 688455680
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007993781344032096,
      "loss": 1.6545,
      "theoretical_loss": 3.8142807811178807,
      "tokens_seen": 688521216
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007993580742226681,
      "loss": 1.5859,
      "theoretical_loss": 3.8142440865285963,
      "tokens_seen": 688586752
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007993380140421263,
      "loss": 1.5753,
      "theoretical_loss": 3.8142073964093046,
      "tokens_seen": 688652288
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007993179538615848,
      "loss": 1.4878,
      "theoretical_loss": 3.814170710759036,
      "tokens_seen": 688717824
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007992978936810432,
      "loss": 1.6496,
      "theoretical_loss": 3.8141340295768202,
      "tokens_seen": 688783360
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007992778335005015,
      "loss": 1.5481,
      "theoretical_loss": 3.814097352861689,
      "tokens_seen": 688848896
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007992577733199599,
      "loss": 1.6138,
      "theoretical_loss": 3.8140606806126733,
      "tokens_seen": 688914432
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007992377131394182,
      "loss": 1.5535,
      "theoretical_loss": 3.8140240128288037,
      "tokens_seen": 688979968
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007992176529588766,
      "loss": 1.6351,
      "theoretical_loss": 3.8139873495091123,
      "tokens_seen": 689045504
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007991975927783351,
      "loss": 1.592,
      "theoretical_loss": 3.8139506906526317,
      "tokens_seen": 689111040
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007991775325977933,
      "loss": 1.5992,
      "theoretical_loss": 3.813914036258393,
      "tokens_seen": 689176576
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007991574724172518,
      "loss": 1.4772,
      "theoretical_loss": 3.81387738632543,
      "tokens_seen": 689242112
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007991374122367102,
      "loss": 1.6113,
      "theoretical_loss": 3.8138407408527755,
      "tokens_seen": 689307648
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007991173520561685,
      "loss": 1.7327,
      "theoretical_loss": 3.8138040998394613,
      "tokens_seen": 689373184
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007990972918756269,
      "loss": 1.5241,
      "theoretical_loss": 3.813767463284522,
      "tokens_seen": 689438720
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007990772316950852,
      "loss": 1.6101,
      "theoretical_loss": 3.8137308311869917,
      "tokens_seen": 689504256
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007990571715145436,
      "loss": 1.6027,
      "theoretical_loss": 3.8136942035459036,
      "tokens_seen": 689569792
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007990371113340021,
      "loss": 1.562,
      "theoretical_loss": 3.813657580360293,
      "tokens_seen": 689635328
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007990170511534603,
      "loss": 1.6389,
      "theoretical_loss": 3.8136209616291934,
      "tokens_seen": 689700864
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007989969909729188,
      "loss": 1.605,
      "theoretical_loss": 3.813584347351641,
      "tokens_seen": 689766400
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007989769307923771,
      "loss": 1.6273,
      "theoretical_loss": 3.81354773752667,
      "tokens_seen": 689831936
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007989568706118355,
      "loss": 1.5705,
      "theoretical_loss": 3.8135111321533173,
      "tokens_seen": 689897472
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007989368104312939,
      "loss": 1.6024,
      "theoretical_loss": 3.813474531230618,
      "tokens_seen": 689963008
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007989167502507522,
      "loss": 1.498,
      "theoretical_loss": 3.813437934757608,
      "tokens_seen": 690028544
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007988966900702106,
      "loss": 1.6946,
      "theoretical_loss": 3.813401342733325,
      "tokens_seen": 690094080
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000798876629889669,
      "loss": 1.5686,
      "theoretical_loss": 3.813364755156804,
      "tokens_seen": 690159616
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007988565697091273,
      "loss": 1.5396,
      "theoretical_loss": 3.8133281720270835,
      "tokens_seen": 690225152
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007988365095285858,
      "loss": 1.5888,
      "theoretical_loss": 3.813291593343201,
      "tokens_seen": 690290688
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007988164493480441,
      "loss": 1.6458,
      "theoretical_loss": 3.8132550191041936,
      "tokens_seen": 690356224
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007987963891675025,
      "loss": 1.6777,
      "theoretical_loss": 3.8132184493090993,
      "tokens_seen": 690421760
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000798776328986961,
      "loss": 1.5544,
      "theoretical_loss": 3.8131818839569562,
      "tokens_seen": 690487296
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007987562688064193,
      "loss": 1.5344,
      "theoretical_loss": 3.8131453230468035,
      "tokens_seen": 690552832
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007987362086258777,
      "loss": 1.6199,
      "theoretical_loss": 3.8131087665776793,
      "tokens_seen": 690618368
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000798716148445336,
      "loss": 1.6303,
      "theoretical_loss": 3.813072214548624,
      "tokens_seen": 690683904
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007986960882647945,
      "loss": 1.678,
      "theoretical_loss": 3.8130356669586765,
      "tokens_seen": 690749440
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007986760280842528,
      "loss": 1.6316,
      "theoretical_loss": 3.8129991238068763,
      "tokens_seen": 690814976
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007986559679037112,
      "loss": 1.5212,
      "theoretical_loss": 3.8129625850922633,
      "tokens_seen": 690880512
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007986359077231695,
      "loss": 1.4729,
      "theoretical_loss": 3.8129260508138785,
      "tokens_seen": 690946048
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007986158475426279,
      "loss": 1.5299,
      "theoretical_loss": 3.8128895209707627,
      "tokens_seen": 691011584
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007985957873620863,
      "loss": 1.7006,
      "theoretical_loss": 3.8128529955619563,
      "tokens_seen": 691077120
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007985757271815447,
      "loss": 1.6765,
      "theoretical_loss": 3.8128164745865014,
      "tokens_seen": 691142656
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000798555667001003,
      "loss": 1.7169,
      "theoretical_loss": 3.8127799580434383,
      "tokens_seen": 691208192
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007985356068204615,
      "loss": 1.4903,
      "theoretical_loss": 3.81274344593181,
      "tokens_seen": 691273728
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007985155466399197,
      "loss": 1.5823,
      "theoretical_loss": 3.8127069382506584,
      "tokens_seen": 691339264
    },
    {
      "epoch": 0.21,
      "objective/train/advantage_avg": 0.495864599943161,
      "objective/train/docs_used": 257040,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4937437772750854,
      "objective/train/original_loss": 1.4937434196472168,
      "objective/train/theoretical_loss": 3.812670434999026,
      "objective/train/tokens_used": 711864800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24615159630775452,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508379936218262,
      "objective/train/weighted_lm_loss": 1.5696989297866821,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.812670434999026,
      "tokens_seen": 691404800
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007984954864593782,
      "loss": 1.5931,
      "theoretical_loss": 3.812670434999026,
      "tokens_seen": 691404800
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007984754262788365,
      "loss": 1.4626,
      "theoretical_loss": 3.8126339361759554,
      "tokens_seen": 691470336
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007984553660982949,
      "loss": 1.5853,
      "theoretical_loss": 3.8125974417804893,
      "tokens_seen": 691535872
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007984353059177533,
      "loss": 1.4963,
      "theoretical_loss": 3.8125609518116717,
      "tokens_seen": 691601408
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007984152457372117,
      "loss": 1.5533,
      "theoretical_loss": 3.8125244662685462,
      "tokens_seen": 691666944
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00079839518555667,
      "loss": 1.5982,
      "theoretical_loss": 3.812487985150157,
      "tokens_seen": 691732480
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007983751253761285,
      "loss": 1.4978,
      "theoretical_loss": 3.8124515084555477,
      "tokens_seen": 691798016
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007983550651955867,
      "loss": 1.482,
      "theoretical_loss": 3.812415036183763,
      "tokens_seen": 691863552
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007983350050150452,
      "loss": 1.631,
      "theoretical_loss": 3.8123785683338487,
      "tokens_seen": 691929088
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007983149448345036,
      "loss": 1.6434,
      "theoretical_loss": 3.8123421049048485,
      "tokens_seen": 691994624
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007982948846539619,
      "loss": 1.5355,
      "theoretical_loss": 3.8123056458958087,
      "tokens_seen": 692060160
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007982748244734203,
      "loss": 1.6391,
      "theoretical_loss": 3.8122691913057754,
      "tokens_seen": 692125696
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007982547642928786,
      "loss": 1.5335,
      "theoretical_loss": 3.8122327411337933,
      "tokens_seen": 692191232
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000798234704112337,
      "loss": 1.6104,
      "theoretical_loss": 3.8121962953789104,
      "tokens_seen": 692256768
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007982146439317955,
      "loss": 1.467,
      "theoretical_loss": 3.812159854040172,
      "tokens_seen": 692322304
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007981945837512537,
      "loss": 1.6437,
      "theoretical_loss": 3.812123417116626,
      "tokens_seen": 692387840
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007981745235707122,
      "loss": 1.5588,
      "theoretical_loss": 3.812086984607319,
      "tokens_seen": 692453376
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007981544633901706,
      "loss": 1.58,
      "theoretical_loss": 3.8120505565112985,
      "tokens_seen": 692518912
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007981344032096289,
      "loss": 1.6004,
      "theoretical_loss": 3.8120141328276125,
      "tokens_seen": 692584448
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007981143430290873,
      "loss": 1.7004,
      "theoretical_loss": 3.8119777135553097,
      "tokens_seen": 692649984
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007980942828485456,
      "loss": 1.5998,
      "theoretical_loss": 3.8119412986934376,
      "tokens_seen": 692715520
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000798074222668004,
      "loss": 1.6219,
      "theoretical_loss": 3.8119048882410453,
      "tokens_seen": 692781056
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007980541624874625,
      "loss": 1.5678,
      "theoretical_loss": 3.8118684821971813,
      "tokens_seen": 692846592
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007980341023069207,
      "loss": 1.5699,
      "theoretical_loss": 3.8118320805608965,
      "tokens_seen": 692912128
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007980140421263792,
      "loss": 1.6347,
      "theoretical_loss": 3.8117956833312383,
      "tokens_seen": 692977664
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007979939819458375,
      "loss": 1.5784,
      "theoretical_loss": 3.8117592905072577,
      "tokens_seen": 693043200
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007979739217652959,
      "loss": 1.6816,
      "theoretical_loss": 3.8117229020880057,
      "tokens_seen": 693108736
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007979538615847543,
      "loss": 1.7214,
      "theoretical_loss": 3.8116865180725314,
      "tokens_seen": 693174272
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007979338014042127,
      "loss": 1.5759,
      "theoretical_loss": 3.811650138459886,
      "tokens_seen": 693239808
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000797913741223671,
      "loss": 1.6032,
      "theoretical_loss": 3.8116137632491207,
      "tokens_seen": 693305344
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007978936810431294,
      "loss": 1.5391,
      "theoretical_loss": 3.811577392439287,
      "tokens_seen": 693370880
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007978736208625877,
      "loss": 1.6395,
      "theoretical_loss": 3.811541026029436,
      "tokens_seen": 693436416
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007978535606820462,
      "loss": 1.6483,
      "theoretical_loss": 3.81150466401862,
      "tokens_seen": 693501952
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007978335005015045,
      "loss": 1.6036,
      "theoretical_loss": 3.8114683064058914,
      "tokens_seen": 693567488
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007978134403209629,
      "loss": 1.545,
      "theoretical_loss": 3.8114319531903025,
      "tokens_seen": 693633024
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007977933801404213,
      "loss": 1.641,
      "theoretical_loss": 3.8113956043709063,
      "tokens_seen": 693698560
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007977733199598797,
      "loss": 1.5923,
      "theoretical_loss": 3.8113592599467556,
      "tokens_seen": 693764096
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000797753259779338,
      "loss": 1.6431,
      "theoretical_loss": 3.811322919916904,
      "tokens_seen": 693829632
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007977331995987964,
      "loss": 1.5936,
      "theoretical_loss": 3.811286584280406,
      "tokens_seen": 693895168
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007977131394182547,
      "loss": 1.6318,
      "theoretical_loss": 3.811250253036314,
      "tokens_seen": 693960704
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007976930792377132,
      "loss": 1.5063,
      "theoretical_loss": 3.811213926183683,
      "tokens_seen": 694026240
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007976730190571715,
      "loss": 1.6774,
      "theoretical_loss": 3.8111776037215686,
      "tokens_seen": 694091776
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007976529588766299,
      "loss": 1.619,
      "theoretical_loss": 3.8111412856490245,
      "tokens_seen": 694157312
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007976328986960882,
      "loss": 1.7793,
      "theoretical_loss": 3.8111049719651056,
      "tokens_seen": 694222848
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007976128385155467,
      "loss": 1.5698,
      "theoretical_loss": 3.811068662668869,
      "tokens_seen": 694288384
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000797592778335005,
      "loss": 1.5048,
      "theoretical_loss": 3.811032357759368,
      "tokens_seen": 694353920
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007975727181544634,
      "loss": 1.5804,
      "theoretical_loss": 3.810996057235661,
      "tokens_seen": 694419456
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007975526579739217,
      "loss": 1.7028,
      "theoretical_loss": 3.810959761096803,
      "tokens_seen": 694484992
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007975325977933801,
      "loss": 1.5144,
      "theoretical_loss": 3.8109234693418514,
      "tokens_seen": 694550528
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007975125376128385,
      "loss": 1.5183,
      "theoretical_loss": 3.8108871819698624,
      "tokens_seen": 694616064
    },
    {
      "epoch": 0.21,
      "objective/train/advantage_avg": 0.493087500333786,
      "objective/train/docs_used": 258341,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5312550067901611,
      "objective/train/original_loss": 1.5312550067901611,
      "objective/train/theoretical_loss": 3.810850898979894,
      "objective/train/tokens_used": 715141600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24364736676216125,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505473613739014,
      "objective/train/weighted_lm_loss": 1.608521580696106,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.810850898979894,
      "tokens_seen": 694681600
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007974924774322969,
      "loss": 1.6878,
      "theoretical_loss": 3.810850898979894,
      "tokens_seen": 694681600
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007974724172517552,
      "loss": 1.5909,
      "theoretical_loss": 3.8108146203710027,
      "tokens_seen": 694747136
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007974523570712137,
      "loss": 1.6253,
      "theoretical_loss": 3.8107783461422473,
      "tokens_seen": 694812672
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000797432296890672,
      "loss": 1.6755,
      "theoretical_loss": 3.8107420762926854,
      "tokens_seen": 694878208
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007974122367101304,
      "loss": 1.5974,
      "theoretical_loss": 3.810705810821375,
      "tokens_seen": 694943744
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007973921765295888,
      "loss": 1.5789,
      "theoretical_loss": 3.8106695497273755,
      "tokens_seen": 695009280
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007973721163490471,
      "loss": 1.6427,
      "theoretical_loss": 3.810633293009746,
      "tokens_seen": 695074816
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007973520561685055,
      "loss": 1.499,
      "theoretical_loss": 3.810597040667545,
      "tokens_seen": 695140352
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000797331995987964,
      "loss": 1.6556,
      "theoretical_loss": 3.8105607926998326,
      "tokens_seen": 695205888
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007973119358074222,
      "loss": 1.5852,
      "theoretical_loss": 3.810524549105668,
      "tokens_seen": 695271424
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007972918756268807,
      "loss": 1.5003,
      "theoretical_loss": 3.810488309884112,
      "tokens_seen": 695336960
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000797271815446339,
      "loss": 1.6606,
      "theoretical_loss": 3.810452075034225,
      "tokens_seen": 695402496
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007972517552657974,
      "loss": 1.5402,
      "theoretical_loss": 3.810415844555067,
      "tokens_seen": 695468032
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007972316950852559,
      "loss": 1.5619,
      "theoretical_loss": 3.8103796184456997,
      "tokens_seen": 695533568
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007972116349047141,
      "loss": 1.4855,
      "theoretical_loss": 3.8103433967051843,
      "tokens_seen": 695599104
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007971915747241726,
      "loss": 1.7278,
      "theoretical_loss": 3.8103071793325825,
      "tokens_seen": 695664640
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000797171514543631,
      "loss": 1.4759,
      "theoretical_loss": 3.8102709663269554,
      "tokens_seen": 695730176
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007971514543630893,
      "loss": 1.5809,
      "theoretical_loss": 3.8102347576873656,
      "tokens_seen": 695795712
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007971313941825477,
      "loss": 1.6217,
      "theoretical_loss": 3.810198553412876,
      "tokens_seen": 695861248
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000797111334002006,
      "loss": 1.5495,
      "theoretical_loss": 3.8101623535025495,
      "tokens_seen": 695926784
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007970912738214644,
      "loss": 1.6009,
      "theoretical_loss": 3.810126157955448,
      "tokens_seen": 695992320
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007970712136409229,
      "loss": 1.4819,
      "theoretical_loss": 3.8100899667706356,
      "tokens_seen": 696057856
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007970511534603811,
      "loss": 1.5304,
      "theoretical_loss": 3.8100537799471756,
      "tokens_seen": 696123392
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007970310932798396,
      "loss": 1.6377,
      "theoretical_loss": 3.810017597484132,
      "tokens_seen": 696188928
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007970110330992979,
      "loss": 1.5886,
      "theoretical_loss": 3.809981419380569,
      "tokens_seen": 696254464
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007969909729187563,
      "loss": 1.6084,
      "theoretical_loss": 3.809945245635552,
      "tokens_seen": 696320000
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007969709127382147,
      "loss": 1.6489,
      "theoretical_loss": 3.8099090762481436,
      "tokens_seen": 696385536
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000796950852557673,
      "loss": 1.5483,
      "theoretical_loss": 3.809872911217411,
      "tokens_seen": 696451072
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007969307923771314,
      "loss": 1.6001,
      "theoretical_loss": 3.809836750542418,
      "tokens_seen": 696516608
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007969107321965898,
      "loss": 1.5508,
      "theoretical_loss": 3.809800594222231,
      "tokens_seen": 696582144
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007968906720160481,
      "loss": 1.4993,
      "theoretical_loss": 3.809764442255916,
      "tokens_seen": 696647680
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007968706118355066,
      "loss": 1.661,
      "theoretical_loss": 3.809728294642539,
      "tokens_seen": 696713216
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007968505516549649,
      "loss": 1.7267,
      "theoretical_loss": 3.8096921513811663,
      "tokens_seen": 696778752
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007968304914744233,
      "loss": 1.4959,
      "theoretical_loss": 3.8096560124708656,
      "tokens_seen": 696844288
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007968104312938817,
      "loss": 1.6139,
      "theoretical_loss": 3.8096198779107024,
      "tokens_seen": 696909824
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007967903711133401,
      "loss": 1.6846,
      "theoretical_loss": 3.809583747699745,
      "tokens_seen": 696975360
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007967703109327984,
      "loss": 1.6259,
      "theoretical_loss": 3.809547621837061,
      "tokens_seen": 697040896
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007967502507522568,
      "loss": 1.5647,
      "theoretical_loss": 3.809511500321719,
      "tokens_seen": 697106432
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007967301905717151,
      "loss": 1.5033,
      "theoretical_loss": 3.809475383152786,
      "tokens_seen": 697171968
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007967101303911736,
      "loss": 1.5079,
      "theoretical_loss": 3.8094392703293307,
      "tokens_seen": 697237504
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007966900702106319,
      "loss": 1.5697,
      "theoretical_loss": 3.809403161850423,
      "tokens_seen": 697303040
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007966700100300903,
      "loss": 1.6956,
      "theoretical_loss": 3.8093670577151304,
      "tokens_seen": 697368576
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007966499498495486,
      "loss": 1.6088,
      "theoretical_loss": 3.8093309579225236,
      "tokens_seen": 697434112
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007966298896690071,
      "loss": 1.5542,
      "theoretical_loss": 3.8092948624716714,
      "tokens_seen": 697499648
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007966098294884654,
      "loss": 1.7591,
      "theoretical_loss": 3.8092587713616446,
      "tokens_seen": 697565184
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007965897693079238,
      "loss": 1.6305,
      "theoretical_loss": 3.809222684591512,
      "tokens_seen": 697630720
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007965697091273821,
      "loss": 1.6408,
      "theoretical_loss": 3.809186602160346,
      "tokens_seen": 697696256
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007965496489468405,
      "loss": 1.5243,
      "theoretical_loss": 3.809150524067215,
      "tokens_seen": 697761792
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007965295887662989,
      "loss": 1.7124,
      "theoretical_loss": 3.809114450311193,
      "tokens_seen": 697827328
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007965095285857573,
      "loss": 1.6639,
      "theoretical_loss": 3.809078380891349,
      "tokens_seen": 697892864
    },
    {
      "epoch": 0.21,
      "objective/train/advantage_avg": 0.49035149812698364,
      "objective/train/docs_used": 259086,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.547105312347412,
      "objective/train/original_loss": 1.5471055507659912,
      "objective/train/theoretical_loss": 3.8090423158067566,
      "objective/train/tokens_used": 718418400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2414310872554779,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050262451171875,
      "objective/train/weighted_lm_loss": 1.62467360496521,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8090423158067566,
      "tokens_seen": 697958400
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007964894684052156,
      "loss": 1.5776,
      "theoretical_loss": 3.8090423158067566,
      "tokens_seen": 697958400
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007964694082246741,
      "loss": 1.5508,
      "theoretical_loss": 3.8090062550564854,
      "tokens_seen": 698023936
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007964493480441324,
      "loss": 1.6278,
      "theoretical_loss": 3.808970198639609,
      "tokens_seen": 698089472
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007964292878635908,
      "loss": 1.4673,
      "theoretical_loss": 3.8089341465552007,
      "tokens_seen": 698155008
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007964092276830492,
      "loss": 1.4976,
      "theoretical_loss": 3.8088980988023318,
      "tokens_seen": 698220544
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007963891675025075,
      "loss": 1.6829,
      "theoretical_loss": 3.808862055380076,
      "tokens_seen": 698286080
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007963691073219659,
      "loss": 1.6115,
      "theoretical_loss": 3.808826016287507,
      "tokens_seen": 698351616
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007963490471414244,
      "loss": 1.5584,
      "theoretical_loss": 3.808789981523698,
      "tokens_seen": 698417152
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007963289869608826,
      "loss": 1.6336,
      "theoretical_loss": 3.808753951087723,
      "tokens_seen": 698482688
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007963089267803411,
      "loss": 1.5842,
      "theoretical_loss": 3.8087179249786565,
      "tokens_seen": 698548224
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007962888665997993,
      "loss": 1.6753,
      "theoretical_loss": 3.8086819031955725,
      "tokens_seen": 698613760
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007962688064192578,
      "loss": 1.4443,
      "theoretical_loss": 3.8086458857375467,
      "tokens_seen": 698679296
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007962487462387162,
      "loss": 1.6275,
      "theoretical_loss": 3.808609872603653,
      "tokens_seen": 698744832
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007962286860581745,
      "loss": 1.6111,
      "theoretical_loss": 3.808573863792968,
      "tokens_seen": 698810368
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007962086258776329,
      "loss": 1.6399,
      "theoretical_loss": 3.8085378593045665,
      "tokens_seen": 698875904
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007961885656970914,
      "loss": 1.6592,
      "theoretical_loss": 3.8085018591375244,
      "tokens_seen": 698941440
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007961685055165496,
      "loss": 1.6648,
      "theoretical_loss": 3.8084658632909183,
      "tokens_seen": 699006976
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007961484453360081,
      "loss": 1.7011,
      "theoretical_loss": 3.808429871763825,
      "tokens_seen": 699072512
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007961283851554663,
      "loss": 1.5506,
      "theoretical_loss": 3.8083938845553202,
      "tokens_seen": 699138048
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007961083249749248,
      "loss": 1.5918,
      "theoretical_loss": 3.808357901664482,
      "tokens_seen": 699203584
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007960882647943832,
      "loss": 1.5598,
      "theoretical_loss": 3.8083219230903875,
      "tokens_seen": 699269120
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007960682046138415,
      "loss": 1.5336,
      "theoretical_loss": 3.808285948832114,
      "tokens_seen": 699334656
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007960481444332999,
      "loss": 1.5881,
      "theoretical_loss": 3.80824997888874,
      "tokens_seen": 699400192
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007960280842527583,
      "loss": 1.6279,
      "theoretical_loss": 3.8082140132593434,
      "tokens_seen": 699465728
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007960080240722166,
      "loss": 1.6427,
      "theoretical_loss": 3.8081780519430026,
      "tokens_seen": 699531264
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007959879638916751,
      "loss": 1.6066,
      "theoretical_loss": 3.808142094938796,
      "tokens_seen": 699596800
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007959679037111333,
      "loss": 1.6084,
      "theoretical_loss": 3.8081061422458036,
      "tokens_seen": 699662336
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007959478435305918,
      "loss": 1.5616,
      "theoretical_loss": 3.808070193863104,
      "tokens_seen": 699727872
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007959277833500501,
      "loss": 1.6628,
      "theoretical_loss": 3.8080342497897774,
      "tokens_seen": 699793408
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007959077231695085,
      "loss": 1.5142,
      "theoretical_loss": 3.807998310024903,
      "tokens_seen": 699858944
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007958876629889669,
      "loss": 1.5766,
      "theoretical_loss": 3.8079623745675613,
      "tokens_seen": 699924480
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007958676028084253,
      "loss": 1.6724,
      "theoretical_loss": 3.807926443416833,
      "tokens_seen": 699990016
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007958475426278836,
      "loss": 1.4754,
      "theoretical_loss": 3.8078905165717987,
      "tokens_seen": 700055552
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007958274824473421,
      "loss": 1.3103,
      "theoretical_loss": 3.8078545940315394,
      "tokens_seen": 700121088
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007958074222668003,
      "loss": 1.6138,
      "theoretical_loss": 3.8078186757951364,
      "tokens_seen": 700186624
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007957873620862588,
      "loss": 1.6303,
      "theoretical_loss": 3.807782761861671,
      "tokens_seen": 700252160
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007957673019057172,
      "loss": 1.6188,
      "theoretical_loss": 3.807746852230226,
      "tokens_seen": 700317696
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007957472417251755,
      "loss": 1.6338,
      "theoretical_loss": 3.8077109468998827,
      "tokens_seen": 700383232
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000795727181544634,
      "loss": 1.6137,
      "theoretical_loss": 3.8076750458697237,
      "tokens_seen": 700448768
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007957071213640923,
      "loss": 1.629,
      "theoretical_loss": 3.807639149138832,
      "tokens_seen": 700514304
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007956870611835507,
      "loss": 1.5426,
      "theoretical_loss": 3.8076032567062903,
      "tokens_seen": 700579840
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000795667001003009,
      "loss": 1.6689,
      "theoretical_loss": 3.807567368571182,
      "tokens_seen": 700645376
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007956469408224675,
      "loss": 1.5117,
      "theoretical_loss": 3.807531484732591,
      "tokens_seen": 700710912
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007956268806419258,
      "loss": 1.6377,
      "theoretical_loss": 3.807495605189601,
      "tokens_seen": 700776448
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007956068204613842,
      "loss": 1.5663,
      "theoretical_loss": 3.8074597299412956,
      "tokens_seen": 700841984
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007955867602808426,
      "loss": 1.5769,
      "theoretical_loss": 3.8074238589867595,
      "tokens_seen": 700907520
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007955667001003009,
      "loss": 1.6036,
      "theoretical_loss": 3.8073879923250775,
      "tokens_seen": 700973056
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007955466399197593,
      "loss": 1.5721,
      "theoretical_loss": 3.807352129955335,
      "tokens_seen": 701038592
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007955265797392177,
      "loss": 1.5065,
      "theoretical_loss": 3.8073162718766165,
      "tokens_seen": 701104128
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000795506519558676,
      "loss": 1.5639,
      "theoretical_loss": 3.8072804180880073,
      "tokens_seen": 701169664
    },
    {
      "epoch": 0.21,
      "objective/train/advantage_avg": 0.49331745505332947,
      "objective/train/docs_used": 260260,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5297163724899292,
      "objective/train/original_loss": 1.5297164916992188,
      "objective/train/theoretical_loss": 3.8072445685885947,
      "objective/train/tokens_used": 721695200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24370309710502625,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505706071853638,
      "objective/train/weighted_lm_loss": 1.6071056127548218,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8072445685885947,
      "tokens_seen": 701235200
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007954864593781345,
      "loss": 1.6031,
      "theoretical_loss": 3.8072445685885947,
      "tokens_seen": 701235200
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007954663991975928,
      "loss": 1.6038,
      "theoretical_loss": 3.8072087233774634,
      "tokens_seen": 701300736
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007954463390170512,
      "loss": 1.5676,
      "theoretical_loss": 3.8071728824537003,
      "tokens_seen": 701366272
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007954262788365096,
      "loss": 1.5613,
      "theoretical_loss": 3.807137045816392,
      "tokens_seen": 701431808
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007954062186559679,
      "loss": 1.6013,
      "theoretical_loss": 3.807101213464625,
      "tokens_seen": 701497344
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007953861584754263,
      "loss": 1.6037,
      "theoretical_loss": 3.807065385397488,
      "tokens_seen": 701562880
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007953660982948848,
      "loss": 1.5803,
      "theoretical_loss": 3.8070295616140664,
      "tokens_seen": 701628416
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000795346038114343,
      "loss": 1.6331,
      "theoretical_loss": 3.806993742113449,
      "tokens_seen": 701693952
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007953259779338015,
      "loss": 1.6372,
      "theoretical_loss": 3.8069579268947242,
      "tokens_seen": 701759488
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007953059177532597,
      "loss": 1.6053,
      "theoretical_loss": 3.80692211595698,
      "tokens_seen": 701825024
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007952858575727182,
      "loss": 1.6005,
      "theoretical_loss": 3.8068863092993057,
      "tokens_seen": 701890560
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007952657973921766,
      "loss": 1.6922,
      "theoretical_loss": 3.8068505069207887,
      "tokens_seen": 701956096
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007952457372116349,
      "loss": 1.6273,
      "theoretical_loss": 3.806814708820519,
      "tokens_seen": 702021632
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007952256770310933,
      "loss": 1.5642,
      "theoretical_loss": 3.8067789149975857,
      "tokens_seen": 702087168
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007952056168505518,
      "loss": 1.5509,
      "theoretical_loss": 3.8067431254510797,
      "tokens_seen": 702152704
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00079518555667001,
      "loss": 1.3745,
      "theoretical_loss": 3.8067073401800897,
      "tokens_seen": 702218240
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007951654964894685,
      "loss": 1.6217,
      "theoretical_loss": 3.806671559183706,
      "tokens_seen": 702283776
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007951454363089267,
      "loss": 1.5893,
      "theoretical_loss": 3.8066357824610204,
      "tokens_seen": 702349312
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007951253761283852,
      "loss": 1.5119,
      "theoretical_loss": 3.806600010011122,
      "tokens_seen": 702414848
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007951053159478436,
      "loss": 1.5991,
      "theoretical_loss": 3.806564241833104,
      "tokens_seen": 702480384
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007950852557673019,
      "loss": 1.5648,
      "theoretical_loss": 3.806528477926056,
      "tokens_seen": 702545920
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007950651955867603,
      "loss": 1.5748,
      "theoretical_loss": 3.8064927182890704,
      "tokens_seen": 702611456
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007950451354062187,
      "loss": 1.732,
      "theoretical_loss": 3.8064569629212395,
      "tokens_seen": 702676992
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000795025075225677,
      "loss": 1.6394,
      "theoretical_loss": 3.8064212118216547,
      "tokens_seen": 702742528
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007950050150451355,
      "loss": 1.609,
      "theoretical_loss": 3.806385464989409,
      "tokens_seen": 702808064
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007949849548645937,
      "loss": 1.5847,
      "theoretical_loss": 3.8063497224235947,
      "tokens_seen": 702873600
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007949648946840522,
      "loss": 1.6652,
      "theoretical_loss": 3.806313984123306,
      "tokens_seen": 702939136
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007949448345035105,
      "loss": 1.6147,
      "theoretical_loss": 3.806278250087635,
      "tokens_seen": 703004672
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007949247743229689,
      "loss": 1.6275,
      "theoretical_loss": 3.806242520315676,
      "tokens_seen": 703070208
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007949047141424273,
      "loss": 1.6169,
      "theoretical_loss": 3.806206794806523,
      "tokens_seen": 703135744
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007948846539618857,
      "loss": 1.6433,
      "theoretical_loss": 3.8061710735592698,
      "tokens_seen": 703201280
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000794864593781344,
      "loss": 1.5227,
      "theoretical_loss": 3.8061353565730105,
      "tokens_seen": 703266816
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007948445336008025,
      "loss": 1.5733,
      "theoretical_loss": 3.806099643846841,
      "tokens_seen": 703332352
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007948244734202607,
      "loss": 1.5473,
      "theoretical_loss": 3.8060639353798553,
      "tokens_seen": 703397888
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007948044132397192,
      "loss": 1.6351,
      "theoretical_loss": 3.8060282311711484,
      "tokens_seen": 703463424
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007947843530591775,
      "loss": 1.5236,
      "theoretical_loss": 3.8059925312198173,
      "tokens_seen": 703528960
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007947642928786359,
      "loss": 1.6539,
      "theoretical_loss": 3.8059568355249564,
      "tokens_seen": 703594496
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007947442326980943,
      "loss": 1.4693,
      "theoretical_loss": 3.8059211440856626,
      "tokens_seen": 703660032
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007947241725175527,
      "loss": 1.5656,
      "theoretical_loss": 3.805885456901032,
      "tokens_seen": 703725568
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000794704112337011,
      "loss": 1.7013,
      "theoretical_loss": 3.8058497739701616,
      "tokens_seen": 703791104
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007946840521564694,
      "loss": 1.602,
      "theoretical_loss": 3.8058140952921478,
      "tokens_seen": 703856640
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007946639919759278,
      "loss": 1.7175,
      "theoretical_loss": 3.805778420866088,
      "tokens_seen": 703922176
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007946439317953862,
      "loss": 1.5338,
      "theoretical_loss": 3.8057427506910795,
      "tokens_seen": 703987712
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007946238716148445,
      "loss": 1.5493,
      "theoretical_loss": 3.8057070847662207,
      "tokens_seen": 704053248
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000794603811434303,
      "loss": 1.5876,
      "theoretical_loss": 3.805671423090609,
      "tokens_seen": 704118784
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007945837512537612,
      "loss": 1.6319,
      "theoretical_loss": 3.8056357656633435,
      "tokens_seen": 704184320
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007945636910732197,
      "loss": 1.5752,
      "theoretical_loss": 3.805600112483522,
      "tokens_seen": 704249856
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000794543630892678,
      "loss": 1.5798,
      "theoretical_loss": 3.805564463550243,
      "tokens_seen": 704315392
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007945235707121364,
      "loss": 1.5116,
      "theoretical_loss": 3.805528818862607,
      "tokens_seen": 704380928
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007945035105315948,
      "loss": 1.614,
      "theoretical_loss": 3.8054931784197126,
      "tokens_seen": 704446464
    },
    {
      "epoch": 0.21,
      "objective/train/advantage_avg": 0.49504390358924866,
      "objective/train/docs_used": 261464,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.364923357963562,
      "objective/train/original_loss": 1.3649234771728516,
      "objective/train/theoretical_loss": 3.8054575422206596,
      "objective/train/tokens_used": 724972000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24523049592971802,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507512092590332,
      "objective/train/weighted_lm_loss": 1.4341381788253784,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 3.8054575422206596,
      "tokens_seen": 704512000
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007944834503510532,
      "loss": 1.5105,
      "theoretical_loss": 3.8054575422206596,
      "tokens_seen": 704512000
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007944633901705115,
      "loss": 1.5119,
      "theoretical_loss": 3.805421910264548,
      "tokens_seen": 704577536
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00079444332998997,
      "loss": 1.6478,
      "theoretical_loss": 3.8053862825504776,
      "tokens_seen": 704643072
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007944232698094282,
      "loss": 1.6431,
      "theoretical_loss": 3.8053506590775497,
      "tokens_seen": 704708608
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007944032096288867,
      "loss": 1.6046,
      "theoretical_loss": 3.805315039844864,
      "tokens_seen": 704774144
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000794383149448345,
      "loss": 1.5347,
      "theoretical_loss": 3.805279424851523,
      "tokens_seen": 704839680
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007943630892678034,
      "loss": 1.7186,
      "theoretical_loss": 3.8052438140966265,
      "tokens_seen": 704905216
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007943430290872618,
      "loss": 1.6004,
      "theoretical_loss": 3.8052082075792777,
      "tokens_seen": 704970752
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007943229689067201,
      "loss": 1.6504,
      "theoretical_loss": 3.8051726052985773,
      "tokens_seen": 705036288
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007943029087261785,
      "loss": 1.5779,
      "theoretical_loss": 3.8051370072536277,
      "tokens_seen": 705101824
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000794282848545637,
      "loss": 1.628,
      "theoretical_loss": 3.8051014134435315,
      "tokens_seen": 705167360
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007942627883650953,
      "loss": 1.5005,
      "theoretical_loss": 3.8050658238673916,
      "tokens_seen": 705232896
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007942427281845537,
      "loss": 1.5776,
      "theoretical_loss": 3.8050302385243104,
      "tokens_seen": 705298432
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007942226680040122,
      "loss": 1.4885,
      "theoretical_loss": 3.804994657413392,
      "tokens_seen": 705363968
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007942026078234704,
      "loss": 1.4819,
      "theoretical_loss": 3.804959080533739,
      "tokens_seen": 705429504
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007941825476429289,
      "loss": 1.5084,
      "theoretical_loss": 3.804923507884456,
      "tokens_seen": 705495040
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007941624874623871,
      "loss": 1.6455,
      "theoretical_loss": 3.804887939464646,
      "tokens_seen": 705560576
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007941424272818456,
      "loss": 1.6027,
      "theoretical_loss": 3.8048523752734145,
      "tokens_seen": 705626112
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.000794122367101304,
      "loss": 1.6124,
      "theoretical_loss": 3.8048168153098656,
      "tokens_seen": 705691648
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007941023069207623,
      "loss": 1.4975,
      "theoretical_loss": 3.8047812595731045,
      "tokens_seen": 705757184
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007940822467402207,
      "loss": 1.6138,
      "theoretical_loss": 3.804745708062236,
      "tokens_seen": 705822720
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007940621865596791,
      "loss": 1.6485,
      "theoretical_loss": 3.804710160776365,
      "tokens_seen": 705888256
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007940421263791374,
      "loss": 1.5516,
      "theoretical_loss": 3.8046746177145985,
      "tokens_seen": 705953792
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007940220661985959,
      "loss": 1.531,
      "theoretical_loss": 3.8046390788760416,
      "tokens_seen": 706019328
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007940020060180541,
      "loss": 1.5917,
      "theoretical_loss": 3.804603544259801,
      "tokens_seen": 706084864
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007939819458375126,
      "loss": 1.6326,
      "theoretical_loss": 3.8045680138649827,
      "tokens_seen": 706150400
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007939618856569709,
      "loss": 1.653,
      "theoretical_loss": 3.804532487690694,
      "tokens_seen": 706215936
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007939418254764293,
      "loss": 1.637,
      "theoretical_loss": 3.8044969657360417,
      "tokens_seen": 706281472
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007939217652958877,
      "loss": 1.5319,
      "theoretical_loss": 3.804461448000133,
      "tokens_seen": 706347008
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007939017051153461,
      "loss": 1.5876,
      "theoretical_loss": 3.804425934482076,
      "tokens_seen": 706412544
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007938816449348044,
      "loss": 1.5274,
      "theoretical_loss": 3.8043904251809786,
      "tokens_seen": 706478080
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007938615847542629,
      "loss": 1.6128,
      "theoretical_loss": 3.804354920095948,
      "tokens_seen": 706543616
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007938415245737211,
      "loss": 1.6492,
      "theoretical_loss": 3.8043194192260934,
      "tokens_seen": 706609152
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007938214643931796,
      "loss": 1.5834,
      "theoretical_loss": 3.804283922570524,
      "tokens_seen": 706674688
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007938014042126379,
      "loss": 1.5575,
      "theoretical_loss": 3.8042484301283475,
      "tokens_seen": 706740224
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007937813440320963,
      "loss": 1.569,
      "theoretical_loss": 3.804212941898674,
      "tokens_seen": 706805760
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007937612838515547,
      "loss": 1.6856,
      "theoretical_loss": 3.8041774578806127,
      "tokens_seen": 706871296
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007937412236710131,
      "loss": 1.6482,
      "theoretical_loss": 3.804141978073274,
      "tokens_seen": 706936832
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007937211634904714,
      "loss": 1.5791,
      "theoretical_loss": 3.8041065024757668,
      "tokens_seen": 707002368
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007937011033099298,
      "loss": 1.5398,
      "theoretical_loss": 3.804071031087202,
      "tokens_seen": 707067904
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007936810431293882,
      "loss": 1.6544,
      "theoretical_loss": 3.8040355639066905,
      "tokens_seen": 707133440
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007936609829488466,
      "loss": 1.6486,
      "theoretical_loss": 3.804000100933343,
      "tokens_seen": 707198976
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007936409227683049,
      "loss": 1.5004,
      "theoretical_loss": 3.8039646421662705,
      "tokens_seen": 707264512
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007936208625877634,
      "loss": 1.563,
      "theoretical_loss": 3.8039291876045844,
      "tokens_seen": 707330048
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007936008024072216,
      "loss": 1.4883,
      "theoretical_loss": 3.803893737247396,
      "tokens_seen": 707395584
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007935807422266801,
      "loss": 1.6022,
      "theoretical_loss": 3.8038582910938183,
      "tokens_seen": 707461120
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007935606820461384,
      "loss": 1.6282,
      "theoretical_loss": 3.8038228491429624,
      "tokens_seen": 707526656
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007935406218655968,
      "loss": 1.5571,
      "theoretical_loss": 3.803787411393942,
      "tokens_seen": 707592192
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007935205616850552,
      "loss": 1.5659,
      "theoretical_loss": 3.8037519778458684,
      "tokens_seen": 707657728
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007935005015045136,
      "loss": 1.5979,
      "theoretical_loss": 3.8037165484978552,
      "tokens_seen": 707723264
    },
    {
      "epoch": 0.21,
      "objective/train/advantage_avg": 0.4892263114452362,
      "objective/train/docs_used": 262675,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.7046771049499512,
      "objective/train/original_loss": 1.7046769857406616,
      "objective/train/theoretical_loss": 3.8036811233490164,
      "objective/train/tokens_used": 728248800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.240013986825943,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501426458358765,
      "objective/train/weighted_lm_loss": 1.7900488376617432,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8036811233490164,
      "tokens_seen": 707788800
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007934804413239719,
      "loss": 1.7129,
      "theoretical_loss": 3.8036811233490164,
      "tokens_seen": 707788800
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007934603811434304,
      "loss": 1.5022,
      "theoretical_loss": 3.8036457023984642,
      "tokens_seen": 707854336
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007934403209628886,
      "loss": 1.6969,
      "theoretical_loss": 3.803610285645314,
      "tokens_seen": 707919872
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007934202607823471,
      "loss": 1.5689,
      "theoretical_loss": 3.803574873088679,
      "tokens_seen": 707985408
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007934002006018054,
      "loss": 1.4729,
      "theoretical_loss": 3.803539464727673,
      "tokens_seen": 708050944
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007933801404212638,
      "loss": 1.6521,
      "theoretical_loss": 3.803504060561412,
      "tokens_seen": 708116480
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007933600802407222,
      "loss": 1.6777,
      "theoretical_loss": 3.80346866058901,
      "tokens_seen": 708182016
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007933400200601805,
      "loss": 1.6004,
      "theoretical_loss": 3.8034332648095823,
      "tokens_seen": 708247552
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007933199598796389,
      "loss": 1.6176,
      "theoretical_loss": 3.803397873222244,
      "tokens_seen": 708313088
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007932998996990974,
      "loss": 1.6163,
      "theoretical_loss": 3.803362485826112,
      "tokens_seen": 708378624
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007932798395185556,
      "loss": 1.6493,
      "theoretical_loss": 3.803327102620301,
      "tokens_seen": 708444160
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007932597793380141,
      "loss": 1.6295,
      "theoretical_loss": 3.8032917236039285,
      "tokens_seen": 708509696
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007932397191574723,
      "loss": 1.567,
      "theoretical_loss": 3.8032563487761095,
      "tokens_seen": 708575232
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007932196589769308,
      "loss": 1.615,
      "theoretical_loss": 3.803220978135962,
      "tokens_seen": 708640768
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007931995987963892,
      "loss": 1.4872,
      "theoretical_loss": 3.8031856116826024,
      "tokens_seen": 708706304
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007931795386158475,
      "loss": 1.6779,
      "theoretical_loss": 3.8031502494151486,
      "tokens_seen": 708771840
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007931594784353059,
      "loss": 1.4168,
      "theoretical_loss": 3.8031148913327177,
      "tokens_seen": 708837376
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007931394182547644,
      "loss": 1.6004,
      "theoretical_loss": 3.803079537434427,
      "tokens_seen": 708902912
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007931193580742226,
      "loss": 1.5426,
      "theoretical_loss": 3.8030441877193963,
      "tokens_seen": 708968448
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007930992978936811,
      "loss": 1.7211,
      "theoretical_loss": 3.803008842186742,
      "tokens_seen": 709033984
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007930792377131393,
      "loss": 1.6251,
      "theoretical_loss": 3.8029735008355843,
      "tokens_seen": 709099520
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007930591775325978,
      "loss": 1.5593,
      "theoretical_loss": 3.8029381636650417,
      "tokens_seen": 709165056
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007930391173520562,
      "loss": 1.5665,
      "theoretical_loss": 3.802902830674233,
      "tokens_seen": 709230592
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007930190571715145,
      "loss": 1.6884,
      "theoretical_loss": 3.802867501862278,
      "tokens_seen": 709296128
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007929989969909729,
      "loss": 1.6235,
      "theoretical_loss": 3.8028321772282965,
      "tokens_seen": 709361664
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007929789368104313,
      "loss": 1.4741,
      "theoretical_loss": 3.802796856771408,
      "tokens_seen": 709427200
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0007929588766298896,
      "loss": 1.6948,
      "theoretical_loss": 3.8027615404907333,
      "tokens_seen": 709492736
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007929388164493481,
      "loss": 1.6068,
      "theoretical_loss": 3.8027262283853926,
      "tokens_seen": 709558272
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007929187562688064,
      "loss": 1.5802,
      "theoretical_loss": 3.8026909204545065,
      "tokens_seen": 709623808
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007928986960882648,
      "loss": 1.6778,
      "theoretical_loss": 3.802655616697197,
      "tokens_seen": 709689344
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007928786359077232,
      "loss": 1.613,
      "theoretical_loss": 3.8026203171125843,
      "tokens_seen": 709754880
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007928585757271815,
      "loss": 1.5604,
      "theoretical_loss": 3.802585021699791,
      "tokens_seen": 709820416
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007928385155466399,
      "loss": 1.6864,
      "theoretical_loss": 3.8025497304579376,
      "tokens_seen": 709885952
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007928184553660983,
      "loss": 1.5349,
      "theoretical_loss": 3.8025144433861477,
      "tokens_seen": 709951488
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007927983951855566,
      "loss": 1.6653,
      "theoretical_loss": 3.802479160483543,
      "tokens_seen": 710017024
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007927783350050151,
      "loss": 1.5369,
      "theoretical_loss": 3.8024438817492463,
      "tokens_seen": 710082560
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007927582748244735,
      "loss": 1.5805,
      "theoretical_loss": 3.80240860718238,
      "tokens_seen": 710148096
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007927382146439318,
      "loss": 1.7039,
      "theoretical_loss": 3.802373336782068,
      "tokens_seen": 710213632
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007927181544633902,
      "loss": 1.7207,
      "theoretical_loss": 3.8023380705474334,
      "tokens_seen": 710279168
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007926980942828486,
      "loss": 1.635,
      "theoretical_loss": 3.8023028084775996,
      "tokens_seen": 710344704
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000792678034102307,
      "loss": 1.6163,
      "theoretical_loss": 3.802267550571691,
      "tokens_seen": 710410240
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007926579739217653,
      "loss": 1.5427,
      "theoretical_loss": 3.8022322968288322,
      "tokens_seen": 710475776
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007926379137412238,
      "loss": 1.7412,
      "theoretical_loss": 3.802197047248147,
      "tokens_seen": 710541312
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000792617853560682,
      "loss": 1.6544,
      "theoretical_loss": 3.8021618018287606,
      "tokens_seen": 710606848
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007925977933801405,
      "loss": 1.6506,
      "theoretical_loss": 3.802126560569798,
      "tokens_seen": 710672384
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007925777331995988,
      "loss": 1.7026,
      "theoretical_loss": 3.802091323470384,
      "tokens_seen": 710737920
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007925576730190572,
      "loss": 1.6655,
      "theoretical_loss": 3.8020560905296445,
      "tokens_seen": 710803456
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007925376128385156,
      "loss": 1.531,
      "theoretical_loss": 3.8020208617467057,
      "tokens_seen": 710868992
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000792517552657974,
      "loss": 1.6783,
      "theoretical_loss": 3.801985637120694,
      "tokens_seen": 710934528
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007924974924774323,
      "loss": 1.6069,
      "theoretical_loss": 3.8019504166507345,
      "tokens_seen": 711000064
    },
    {
      "epoch": 0.22,
      "objective/train/advantage_avg": 0.491804838180542,
      "objective/train/docs_used": 263840,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5849299430847168,
      "objective/train/original_loss": 1.5849299430847168,
      "objective/train/theoretical_loss": 3.801915200335954,
      "objective/train/tokens_used": 731525600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24257409572601318,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050413727760315,
      "objective/train/weighted_lm_loss": 1.6650092601776123,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.801915200335954,
      "tokens_seen": 711065600
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007924774322968908,
      "loss": 1.6198,
      "theoretical_loss": 3.801915200335954,
      "tokens_seen": 711065600
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000792457372116349,
      "loss": 1.4927,
      "theoretical_loss": 3.8018799881754806,
      "tokens_seen": 711131136
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007924373119358075,
      "loss": 1.6069,
      "theoretical_loss": 3.801844780168441,
      "tokens_seen": 711196672
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007924172517552658,
      "loss": 1.6746,
      "theoretical_loss": 3.801809576313962,
      "tokens_seen": 711262208
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007923971915747242,
      "loss": 1.5904,
      "theoretical_loss": 3.8017743766111716,
      "tokens_seen": 711327744
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007923771313941826,
      "loss": 1.6472,
      "theoretical_loss": 3.8017391810591974,
      "tokens_seen": 711393280
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007923570712136409,
      "loss": 1.6637,
      "theoretical_loss": 3.8017039896571685,
      "tokens_seen": 711458816
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007923370110330993,
      "loss": 1.6714,
      "theoretical_loss": 3.8016688024042127,
      "tokens_seen": 711524352
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007923169508525578,
      "loss": 1.6044,
      "theoretical_loss": 3.8016336192994586,
      "tokens_seen": 711589888
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000792296890672016,
      "loss": 1.4999,
      "theoretical_loss": 3.801598440342036,
      "tokens_seen": 711655424
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007922768304914745,
      "loss": 1.5379,
      "theoretical_loss": 3.8015632655310734,
      "tokens_seen": 711720960
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007922567703109327,
      "loss": 1.5861,
      "theoretical_loss": 3.8015280948657004,
      "tokens_seen": 711786496
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007922367101303912,
      "loss": 1.5943,
      "theoretical_loss": 3.801492928345047,
      "tokens_seen": 711852032
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007922166499498496,
      "loss": 1.5746,
      "theoretical_loss": 3.8014577659682436,
      "tokens_seen": 711917568
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007921965897693079,
      "loss": 1.7062,
      "theoretical_loss": 3.8014226077344198,
      "tokens_seen": 711983104
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007921765295887663,
      "loss": 1.5825,
      "theoretical_loss": 3.8013874536427066,
      "tokens_seen": 712048640
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007921564694082248,
      "loss": 1.5413,
      "theoretical_loss": 3.8013523036922345,
      "tokens_seen": 712114176
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000792136409227683,
      "loss": 1.5859,
      "theoretical_loss": 3.801317157882135,
      "tokens_seen": 712179712
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007921163490471415,
      "loss": 1.5603,
      "theoretical_loss": 3.8012820162115393,
      "tokens_seen": 712245248
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007920962888665997,
      "loss": 1.6049,
      "theoretical_loss": 3.8012468786795788,
      "tokens_seen": 712310784
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007920762286860582,
      "loss": 1.485,
      "theoretical_loss": 3.8012117452853857,
      "tokens_seen": 712376320
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007920561685055166,
      "loss": 1.548,
      "theoretical_loss": 3.8011766160280924,
      "tokens_seen": 712441856
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007920361083249749,
      "loss": 1.5592,
      "theoretical_loss": 3.801141490906831,
      "tokens_seen": 712507392
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007920160481444333,
      "loss": 1.5791,
      "theoretical_loss": 3.801106369920734,
      "tokens_seen": 712572928
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007919959879638917,
      "loss": 1.5164,
      "theoretical_loss": 3.801071253068934,
      "tokens_seen": 712638464
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00079197592778335,
      "loss": 1.6865,
      "theoretical_loss": 3.801036140350565,
      "tokens_seen": 712704000
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007919558676028085,
      "loss": 1.4819,
      "theoretical_loss": 3.80100103176476,
      "tokens_seen": 712769536
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007919358074222668,
      "loss": 1.6839,
      "theoretical_loss": 3.800965927310653,
      "tokens_seen": 712835072
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007919157472417252,
      "loss": 1.5149,
      "theoretical_loss": 3.8009308269873774,
      "tokens_seen": 712900608
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007918956870611836,
      "loss": 1.6399,
      "theoretical_loss": 3.8008957307940685,
      "tokens_seen": 712966144
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000791875626880642,
      "loss": 1.7079,
      "theoretical_loss": 3.8008606387298594,
      "tokens_seen": 713031680
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007918555667001003,
      "loss": 1.5256,
      "theoretical_loss": 3.800825550793886,
      "tokens_seen": 713097216
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007918355065195587,
      "loss": 1.4973,
      "theoretical_loss": 3.8007904669852834,
      "tokens_seen": 713162752
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000791815446339017,
      "loss": 1.5021,
      "theoretical_loss": 3.8007553873031856,
      "tokens_seen": 713228288
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007917953861584755,
      "loss": 1.629,
      "theoretical_loss": 3.80072031174673,
      "tokens_seen": 713293824
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007917753259779338,
      "loss": 1.6775,
      "theoretical_loss": 3.80068524031505,
      "tokens_seen": 713359360
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007917552657973922,
      "loss": 1.5771,
      "theoretical_loss": 3.8006501730072837,
      "tokens_seen": 713424896
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007917352056168505,
      "loss": 1.6107,
      "theoretical_loss": 3.8006151098225667,
      "tokens_seen": 713490432
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000791715145436309,
      "loss": 1.6098,
      "theoretical_loss": 3.800580050760036,
      "tokens_seen": 713555968
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007916950852557673,
      "loss": 1.4269,
      "theoretical_loss": 3.8005449958188278,
      "tokens_seen": 713621504
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007916750250752257,
      "loss": 1.5848,
      "theoretical_loss": 3.8005099449980793,
      "tokens_seen": 713687040
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000791654964894684,
      "loss": 1.6125,
      "theoretical_loss": 3.8004748982969283,
      "tokens_seen": 713752576
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007916349047141424,
      "loss": 1.5103,
      "theoretical_loss": 3.800439855714512,
      "tokens_seen": 713818112
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007916148445336008,
      "loss": 1.6249,
      "theoretical_loss": 3.8004048172499685,
      "tokens_seen": 713883648
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007915947843530592,
      "loss": 1.6413,
      "theoretical_loss": 3.800369782902436,
      "tokens_seen": 713949184
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007915747241725175,
      "loss": 1.564,
      "theoretical_loss": 3.800334752671053,
      "tokens_seen": 714014720
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000791554663991976,
      "loss": 1.6036,
      "theoretical_loss": 3.8002997265549574,
      "tokens_seen": 714080256
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007915346038114343,
      "loss": 1.6338,
      "theoretical_loss": 3.800264704553289,
      "tokens_seen": 714145792
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007915145436308927,
      "loss": 1.5818,
      "theoretical_loss": 3.800229686665187,
      "tokens_seen": 714211328
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000791494483450351,
      "loss": 1.6419,
      "theoretical_loss": 3.80019467288979,
      "tokens_seen": 714276864
    },
    {
      "epoch": 0.22,
      "objective/train/advantage_avg": 0.49227601289749146,
      "objective/train/docs_used": 265036,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4128671884536743,
      "objective/train/original_loss": 1.4128673076629639,
      "objective/train/theoretical_loss": 3.8001596632262387,
      "objective/train/tokens_used": 734802400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24258217215538025,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050460696220398,
      "objective/train/weighted_lm_loss": 1.4842172861099243,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.8001596632262387,
      "tokens_seen": 714342400
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007914744232698094,
      "loss": 1.6004,
      "theoretical_loss": 3.8001596632262387,
      "tokens_seen": 714342400
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007914543630892678,
      "loss": 1.6077,
      "theoretical_loss": 3.8001246576736722,
      "tokens_seen": 714407936
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007914343029087262,
      "loss": 1.508,
      "theoretical_loss": 3.800089656231231,
      "tokens_seen": 714473472
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007914142427281845,
      "loss": 1.6235,
      "theoretical_loss": 3.800054658898056,
      "tokens_seen": 714539008
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000791394182547643,
      "loss": 1.5177,
      "theoretical_loss": 3.8000196656732874,
      "tokens_seen": 714604544
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007913741223671012,
      "loss": 1.5371,
      "theoretical_loss": 3.7999846765560665,
      "tokens_seen": 714670080
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007913540621865597,
      "loss": 1.5648,
      "theoretical_loss": 3.7999496915455344,
      "tokens_seen": 714735616
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007913340020060181,
      "loss": 1.5554,
      "theoretical_loss": 3.7999147106408326,
      "tokens_seen": 714801152
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007913139418254764,
      "loss": 1.6221,
      "theoretical_loss": 3.7998797338411032,
      "tokens_seen": 714866688
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007912938816449348,
      "loss": 1.5618,
      "theoretical_loss": 3.7998447611454873,
      "tokens_seen": 714932224
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007912738214643931,
      "loss": 1.4989,
      "theoretical_loss": 3.7998097925531282,
      "tokens_seen": 714997760
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007912537612838516,
      "loss": 1.762,
      "theoretical_loss": 3.7997748280631685,
      "tokens_seen": 715063296
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00079123370110331,
      "loss": 1.5914,
      "theoretical_loss": 3.7997398676747496,
      "tokens_seen": 715128832
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007912136409227683,
      "loss": 1.6376,
      "theoretical_loss": 3.7997049113870163,
      "tokens_seen": 715194368
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007911935807422267,
      "loss": 1.5687,
      "theoretical_loss": 3.799669959199111,
      "tokens_seen": 715259904
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007911735205616852,
      "loss": 1.5701,
      "theoretical_loss": 3.799635011110177,
      "tokens_seen": 715325440
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007911534603811434,
      "loss": 1.6408,
      "theoretical_loss": 3.7996000671193593,
      "tokens_seen": 715390976
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007911334002006019,
      "loss": 1.6203,
      "theoretical_loss": 3.799565127225801,
      "tokens_seen": 715456512
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007911133400200601,
      "loss": 1.6039,
      "theoretical_loss": 3.799530191428646,
      "tokens_seen": 715522048
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007910932798395186,
      "loss": 1.5181,
      "theoretical_loss": 3.79949525972704,
      "tokens_seen": 715587584
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000791073219658977,
      "loss": 1.5076,
      "theoretical_loss": 3.7994603321201277,
      "tokens_seen": 715653120
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007910531594784353,
      "loss": 1.6501,
      "theoretical_loss": 3.7994254086070534,
      "tokens_seen": 715718656
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007910330992978937,
      "loss": 1.5293,
      "theoretical_loss": 3.799390489186963,
      "tokens_seen": 715784192
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007910130391173521,
      "loss": 1.5227,
      "theoretical_loss": 3.799355573859003,
      "tokens_seen": 715849728
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007909929789368104,
      "loss": 1.6207,
      "theoretical_loss": 3.7993206626223177,
      "tokens_seen": 715915264
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007909729187562689,
      "loss": 1.6393,
      "theoretical_loss": 3.7992857554760544,
      "tokens_seen": 715980800
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007909528585757272,
      "loss": 1.5905,
      "theoretical_loss": 3.799250852419359,
      "tokens_seen": 716046336
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007909327983951856,
      "loss": 1.5667,
      "theoretical_loss": 3.7992159534513776,
      "tokens_seen": 716111872
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000790912738214644,
      "loss": 1.5622,
      "theoretical_loss": 3.799181058571258,
      "tokens_seen": 716177408
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007908926780341024,
      "loss": 1.553,
      "theoretical_loss": 3.7991461677781473,
      "tokens_seen": 716242944
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007908726178535607,
      "loss": 1.5286,
      "theoretical_loss": 3.7991112810711924,
      "tokens_seen": 716308480
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007908525576730191,
      "loss": 1.4533,
      "theoretical_loss": 3.7990763984495413,
      "tokens_seen": 716374016
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007908324974924774,
      "loss": 1.5553,
      "theoretical_loss": 3.7990415199123424,
      "tokens_seen": 716439552
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007908124373119359,
      "loss": 1.6166,
      "theoretical_loss": 3.799006645458743,
      "tokens_seen": 716505088
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007907923771313942,
      "loss": 1.546,
      "theoretical_loss": 3.798971775087892,
      "tokens_seen": 716570624
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007907723169508526,
      "loss": 1.5911,
      "theoretical_loss": 3.798936908798938,
      "tokens_seen": 716636160
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007907522567703109,
      "loss": 1.7353,
      "theoretical_loss": 3.79890204659103,
      "tokens_seen": 716701696
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007907321965897694,
      "loss": 1.5854,
      "theoretical_loss": 3.7988671884633174,
      "tokens_seen": 716767232
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007907121364092277,
      "loss": 1.5094,
      "theoretical_loss": 3.79883233441495,
      "tokens_seen": 716832768
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007906920762286861,
      "loss": 1.5658,
      "theoretical_loss": 3.7987974844450756,
      "tokens_seen": 716898304
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007906720160481444,
      "loss": 1.6501,
      "theoretical_loss": 3.7987626385528466,
      "tokens_seen": 716963840
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007906519558676028,
      "loss": 1.6031,
      "theoretical_loss": 3.7987277967374125,
      "tokens_seen": 717029376
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007906318956870612,
      "loss": 1.6597,
      "theoretical_loss": 3.7986929589979224,
      "tokens_seen": 717094912
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007906118355065196,
      "loss": 1.6718,
      "theoretical_loss": 3.798658125333529,
      "tokens_seen": 717160448
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007905917753259779,
      "loss": 1.6507,
      "theoretical_loss": 3.798623295743382,
      "tokens_seen": 717225984
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007905717151454364,
      "loss": 1.6237,
      "theoretical_loss": 3.7985884702266333,
      "tokens_seen": 717291520
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007905516549648947,
      "loss": 1.6251,
      "theoretical_loss": 3.7985536487824336,
      "tokens_seen": 717357056
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007905315947843531,
      "loss": 1.5968,
      "theoretical_loss": 3.798518831409936,
      "tokens_seen": 717422592
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007905115346038114,
      "loss": 1.7573,
      "theoretical_loss": 3.798484018108291,
      "tokens_seen": 717488128
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007904914744232698,
      "loss": 1.5381,
      "theoretical_loss": 3.7984492088766526,
      "tokens_seen": 717553664
    },
    {
      "epoch": 0.22,
      "objective/train/advantage_avg": 0.4918719232082367,
      "objective/train/docs_used": 266305,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5350476503372192,
      "objective/train/original_loss": 1.5350477695465088,
      "objective/train/theoretical_loss": 3.7984144037141716,
      "objective/train/tokens_used": 738079200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24277284741401672,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504213571548462,
      "objective/train/weighted_lm_loss": 1.612541675567627,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.7984144037141716,
      "tokens_seen": 717619200
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007904714142427282,
      "loss": 1.5408,
      "theoretical_loss": 3.7984144037141716,
      "tokens_seen": 717619200
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007904513540621866,
      "loss": 1.6247,
      "theoretical_loss": 3.7983796026200016,
      "tokens_seen": 717684736
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007904312938816449,
      "loss": 1.6049,
      "theoretical_loss": 3.7983448055932953,
      "tokens_seen": 717750272
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007904112337011034,
      "loss": 1.7175,
      "theoretical_loss": 3.7983100126332063,
      "tokens_seen": 717815808
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007903911735205616,
      "loss": 1.7088,
      "theoretical_loss": 3.798275223738888,
      "tokens_seen": 717881344
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007903711133400201,
      "loss": 1.5637,
      "theoretical_loss": 3.798240438909495,
      "tokens_seen": 717946880
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007903510531594785,
      "loss": 1.6292,
      "theoretical_loss": 3.79820565814418,
      "tokens_seen": 718012416
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007903309929789368,
      "loss": 1.5838,
      "theoretical_loss": 3.798170881442098,
      "tokens_seen": 718077952
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007903109327983952,
      "loss": 1.558,
      "theoretical_loss": 3.798136108802403,
      "tokens_seen": 718143488
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007902908726178535,
      "loss": 1.5955,
      "theoretical_loss": 3.79810134022425,
      "tokens_seen": 718209024
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007902708124373119,
      "loss": 1.5749,
      "theoretical_loss": 3.798066575706795,
      "tokens_seen": 718274560
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007902507522567704,
      "loss": 1.6152,
      "theoretical_loss": 3.798031815249192,
      "tokens_seen": 718340096
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007902306920762286,
      "loss": 1.5295,
      "theoretical_loss": 3.797997058850598,
      "tokens_seen": 718405632
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007902106318956871,
      "loss": 1.6052,
      "theoretical_loss": 3.7979623065101666,
      "tokens_seen": 718471168
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007901905717151455,
      "loss": 1.5276,
      "theoretical_loss": 3.797927558227056,
      "tokens_seen": 718536704
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007901705115346038,
      "loss": 1.468,
      "theoretical_loss": 3.797892814000422,
      "tokens_seen": 718602240
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007901504513540622,
      "loss": 1.5467,
      "theoretical_loss": 3.7978580738294205,
      "tokens_seen": 718667776
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007901303911735205,
      "loss": 1.5773,
      "theoretical_loss": 3.7978233377132087,
      "tokens_seen": 718733312
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007901103309929789,
      "loss": 1.5912,
      "theoretical_loss": 3.7977886056509433,
      "tokens_seen": 718798848
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007900902708124374,
      "loss": 1.5103,
      "theoretical_loss": 3.7977538776417825,
      "tokens_seen": 718864384
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007900702106318956,
      "loss": 1.6416,
      "theoretical_loss": 3.7977191536848833,
      "tokens_seen": 718929920
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007900501504513541,
      "loss": 1.5653,
      "theoretical_loss": 3.797684433779404,
      "tokens_seen": 718995456
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007900300902708124,
      "loss": 1.5652,
      "theoretical_loss": 3.797649717924502,
      "tokens_seen": 719060992
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007900100300902708,
      "loss": 1.5965,
      "theoretical_loss": 3.7976150061193352,
      "tokens_seen": 719126528
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007899899699097292,
      "loss": 1.5144,
      "theoretical_loss": 3.7975802983630635,
      "tokens_seen": 719192064
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007899699097291876,
      "loss": 1.6968,
      "theoretical_loss": 3.797545594654845,
      "tokens_seen": 719257600
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007899498495486459,
      "loss": 1.5588,
      "theoretical_loss": 3.797510894993839,
      "tokens_seen": 719323136
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007899297893681044,
      "loss": 1.5348,
      "theoretical_loss": 3.7974761993792048,
      "tokens_seen": 719388672
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007899097291875626,
      "loss": 1.5873,
      "theoretical_loss": 3.797441507810102,
      "tokens_seen": 719454208
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007898896690070211,
      "loss": 1.6881,
      "theoretical_loss": 3.7974068202856905,
      "tokens_seen": 719519744
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007898696088264794,
      "loss": 1.4768,
      "theoretical_loss": 3.79737213680513,
      "tokens_seen": 719585280
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007898495486459378,
      "loss": 1.5679,
      "theoretical_loss": 3.797337457367581,
      "tokens_seen": 719650816
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007898294884653962,
      "loss": 1.6461,
      "theoretical_loss": 3.7973027819722045,
      "tokens_seen": 719716352
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007898094282848546,
      "loss": 1.587,
      "theoretical_loss": 3.797268110618161,
      "tokens_seen": 719781888
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000789789368104313,
      "loss": 1.5568,
      "theoretical_loss": 3.797233443304612,
      "tokens_seen": 719847424
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007897693079237713,
      "loss": 1.6034,
      "theoretical_loss": 3.797198780030718,
      "tokens_seen": 719912960
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007897492477432298,
      "loss": 1.5823,
      "theoretical_loss": 3.7971641207956415,
      "tokens_seen": 719978496
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007897291875626881,
      "loss": 1.6019,
      "theoretical_loss": 3.7971294655985437,
      "tokens_seen": 720044032
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007897091273821465,
      "loss": 1.6681,
      "theoretical_loss": 3.7970948144385868,
      "tokens_seen": 720109568
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007896890672016048,
      "loss": 1.635,
      "theoretical_loss": 3.7970601673149336,
      "tokens_seen": 720175104
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007896690070210632,
      "loss": 1.5091,
      "theoretical_loss": 3.797025524226746,
      "tokens_seen": 720240640
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007896489468405216,
      "loss": 1.5042,
      "theoretical_loss": 3.7969908851731873,
      "tokens_seen": 720306176
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00078962888665998,
      "loss": 1.505,
      "theoretical_loss": 3.796956250153421,
      "tokens_seen": 720371712
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007896088264794383,
      "loss": 1.579,
      "theoretical_loss": 3.7969216191666098,
      "tokens_seen": 720437248
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007895887662988968,
      "loss": 1.6215,
      "theoretical_loss": 3.796886992211917,
      "tokens_seen": 720502784
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007895687061183551,
      "loss": 1.5679,
      "theoretical_loss": 3.796852369288507,
      "tokens_seen": 720568320
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007895486459378135,
      "loss": 1.6324,
      "theoretical_loss": 3.796817750395544,
      "tokens_seen": 720633856
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007895285857572719,
      "loss": 1.557,
      "theoretical_loss": 3.796783135532192,
      "tokens_seen": 720699392
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007895085255767302,
      "loss": 1.6823,
      "theoretical_loss": 3.796748524697616,
      "tokens_seen": 720764928
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007894884653961886,
      "loss": 1.455,
      "theoretical_loss": 3.7967139178909806,
      "tokens_seen": 720830464
    },
    {
      "epoch": 0.22,
      "objective/train/advantage_avg": 0.494853138923645,
      "objective/train/docs_used": 267624,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4048925638198853,
      "objective/train/original_loss": 1.4048924446105957,
      "objective/train/theoretical_loss": 3.7966793151114504,
      "objective/train/tokens_used": 741356000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24536572396755219,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507327318191528,
      "objective/train/weighted_lm_loss": 1.4761347770690918,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.7966793151114504,
      "tokens_seen": 720896000
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000789468405215647,
      "loss": 1.4701,
      "theoretical_loss": 3.7966793151114504,
      "tokens_seen": 720896000
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007894483450351053,
      "loss": 1.5125,
      "theoretical_loss": 3.7966447163581916,
      "tokens_seen": 720961536
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007894282848545638,
      "loss": 1.5695,
      "theoretical_loss": 3.796610121630369,
      "tokens_seen": 721027072
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000789408224674022,
      "loss": 1.4918,
      "theoretical_loss": 3.796575530927149,
      "tokens_seen": 721092608
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007893881644934805,
      "loss": 1.5916,
      "theoretical_loss": 3.796540944247698,
      "tokens_seen": 721158144
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007893681043129389,
      "loss": 1.5412,
      "theoretical_loss": 3.7965063615911814,
      "tokens_seen": 721223680
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007893480441323972,
      "loss": 1.5931,
      "theoretical_loss": 3.7964717829567665,
      "tokens_seen": 721289216
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007893279839518556,
      "loss": 1.5241,
      "theoretical_loss": 3.7964372083436198,
      "tokens_seen": 721354752
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007893079237713139,
      "loss": 1.6341,
      "theoretical_loss": 3.796402637750908,
      "tokens_seen": 721420288
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007892878635907723,
      "loss": 1.5543,
      "theoretical_loss": 3.796368071177799,
      "tokens_seen": 721485824
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007892678034102308,
      "loss": 1.593,
      "theoretical_loss": 3.7963335086234604,
      "tokens_seen": 721551360
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000789247743229689,
      "loss": 1.6921,
      "theoretical_loss": 3.79629895008706,
      "tokens_seen": 721616896
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007892276830491475,
      "loss": 1.6719,
      "theoretical_loss": 3.796264395567766,
      "tokens_seen": 721682432
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007892076228686059,
      "loss": 1.6716,
      "theoretical_loss": 3.7962298450647456,
      "tokens_seen": 721747968
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007891875626880642,
      "loss": 1.5506,
      "theoretical_loss": 3.796195298577169,
      "tokens_seen": 721813504
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007891675025075226,
      "loss": 1.5707,
      "theoretical_loss": 3.7961607561042037,
      "tokens_seen": 721879040
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000789147442326981,
      "loss": 1.5197,
      "theoretical_loss": 3.7961262176450195,
      "tokens_seen": 721944576
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007891273821464393,
      "loss": 1.5943,
      "theoretical_loss": 3.796091683198785,
      "tokens_seen": 722010112
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007891073219658978,
      "loss": 1.6489,
      "theoretical_loss": 3.7960571527646714,
      "tokens_seen": 722075648
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000789087261785356,
      "loss": 1.5536,
      "theoretical_loss": 3.7960226263418466,
      "tokens_seen": 722141184
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007890672016048145,
      "loss": 1.6451,
      "theoretical_loss": 3.795988103929482,
      "tokens_seen": 722206720
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007890471414242728,
      "loss": 1.5525,
      "theoretical_loss": 3.7959535855267466,
      "tokens_seen": 722272256
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007890270812437312,
      "loss": 1.6085,
      "theoretical_loss": 3.795919071132812,
      "tokens_seen": 722337792
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007890070210631896,
      "loss": 1.4768,
      "theoretical_loss": 3.795884560746849,
      "tokens_seen": 722403328
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000788986960882648,
      "loss": 1.4919,
      "theoretical_loss": 3.7958500543680276,
      "tokens_seen": 722468864
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007889669007021063,
      "loss": 1.5701,
      "theoretical_loss": 3.7958155519955206,
      "tokens_seen": 722534400
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007889468405215648,
      "loss": 1.6325,
      "theoretical_loss": 3.795781053628498,
      "tokens_seen": 722599936
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000788926780341023,
      "loss": 1.5852,
      "theoretical_loss": 3.7957465592661324,
      "tokens_seen": 722665472
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007889067201604815,
      "loss": 1.5534,
      "theoretical_loss": 3.795712068907596,
      "tokens_seen": 722731008
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007888866599799398,
      "loss": 1.7387,
      "theoretical_loss": 3.7956775825520603,
      "tokens_seen": 722796544
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007888665997993982,
      "loss": 1.6786,
      "theoretical_loss": 3.7956431001986988,
      "tokens_seen": 722862080
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007888465396188566,
      "loss": 1.6192,
      "theoretical_loss": 3.7956086218466836,
      "tokens_seen": 722927616
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000788826479438315,
      "loss": 1.5977,
      "theoretical_loss": 3.795574147495188,
      "tokens_seen": 722993152
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007888064192577733,
      "loss": 1.5721,
      "theoretical_loss": 3.7955396771433847,
      "tokens_seen": 723058688
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007887863590772317,
      "loss": 1.615,
      "theoretical_loss": 3.7955052107904477,
      "tokens_seen": 723124224
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00078876629889669,
      "loss": 1.6817,
      "theoretical_loss": 3.7954707484355508,
      "tokens_seen": 723189760
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007887462387161485,
      "loss": 1.6336,
      "theoretical_loss": 3.795436290077868,
      "tokens_seen": 723255296
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007887261785356068,
      "loss": 1.6542,
      "theoretical_loss": 3.7954018357165737,
      "tokens_seen": 723320832
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007887061183550652,
      "loss": 1.6355,
      "theoretical_loss": 3.7953673853508416,
      "tokens_seen": 723386368
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007886860581745235,
      "loss": 1.6461,
      "theoretical_loss": 3.795332938979847,
      "tokens_seen": 723451904
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000788665997993982,
      "loss": 1.5919,
      "theoretical_loss": 3.795298496602765,
      "tokens_seen": 723517440
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007886459378134403,
      "loss": 1.56,
      "theoretical_loss": 3.7952640582187707,
      "tokens_seen": 723582976
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007886258776328987,
      "loss": 1.462,
      "theoretical_loss": 3.795229623827039,
      "tokens_seen": 723648512
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007886058174523571,
      "loss": 1.6575,
      "theoretical_loss": 3.7951951934267463,
      "tokens_seen": 723714048
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007885857572718155,
      "loss": 1.564,
      "theoretical_loss": 3.795160767017068,
      "tokens_seen": 723779584
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007885656970912738,
      "loss": 1.545,
      "theoretical_loss": 3.795126344597181,
      "tokens_seen": 723845120
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007885456369107323,
      "loss": 1.5575,
      "theoretical_loss": 3.7950919261662612,
      "tokens_seen": 723910656
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007885255767301905,
      "loss": 1.6048,
      "theoretical_loss": 3.7950575117234857,
      "tokens_seen": 723976192
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000788505516549649,
      "loss": 1.6036,
      "theoretical_loss": 3.795023101268031,
      "tokens_seen": 724041728
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007884854563691073,
      "loss": 1.5814,
      "theoretical_loss": 3.7949886947990743,
      "tokens_seen": 724107264
    },
    {
      "epoch": 0.22,
      "objective/train/advantage_avg": 0.4941313862800598,
      "objective/train/docs_used": 268786,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5286294221878052,
      "objective/train/original_loss": 1.5286293029785156,
      "objective/train/theoretical_loss": 3.7949542923157935,
      "objective/train/tokens_used": 744632800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24435481429100037,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050655484199524,
      "objective/train/weighted_lm_loss": 1.6058937311172485,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.7949542923157935,
      "tokens_seen": 724172800
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007884653961885657,
      "loss": 1.6391,
      "theoretical_loss": 3.7949542923157935,
      "tokens_seen": 724172800
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007884453360080241,
      "loss": 1.6482,
      "theoretical_loss": 3.794919893817366,
      "tokens_seen": 724238336
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007884252758274824,
      "loss": 1.8053,
      "theoretical_loss": 3.7948854993029695,
      "tokens_seen": 724303872
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007884052156469408,
      "loss": 1.5236,
      "theoretical_loss": 3.7948511087717822,
      "tokens_seen": 724369408
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007883851554663993,
      "loss": 1.5501,
      "theoretical_loss": 3.7948167222229827,
      "tokens_seen": 724434944
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007883650952858575,
      "loss": 1.6084,
      "theoretical_loss": 3.7947823396557494,
      "tokens_seen": 724500480
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000788345035105316,
      "loss": 1.5137,
      "theoretical_loss": 3.7947479610692616,
      "tokens_seen": 724566016
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007883249749247742,
      "loss": 1.5571,
      "theoretical_loss": 3.7947135864626977,
      "tokens_seen": 724631552
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007883049147442327,
      "loss": 1.6306,
      "theoretical_loss": 3.794679215835238,
      "tokens_seen": 724697088
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007882848545636912,
      "loss": 1.49,
      "theoretical_loss": 3.7946448491860614,
      "tokens_seen": 724762624
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007882647943831494,
      "loss": 1.6241,
      "theoretical_loss": 3.794610486514348,
      "tokens_seen": 724828160
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007882447342026079,
      "loss": 1.5668,
      "theoretical_loss": 3.794576127819278,
      "tokens_seen": 724893696
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007882246740220663,
      "loss": 1.5605,
      "theoretical_loss": 3.7945417731000317,
      "tokens_seen": 724959232
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007882046138415246,
      "loss": 1.6471,
      "theoretical_loss": 3.7945074223557893,
      "tokens_seen": 725024768
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000788184553660983,
      "loss": 1.5899,
      "theoretical_loss": 3.7944730755857323,
      "tokens_seen": 725090304
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007881644934804413,
      "loss": 1.6219,
      "theoretical_loss": 3.794438732789041,
      "tokens_seen": 725155840
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007881444332998997,
      "loss": 1.5868,
      "theoretical_loss": 3.7944043939648973,
      "tokens_seen": 725221376
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007881243731193582,
      "loss": 1.5024,
      "theoretical_loss": 3.7943700591124827,
      "tokens_seen": 725286912
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007881043129388164,
      "loss": 1.5052,
      "theoretical_loss": 3.794335728230979,
      "tokens_seen": 725352448
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007880842527582749,
      "loss": 1.6316,
      "theoretical_loss": 3.7943014013195677,
      "tokens_seen": 725417984
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007880641925777332,
      "loss": 1.5751,
      "theoretical_loss": 3.794267078377432,
      "tokens_seen": 725483520
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007880441323971916,
      "loss": 1.5446,
      "theoretical_loss": 3.7942327594037537,
      "tokens_seen": 725549056
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00078802407221665,
      "loss": 1.5184,
      "theoretical_loss": 3.7941984443977157,
      "tokens_seen": 725614592
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007880040120361084,
      "loss": 1.4761,
      "theoretical_loss": 3.7941641333585014,
      "tokens_seen": 725680128
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007879839518555667,
      "loss": 1.5728,
      "theoretical_loss": 3.794129826285294,
      "tokens_seen": 725745664
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007879638916750252,
      "loss": 1.4827,
      "theoretical_loss": 3.794095523177276,
      "tokens_seen": 725811200
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007879438314944834,
      "loss": 1.5944,
      "theoretical_loss": 3.7940612240336327,
      "tokens_seen": 725876736
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007879237713139419,
      "loss": 1.5847,
      "theoretical_loss": 3.794026928853547,
      "tokens_seen": 725942272
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007879037111334002,
      "loss": 1.579,
      "theoretical_loss": 3.7939926376362036,
      "tokens_seen": 726007808
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007878836509528586,
      "loss": 1.5409,
      "theoretical_loss": 3.7939583503807865,
      "tokens_seen": 726073344
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000787863590772317,
      "loss": 1.632,
      "theoretical_loss": 3.793924067086481,
      "tokens_seen": 726138880
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007878435305917754,
      "loss": 1.5681,
      "theoretical_loss": 3.7938897877524713,
      "tokens_seen": 726204416
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007878234704112337,
      "loss": 1.6034,
      "theoretical_loss": 3.793855512377943,
      "tokens_seen": 726269952
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007878034102306921,
      "loss": 1.6,
      "theoretical_loss": 3.7938212409620817,
      "tokens_seen": 726335488
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007877833500501504,
      "loss": 1.5289,
      "theoretical_loss": 3.793786973504073,
      "tokens_seen": 726401024
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007877632898696089,
      "loss": 1.6196,
      "theoretical_loss": 3.7937527100031025,
      "tokens_seen": 726466560
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007877432296890672,
      "loss": 1.6569,
      "theoretical_loss": 3.793718450458357,
      "tokens_seen": 726532096
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007877231695085256,
      "loss": 1.5011,
      "theoretical_loss": 3.793684194869022,
      "tokens_seen": 726597632
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007877031093279839,
      "loss": 1.4242,
      "theoretical_loss": 3.7936499432342847,
      "tokens_seen": 726663168
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007876830491474424,
      "loss": 1.6393,
      "theoretical_loss": 3.7936156955533322,
      "tokens_seen": 726728704
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007876629889669007,
      "loss": 1.5394,
      "theoretical_loss": 3.793581451825351,
      "tokens_seen": 726794240
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007876429287863591,
      "loss": 1.5205,
      "theoretical_loss": 3.7935472120495284,
      "tokens_seen": 726859776
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007876228686058175,
      "loss": 1.5882,
      "theoretical_loss": 3.7935129762250526,
      "tokens_seen": 726925312
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007876028084252759,
      "loss": 1.5245,
      "theoretical_loss": 3.7934787443511113,
      "tokens_seen": 726990848
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007875827482447342,
      "loss": 1.6943,
      "theoretical_loss": 3.7934445164268924,
      "tokens_seen": 727056384
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007875626880641927,
      "loss": 1.7291,
      "theoretical_loss": 3.7934102924515845,
      "tokens_seen": 727121920
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007875426278836509,
      "loss": 1.5935,
      "theoretical_loss": 3.7933760724243752,
      "tokens_seen": 727187456
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007875225677031094,
      "loss": 1.6417,
      "theoretical_loss": 3.7933418563444543,
      "tokens_seen": 727252992
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007875025075225677,
      "loss": 1.5997,
      "theoretical_loss": 3.7933076442110103,
      "tokens_seen": 727318528
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007874824473420261,
      "loss": 1.5442,
      "theoretical_loss": 3.793273436023233,
      "tokens_seen": 727384064
    },
    {
      "epoch": 0.22,
      "objective/train/advantage_avg": 0.4955691993236542,
      "objective/train/docs_used": 269904,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5276315212249756,
      "objective/train/original_loss": 1.5276315212249756,
      "objective/train/theoretical_loss": 3.7932392317803116,
      "objective/train/tokens_used": 747909600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24581311643123627,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508066415786743,
      "objective/train/weighted_lm_loss": 1.605190396308899,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7932392317803116,
      "tokens_seen": 727449600
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007874623871614845,
      "loss": 1.6204,
      "theoretical_loss": 3.7932392317803116,
      "tokens_seen": 727449600
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007874423269809428,
      "loss": 1.4769,
      "theoretical_loss": 3.793205031481436,
      "tokens_seen": 727515136
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007874222668004012,
      "loss": 1.6299,
      "theoretical_loss": 3.793170835125796,
      "tokens_seen": 727580672
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007874022066198597,
      "loss": 1.5609,
      "theoretical_loss": 3.7931366427125814,
      "tokens_seen": 727646208
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007873821464393179,
      "loss": 1.5415,
      "theoretical_loss": 3.7931024542409837,
      "tokens_seen": 727711744
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007873620862587764,
      "loss": 1.6417,
      "theoretical_loss": 3.7930682697101927,
      "tokens_seen": 727777280
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007873420260782346,
      "loss": 1.5456,
      "theoretical_loss": 3.7930340891194,
      "tokens_seen": 727842816
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007873219658976931,
      "loss": 1.6418,
      "theoretical_loss": 3.7929999124677964,
      "tokens_seen": 727908352
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007873019057171515,
      "loss": 1.4928,
      "theoretical_loss": 3.7929657397545733,
      "tokens_seen": 727973888
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007872818455366098,
      "loss": 1.5043,
      "theoretical_loss": 3.792931570978922,
      "tokens_seen": 728039424
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007872617853560682,
      "loss": 1.521,
      "theoretical_loss": 3.792897406140036,
      "tokens_seen": 728104960
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007872417251755267,
      "loss": 1.5265,
      "theoretical_loss": 3.792863245237106,
      "tokens_seen": 728170496
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007872216649949849,
      "loss": 1.6199,
      "theoretical_loss": 3.792829088269324,
      "tokens_seen": 728236032
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007872016048144434,
      "loss": 1.5429,
      "theoretical_loss": 3.792794935235884,
      "tokens_seen": 728301568
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007871815446339016,
      "loss": 1.5364,
      "theoretical_loss": 3.792760786135978,
      "tokens_seen": 728367104
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007871614844533601,
      "loss": 1.517,
      "theoretical_loss": 3.792726640968799,
      "tokens_seen": 728432640
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007871414242728185,
      "loss": 1.6017,
      "theoretical_loss": 3.792692499733541,
      "tokens_seen": 728498176
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007871213640922768,
      "loss": 1.6461,
      "theoretical_loss": 3.792658362429397,
      "tokens_seen": 728563712
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007871013039117352,
      "loss": 1.5711,
      "theoretical_loss": 3.792624229055561,
      "tokens_seen": 728629248
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007870812437311936,
      "loss": 1.5317,
      "theoretical_loss": 3.7925900996112274,
      "tokens_seen": 728694784
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007870611835506519,
      "loss": 1.5604,
      "theoretical_loss": 3.7925559740955896,
      "tokens_seen": 728760320
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007870411233701104,
      "loss": 1.574,
      "theoretical_loss": 3.792521852507843,
      "tokens_seen": 728825856
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007870210631895686,
      "loss": 1.5346,
      "theoretical_loss": 3.7924877348471817,
      "tokens_seen": 728891392
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007870010030090271,
      "loss": 1.5784,
      "theoretical_loss": 3.7924536211128013,
      "tokens_seen": 728956928
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007869809428284854,
      "loss": 1.5469,
      "theoretical_loss": 3.7924195113038968,
      "tokens_seen": 729022464
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007869608826479438,
      "loss": 1.5797,
      "theoretical_loss": 3.7923854054196635,
      "tokens_seen": 729088000
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007869408224674022,
      "loss": 1.6433,
      "theoretical_loss": 3.792351303459297,
      "tokens_seen": 729153536
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007869207622868606,
      "loss": 1.5127,
      "theoretical_loss": 3.792317205421994,
      "tokens_seen": 729219072
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007869007021063189,
      "loss": 1.6056,
      "theoretical_loss": 3.7922831113069493,
      "tokens_seen": 729284608
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007868806419257774,
      "loss": 1.5127,
      "theoretical_loss": 3.792249021113361,
      "tokens_seen": 729350144
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007868605817452357,
      "loss": 1.5448,
      "theoretical_loss": 3.792214934840424,
      "tokens_seen": 729415680
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007868405215646941,
      "loss": 1.4949,
      "theoretical_loss": 3.792180852487337,
      "tokens_seen": 729481216
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007868204613841524,
      "loss": 1.5868,
      "theoretical_loss": 3.792146774053296,
      "tokens_seen": 729546752
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007868004012036108,
      "loss": 1.667,
      "theoretical_loss": 3.7921126995374985,
      "tokens_seen": 729612288
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007867803410230693,
      "loss": 1.4273,
      "theoretical_loss": 3.7920786289391417,
      "tokens_seen": 729677824
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007867602808425276,
      "loss": 1.5585,
      "theoretical_loss": 3.792044562257425,
      "tokens_seen": 729743360
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000786740220661986,
      "loss": 1.5144,
      "theoretical_loss": 3.792010499491545,
      "tokens_seen": 729808896
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007867201604814443,
      "loss": 1.4325,
      "theoretical_loss": 3.7919764406407,
      "tokens_seen": 729874432
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007867001003009028,
      "loss": 1.5243,
      "theoretical_loss": 3.7919423857040897,
      "tokens_seen": 729939968
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007866800401203611,
      "loss": 1.4948,
      "theoretical_loss": 3.7919083346809117,
      "tokens_seen": 730005504
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007866599799398195,
      "loss": 1.5905,
      "theoretical_loss": 3.7918742875703657,
      "tokens_seen": 730071040
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007866399197592779,
      "loss": 1.5229,
      "theoretical_loss": 3.791840244371651,
      "tokens_seen": 730136576
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007866198595787363,
      "loss": 1.5017,
      "theoretical_loss": 3.7918062050839665,
      "tokens_seen": 730202112
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007865997993981946,
      "loss": 1.6211,
      "theoretical_loss": 3.7917721697065123,
      "tokens_seen": 730267648
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000786579739217653,
      "loss": 1.5065,
      "theoretical_loss": 3.7917381382384883,
      "tokens_seen": 730333184
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007865596790371113,
      "loss": 1.6059,
      "theoretical_loss": 3.791704110679095,
      "tokens_seen": 730398720
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007865396188565698,
      "loss": 1.4476,
      "theoretical_loss": 3.791670087027532,
      "tokens_seen": 730464256
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007865195586760281,
      "loss": 1.4362,
      "theoretical_loss": 3.7916360672830014,
      "tokens_seen": 730529792
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007864994984954865,
      "loss": 1.6911,
      "theoretical_loss": 3.791602051444703,
      "tokens_seen": 730595328
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007864794383149449,
      "loss": 1.5424,
      "theoretical_loss": 3.7915680395118376,
      "tokens_seen": 730660864
    },
    {
      "epoch": 0.22,
      "objective/train/advantage_avg": 0.4948250651359558,
      "objective/train/docs_used": 271086,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.462997555732727,
      "objective/train/original_loss": 1.4629974365234375,
      "objective/train/theoretical_loss": 3.7915340314836077,
      "objective/train/tokens_used": 751186400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24516312777996063,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507289171218872,
      "objective/train/weighted_lm_loss": 1.5371520519256592,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.7915340314836077,
      "tokens_seen": 730726400
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007864593781344032,
      "loss": 1.5211,
      "theoretical_loss": 3.7915340314836077,
      "tokens_seen": 730726400
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007864393179538616,
      "loss": 1.4593,
      "theoretical_loss": 3.791500027359214,
      "tokens_seen": 730791936
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007864192577733201,
      "loss": 1.5354,
      "theoretical_loss": 3.791466027137859,
      "tokens_seen": 730857472
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007863991975927783,
      "loss": 1.5912,
      "theoretical_loss": 3.7914320308187444,
      "tokens_seen": 730923008
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007863791374122368,
      "loss": 1.5214,
      "theoretical_loss": 3.7913980384010726,
      "tokens_seen": 730988544
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000786359077231695,
      "loss": 1.5476,
      "theoretical_loss": 3.791364049884046,
      "tokens_seen": 731054080
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007863390170511535,
      "loss": 1.6559,
      "theoretical_loss": 3.7913300652668678,
      "tokens_seen": 731119616
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007863189568706119,
      "loss": 1.588,
      "theoretical_loss": 3.791296084548741,
      "tokens_seen": 731185152
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007862988966900702,
      "loss": 1.4801,
      "theoretical_loss": 3.791262107728868,
      "tokens_seen": 731250688
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007862788365095286,
      "loss": 1.6053,
      "theoretical_loss": 3.7912281348064534,
      "tokens_seen": 731316224
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007862587763289871,
      "loss": 1.5866,
      "theoretical_loss": 3.7911941657807002,
      "tokens_seen": 731381760
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007862387161484453,
      "loss": 1.5112,
      "theoretical_loss": 3.791160200650812,
      "tokens_seen": 731447296
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007862186559679038,
      "loss": 1.615,
      "theoretical_loss": 3.7911262394159944,
      "tokens_seen": 731512832
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000786198595787362,
      "loss": 1.4974,
      "theoretical_loss": 3.7910922820754505,
      "tokens_seen": 731578368
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007861785356068205,
      "loss": 1.4542,
      "theoretical_loss": 3.7910583286283854,
      "tokens_seen": 731643904
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007861584754262789,
      "loss": 1.5288,
      "theoretical_loss": 3.7910243790740044,
      "tokens_seen": 731709440
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007861384152457372,
      "loss": 1.5616,
      "theoretical_loss": 3.790990433411512,
      "tokens_seen": 731774976
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007861183550651956,
      "loss": 1.5907,
      "theoretical_loss": 3.790956491640114,
      "tokens_seen": 731840512
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000786098294884654,
      "loss": 1.6463,
      "theoretical_loss": 3.7909225537590157,
      "tokens_seen": 731906048
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007860782347041123,
      "loss": 1.4904,
      "theoretical_loss": 3.7908886197674225,
      "tokens_seen": 731971584
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007860581745235708,
      "loss": 1.4949,
      "theoretical_loss": 3.7908546896645414,
      "tokens_seen": 732037120
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000786038114343029,
      "loss": 1.5844,
      "theoretical_loss": 3.7908207634495783,
      "tokens_seen": 732102656
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007860180541624875,
      "loss": 1.6109,
      "theoretical_loss": 3.790786841121739,
      "tokens_seen": 732168192
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007859979939819458,
      "loss": 1.5295,
      "theoretical_loss": 3.790752922680232,
      "tokens_seen": 732233728
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007859779338014042,
      "loss": 1.558,
      "theoretical_loss": 3.7907190081242623,
      "tokens_seen": 732299264
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007859578736208626,
      "loss": 1.6146,
      "theoretical_loss": 3.7906850974530384,
      "tokens_seen": 732364800
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000785937813440321,
      "loss": 1.4625,
      "theoretical_loss": 3.7906511906657676,
      "tokens_seen": 732430336
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007859177532597793,
      "loss": 1.5532,
      "theoretical_loss": 3.7906172877616573,
      "tokens_seen": 732495872
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007858976930792378,
      "loss": 1.5032,
      "theoretical_loss": 3.790583388739915,
      "tokens_seen": 732561408
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007858776328986961,
      "loss": 1.5335,
      "theoretical_loss": 3.79054949359975,
      "tokens_seen": 732626944
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007858575727181545,
      "loss": 1.5711,
      "theoretical_loss": 3.7905156023403697,
      "tokens_seen": 732692480
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007858375125376128,
      "loss": 1.4873,
      "theoretical_loss": 3.790481714960983,
      "tokens_seen": 732758016
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007858174523570712,
      "loss": 1.4483,
      "theoretical_loss": 3.7904478314607988,
      "tokens_seen": 732823552
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007857973921765296,
      "loss": 1.7101,
      "theoretical_loss": 3.7904139518390263,
      "tokens_seen": 732889088
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000785777331995988,
      "loss": 1.6459,
      "theoretical_loss": 3.7903800760948743,
      "tokens_seen": 732954624
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007857572718154463,
      "loss": 1.5922,
      "theoretical_loss": 3.790346204227553,
      "tokens_seen": 733020160
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007857372116349047,
      "loss": 1.4977,
      "theoretical_loss": 3.790312336236272,
      "tokens_seen": 733085696
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007857171514543631,
      "loss": 1.5984,
      "theoretical_loss": 3.7902784721202414,
      "tokens_seen": 733151232
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007856970912738215,
      "loss": 1.5617,
      "theoretical_loss": 3.790244611878671,
      "tokens_seen": 733216768
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007856770310932798,
      "loss": 1.5315,
      "theoretical_loss": 3.790210755510771,
      "tokens_seen": 733282304
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007856569709127383,
      "loss": 1.6488,
      "theoretical_loss": 3.7901769030157535,
      "tokens_seen": 733347840
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007856369107321965,
      "loss": 1.5304,
      "theoretical_loss": 3.7901430543928276,
      "tokens_seen": 733413376
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000785616850551655,
      "loss": 1.5984,
      "theoretical_loss": 3.790109209641206,
      "tokens_seen": 733478912
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007855967903711133,
      "loss": 1.6362,
      "theoretical_loss": 3.7900753687600996,
      "tokens_seen": 733544448
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007855767301905717,
      "loss": 1.5384,
      "theoretical_loss": 3.7900415317487193,
      "tokens_seen": 733609984
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007855566700100301,
      "loss": 1.5263,
      "theoretical_loss": 3.790007698606278,
      "tokens_seen": 733675520
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007855366098294885,
      "loss": 1.4937,
      "theoretical_loss": 3.7899738693319875,
      "tokens_seen": 733741056
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007855165496489468,
      "loss": 1.588,
      "theoretical_loss": 3.7899400439250597,
      "tokens_seen": 733806592
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007854964894684053,
      "loss": 1.5583,
      "theoretical_loss": 3.7899062223847078,
      "tokens_seen": 733872128
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007854764292878635,
      "loss": 1.629,
      "theoretical_loss": 3.7898724047101444,
      "tokens_seen": 733937664
    },
    {
      "epoch": 0.22,
      "objective/train/advantage_avg": 0.4948802888393402,
      "objective/train/docs_used": 272255,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5846424102783203,
      "objective/train/original_loss": 1.5846424102783203,
      "objective/train/theoretical_loss": 3.7898385909005814,
      "objective/train/tokens_used": 754463200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24517832696437836,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050734519958496,
      "objective/train/weighted_lm_loss": 1.6648907661437988,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7898385909005814,
      "tokens_seen": 734003200
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000785456369107322,
      "loss": 1.7021,
      "theoretical_loss": 3.7898385909005814,
      "tokens_seen": 734003200
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007854363089267803,
      "loss": 1.6323,
      "theoretical_loss": 3.7898047809552335,
      "tokens_seen": 734068736
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007854162487462387,
      "loss": 1.3808,
      "theoretical_loss": 3.789770974873314,
      "tokens_seen": 734134272
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007853961885656971,
      "loss": 1.581,
      "theoretical_loss": 3.7897371726540356,
      "tokens_seen": 734199808
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007853761283851554,
      "loss": 1.6463,
      "theoretical_loss": 3.7897033742966135,
      "tokens_seen": 734265344
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007853560682046138,
      "loss": 1.547,
      "theoretical_loss": 3.7896695798002606,
      "tokens_seen": 734330880
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007853360080240723,
      "loss": 1.5695,
      "theoretical_loss": 3.7896357891641923,
      "tokens_seen": 734396416
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007853159478435305,
      "loss": 1.5675,
      "theoretical_loss": 3.789602002387623,
      "tokens_seen": 734461952
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000785295887662989,
      "loss": 1.5541,
      "theoretical_loss": 3.789568219469767,
      "tokens_seen": 734527488
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007852758274824475,
      "loss": 1.5826,
      "theoretical_loss": 3.78953444040984,
      "tokens_seen": 734593024
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007852557673019057,
      "loss": 1.7494,
      "theoretical_loss": 3.789500665207057,
      "tokens_seen": 734658560
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007852357071213642,
      "loss": 1.5821,
      "theoretical_loss": 3.7894668938606335,
      "tokens_seen": 734724096
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007852156469408224,
      "loss": 1.5327,
      "theoretical_loss": 3.789433126369786,
      "tokens_seen": 734789632
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007851955867602809,
      "loss": 1.644,
      "theoretical_loss": 3.7893993627337297,
      "tokens_seen": 734855168
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007851755265797393,
      "loss": 1.5004,
      "theoretical_loss": 3.7893656029516807,
      "tokens_seen": 734920704
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007851554663991976,
      "loss": 1.5558,
      "theoretical_loss": 3.789331847022856,
      "tokens_seen": 734986240
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000785135406218656,
      "loss": 1.4657,
      "theoretical_loss": 3.7892980949464716,
      "tokens_seen": 735051776
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007851153460381144,
      "loss": 1.6541,
      "theoretical_loss": 3.7892643467217457,
      "tokens_seen": 735117312
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007850952858575727,
      "loss": 1.5398,
      "theoretical_loss": 3.789230602347894,
      "tokens_seen": 735182848
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007850752256770312,
      "loss": 1.5848,
      "theoretical_loss": 3.789196861824135,
      "tokens_seen": 735248384
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007850551654964894,
      "loss": 1.5054,
      "theoretical_loss": 3.7891631251496856,
      "tokens_seen": 735313920
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007850351053159479,
      "loss": 1.6527,
      "theoretical_loss": 3.7891293923237637,
      "tokens_seen": 735379456
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007850150451354062,
      "loss": 1.4059,
      "theoretical_loss": 3.789095663345588,
      "tokens_seen": 735444992
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007849949849548646,
      "loss": 1.5884,
      "theoretical_loss": 3.7890619382143758,
      "tokens_seen": 735510528
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000784974924774323,
      "loss": 1.5862,
      "theoretical_loss": 3.7890282169293465,
      "tokens_seen": 735576064
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007849548645937814,
      "loss": 1.5663,
      "theoretical_loss": 3.7889944994897182,
      "tokens_seen": 735641600
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007849348044132397,
      "loss": 1.5345,
      "theoretical_loss": 3.7889607858947105,
      "tokens_seen": 735707136
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007849147442326982,
      "loss": 1.4874,
      "theoretical_loss": 3.788927076143542,
      "tokens_seen": 735772672
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007848946840521565,
      "loss": 1.5477,
      "theoretical_loss": 3.7888933702354324,
      "tokens_seen": 735838208
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007848746238716149,
      "loss": 1.6029,
      "theoretical_loss": 3.788859668169601,
      "tokens_seen": 735903744
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007848545636910732,
      "loss": 1.537,
      "theoretical_loss": 3.788825969945268,
      "tokens_seen": 735969280
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007848345035105317,
      "loss": 1.6824,
      "theoretical_loss": 3.7887922755616543,
      "tokens_seen": 736034816
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00078481444332999,
      "loss": 1.5639,
      "theoretical_loss": 3.7887585850179786,
      "tokens_seen": 736100352
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007847943831494484,
      "loss": 1.5474,
      "theoretical_loss": 3.788724898313463,
      "tokens_seen": 736165888
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007847743229689067,
      "loss": 1.5166,
      "theoretical_loss": 3.7886912154473276,
      "tokens_seen": 736231424
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007847542627883651,
      "loss": 1.6416,
      "theoretical_loss": 3.788657536418793,
      "tokens_seen": 736296960
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007847342026078235,
      "loss": 1.6185,
      "theoretical_loss": 3.788623861227081,
      "tokens_seen": 736362496
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007847141424272819,
      "loss": 1.4878,
      "theoretical_loss": 3.7885901898714134,
      "tokens_seen": 736428032
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007846940822467402,
      "loss": 1.509,
      "theoretical_loss": 3.788556522351011,
      "tokens_seen": 736493568
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007846740220661987,
      "loss": 1.5854,
      "theoretical_loss": 3.7885228586650967,
      "tokens_seen": 736559104
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007846539618856569,
      "loss": 1.4612,
      "theoretical_loss": 3.7884891988128926,
      "tokens_seen": 736624640
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007846339017051154,
      "loss": 1.5634,
      "theoretical_loss": 3.7884555427936197,
      "tokens_seen": 736690176
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007846138415245737,
      "loss": 1.6067,
      "theoretical_loss": 3.788421890606502,
      "tokens_seen": 736755712
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007845937813440321,
      "loss": 1.5106,
      "theoretical_loss": 3.788388242250762,
      "tokens_seen": 736821248
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007845737211634905,
      "loss": 1.6005,
      "theoretical_loss": 3.7883545977256228,
      "tokens_seen": 736886784
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007845536609829489,
      "loss": 1.6012,
      "theoretical_loss": 3.788320957030307,
      "tokens_seen": 736952320
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007845336008024072,
      "loss": 1.569,
      "theoretical_loss": 3.7882873201640397,
      "tokens_seen": 737017856
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007845135406218657,
      "loss": 1.6627,
      "theoretical_loss": 3.7882536871260433,
      "tokens_seen": 737083392
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007844934804413239,
      "loss": 1.4554,
      "theoretical_loss": 3.7882200579155416,
      "tokens_seen": 737148928
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007844734202607824,
      "loss": 1.5707,
      "theoretical_loss": 3.7881864325317602,
      "tokens_seen": 737214464
    },
    {
      "epoch": 0.22,
      "objective/train/advantage_avg": 0.4932360053062439,
      "objective/train/docs_used": 273341,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5445104837417603,
      "objective/train/original_loss": 1.5445103645324707,
      "objective/train/theoretical_loss": 3.788152810973922,
      "objective/train/tokens_used": 757740000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24367868900299072,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505623817443848,
      "objective/train/weighted_lm_loss": 1.6224743127822876,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.788152810973922,
      "tokens_seen": 737280000
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007844533600802407,
      "loss": 1.6035,
      "theoretical_loss": 3.788152810973922,
      "tokens_seen": 737280000
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007844332998996991,
      "loss": 1.6159,
      "theoretical_loss": 3.788119193241253,
      "tokens_seen": 737345536
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007844132397191575,
      "loss": 1.589,
      "theoretical_loss": 3.788085579332977,
      "tokens_seen": 737411072
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007843931795386158,
      "loss": 1.4896,
      "theoretical_loss": 3.788051969248319,
      "tokens_seen": 737476608
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007843731193580742,
      "loss": 1.622,
      "theoretical_loss": 3.788018362986506,
      "tokens_seen": 737542144
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007843530591775327,
      "loss": 1.5806,
      "theoretical_loss": 3.7879847605467614,
      "tokens_seen": 737607680
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007843329989969909,
      "loss": 1.5376,
      "theoretical_loss": 3.787951161928312,
      "tokens_seen": 737673216
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007843129388164494,
      "loss": 1.5502,
      "theoretical_loss": 3.7879175671303846,
      "tokens_seen": 737738752
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007842928786359078,
      "loss": 1.6068,
      "theoretical_loss": 3.7878839761522043,
      "tokens_seen": 737804288
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007842728184553661,
      "loss": 1.5394,
      "theoretical_loss": 3.7878503889929975,
      "tokens_seen": 737869824
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007842527582748245,
      "loss": 1.6246,
      "theoretical_loss": 3.7878168056519916,
      "tokens_seen": 737935360
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007842326980942828,
      "loss": 1.5885,
      "theoretical_loss": 3.787783226128413,
      "tokens_seen": 738000896
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007842126379137412,
      "loss": 1.4953,
      "theoretical_loss": 3.787749650421489,
      "tokens_seen": 738066432
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007841925777331997,
      "loss": 1.5716,
      "theoretical_loss": 3.7877160785304476,
      "tokens_seen": 738131968
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007841725175526579,
      "loss": 1.5421,
      "theoretical_loss": 3.787682510454515,
      "tokens_seen": 738197504
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007841524573721164,
      "loss": 1.5583,
      "theoretical_loss": 3.78764894619292,
      "tokens_seen": 738263040
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007841323971915747,
      "loss": 1.654,
      "theoretical_loss": 3.78761538574489,
      "tokens_seen": 738328576
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007841123370110331,
      "loss": 1.5801,
      "theoretical_loss": 3.787581829109654,
      "tokens_seen": 738394112
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007840922768304915,
      "loss": 1.4853,
      "theoretical_loss": 3.7875482762864405,
      "tokens_seen": 738459648
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007840722166499498,
      "loss": 1.5361,
      "theoretical_loss": 3.787514727274478,
      "tokens_seen": 738525184
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007840521564694082,
      "loss": 1.5535,
      "theoretical_loss": 3.7874811820729946,
      "tokens_seen": 738590720
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007840320962888666,
      "loss": 1.4369,
      "theoretical_loss": 3.7874476406812203,
      "tokens_seen": 738656256
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007840120361083249,
      "loss": 1.578,
      "theoretical_loss": 3.787414103098384,
      "tokens_seen": 738721792
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007839919759277834,
      "loss": 1.4614,
      "theoretical_loss": 3.787380569323716,
      "tokens_seen": 738787328
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007839719157472417,
      "loss": 1.5497,
      "theoretical_loss": 3.7873470393564457,
      "tokens_seen": 738852864
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007839518555667001,
      "loss": 1.6377,
      "theoretical_loss": 3.7873135131958033,
      "tokens_seen": 738918400
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007839317953861585,
      "loss": 1.5297,
      "theoretical_loss": 3.7872799908410193,
      "tokens_seen": 738983936
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007839117352056169,
      "loss": 1.512,
      "theoretical_loss": 3.7872464722913235,
      "tokens_seen": 739049472
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007838916750250752,
      "loss": 1.5288,
      "theoretical_loss": 3.787212957545947,
      "tokens_seen": 739115008
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007838716148445336,
      "loss": 1.6443,
      "theoretical_loss": 3.787179446604121,
      "tokens_seen": 739180544
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007838515546639919,
      "loss": 1.4683,
      "theoretical_loss": 3.787145939465076,
      "tokens_seen": 739246080
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007838314944834504,
      "loss": 1.7587,
      "theoretical_loss": 3.7871124361280444,
      "tokens_seen": 739311616
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007838114343029087,
      "loss": 1.5525,
      "theoretical_loss": 3.7870789365922564,
      "tokens_seen": 739377152
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007837913741223671,
      "loss": 1.4346,
      "theoretical_loss": 3.7870454408569456,
      "tokens_seen": 739442688
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007837713139418255,
      "loss": 1.4761,
      "theoretical_loss": 3.7870119489213425,
      "tokens_seen": 739508224
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007837512537612839,
      "loss": 1.5762,
      "theoretical_loss": 3.78697846078468,
      "tokens_seen": 739573760
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007837311935807423,
      "loss": 1.7249,
      "theoretical_loss": 3.7869449764461915,
      "tokens_seen": 739639296
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007837111334002006,
      "loss": 1.5301,
      "theoretical_loss": 3.7869114959051084,
      "tokens_seen": 739704832
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007836910732196591,
      "loss": 1.594,
      "theoretical_loss": 3.786878019160664,
      "tokens_seen": 739770368
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007836710130391173,
      "loss": 1.5106,
      "theoretical_loss": 3.786844546212092,
      "tokens_seen": 739835904
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007836509528585758,
      "loss": 1.5896,
      "theoretical_loss": 3.7868110770586254,
      "tokens_seen": 739901440
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007836308926780341,
      "loss": 1.4769,
      "theoretical_loss": 3.7867776116994976,
      "tokens_seen": 739966976
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007836108324974925,
      "loss": 1.5966,
      "theoretical_loss": 3.7867441501339427,
      "tokens_seen": 740032512
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007835907723169509,
      "loss": 1.6182,
      "theoretical_loss": 3.7867106923611953,
      "tokens_seen": 740098048
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007835707121364093,
      "loss": 1.5045,
      "theoretical_loss": 3.7866772383804888,
      "tokens_seen": 740163584
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007835506519558676,
      "loss": 1.6298,
      "theoretical_loss": 3.786643788191058,
      "tokens_seen": 740229120
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007835305917753261,
      "loss": 1.566,
      "theoretical_loss": 3.7866103417921373,
      "tokens_seen": 740294656
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007835105315947843,
      "loss": 1.7352,
      "theoretical_loss": 3.786576899182963,
      "tokens_seen": 740360192
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007834904714142428,
      "loss": 1.5456,
      "theoretical_loss": 3.7865434603627692,
      "tokens_seen": 740425728
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007834704112337011,
      "loss": 1.6428,
      "theoretical_loss": 3.7865100253307915,
      "tokens_seen": 740491264
    },
    {
      "epoch": 0.22,
      "objective/train/advantage_avg": 0.4925503432750702,
      "objective/train/docs_used": 274540,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6329230070114136,
      "objective/train/original_loss": 1.632922887802124,
      "objective/train/theoretical_loss": 3.786476594086265,
      "objective/train/tokens_used": 761016800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2432449609041214,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504916906356812,
      "objective/train/weighted_lm_loss": 1.7151329517364502,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.786476594086265,
      "tokens_seen": 740556800
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007834503510531595,
      "loss": 1.6719,
      "theoretical_loss": 3.786476594086265,
      "tokens_seen": 740556800
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007834302908726179,
      "loss": 1.5262,
      "theoretical_loss": 3.786443166628427,
      "tokens_seen": 740622336
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007834102306920762,
      "loss": 1.5985,
      "theoretical_loss": 3.7864097429565113,
      "tokens_seen": 740687872
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007833901705115346,
      "loss": 1.5076,
      "theoretical_loss": 3.7863763230697565,
      "tokens_seen": 740753408
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007833701103309931,
      "loss": 1.5015,
      "theoretical_loss": 3.7863429069673984,
      "tokens_seen": 740818944
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007833500501504513,
      "loss": 1.5888,
      "theoretical_loss": 3.786309494648673,
      "tokens_seen": 740884480
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007833299899699098,
      "loss": 1.4897,
      "theoretical_loss": 3.7862760861128173,
      "tokens_seen": 740950016
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007833099297893682,
      "loss": 1.4225,
      "theoretical_loss": 3.7862426813590697,
      "tokens_seen": 741015552
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007832898696088265,
      "loss": 1.7624,
      "theoretical_loss": 3.7862092803866663,
      "tokens_seen": 741081088
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007832698094282849,
      "loss": 1.4397,
      "theoretical_loss": 3.7861758831948453,
      "tokens_seen": 741146624
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007832497492477432,
      "loss": 1.4616,
      "theoretical_loss": 3.7861424897828453,
      "tokens_seen": 741212160
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007832296890672016,
      "loss": 1.5894,
      "theoretical_loss": 3.7861091001499023,
      "tokens_seen": 741277696
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007832096288866601,
      "loss": 1.5136,
      "theoretical_loss": 3.786075714295257,
      "tokens_seen": 741343232
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007831895687061183,
      "loss": 1.5964,
      "theoretical_loss": 3.786042332218146,
      "tokens_seen": 741408768
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007831695085255768,
      "loss": 1.6782,
      "theoretical_loss": 3.7860089539178086,
      "tokens_seen": 741474304
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000783149448345035,
      "loss": 1.5885,
      "theoretical_loss": 3.785975579393485,
      "tokens_seen": 741539840
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007831293881644935,
      "loss": 1.5772,
      "theoretical_loss": 3.7859422086444123,
      "tokens_seen": 741605376
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007831093279839519,
      "loss": 1.5754,
      "theoretical_loss": 3.785908841669831,
      "tokens_seen": 741670912
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007830892678034102,
      "loss": 1.5719,
      "theoretical_loss": 3.7858754784689808,
      "tokens_seen": 741736448
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007830692076228686,
      "loss": 1.5013,
      "theoretical_loss": 3.785842119041101,
      "tokens_seen": 741801984
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000783049147442327,
      "loss": 1.6325,
      "theoretical_loss": 3.7858087633854325,
      "tokens_seen": 741867520
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007830290872617853,
      "loss": 1.5484,
      "theoretical_loss": 3.7857754115012145,
      "tokens_seen": 741933056
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007830090270812438,
      "loss": 1.5839,
      "theoretical_loss": 3.785742063387688,
      "tokens_seen": 741998592
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007829889669007021,
      "loss": 1.4878,
      "theoretical_loss": 3.785708719044094,
      "tokens_seen": 742064128
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007829689067201605,
      "loss": 1.5863,
      "theoretical_loss": 3.785675378469673,
      "tokens_seen": 742129664
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007829488465396189,
      "loss": 1.5131,
      "theoretical_loss": 3.7856420416636665,
      "tokens_seen": 742195200
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007829287863590773,
      "loss": 1.5166,
      "theoretical_loss": 3.7856087086253156,
      "tokens_seen": 742260736
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007829087261785356,
      "loss": 1.633,
      "theoretical_loss": 3.785575379353862,
      "tokens_seen": 742326272
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000782888665997994,
      "loss": 1.5373,
      "theoretical_loss": 3.7855420538485474,
      "tokens_seen": 742391808
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0007828686058174523,
      "loss": 1.5729,
      "theoretical_loss": 3.785508732108614,
      "tokens_seen": 742457344
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007828485456369108,
      "loss": 1.6371,
      "theoretical_loss": 3.7854754141333036,
      "tokens_seen": 742522880
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007828284854563691,
      "loss": 1.549,
      "theoretical_loss": 3.785442099921859,
      "tokens_seen": 742588416
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007828084252758275,
      "loss": 1.5941,
      "theoretical_loss": 3.7854087894735233,
      "tokens_seen": 742653952
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007827883650952858,
      "loss": 1.6761,
      "theoretical_loss": 3.785375482787538,
      "tokens_seen": 742719488
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007827683049147443,
      "loss": 1.3576,
      "theoretical_loss": 3.7853421798631484,
      "tokens_seen": 742785024
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007827482447342026,
      "loss": 1.4054,
      "theoretical_loss": 3.7853088806995965,
      "tokens_seen": 742850560
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000782728184553661,
      "loss": 1.6165,
      "theoretical_loss": 3.7852755852961257,
      "tokens_seen": 742916096
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007827081243731193,
      "loss": 1.5974,
      "theoretical_loss": 3.7852422936519803,
      "tokens_seen": 742981632
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007826880641925777,
      "loss": 1.6086,
      "theoretical_loss": 3.785209005766404,
      "tokens_seen": 743047168
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007826680040120361,
      "loss": 1.5917,
      "theoretical_loss": 3.785175721638641,
      "tokens_seen": 743112704
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007826479438314945,
      "loss": 1.5793,
      "theoretical_loss": 3.785142441267936,
      "tokens_seen": 743178240
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007826278836509528,
      "loss": 1.4332,
      "theoretical_loss": 3.7851091646535333,
      "tokens_seen": 743243776
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007826078234704113,
      "loss": 1.5988,
      "theoretical_loss": 3.785075891794678,
      "tokens_seen": 743309312
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007825877632898696,
      "loss": 1.5703,
      "theoretical_loss": 3.7850426226906153,
      "tokens_seen": 743374848
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000782567703109328,
      "loss": 1.6676,
      "theoretical_loss": 3.7850093573405905,
      "tokens_seen": 743440384
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007825476429287864,
      "loss": 1.5384,
      "theoretical_loss": 3.784976095743849,
      "tokens_seen": 743505920
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007825275827482447,
      "loss": 1.5463,
      "theoretical_loss": 3.784942837899637,
      "tokens_seen": 743571456
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007825075225677031,
      "loss": 1.4261,
      "theoretical_loss": 3.7849095838071993,
      "tokens_seen": 743636992
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007824874623871616,
      "loss": 1.6289,
      "theoretical_loss": 3.7848763334657827,
      "tokens_seen": 743702528
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007824674022066198,
      "loss": 1.7337,
      "theoretical_loss": 3.784843086874634,
      "tokens_seen": 743768064
    },
    {
      "epoch": 0.23,
      "objective/train/advantage_avg": 0.4921451508998871,
      "objective/train/docs_used": 275779,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5914231538772583,
      "objective/train/original_loss": 1.5914231538772583,
      "objective/train/theoretical_loss": 3.784809844033,
      "objective/train/tokens_used": 764293600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2425653487443924,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504475831985474,
      "objective/train/weighted_lm_loss": 1.6718097925186157,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.784809844033,
      "tokens_seen": 743833600
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007824473420260783,
      "loss": 1.5922,
      "theoretical_loss": 3.784809844033,
      "tokens_seen": 743833600
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007824272818455365,
      "loss": 1.5459,
      "theoretical_loss": 3.7847766049401264,
      "tokens_seen": 743899136
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000782407221664995,
      "loss": 1.6481,
      "theoretical_loss": 3.7847433695952617,
      "tokens_seen": 743964672
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007823871614844534,
      "loss": 1.6173,
      "theoretical_loss": 3.784710137997652,
      "tokens_seen": 744030208
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007823671013039117,
      "loss": 1.6636,
      "theoretical_loss": 3.784676910146545,
      "tokens_seen": 744095744
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007823470411233701,
      "loss": 1.515,
      "theoretical_loss": 3.7846436860411883,
      "tokens_seen": 744161280
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007823269809428286,
      "loss": 1.5366,
      "theoretical_loss": 3.7846104656808306,
      "tokens_seen": 744226816
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007823069207622868,
      "loss": 1.5114,
      "theoretical_loss": 3.7845772490647196,
      "tokens_seen": 744292352
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007822868605817453,
      "loss": 1.5009,
      "theoretical_loss": 3.784544036192103,
      "tokens_seen": 744357888
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007822668004012036,
      "loss": 1.523,
      "theoretical_loss": 3.7845108270622307,
      "tokens_seen": 744423424
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000782246740220662,
      "loss": 1.5747,
      "theoretical_loss": 3.7844776216743505,
      "tokens_seen": 744488960
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007822266800401205,
      "loss": 1.5068,
      "theoretical_loss": 3.784444420027712,
      "tokens_seen": 744554496
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007822066198595787,
      "loss": 1.5655,
      "theoretical_loss": 3.7844112221215633,
      "tokens_seen": 744620032
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007821865596790372,
      "loss": 1.5846,
      "theoretical_loss": 3.7843780279551558,
      "tokens_seen": 744685568
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007821664994984955,
      "loss": 1.5584,
      "theoretical_loss": 3.784344837527737,
      "tokens_seen": 744751104
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007821464393179539,
      "loss": 1.5834,
      "theoretical_loss": 3.784311650838558,
      "tokens_seen": 744816640
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007821263791374123,
      "loss": 1.5561,
      "theoretical_loss": 3.784278467886869,
      "tokens_seen": 744882176
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007821063189568706,
      "loss": 1.4605,
      "theoretical_loss": 3.7842452886719196,
      "tokens_seen": 744947712
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000782086258776329,
      "loss": 1.533,
      "theoretical_loss": 3.784212113192961,
      "tokens_seen": 745013248
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007820661985957874,
      "loss": 1.5748,
      "theoretical_loss": 3.7841789414492437,
      "tokens_seen": 745078784
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007820461384152457,
      "loss": 1.518,
      "theoretical_loss": 3.7841457734400183,
      "tokens_seen": 745144320
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007820260782347042,
      "loss": 1.3898,
      "theoretical_loss": 3.7841126091645365,
      "tokens_seen": 745209856
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007820060180541625,
      "loss": 1.752,
      "theoretical_loss": 3.7840794486220495,
      "tokens_seen": 745275392
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007819859578736209,
      "loss": 1.5674,
      "theoretical_loss": 3.784046291811809,
      "tokens_seen": 745340928
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007819658976930793,
      "loss": 1.5495,
      "theoretical_loss": 3.7840131387330667,
      "tokens_seen": 745406464
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007819458375125377,
      "loss": 1.5296,
      "theoretical_loss": 3.7839799893850743,
      "tokens_seen": 745472000
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000781925777331996,
      "loss": 1.6125,
      "theoretical_loss": 3.783946843767084,
      "tokens_seen": 745537536
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007819057171514544,
      "loss": 1.6196,
      "theoretical_loss": 3.7839137018783493,
      "tokens_seen": 745603072
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007818856569709127,
      "loss": 1.513,
      "theoretical_loss": 3.7838805637181223,
      "tokens_seen": 745668608
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007818655967903712,
      "loss": 1.6035,
      "theoretical_loss": 3.7838474292856557,
      "tokens_seen": 745734144
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007818455366098295,
      "loss": 1.5825,
      "theoretical_loss": 3.783814298580203,
      "tokens_seen": 745799680
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007818254764292879,
      "loss": 1.5645,
      "theoretical_loss": 3.783781171601017,
      "tokens_seen": 745865216
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007818054162487462,
      "loss": 1.5189,
      "theoretical_loss": 3.783748048347351,
      "tokens_seen": 745930752
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007817853560682047,
      "loss": 1.6216,
      "theoretical_loss": 3.78371492881846,
      "tokens_seen": 745996288
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000781765295887663,
      "loss": 1.5186,
      "theoretical_loss": 3.7836818130135974,
      "tokens_seen": 746061824
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007817452357071214,
      "loss": 1.5997,
      "theoretical_loss": 3.7836487009320168,
      "tokens_seen": 746127360
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007817251755265797,
      "loss": 1.4944,
      "theoretical_loss": 3.783615592572973,
      "tokens_seen": 746192896
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007817051153460381,
      "loss": 1.5422,
      "theoretical_loss": 3.7835824879357207,
      "tokens_seen": 746258432
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007816850551654965,
      "loss": 1.4663,
      "theoretical_loss": 3.783549387019515,
      "tokens_seen": 746323968
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007816649949849549,
      "loss": 1.5202,
      "theoretical_loss": 3.78351628982361,
      "tokens_seen": 746389504
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007816449348044132,
      "loss": 1.6376,
      "theoretical_loss": 3.7834831963472624,
      "tokens_seen": 746455040
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007816248746238717,
      "loss": 1.5701,
      "theoretical_loss": 3.7834501065897266,
      "tokens_seen": 746520576
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00078160481444333,
      "loss": 1.5658,
      "theoretical_loss": 3.7834170205502584,
      "tokens_seen": 746586112
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007815847542627884,
      "loss": 1.593,
      "theoretical_loss": 3.7833839382281145,
      "tokens_seen": 746651648
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007815646940822468,
      "loss": 1.5813,
      "theoretical_loss": 3.7833508596225496,
      "tokens_seen": 746717184
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007815446339017051,
      "loss": 1.4467,
      "theoretical_loss": 3.783317784732821,
      "tokens_seen": 746782720
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007815245737211635,
      "loss": 1.5388,
      "theoretical_loss": 3.783284713558186,
      "tokens_seen": 746848256
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000781504513540622,
      "loss": 1.6172,
      "theoretical_loss": 3.7832516460978995,
      "tokens_seen": 746913792
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007814844533600802,
      "loss": 1.5129,
      "theoretical_loss": 3.7832185823512194,
      "tokens_seen": 746979328
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007814643931795387,
      "loss": 1.5878,
      "theoretical_loss": 3.783185522317403,
      "tokens_seen": 747044864
    },
    {
      "epoch": 0.23,
      "objective/train/advantage_avg": 0.48972252011299133,
      "objective/train/docs_used": 277055,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4956200122833252,
      "objective/train/original_loss": 1.4956200122833252,
      "objective/train/theoretical_loss": 3.783152465995708,
      "objective/train/tokens_used": 767570400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24053406715393066,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501949787139893,
      "objective/train/weighted_lm_loss": 1.5706262588500977,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.783152465995708,
      "tokens_seen": 747110400
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007814443329989969,
      "loss": 1.648,
      "theoretical_loss": 3.783152465995708,
      "tokens_seen": 747110400
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007814242728184554,
      "loss": 1.4789,
      "theoretical_loss": 3.783119413385392,
      "tokens_seen": 747175936
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007814042126379138,
      "loss": 1.5212,
      "theoretical_loss": 3.783086364485712,
      "tokens_seen": 747241472
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007813841524573721,
      "loss": 1.6248,
      "theoretical_loss": 3.7830533192959264,
      "tokens_seen": 747307008
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007813640922768305,
      "loss": 1.6118,
      "theoretical_loss": 3.7830202778152935,
      "tokens_seen": 747372544
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000781344032096289,
      "loss": 1.4851,
      "theoretical_loss": 3.782987240043072,
      "tokens_seen": 747438080
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007813239719157472,
      "loss": 1.5586,
      "theoretical_loss": 3.7829542059785206,
      "tokens_seen": 747503616
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007813039117352057,
      "loss": 1.4784,
      "theoretical_loss": 3.7829211756208974,
      "tokens_seen": 747569152
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007812838515546639,
      "loss": 1.5165,
      "theoretical_loss": 3.7828881489694632,
      "tokens_seen": 747634688
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007812637913741224,
      "loss": 1.561,
      "theoretical_loss": 3.7828551260234753,
      "tokens_seen": 747700224
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007812437311935808,
      "loss": 1.4978,
      "theoretical_loss": 3.782822106782195,
      "tokens_seen": 747765760
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007812236710130391,
      "loss": 1.4905,
      "theoretical_loss": 3.7827890912448803,
      "tokens_seen": 747831296
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007812036108324975,
      "loss": 1.5409,
      "theoretical_loss": 3.7827560794107926,
      "tokens_seen": 747896832
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007811835506519559,
      "loss": 1.5891,
      "theoretical_loss": 3.782723071279192,
      "tokens_seen": 747962368
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007811634904714142,
      "loss": 1.4398,
      "theoretical_loss": 3.782690066849338,
      "tokens_seen": 748027904
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007811434302908727,
      "loss": 1.6519,
      "theoretical_loss": 3.7826570661204917,
      "tokens_seen": 748093440
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007811233701103309,
      "loss": 1.4951,
      "theoretical_loss": 3.7826240690919137,
      "tokens_seen": 748158976
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007811033099297894,
      "loss": 1.5375,
      "theoretical_loss": 3.7825910757628654,
      "tokens_seen": 748224512
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007810832497492477,
      "loss": 1.7363,
      "theoretical_loss": 3.7825580861326085,
      "tokens_seen": 748290048
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007810631895687061,
      "loss": 1.5526,
      "theoretical_loss": 3.7825251002004032,
      "tokens_seen": 748355584
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007810431293881645,
      "loss": 1.5729,
      "theoretical_loss": 3.7824921179655115,
      "tokens_seen": 748421120
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007810230692076229,
      "loss": 1.5789,
      "theoretical_loss": 3.7824591394271962,
      "tokens_seen": 748486656
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007810030090270812,
      "loss": 1.557,
      "theoretical_loss": 3.782426164584718,
      "tokens_seen": 748552192
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007809829488465397,
      "loss": 1.5806,
      "theoretical_loss": 3.7823931934373407,
      "tokens_seen": 748617728
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000780962888665998,
      "loss": 1.5143,
      "theoretical_loss": 3.782360225984325,
      "tokens_seen": 748683264
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007809428284854564,
      "loss": 1.6163,
      "theoretical_loss": 3.7823272622249355,
      "tokens_seen": 748748800
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007809227683049147,
      "loss": 1.4267,
      "theoretical_loss": 3.7822943021584337,
      "tokens_seen": 748814336
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007809027081243731,
      "loss": 1.4885,
      "theoretical_loss": 3.782261345784084,
      "tokens_seen": 748879872
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007808826479438315,
      "loss": 1.6731,
      "theoretical_loss": 3.782228393101149,
      "tokens_seen": 748945408
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007808625877632899,
      "loss": 1.5349,
      "theoretical_loss": 3.7821954441088916,
      "tokens_seen": 749010944
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007808425275827482,
      "loss": 1.5763,
      "theoretical_loss": 3.782162498806577,
      "tokens_seen": 749076480
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007808224674022066,
      "loss": 1.5453,
      "theoretical_loss": 3.782129557193468,
      "tokens_seen": 749142016
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000780802407221665,
      "loss": 1.5816,
      "theoretical_loss": 3.78209661926883,
      "tokens_seen": 749207552
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007807823470411234,
      "loss": 1.4659,
      "theoretical_loss": 3.7820636850319262,
      "tokens_seen": 749273088
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007807622868605818,
      "loss": 1.5221,
      "theoretical_loss": 3.782030754482022,
      "tokens_seen": 749338624
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007807422266800401,
      "loss": 1.6097,
      "theoretical_loss": 3.781997827618382,
      "tokens_seen": 749404160
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007807221664994985,
      "loss": 1.6654,
      "theoretical_loss": 3.781964904440271,
      "tokens_seen": 749469696
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007807021063189569,
      "loss": 1.6082,
      "theoretical_loss": 3.7819319849469544,
      "tokens_seen": 749535232
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007806820461384153,
      "loss": 1.556,
      "theoretical_loss": 3.7818990691376984,
      "tokens_seen": 749600768
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007806619859578736,
      "loss": 1.6522,
      "theoretical_loss": 3.7818661570117675,
      "tokens_seen": 749666304
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007806419257773321,
      "loss": 1.6318,
      "theoretical_loss": 3.7818332485684283,
      "tokens_seen": 749731840
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007806218655967904,
      "loss": 1.6752,
      "theoretical_loss": 3.7818003438069465,
      "tokens_seen": 749797376
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007806018054162488,
      "loss": 1.4901,
      "theoretical_loss": 3.781767442726589,
      "tokens_seen": 749862912
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007805817452357072,
      "loss": 1.5268,
      "theoretical_loss": 3.7817345453266213,
      "tokens_seen": 749928448
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007805616850551655,
      "loss": 1.5535,
      "theoretical_loss": 3.781701651606311,
      "tokens_seen": 749993984
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007805416248746239,
      "loss": 1.5772,
      "theoretical_loss": 3.7816687615649256,
      "tokens_seen": 750059520
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007805215646940824,
      "loss": 1.5844,
      "theoretical_loss": 3.7816358752017303,
      "tokens_seen": 750125056
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007805015045135406,
      "loss": 1.649,
      "theoretical_loss": 3.7816029925159946,
      "tokens_seen": 750190592
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007804814443329991,
      "loss": 1.6263,
      "theoretical_loss": 3.7815701135069846,
      "tokens_seen": 750256128
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007804613841524573,
      "loss": 1.5047,
      "theoretical_loss": 3.7815372381739687,
      "tokens_seen": 750321664
    },
    {
      "epoch": 0.23,
      "objective/train/advantage_avg": 0.4931403398513794,
      "objective/train/docs_used": 278273,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.431071400642395,
      "objective/train/original_loss": 1.4310712814331055,
      "objective/train/theoretical_loss": 3.7815043665162147,
      "objective/train/tokens_used": 770847200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24355730414390564,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050552248954773,
      "objective/train/weighted_lm_loss": 1.5032776594161987,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7815043665162147,
      "tokens_seen": 750387200
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007804413239719158,
      "loss": 1.5337,
      "theoretical_loss": 3.7815043665162147,
      "tokens_seen": 750387200
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007804212637913742,
      "loss": 1.557,
      "theoretical_loss": 3.7814714985329907,
      "tokens_seen": 750452736
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007804012036108325,
      "loss": 1.594,
      "theoretical_loss": 3.7814386342235653,
      "tokens_seen": 750518272
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007803811434302909,
      "loss": 1.4072,
      "theoretical_loss": 3.781405773587207,
      "tokens_seen": 750583808
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007803610832497494,
      "loss": 1.5198,
      "theoretical_loss": 3.7813729166231855,
      "tokens_seen": 750649344
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007803410230692076,
      "loss": 1.4929,
      "theoretical_loss": 3.7813400633307683,
      "tokens_seen": 750714880
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007803209628886661,
      "loss": 1.442,
      "theoretical_loss": 3.7813072137092254,
      "tokens_seen": 750780416
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007803009027081243,
      "loss": 1.4845,
      "theoretical_loss": 3.781274367757826,
      "tokens_seen": 750845952
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007802808425275828,
      "loss": 1.6004,
      "theoretical_loss": 3.7812415254758407,
      "tokens_seen": 750911488
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007802607823470412,
      "loss": 1.5159,
      "theoretical_loss": 3.7812086868625387,
      "tokens_seen": 750977024
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007802407221664995,
      "loss": 1.461,
      "theoretical_loss": 3.7811758519171894,
      "tokens_seen": 751042560
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007802206619859579,
      "loss": 1.6043,
      "theoretical_loss": 3.781143020639064,
      "tokens_seen": 751108096
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007802006018054163,
      "loss": 1.5826,
      "theoretical_loss": 3.781110193027433,
      "tokens_seen": 751173632
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007801805416248746,
      "loss": 1.6208,
      "theoretical_loss": 3.781077369081567,
      "tokens_seen": 751239168
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007801604814443331,
      "loss": 1.5903,
      "theoretical_loss": 3.781044548800736,
      "tokens_seen": 751304704
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007801404212637913,
      "loss": 1.6331,
      "theoretical_loss": 3.7810117321842123,
      "tokens_seen": 751370240
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007801203610832498,
      "loss": 1.6124,
      "theoretical_loss": 3.7809789192312673,
      "tokens_seen": 751435776
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007801003009027081,
      "loss": 1.598,
      "theoretical_loss": 3.780946109941171,
      "tokens_seen": 751501312
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007800802407221665,
      "loss": 1.5621,
      "theoretical_loss": 3.7809133043131973,
      "tokens_seen": 751566848
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007800601805416249,
      "loss": 1.5991,
      "theoretical_loss": 3.7808805023466165,
      "tokens_seen": 751632384
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007800401203610833,
      "loss": 1.6092,
      "theoretical_loss": 3.780847704040702,
      "tokens_seen": 751697920
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007800200601805416,
      "loss": 1.5308,
      "theoretical_loss": 3.780814909394725,
      "tokens_seen": 751763456
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007800000000000001,
      "loss": 1.6001,
      "theoretical_loss": 3.7807821184079584,
      "tokens_seen": 751828992
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007799799398194583,
      "loss": 1.4683,
      "theoretical_loss": 3.780749331079676,
      "tokens_seen": 751894528
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007799598796389168,
      "loss": 1.5728,
      "theoretical_loss": 3.780716547409149,
      "tokens_seen": 751960064
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007799398194583751,
      "loss": 1.5938,
      "theoretical_loss": 3.7806837673956526,
      "tokens_seen": 752025600
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007799197592778335,
      "loss": 1.6175,
      "theoretical_loss": 3.780650991038459,
      "tokens_seen": 752091136
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007798996990972919,
      "loss": 1.565,
      "theoretical_loss": 3.7806182183368415,
      "tokens_seen": 752156672
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007798796389167503,
      "loss": 1.5365,
      "theoretical_loss": 3.780585449290075,
      "tokens_seen": 752222208
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007798595787362086,
      "loss": 1.5461,
      "theoretical_loss": 3.7805526838974326,
      "tokens_seen": 752287744
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000779839518555667,
      "loss": 1.6144,
      "theoretical_loss": 3.7805199221581893,
      "tokens_seen": 752353280
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007798194583751254,
      "loss": 1.4486,
      "theoretical_loss": 3.7804871640716193,
      "tokens_seen": 752418816
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007797993981945838,
      "loss": 1.6904,
      "theoretical_loss": 3.780454409636997,
      "tokens_seen": 752484352
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007797793380140421,
      "loss": 1.5733,
      "theoretical_loss": 3.780421658853597,
      "tokens_seen": 752549888
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007797592778335005,
      "loss": 1.731,
      "theoretical_loss": 3.7803889117206957,
      "tokens_seen": 752615424
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007797392176529588,
      "loss": 1.5194,
      "theoretical_loss": 3.780356168237567,
      "tokens_seen": 752680960
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007797191574724173,
      "loss": 1.6039,
      "theoretical_loss": 3.780323428403487,
      "tokens_seen": 752746496
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007796990972918756,
      "loss": 1.6902,
      "theoretical_loss": 3.780290692217731,
      "tokens_seen": 752812032
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000779679037111334,
      "loss": 1.5576,
      "theoretical_loss": 3.7802579596795756,
      "tokens_seen": 752877568
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007796589769307924,
      "loss": 1.454,
      "theoretical_loss": 3.7802252307882966,
      "tokens_seen": 752943104
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007796389167502508,
      "loss": 1.465,
      "theoretical_loss": 3.7801925055431695,
      "tokens_seen": 753008640
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007796188565697091,
      "loss": 1.6124,
      "theoretical_loss": 3.780159783943472,
      "tokens_seen": 753074176
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007795987963891676,
      "loss": 1.5425,
      "theoretical_loss": 3.7801270659884807,
      "tokens_seen": 753139712
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007795787362086258,
      "loss": 1.6054,
      "theoretical_loss": 3.7800943516774717,
      "tokens_seen": 753205248
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007795586760280843,
      "loss": 1.559,
      "theoretical_loss": 3.7800616410097225,
      "tokens_seen": 753270784
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007795386158475426,
      "loss": 1.5107,
      "theoretical_loss": 3.780028933984511,
      "tokens_seen": 753336320
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000779518555667001,
      "loss": 1.6181,
      "theoretical_loss": 3.7799962306011143,
      "tokens_seen": 753401856
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007794984954864594,
      "loss": 1.6451,
      "theoretical_loss": 3.77996353085881,
      "tokens_seen": 753467392
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007794784353059177,
      "loss": 1.627,
      "theoretical_loss": 3.7799308347568763,
      "tokens_seen": 753532928
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007794583751253761,
      "loss": 1.563,
      "theoretical_loss": 3.7798981422945914,
      "tokens_seen": 753598464
    },
    {
      "epoch": 0.23,
      "objective/train/advantage_avg": 0.48776188492774963,
      "objective/train/docs_used": 279492,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5366716384887695,
      "objective/train/original_loss": 1.5366718769073486,
      "objective/train/theoretical_loss": 3.779865453471234,
      "objective/train/tokens_used": 774124000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2389141321182251,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0499906539916992,
      "objective/train/weighted_lm_loss": 1.6135380268096924,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.779865453471234,
      "tokens_seen": 753664000
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007794383149448346,
      "loss": 1.5559,
      "theoretical_loss": 3.779865453471234,
      "tokens_seen": 753664000
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007794182547642928,
      "loss": 1.5738,
      "theoretical_loss": 3.7798327682860817,
      "tokens_seen": 753729536
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007793981945837513,
      "loss": 1.5779,
      "theoretical_loss": 3.779800086738414,
      "tokens_seen": 753795072
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007793781344032095,
      "loss": 1.5429,
      "theoretical_loss": 3.77976740882751,
      "tokens_seen": 753860608
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000779358074222668,
      "loss": 1.5679,
      "theoretical_loss": 3.7797347345526484,
      "tokens_seen": 753926144
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007793380140421264,
      "loss": 1.5778,
      "theoretical_loss": 3.779702063913109,
      "tokens_seen": 753991680
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007793179538615847,
      "loss": 1.5505,
      "theoretical_loss": 3.7796693969081714,
      "tokens_seen": 754057216
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007792978936810431,
      "loss": 1.6022,
      "theoretical_loss": 3.7796367335371155,
      "tokens_seen": 754122752
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007792778335005016,
      "loss": 1.5268,
      "theoretical_loss": 3.7796040737992205,
      "tokens_seen": 754188288
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007792577733199599,
      "loss": 1.5553,
      "theoretical_loss": 3.7795714176937674,
      "tokens_seen": 754253824
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007792377131394183,
      "loss": 1.6663,
      "theoretical_loss": 3.779538765220037,
      "tokens_seen": 754319360
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007792176529588767,
      "loss": 1.6113,
      "theoretical_loss": 3.779506116377309,
      "tokens_seen": 754384896
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000779197592778335,
      "loss": 1.4958,
      "theoretical_loss": 3.7794734711648648,
      "tokens_seen": 754450432
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007791775325977935,
      "loss": 1.6287,
      "theoretical_loss": 3.7794408295819855,
      "tokens_seen": 754515968
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007791574724172517,
      "loss": 1.5711,
      "theoretical_loss": 3.779408191627952,
      "tokens_seen": 754581504
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007791374122367102,
      "loss": 1.5933,
      "theoretical_loss": 3.779375557302046,
      "tokens_seen": 754647040
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007791173520561685,
      "loss": 1.6062,
      "theoretical_loss": 3.779342926603549,
      "tokens_seen": 754712576
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007790972918756269,
      "loss": 1.5012,
      "theoretical_loss": 3.779310299531743,
      "tokens_seen": 754778112
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007790772316950853,
      "loss": 1.5742,
      "theoretical_loss": 3.7792776760859104,
      "tokens_seen": 754843648
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007790571715145437,
      "loss": 1.6304,
      "theoretical_loss": 3.7792450562653324,
      "tokens_seen": 754909184
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000779037111334002,
      "loss": 1.5054,
      "theoretical_loss": 3.7792124400692924,
      "tokens_seen": 754974720
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007790170511534605,
      "loss": 1.4889,
      "theoretical_loss": 3.7791798274970727,
      "tokens_seen": 755040256
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007789969909729187,
      "loss": 1.5854,
      "theoretical_loss": 3.7791472185479567,
      "tokens_seen": 755105792
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007789769307923772,
      "loss": 1.568,
      "theoretical_loss": 3.7791146132212274,
      "tokens_seen": 755171328
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007789568706118355,
      "loss": 1.5027,
      "theoretical_loss": 3.7790820115161674,
      "tokens_seen": 755236864
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007789368104312939,
      "loss": 1.5203,
      "theoretical_loss": 3.7790494134320607,
      "tokens_seen": 755302400
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007789167502507523,
      "loss": 1.6215,
      "theoretical_loss": 3.779016818968191,
      "tokens_seen": 755367936
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007788966900702107,
      "loss": 1.5524,
      "theoretical_loss": 3.778984228123842,
      "tokens_seen": 755433472
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000778876629889669,
      "loss": 1.683,
      "theoretical_loss": 3.778951640898298,
      "tokens_seen": 755499008
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007788565697091274,
      "loss": 1.6198,
      "theoretical_loss": 3.7789190572908433,
      "tokens_seen": 755564544
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007788365095285858,
      "loss": 1.5779,
      "theoretical_loss": 3.7788864773007624,
      "tokens_seen": 755630080
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007788164493480442,
      "loss": 1.5067,
      "theoretical_loss": 3.7788539009273405,
      "tokens_seen": 755695616
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007787963891675025,
      "loss": 1.6102,
      "theoretical_loss": 3.7788213281698617,
      "tokens_seen": 755761152
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000778776328986961,
      "loss": 1.5208,
      "theoretical_loss": 3.778788759027611,
      "tokens_seen": 755826688
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007787562688064192,
      "loss": 1.58,
      "theoretical_loss": 3.7787561934998743,
      "tokens_seen": 755892224
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007787362086258777,
      "loss": 1.5076,
      "theoretical_loss": 3.7787236315859376,
      "tokens_seen": 755957760
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000778716148445336,
      "loss": 1.5129,
      "theoretical_loss": 3.778691073285086,
      "tokens_seen": 756023296
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007786960882647944,
      "loss": 1.602,
      "theoretical_loss": 3.778658518596605,
      "tokens_seen": 756088832
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007786760280842528,
      "loss": 1.6815,
      "theoretical_loss": 3.7786259675197815,
      "tokens_seen": 756154368
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007786559679037112,
      "loss": 1.6083,
      "theoretical_loss": 3.7785934200539018,
      "tokens_seen": 756219904
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007786359077231695,
      "loss": 1.5432,
      "theoretical_loss": 3.7785608761982523,
      "tokens_seen": 756285440
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000778615847542628,
      "loss": 1.4662,
      "theoretical_loss": 3.778528335952119,
      "tokens_seen": 756350976
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007785957873620862,
      "loss": 1.643,
      "theoretical_loss": 3.7784957993147907,
      "tokens_seen": 756416512
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007785757271815447,
      "loss": 1.4588,
      "theoretical_loss": 3.7784632662855526,
      "tokens_seen": 756482048
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000778555667001003,
      "loss": 1.5764,
      "theoretical_loss": 3.7784307368636934,
      "tokens_seen": 756547584
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007785356068204614,
      "loss": 1.5294,
      "theoretical_loss": 3.7783982110484997,
      "tokens_seen": 756613120
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007785155466399198,
      "loss": 1.4983,
      "theoretical_loss": 3.77836568883926,
      "tokens_seen": 756678656
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007784954864593781,
      "loss": 1.5932,
      "theoretical_loss": 3.778333170235262,
      "tokens_seen": 756744192
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007784754262788365,
      "loss": 1.4658,
      "theoretical_loss": 3.7783006552357934,
      "tokens_seen": 756809728
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000778455366098295,
      "loss": 1.6186,
      "theoretical_loss": 3.7782681438401435,
      "tokens_seen": 756875264
    },
    {
      "epoch": 0.23,
      "objective/train/advantage_avg": 0.49409666657447815,
      "objective/train/docs_used": 280659,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6078691482543945,
      "objective/train/original_loss": 1.607869029045105,
      "objective/train/theoretical_loss": 3.7782356360476,
      "objective/train/tokens_used": 777400800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24458207190036774,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050653100013733,
      "objective/train/weighted_lm_loss": 1.6892685890197754,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7782356360476,
      "tokens_seen": 756940800
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007784353059177532,
      "loss": 1.6037,
      "theoretical_loss": 3.7782356360476,
      "tokens_seen": 756940800
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007784152457372117,
      "loss": 1.555,
      "theoretical_loss": 3.7782031318574525,
      "tokens_seen": 757006336
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007783951855566699,
      "loss": 1.6298,
      "theoretical_loss": 3.7781706312689893,
      "tokens_seen": 757071872
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007783751253761284,
      "loss": 1.4833,
      "theoretical_loss": 3.7781381342815,
      "tokens_seen": 757137408
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007783550651955868,
      "loss": 1.6445,
      "theoretical_loss": 3.7781056408942737,
      "tokens_seen": 757202944
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007783350050150451,
      "loss": 1.5297,
      "theoretical_loss": 3.7780731511066006,
      "tokens_seen": 757268480
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007783149448345035,
      "loss": 1.6313,
      "theoretical_loss": 3.7780406649177696,
      "tokens_seen": 757334016
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000778294884653962,
      "loss": 1.5451,
      "theoretical_loss": 3.778008182327071,
      "tokens_seen": 757399552
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007782748244734202,
      "loss": 1.5548,
      "theoretical_loss": 3.7779757033337953,
      "tokens_seen": 757465088
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007782547642928787,
      "loss": 1.5992,
      "theoretical_loss": 3.7779432279372323,
      "tokens_seen": 757530624
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000778234704112337,
      "loss": 1.5623,
      "theoretical_loss": 3.777910756136673,
      "tokens_seen": 757596160
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007782146439317954,
      "loss": 1.577,
      "theoretical_loss": 3.7778782879314083,
      "tokens_seen": 757661696
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007781945837512538,
      "loss": 1.5889,
      "theoretical_loss": 3.777845823320729,
      "tokens_seen": 757727232
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007781745235707121,
      "loss": 1.5361,
      "theoretical_loss": 3.7778133623039265,
      "tokens_seen": 757792768
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007781544633901705,
      "loss": 1.6085,
      "theoretical_loss": 3.777780904880292,
      "tokens_seen": 757858304
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007781344032096289,
      "loss": 1.5885,
      "theoretical_loss": 3.777748451049117,
      "tokens_seen": 757923840
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007781143430290872,
      "loss": 1.5951,
      "theoretical_loss": 3.7777160008096935,
      "tokens_seen": 757989376
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007780942828485457,
      "loss": 1.6325,
      "theoretical_loss": 3.777683554161313,
      "tokens_seen": 758054912
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000778074222668004,
      "loss": 1.6196,
      "theoretical_loss": 3.777651111103269,
      "tokens_seen": 758120448
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007780541624874624,
      "loss": 1.4281,
      "theoretical_loss": 3.777618671634852,
      "tokens_seen": 758185984
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007780341023069208,
      "loss": 1.5615,
      "theoretical_loss": 3.7775862357553565,
      "tokens_seen": 758251520
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007780140421263791,
      "loss": 1.5671,
      "theoretical_loss": 3.777553803464074,
      "tokens_seen": 758317056
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007779939819458375,
      "loss": 1.677,
      "theoretical_loss": 3.777521374760298,
      "tokens_seen": 758382592
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007779739217652959,
      "loss": 1.6055,
      "theoretical_loss": 3.7774889496433213,
      "tokens_seen": 758448128
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007779538615847542,
      "loss": 1.5473,
      "theoretical_loss": 3.777456528112438,
      "tokens_seen": 758513664
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007779338014042127,
      "loss": 1.6224,
      "theoretical_loss": 3.7774241101669417,
      "tokens_seen": 758579200
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000777913741223671,
      "loss": 1.7099,
      "theoretical_loss": 3.7773916958061253,
      "tokens_seen": 758644736
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007778936810431294,
      "loss": 1.6946,
      "theoretical_loss": 3.7773592850292834,
      "tokens_seen": 758710272
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007778736208625877,
      "loss": 1.7443,
      "theoretical_loss": 3.7773268778357103,
      "tokens_seen": 758775808
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007778535606820462,
      "loss": 1.5822,
      "theoretical_loss": 3.7772944742247,
      "tokens_seen": 758841344
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007778335005015045,
      "loss": 1.5798,
      "theoretical_loss": 3.777262074195548,
      "tokens_seen": 758906880
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007778134403209629,
      "loss": 1.4893,
      "theoretical_loss": 3.7772296777475476,
      "tokens_seen": 758972416
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007777933801404212,
      "loss": 1.6294,
      "theoretical_loss": 3.7771972848799953,
      "tokens_seen": 759037952
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007777733199598796,
      "loss": 1.6216,
      "theoretical_loss": 3.7771648955921853,
      "tokens_seen": 759103488
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007777532597793381,
      "loss": 1.6773,
      "theoretical_loss": 3.777132509883413,
      "tokens_seen": 759169024
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007777331995987964,
      "loss": 1.4738,
      "theoretical_loss": 3.777100127752975,
      "tokens_seen": 759234560
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007777131394182548,
      "loss": 1.6619,
      "theoretical_loss": 3.7770677492001665,
      "tokens_seen": 759300096
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007776930792377132,
      "loss": 1.6118,
      "theoretical_loss": 3.7770353742242833,
      "tokens_seen": 759365632
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007776730190571716,
      "loss": 1.624,
      "theoretical_loss": 3.7770030028246215,
      "tokens_seen": 759431168
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007776529588766299,
      "loss": 1.6658,
      "theoretical_loss": 3.7769706350004784,
      "tokens_seen": 759496704
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007776328986960884,
      "loss": 1.5252,
      "theoretical_loss": 3.776938270751149,
      "tokens_seen": 759562240
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007776128385155466,
      "loss": 1.5499,
      "theoretical_loss": 3.7769059100759317,
      "tokens_seen": 759627776
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007775927783350051,
      "loss": 1.5684,
      "theoretical_loss": 3.7768735529741226,
      "tokens_seen": 759693312
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007775727181544634,
      "loss": 1.492,
      "theoretical_loss": 3.7768411994450193,
      "tokens_seen": 759758848
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007775526579739218,
      "loss": 1.458,
      "theoretical_loss": 3.776808849487919,
      "tokens_seen": 759824384
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007775325977933802,
      "loss": 1.5818,
      "theoretical_loss": 3.7767765031021194,
      "tokens_seen": 759889920
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007775125376128385,
      "loss": 1.6171,
      "theoretical_loss": 3.776744160286918,
      "tokens_seen": 759955456
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007774924774322969,
      "loss": 1.6152,
      "theoretical_loss": 3.776711821041613,
      "tokens_seen": 760020992
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007774724172517554,
      "loss": 1.6122,
      "theoretical_loss": 3.776679485365503,
      "tokens_seen": 760086528
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007774523570712136,
      "loss": 1.736,
      "theoretical_loss": 3.7766471532578856,
      "tokens_seen": 760152064
    },
    {
      "epoch": 0.23,
      "objective/train/advantage_avg": 0.4914008378982544,
      "objective/train/docs_used": 281753,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5565385818481445,
      "objective/train/original_loss": 1.5565383434295654,
      "objective/train/theoretical_loss": 3.77661482471806,
      "objective/train/tokens_used": 780677600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24271810054779053,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050373911857605,
      "objective/train/weighted_lm_loss": 1.6349278688430786,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.77661482471806,
      "tokens_seen": 760217600
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007774322968906721,
      "loss": 1.5703,
      "theoretical_loss": 3.77661482471806,
      "tokens_seen": 760217600
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007774122367101303,
      "loss": 1.5262,
      "theoretical_loss": 3.7765824997453246,
      "tokens_seen": 760283136
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007773921765295888,
      "loss": 1.755,
      "theoretical_loss": 3.7765501783389785,
      "tokens_seen": 760348672
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007773721163490472,
      "loss": 1.5408,
      "theoretical_loss": 3.7765178604983216,
      "tokens_seen": 760414208
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007773520561685055,
      "loss": 1.6734,
      "theoretical_loss": 3.776485546222651,
      "tokens_seen": 760479744
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007773319959879639,
      "loss": 1.5087,
      "theoretical_loss": 3.776453235511269,
      "tokens_seen": 760545280
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007773119358074224,
      "loss": 1.6407,
      "theoretical_loss": 3.776420928363475,
      "tokens_seen": 760610816
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007772918756268806,
      "loss": 1.5477,
      "theoretical_loss": 3.7763886247785665,
      "tokens_seen": 760676352
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007772718154463391,
      "loss": 1.6531,
      "theoretical_loss": 3.776356324755847,
      "tokens_seen": 760741888
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007772517552657973,
      "loss": 1.4913,
      "theoretical_loss": 3.776324028294614,
      "tokens_seen": 760807424
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007772316950852558,
      "loss": 1.5826,
      "theoretical_loss": 3.77629173539417,
      "tokens_seen": 760872960
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007772116349047142,
      "loss": 1.6504,
      "theoretical_loss": 3.776259446053815,
      "tokens_seen": 760938496
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007771915747241725,
      "loss": 1.4761,
      "theoretical_loss": 3.7762271602728497,
      "tokens_seen": 761004032
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007771715145436309,
      "loss": 1.5062,
      "theoretical_loss": 3.7761948780505756,
      "tokens_seen": 761069568
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007771514543630893,
      "loss": 1.5518,
      "theoretical_loss": 3.776162599386294,
      "tokens_seen": 761135104
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007771313941825476,
      "loss": 1.639,
      "theoretical_loss": 3.7761303242793067,
      "tokens_seen": 761200640
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007771113340020061,
      "loss": 1.6436,
      "theoretical_loss": 3.7760980527289156,
      "tokens_seen": 761266176
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007770912738214644,
      "loss": 1.517,
      "theoretical_loss": 3.7760657847344214,
      "tokens_seen": 761331712
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007770712136409228,
      "loss": 1.5413,
      "theoretical_loss": 3.776033520295128,
      "tokens_seen": 761397248
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007770511534603812,
      "loss": 1.5092,
      "theoretical_loss": 3.7760012594103367,
      "tokens_seen": 761462784
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007770310932798395,
      "loss": 1.5427,
      "theoretical_loss": 3.77596900207935,
      "tokens_seen": 761528320
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007770110330992979,
      "loss": 1.4261,
      "theoretical_loss": 3.7759367483014707,
      "tokens_seen": 761593856
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007769909729187563,
      "loss": 1.4987,
      "theoretical_loss": 3.7759044980760024,
      "tokens_seen": 761659392
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007769709127382146,
      "loss": 1.5964,
      "theoretical_loss": 3.775872251402247,
      "tokens_seen": 761724928
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007769508525576731,
      "loss": 1.6302,
      "theoretical_loss": 3.775840008279509,
      "tokens_seen": 761790464
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007769307923771314,
      "loss": 1.586,
      "theoretical_loss": 3.7758077687070912,
      "tokens_seen": 761856000
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007769107321965898,
      "loss": 1.5466,
      "theoretical_loss": 3.775775532684298,
      "tokens_seen": 761921536
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007768906720160481,
      "loss": 1.558,
      "theoretical_loss": 3.775743300210433,
      "tokens_seen": 761987072
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007768706118355066,
      "loss": 1.6592,
      "theoretical_loss": 3.7757110712847997,
      "tokens_seen": 762052608
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007768505516549649,
      "loss": 1.6957,
      "theoretical_loss": 3.775678845906703,
      "tokens_seen": 762118144
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007768304914744233,
      "loss": 1.5502,
      "theoretical_loss": 3.7756466240754474,
      "tokens_seen": 762183680
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007768104312938816,
      "loss": 1.5597,
      "theoretical_loss": 3.7756144057903374,
      "tokens_seen": 762249216
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00077679037111334,
      "loss": 1.5967,
      "theoretical_loss": 3.775582191050678,
      "tokens_seen": 762314752
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007767703109327984,
      "loss": 1.5695,
      "theoretical_loss": 3.775549979855774,
      "tokens_seen": 762380288
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007767502507522568,
      "loss": 1.6352,
      "theoretical_loss": 3.775517772204932,
      "tokens_seen": 762445824
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007767301905717151,
      "loss": 1.5758,
      "theoretical_loss": 3.775485568097455,
      "tokens_seen": 762511360
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007767101303911736,
      "loss": 1.5208,
      "theoretical_loss": 3.775453367532651,
      "tokens_seen": 762576896
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007766900702106319,
      "loss": 1.5151,
      "theoretical_loss": 3.7754211705098246,
      "tokens_seen": 762642432
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007766700100300903,
      "loss": 1.6954,
      "theoretical_loss": 3.775388977028283,
      "tokens_seen": 762707968
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007766499498495486,
      "loss": 1.5447,
      "theoretical_loss": 3.7753567870873317,
      "tokens_seen": 762773504
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000776629889669007,
      "loss": 1.5751,
      "theoretical_loss": 3.775324600686276,
      "tokens_seen": 762839040
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007766098294884654,
      "loss": 1.5869,
      "theoretical_loss": 3.775292417824425,
      "tokens_seen": 762904576
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007765897693079238,
      "loss": 1.6638,
      "theoretical_loss": 3.775260238501084,
      "tokens_seen": 762970112
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007765697091273821,
      "loss": 1.451,
      "theoretical_loss": 3.77522806271556,
      "tokens_seen": 763035648
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007765496489468406,
      "loss": 1.5693,
      "theoretical_loss": 3.7751958904671614,
      "tokens_seen": 763101184
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007765295887662988,
      "loss": 1.5979,
      "theoretical_loss": 3.7751637217551943,
      "tokens_seen": 763166720
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007765095285857573,
      "loss": 1.6288,
      "theoretical_loss": 3.775131556578967,
      "tokens_seen": 763232256
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007764894684052157,
      "loss": 1.5196,
      "theoretical_loss": 3.7750993949377865,
      "tokens_seen": 763297792
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000776469408224674,
      "loss": 1.6061,
      "theoretical_loss": 3.7750672368309623,
      "tokens_seen": 763363328
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007764493480441324,
      "loss": 1.7025,
      "theoretical_loss": 3.7750350822578014,
      "tokens_seen": 763428864
    },
    {
      "epoch": 0.23,
      "objective/train/advantage_avg": 0.4942236542701721,
      "objective/train/docs_used": 282838,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4019079208374023,
      "objective/train/original_loss": 1.4019079208374023,
      "objective/train/theoretical_loss": 3.775002931217613,
      "objective/train/tokens_used": 783954400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24450983107089996,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506654977798462,
      "objective/train/weighted_lm_loss": 1.4728753566741943,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.775002931217613,
      "tokens_seen": 763494400
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007764292878635907,
      "loss": 1.5848,
      "theoretical_loss": 3.775002931217613,
      "tokens_seen": 763494400
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007764092276830491,
      "loss": 1.6831,
      "theoretical_loss": 3.7749707837097057,
      "tokens_seen": 763559936
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007763891675025076,
      "loss": 1.6031,
      "theoretical_loss": 3.7749386397333873,
      "tokens_seen": 763625472
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007763691073219658,
      "loss": 1.5377,
      "theoretical_loss": 3.774906499287968,
      "tokens_seen": 763691008
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007763490471414243,
      "loss": 1.5972,
      "theoretical_loss": 3.7748743623727554,
      "tokens_seen": 763756544
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007763289869608827,
      "loss": 1.6002,
      "theoretical_loss": 3.774842228987061,
      "tokens_seen": 763822080
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000776308926780341,
      "loss": 1.5737,
      "theoretical_loss": 3.774810099130193,
      "tokens_seen": 763887616
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007762888665997994,
      "loss": 1.5427,
      "theoretical_loss": 3.7747779728014614,
      "tokens_seen": 763953152
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007762688064192577,
      "loss": 1.613,
      "theoretical_loss": 3.7747458500001763,
      "tokens_seen": 764018688
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007762487462387162,
      "loss": 1.5593,
      "theoretical_loss": 3.7747137307256478,
      "tokens_seen": 764084224
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007762286860581746,
      "loss": 1.6187,
      "theoretical_loss": 3.7746816149771862,
      "tokens_seen": 764149760
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007762086258776329,
      "loss": 1.6038,
      "theoretical_loss": 3.774649502754102,
      "tokens_seen": 764215296
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007761885656970913,
      "loss": 1.6789,
      "theoretical_loss": 3.7746173940557064,
      "tokens_seen": 764280832
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007761685055165497,
      "loss": 1.3961,
      "theoretical_loss": 3.77458528888131,
      "tokens_seen": 764346368
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000776148445336008,
      "loss": 1.556,
      "theoretical_loss": 3.7745531872302234,
      "tokens_seen": 764411904
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007761283851554665,
      "loss": 1.5464,
      "theoretical_loss": 3.774521089101759,
      "tokens_seen": 764477440
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007761083249749248,
      "loss": 1.5297,
      "theoretical_loss": 3.7744889944952273,
      "tokens_seen": 764542976
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007760882647943832,
      "loss": 1.6049,
      "theoretical_loss": 3.774456903409941,
      "tokens_seen": 764608512
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007760682046138416,
      "loss": 1.6316,
      "theoretical_loss": 3.774424815845211,
      "tokens_seen": 764674048
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007760481444333,
      "loss": 1.6614,
      "theoretical_loss": 3.7743927318003503,
      "tokens_seen": 764739584
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007760280842527583,
      "loss": 1.5766,
      "theoretical_loss": 3.7743606512746704,
      "tokens_seen": 764805120
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007760080240722167,
      "loss": 1.5469,
      "theoretical_loss": 3.774328574267485,
      "tokens_seen": 764870656
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000775987963891675,
      "loss": 1.5127,
      "theoretical_loss": 3.774296500778105,
      "tokens_seen": 764936192
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007759679037111335,
      "loss": 1.6007,
      "theoretical_loss": 3.774264430805845,
      "tokens_seen": 765001728
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007759478435305918,
      "loss": 1.6396,
      "theoretical_loss": 3.7742323643500164,
      "tokens_seen": 765067264
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007759277833500502,
      "loss": 1.5566,
      "theoretical_loss": 3.774200301409934,
      "tokens_seen": 765132800
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007759077231695085,
      "loss": 1.5469,
      "theoretical_loss": 3.77416824198491,
      "tokens_seen": 765198336
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000775887662988967,
      "loss": 1.4872,
      "theoretical_loss": 3.774136186074259,
      "tokens_seen": 765263872
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007758676028084253,
      "loss": 1.4994,
      "theoretical_loss": 3.7741041336772945,
      "tokens_seen": 765329408
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007758475426278837,
      "loss": 1.594,
      "theoretical_loss": 3.7740720847933305,
      "tokens_seen": 765394944
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000775827482447342,
      "loss": 1.5735,
      "theoretical_loss": 3.7740400394216813,
      "tokens_seen": 765460480
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007758074222668004,
      "loss": 1.6261,
      "theoretical_loss": 3.774007997561661,
      "tokens_seen": 765526016
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007757873620862588,
      "loss": 1.6801,
      "theoretical_loss": 3.773975959212585,
      "tokens_seen": 765591552
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007757673019057172,
      "loss": 1.5897,
      "theoretical_loss": 3.773943924373767,
      "tokens_seen": 765657088
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007757472417251755,
      "loss": 1.6527,
      "theoretical_loss": 3.7739118930445223,
      "tokens_seen": 765722624
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000775727181544634,
      "loss": 1.426,
      "theoretical_loss": 3.7738798652241665,
      "tokens_seen": 765788160
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007757071213640923,
      "loss": 1.7207,
      "theoretical_loss": 3.773847840912015,
      "tokens_seen": 765853696
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007756870611835507,
      "loss": 1.6542,
      "theoretical_loss": 3.773815820107383,
      "tokens_seen": 765919232
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000775667001003009,
      "loss": 1.6583,
      "theoretical_loss": 3.7737838028095867,
      "tokens_seen": 765984768
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007756469408224674,
      "loss": 1.5591,
      "theoretical_loss": 3.7737517890179415,
      "tokens_seen": 766050304
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007756268806419258,
      "loss": 1.5703,
      "theoretical_loss": 3.7737197787317642,
      "tokens_seen": 766115840
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007756068204613842,
      "loss": 1.608,
      "theoretical_loss": 3.77368777195037,
      "tokens_seen": 766181376
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007755867602808425,
      "loss": 1.518,
      "theoretical_loss": 3.773655768673077,
      "tokens_seen": 766246912
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000775566700100301,
      "loss": 1.5365,
      "theoretical_loss": 3.773623768899201,
      "tokens_seen": 766312448
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007755466399197592,
      "loss": 1.6765,
      "theoretical_loss": 3.7735917726280586,
      "tokens_seen": 766377984
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007755265797392177,
      "loss": 1.5477,
      "theoretical_loss": 3.7735597798589673,
      "tokens_seen": 766443520
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007755065195586761,
      "loss": 1.7284,
      "theoretical_loss": 3.7735277905912445,
      "tokens_seen": 766509056
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007754864593781344,
      "loss": 1.5645,
      "theoretical_loss": 3.7734958048242078,
      "tokens_seen": 766574592
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007754663991975928,
      "loss": 1.5891,
      "theoretical_loss": 3.773463822557175,
      "tokens_seen": 766640128
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007754463390170511,
      "loss": 1.6468,
      "theoretical_loss": 3.7734318437894627,
      "tokens_seen": 766705664
    },
    {
      "epoch": 0.23,
      "objective/train/advantage_avg": 0.49038955569267273,
      "objective/train/docs_used": 283988,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4985244274139404,
      "objective/train/original_loss": 1.4985241889953613,
      "objective/train/theoretical_loss": 3.773399868520391,
      "objective/train/tokens_used": 787231200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24098964035511017,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502640008926392,
      "objective/train/weighted_lm_loss": 1.5738970041275024,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.773399868520391,
      "tokens_seen": 766771200
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007754262788365095,
      "loss": 1.526,
      "theoretical_loss": 3.773399868520391,
      "tokens_seen": 766771200
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000775406218655968,
      "loss": 1.6056,
      "theoretical_loss": 3.773367896749276,
      "tokens_seen": 766836736
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007753861584754262,
      "loss": 1.497,
      "theoretical_loss": 3.773335928475438,
      "tokens_seen": 766902272
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007753660982948847,
      "loss": 1.6116,
      "theoretical_loss": 3.7733039636981944,
      "tokens_seen": 766967808
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007753460381143431,
      "loss": 1.6268,
      "theoretical_loss": 3.7732720024168644,
      "tokens_seen": 767033344
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007753259779338014,
      "loss": 1.5227,
      "theoretical_loss": 3.773240044630768,
      "tokens_seen": 767098880
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007753059177532598,
      "loss": 1.5316,
      "theoretical_loss": 3.7732080903392227,
      "tokens_seen": 767164416
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007752858575727181,
      "loss": 1.6101,
      "theoretical_loss": 3.7731761395415484,
      "tokens_seen": 767229952
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007752657973921765,
      "loss": 1.5741,
      "theoretical_loss": 3.773144192237065,
      "tokens_seen": 767295488
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000775245737211635,
      "loss": 1.5869,
      "theoretical_loss": 3.773112248425093,
      "tokens_seen": 767361024
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007752256770310932,
      "loss": 1.4962,
      "theoretical_loss": 3.773080308104951,
      "tokens_seen": 767426560
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007752056168505517,
      "loss": 1.6621,
      "theoretical_loss": 3.77304837127596,
      "tokens_seen": 767492096
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00077518555667001,
      "loss": 1.578,
      "theoretical_loss": 3.7730164379374402,
      "tokens_seen": 767557632
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007751654964894684,
      "loss": 1.5545,
      "theoretical_loss": 3.7729845080887117,
      "tokens_seen": 767623168
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007751454363089268,
      "loss": 1.5694,
      "theoretical_loss": 3.7729525817290956,
      "tokens_seen": 767688704
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007751253761283852,
      "loss": 1.5779,
      "theoretical_loss": 3.772920658857913,
      "tokens_seen": 767754240
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007751053159478435,
      "loss": 1.5634,
      "theoretical_loss": 3.772888739474485,
      "tokens_seen": 767819776
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000775085255767302,
      "loss": 1.5747,
      "theoretical_loss": 3.7728568235781323,
      "tokens_seen": 767885312
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007750651955867602,
      "loss": 1.5485,
      "theoretical_loss": 3.7728249111681773,
      "tokens_seen": 767950848
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007750451354062187,
      "loss": 1.5458,
      "theoretical_loss": 3.7727930022439407,
      "tokens_seen": 768016384
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000775025075225677,
      "loss": 1.5263,
      "theoretical_loss": 3.772761096804745,
      "tokens_seen": 768081920
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007750050150451354,
      "loss": 1.526,
      "theoretical_loss": 3.772729194849912,
      "tokens_seen": 768147456
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007749849548645938,
      "loss": 1.5331,
      "theoretical_loss": 3.7726972963787646,
      "tokens_seen": 768212992
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007749648946840522,
      "loss": 1.5721,
      "theoretical_loss": 3.7726654013906242,
      "tokens_seen": 768278528
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007749448345035105,
      "loss": 1.6389,
      "theoretical_loss": 3.7726335098848143,
      "tokens_seen": 768344064
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007749247743229689,
      "loss": 1.6416,
      "theoretical_loss": 3.772601621860657,
      "tokens_seen": 768409600
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007749047141424272,
      "loss": 1.5411,
      "theoretical_loss": 3.772569737317476,
      "tokens_seen": 768475136
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007748846539618857,
      "loss": 1.5449,
      "theoretical_loss": 3.7725378562545937,
      "tokens_seen": 768540672
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000774864593781344,
      "loss": 1.5285,
      "theoretical_loss": 3.772505978671334,
      "tokens_seen": 768606208
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007748445336008024,
      "loss": 1.5871,
      "theoretical_loss": 3.7724741045670207,
      "tokens_seen": 768671744
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007748244734202607,
      "loss": 1.6613,
      "theoretical_loss": 3.7724422339409776,
      "tokens_seen": 768737280
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007748044132397192,
      "loss": 1.699,
      "theoretical_loss": 3.7724103667925277,
      "tokens_seen": 768802816
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007747843530591775,
      "loss": 1.614,
      "theoretical_loss": 3.772378503120996,
      "tokens_seen": 768868352
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007747642928786359,
      "loss": 1.485,
      "theoretical_loss": 3.7723466429257067,
      "tokens_seen": 768933888
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007747442326980944,
      "loss": 1.7294,
      "theoretical_loss": 3.772314786205984,
      "tokens_seen": 768999424
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007747241725175527,
      "loss": 1.6087,
      "theoretical_loss": 3.772282932961153,
      "tokens_seen": 769064960
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007747041123370111,
      "loss": 1.5298,
      "theoretical_loss": 3.7722510831905387,
      "tokens_seen": 769130496
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007746840521564694,
      "loss": 1.593,
      "theoretical_loss": 3.7722192368934655,
      "tokens_seen": 769196032
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007746639919759278,
      "loss": 1.6192,
      "theoretical_loss": 3.772187394069259,
      "tokens_seen": 769261568
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007746439317953862,
      "loss": 1.6289,
      "theoretical_loss": 3.772155554717245,
      "tokens_seen": 769327104
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007746238716148446,
      "loss": 1.5596,
      "theoretical_loss": 3.7721237188367494,
      "tokens_seen": 769392640
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007746038114343029,
      "loss": 1.5476,
      "theoretical_loss": 3.772091886427097,
      "tokens_seen": 769458176
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007745837512537614,
      "loss": 1.501,
      "theoretical_loss": 3.7720600574876144,
      "tokens_seen": 769523712
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007745636910732196,
      "loss": 1.5538,
      "theoretical_loss": 3.772028232017628,
      "tokens_seen": 769589248
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007745436308926781,
      "loss": 1.6311,
      "theoretical_loss": 3.771996410016464,
      "tokens_seen": 769654784
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007745235707121365,
      "loss": 1.4742,
      "theoretical_loss": 3.771964591483449,
      "tokens_seen": 769720320
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007745035105315948,
      "loss": 1.5959,
      "theoretical_loss": 3.7719327764179096,
      "tokens_seen": 769785856
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007744834503510532,
      "loss": 1.5834,
      "theoretical_loss": 3.771900964819173,
      "tokens_seen": 769851392
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007744633901705115,
      "loss": 1.4699,
      "theoretical_loss": 3.7718691566865665,
      "tokens_seen": 769916928
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007744433299899699,
      "loss": 1.5936,
      "theoretical_loss": 3.771837352019417,
      "tokens_seen": 769982464
    },
    {
      "epoch": 0.23,
      "objective/train/advantage_avg": 0.49243083596229553,
      "objective/train/docs_used": 285055,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5287210941314697,
      "objective/train/original_loss": 1.5287210941314697,
      "objective/train/theoretical_loss": 3.7718055508170525,
      "objective/train/tokens_used": 790508000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24285641312599182,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504777431488037,
      "objective/train/weighted_lm_loss": 1.605863094329834,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7718055508170525,
      "tokens_seen": 770048000
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007744232698094284,
      "loss": 1.5967,
      "theoretical_loss": 3.7718055508170525,
      "tokens_seen": 770048000
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007744032096288866,
      "loss": 1.5852,
      "theoretical_loss": 3.7717737530788007,
      "tokens_seen": 770113536
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007743831494483451,
      "loss": 1.5249,
      "theoretical_loss": 3.7717419588039887,
      "tokens_seen": 770179072
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007743630892678035,
      "loss": 1.584,
      "theoretical_loss": 3.771710167991946,
      "tokens_seen": 770244608
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007743430290872618,
      "loss": 1.6257,
      "theoretical_loss": 3.7716783806419993,
      "tokens_seen": 770310144
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007743229689067202,
      "loss": 1.4659,
      "theoretical_loss": 3.7716465967534782,
      "tokens_seen": 770375680
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007743029087261785,
      "loss": 1.5641,
      "theoretical_loss": 3.7716148163257115,
      "tokens_seen": 770441216
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007742828485456369,
      "loss": 1.4458,
      "theoretical_loss": 3.7715830393580267,
      "tokens_seen": 770506752
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007742627883650954,
      "loss": 1.5398,
      "theoretical_loss": 3.7715512658497543,
      "tokens_seen": 770572288
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007742427281845536,
      "loss": 1.5322,
      "theoretical_loss": 3.7715194958002227,
      "tokens_seen": 770637824
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007742226680040121,
      "loss": 1.5313,
      "theoretical_loss": 3.7714877292087623,
      "tokens_seen": 770703360
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007742026078234704,
      "loss": 1.5633,
      "theoretical_loss": 3.7714559660747007,
      "tokens_seen": 770768896
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007741825476429288,
      "loss": 1.5348,
      "theoretical_loss": 3.77142420639737,
      "tokens_seen": 770834432
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007741624874623872,
      "loss": 1.6001,
      "theoretical_loss": 3.7713924501760983,
      "tokens_seen": 770899968
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007741424272818456,
      "loss": 1.5319,
      "theoretical_loss": 3.7713606974102167,
      "tokens_seen": 770965504
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007741223671013039,
      "loss": 1.6015,
      "theoretical_loss": 3.7713289480990557,
      "tokens_seen": 771031040
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007741023069207624,
      "loss": 1.6177,
      "theoretical_loss": 3.771297202241945,
      "tokens_seen": 771096576
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007740822467402206,
      "loss": 1.6149,
      "theoretical_loss": 3.7712654598382165,
      "tokens_seen": 771162112
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007740621865596791,
      "loss": 1.6909,
      "theoretical_loss": 3.7712337208872,
      "tokens_seen": 771227648
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007740421263791374,
      "loss": 1.4497,
      "theoretical_loss": 3.771201985388227,
      "tokens_seen": 771293184
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007740220661985958,
      "loss": 1.5794,
      "theoretical_loss": 3.771170253340629,
      "tokens_seen": 771358720
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007740020060180542,
      "loss": 1.5743,
      "theoretical_loss": 3.771138524743737,
      "tokens_seen": 771424256
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007739819458375126,
      "loss": 1.5666,
      "theoretical_loss": 3.7711067995968826,
      "tokens_seen": 771489792
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007739618856569709,
      "loss": 1.553,
      "theoretical_loss": 3.7710750778993987,
      "tokens_seen": 771555328
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007739418254764293,
      "loss": 1.6214,
      "theoretical_loss": 3.771043359650616,
      "tokens_seen": 771620864
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007739217652958876,
      "loss": 1.5657,
      "theoretical_loss": 3.7710116448498674,
      "tokens_seen": 771686400
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007739017051153461,
      "loss": 1.7227,
      "theoretical_loss": 3.770979933496485,
      "tokens_seen": 771751936
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007738816449348044,
      "loss": 1.5689,
      "theoretical_loss": 3.7709482255898017,
      "tokens_seen": 771817472
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007738615847542628,
      "loss": 1.5229,
      "theoretical_loss": 3.77091652112915,
      "tokens_seen": 771883008
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007738415245737211,
      "loss": 1.5584,
      "theoretical_loss": 3.770884820113863,
      "tokens_seen": 771948544
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007738214643931796,
      "loss": 1.6013,
      "theoretical_loss": 3.770853122543274,
      "tokens_seen": 772014080
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007738014042126379,
      "loss": 1.5301,
      "theoretical_loss": 3.770821428416716,
      "tokens_seen": 772079616
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007737813440320963,
      "loss": 1.586,
      "theoretical_loss": 3.770789737733522,
      "tokens_seen": 772145152
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007737612838515547,
      "loss": 1.5376,
      "theoretical_loss": 3.770758050493027,
      "tokens_seen": 772210688
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007737412236710131,
      "loss": 1.4015,
      "theoretical_loss": 3.770726366694564,
      "tokens_seen": 772276224
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007737211634904714,
      "loss": 1.7061,
      "theoretical_loss": 3.7706946863374666,
      "tokens_seen": 772341760
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007737011033099298,
      "loss": 1.5885,
      "theoretical_loss": 3.7706630094210705,
      "tokens_seen": 772407296
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007736810431293881,
      "loss": 1.4734,
      "theoretical_loss": 3.770631335944709,
      "tokens_seen": 772472832
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007736609829488466,
      "loss": 1.5977,
      "theoretical_loss": 3.7705996659077172,
      "tokens_seen": 772538368
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007736409227683049,
      "loss": 1.4527,
      "theoretical_loss": 3.770567999309429,
      "tokens_seen": 772603904
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007736208625877633,
      "loss": 1.6232,
      "theoretical_loss": 3.7705363361491813,
      "tokens_seen": 772669440
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007736008024072217,
      "loss": 1.4619,
      "theoretical_loss": 3.770504676426307,
      "tokens_seen": 772734976
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00077358074222668,
      "loss": 1.6466,
      "theoretical_loss": 3.770473020140143,
      "tokens_seen": 772800512
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007735606820461384,
      "loss": 1.5803,
      "theoretical_loss": 3.7704413672900245,
      "tokens_seen": 772866048
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007735406218655969,
      "loss": 1.6873,
      "theoretical_loss": 3.770409717875287,
      "tokens_seen": 772931584
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007735205616850551,
      "loss": 1.5716,
      "theoretical_loss": 3.7703780718952666,
      "tokens_seen": 772997120
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007735005015045136,
      "loss": 1.76,
      "theoretical_loss": 3.770346429349299,
      "tokens_seen": 773062656
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007734804413239718,
      "loss": 1.6539,
      "theoretical_loss": 3.7703147902367213,
      "tokens_seen": 773128192
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007734603811434303,
      "loss": 1.5515,
      "theoretical_loss": 3.770283154556869,
      "tokens_seen": 773193728
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007734403209628887,
      "loss": 1.4677,
      "theoretical_loss": 3.7702515223090796,
      "tokens_seen": 773259264
    },
    {
      "epoch": 0.23,
      "objective/train/advantage_avg": 0.4907318353652954,
      "objective/train/docs_used": 286255,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4055989980697632,
      "objective/train/original_loss": 1.4055989980697632,
      "objective/train/theoretical_loss": 3.7702198934926896,
      "objective/train/tokens_used": 793784800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24155806005001068,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503010749816895,
      "objective/train/weighted_lm_loss": 1.4761121273040771,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7702198934926896,
      "tokens_seen": 773324800
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000773420260782347,
      "loss": 1.5701,
      "theoretical_loss": 3.7702198934926896,
      "tokens_seen": 773324800
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007734002006018054,
      "loss": 1.5682,
      "theoretical_loss": 3.770188268107036,
      "tokens_seen": 773390336
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007733801404212639,
      "loss": 1.516,
      "theoretical_loss": 3.7701566461514555,
      "tokens_seen": 773455872
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007733600802407221,
      "loss": 1.6158,
      "theoretical_loss": 3.7701250276252862,
      "tokens_seen": 773521408
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007733400200601806,
      "loss": 1.5474,
      "theoretical_loss": 3.7700934125278653,
      "tokens_seen": 773586944
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007733199598796388,
      "loss": 1.6568,
      "theoretical_loss": 3.7700618008585316,
      "tokens_seen": 773652480
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007732998996990973,
      "loss": 1.6022,
      "theoretical_loss": 3.7700301926166215,
      "tokens_seen": 773718016
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007732798395185557,
      "loss": 1.6433,
      "theoretical_loss": 3.7699985878014735,
      "tokens_seen": 773783552
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000773259779338014,
      "loss": 1.619,
      "theoretical_loss": 3.7699669864124266,
      "tokens_seen": 773849088
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007732397191574725,
      "loss": 1.6331,
      "theoretical_loss": 3.7699353884488187,
      "tokens_seen": 773914624
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007732196589769308,
      "loss": 1.5911,
      "theoretical_loss": 3.7699037939099886,
      "tokens_seen": 773980160
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007731995987963892,
      "loss": 1.5263,
      "theoretical_loss": 3.769872202795275,
      "tokens_seen": 774045696
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007731795386158476,
      "loss": 1.518,
      "theoretical_loss": 3.769840615104018,
      "tokens_seen": 774111232
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000773159478435306,
      "loss": 1.729,
      "theoretical_loss": 3.769809030835555,
      "tokens_seen": 774176768
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007731394182547643,
      "loss": 1.5644,
      "theoretical_loss": 3.769777449989227,
      "tokens_seen": 774242304
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007731193580742228,
      "loss": 1.7005,
      "theoretical_loss": 3.7697458725643727,
      "tokens_seen": 774307840
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000773099297893681,
      "loss": 1.5545,
      "theoretical_loss": 3.7697142985603325,
      "tokens_seen": 774373376
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007730792377131395,
      "loss": 1.5215,
      "theoretical_loss": 3.7696827279764458,
      "tokens_seen": 774438912
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007730591775325978,
      "loss": 1.6622,
      "theoretical_loss": 3.7696511608120526,
      "tokens_seen": 774504448
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007730391173520562,
      "loss": 1.6212,
      "theoretical_loss": 3.769619597066494,
      "tokens_seen": 774569984
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007730190571715146,
      "loss": 1.6814,
      "theoretical_loss": 3.76958803673911,
      "tokens_seen": 774635520
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000772998996990973,
      "loss": 1.5643,
      "theoretical_loss": 3.7695564798292414,
      "tokens_seen": 774701056
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007729789368104313,
      "loss": 1.6888,
      "theoretical_loss": 3.7695249263362287,
      "tokens_seen": 774766592
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007729588766298897,
      "loss": 1.6246,
      "theoretical_loss": 3.7694933762594136,
      "tokens_seen": 774832128
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000772938816449348,
      "loss": 1.6038,
      "theoretical_loss": 3.7694618295981375,
      "tokens_seen": 774897664
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007729187562688065,
      "loss": 1.6144,
      "theoretical_loss": 3.7694302863517413,
      "tokens_seen": 774963200
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007728986960882648,
      "loss": 1.5243,
      "theoretical_loss": 3.769398746519566,
      "tokens_seen": 775028736
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007728786359077232,
      "loss": 1.6769,
      "theoretical_loss": 3.7693672101009543,
      "tokens_seen": 775094272
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007728585757271815,
      "loss": 1.4813,
      "theoretical_loss": 3.769335677095248,
      "tokens_seen": 775159808
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00077283851554664,
      "loss": 1.6112,
      "theoretical_loss": 3.76930414750179,
      "tokens_seen": 775225344
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007728184553660983,
      "loss": 1.7433,
      "theoretical_loss": 3.769272621319921,
      "tokens_seen": 775290880
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007727983951855567,
      "loss": 1.5418,
      "theoretical_loss": 3.7692410985489846,
      "tokens_seen": 775356416
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007727783350050151,
      "loss": 1.6268,
      "theoretical_loss": 3.769209579188323,
      "tokens_seen": 775421952
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0007727582748244735,
      "loss": 1.5555,
      "theoretical_loss": 3.7691780632372796,
      "tokens_seen": 775487488
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007727382146439318,
      "loss": 1.5233,
      "theoretical_loss": 3.769146550695197,
      "tokens_seen": 775553024
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007727181544633902,
      "loss": 1.656,
      "theoretical_loss": 3.769115041561419,
      "tokens_seen": 775618560
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007726980942828485,
      "loss": 1.5705,
      "theoretical_loss": 3.7690835358352883,
      "tokens_seen": 775684096
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000772678034102307,
      "loss": 1.5781,
      "theoretical_loss": 3.7690520335161493,
      "tokens_seen": 775749632
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007726579739217653,
      "loss": 1.7681,
      "theoretical_loss": 3.769020534603345,
      "tokens_seen": 775815168
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007726379137412237,
      "loss": 1.6432,
      "theoretical_loss": 3.76898903909622,
      "tokens_seen": 775880704
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007726178535606821,
      "loss": 1.5245,
      "theoretical_loss": 3.7689575469941183,
      "tokens_seen": 775946240
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007725977933801404,
      "loss": 1.4918,
      "theoretical_loss": 3.768926058296384,
      "tokens_seen": 776011776
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007725777331995988,
      "loss": 1.6071,
      "theoretical_loss": 3.768894573002362,
      "tokens_seen": 776077312
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007725576730190573,
      "loss": 1.6793,
      "theoretical_loss": 3.7688630911113967,
      "tokens_seen": 776142848
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007725376128385155,
      "loss": 1.6084,
      "theoretical_loss": 3.768831612622833,
      "tokens_seen": 776208384
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000772517552657974,
      "loss": 1.4955,
      "theoretical_loss": 3.768800137536016,
      "tokens_seen": 776273920
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007724974924774322,
      "loss": 1.5472,
      "theoretical_loss": 3.768768665850291,
      "tokens_seen": 776339456
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007724774322968907,
      "loss": 1.7013,
      "theoretical_loss": 3.7687371975650032,
      "tokens_seen": 776404992
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007724573721163491,
      "loss": 1.5474,
      "theoretical_loss": 3.7687057326794986,
      "tokens_seen": 776470528
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007724373119358074,
      "loss": 1.5775,
      "theoretical_loss": 3.768674271193123,
      "tokens_seen": 776536064
    },
    {
      "epoch": 0.24,
      "objective/train/advantage_avg": 0.494631826877594,
      "objective/train/docs_used": 287303,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5359113216400146,
      "objective/train/original_loss": 1.5359114408493042,
      "objective/train/theoretical_loss": 3.768642813105222,
      "objective/train/tokens_used": 797061600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24497190117835999,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050708532333374,
      "objective/train/weighted_lm_loss": 1.6136270761489868,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.768642813105222,
      "tokens_seen": 776601600
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007724172517552658,
      "loss": 1.6239,
      "theoretical_loss": 3.768642813105222,
      "tokens_seen": 776601600
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007723971915747243,
      "loss": 1.8133,
      "theoretical_loss": 3.768611358415142,
      "tokens_seen": 776667136
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007723771313941825,
      "loss": 1.5074,
      "theoretical_loss": 3.768579907122229,
      "tokens_seen": 776732672
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000772357071213641,
      "loss": 1.6204,
      "theoretical_loss": 3.76854845922583,
      "tokens_seen": 776798208
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007723370110330992,
      "loss": 1.6579,
      "theoretical_loss": 3.7685170147252913,
      "tokens_seen": 776863744
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007723169508525577,
      "loss": 1.5122,
      "theoretical_loss": 3.76848557361996,
      "tokens_seen": 776929280
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007722968906720161,
      "loss": 1.5331,
      "theoretical_loss": 3.768454135909183,
      "tokens_seen": 776994816
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007722768304914744,
      "loss": 1.5532,
      "theoretical_loss": 3.768422701592308,
      "tokens_seen": 777060352
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007722567703109328,
      "loss": 1.6846,
      "theoretical_loss": 3.7683912706686815,
      "tokens_seen": 777125888
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007722367101303912,
      "loss": 1.5966,
      "theoretical_loss": 3.7683598431376524,
      "tokens_seen": 777191424
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007722166499498495,
      "loss": 1.484,
      "theoretical_loss": 3.768328418998567,
      "tokens_seen": 777256960
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000772196589769308,
      "loss": 1.6011,
      "theoretical_loss": 3.7682969982507744,
      "tokens_seen": 777322496
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007721765295887662,
      "loss": 1.5699,
      "theoretical_loss": 3.768265580893622,
      "tokens_seen": 777388032
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007721564694082247,
      "loss": 1.6059,
      "theoretical_loss": 3.7682341669264585,
      "tokens_seen": 777453568
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000772136409227683,
      "loss": 1.5105,
      "theoretical_loss": 3.7682027563486327,
      "tokens_seen": 777519104
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007721163490471414,
      "loss": 1.656,
      "theoretical_loss": 3.7681713491594926,
      "tokens_seen": 777584640
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007720962888665998,
      "loss": 1.6417,
      "theoretical_loss": 3.7681399453583877,
      "tokens_seen": 777650176
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007720762286860582,
      "loss": 1.6236,
      "theoretical_loss": 3.768108544944666,
      "tokens_seen": 777715712
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007720561685055165,
      "loss": 1.5793,
      "theoretical_loss": 3.768077147917678,
      "tokens_seen": 777781248
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000772036108324975,
      "loss": 1.5542,
      "theoretical_loss": 3.7680457542767725,
      "tokens_seen": 777846784
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007720160481444333,
      "loss": 1.65,
      "theoretical_loss": 3.7680143640212993,
      "tokens_seen": 777912320
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007719959879638917,
      "loss": 1.5318,
      "theoretical_loss": 3.7679829771506075,
      "tokens_seen": 777977856
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00077197592778335,
      "loss": 1.4914,
      "theoretical_loss": 3.7679515936640477,
      "tokens_seen": 778043392
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007719558676028084,
      "loss": 1.6576,
      "theoretical_loss": 3.76792021356097,
      "tokens_seen": 778108928
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007719358074222668,
      "loss": 1.5573,
      "theoretical_loss": 3.767888836840724,
      "tokens_seen": 778174464
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007719157472417252,
      "loss": 1.4435,
      "theoretical_loss": 3.7678574635026614,
      "tokens_seen": 778240000
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007718956870611835,
      "loss": 1.4878,
      "theoretical_loss": 3.7678260935461316,
      "tokens_seen": 778305536
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007718756268806419,
      "loss": 1.4595,
      "theoretical_loss": 3.7677947269704863,
      "tokens_seen": 778371072
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007718555667001003,
      "loss": 1.4134,
      "theoretical_loss": 3.7677633637750763,
      "tokens_seen": 778436608
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007718355065195587,
      "loss": 1.5421,
      "theoretical_loss": 3.767732003959252,
      "tokens_seen": 778502144
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000771815446339017,
      "loss": 1.5785,
      "theoretical_loss": 3.767700647522366,
      "tokens_seen": 778567680
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007717953861584755,
      "loss": 1.5411,
      "theoretical_loss": 3.7676692944637695,
      "tokens_seen": 778633216
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007717753259779337,
      "loss": 1.5445,
      "theoretical_loss": 3.767637944782814,
      "tokens_seen": 778698752
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007717552657973922,
      "loss": 1.4042,
      "theoretical_loss": 3.767606598478851,
      "tokens_seen": 778764288
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007717352056168507,
      "loss": 1.464,
      "theoretical_loss": 3.7675752555512334,
      "tokens_seen": 778829824
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007717151454363089,
      "loss": 1.6121,
      "theoretical_loss": 3.7675439159993136,
      "tokens_seen": 778895360
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007716950852557674,
      "loss": 1.6141,
      "theoretical_loss": 3.7675125798224425,
      "tokens_seen": 778960896
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007716750250752257,
      "loss": 1.5378,
      "theoretical_loss": 3.7674812470199743,
      "tokens_seen": 779026432
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007716549648946841,
      "loss": 1.5599,
      "theoretical_loss": 3.7674499175912617,
      "tokens_seen": 779091968
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007716349047141425,
      "loss": 1.5268,
      "theoretical_loss": 3.767418591535657,
      "tokens_seen": 779157504
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007716148445336008,
      "loss": 1.5934,
      "theoretical_loss": 3.7673872688525134,
      "tokens_seen": 779223040
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007715947843530592,
      "loss": 1.6528,
      "theoretical_loss": 3.7673559495411846,
      "tokens_seen": 779288576
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007715747241725177,
      "loss": 1.4885,
      "theoretical_loss": 3.767324633601024,
      "tokens_seen": 779354112
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007715546639919759,
      "loss": 1.5787,
      "theoretical_loss": 3.767293321031385,
      "tokens_seen": 779419648
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007715346038114344,
      "loss": 1.5581,
      "theoretical_loss": 3.7672620118316225,
      "tokens_seen": 779485184
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007715145436308926,
      "loss": 1.5708,
      "theoretical_loss": 3.7672307060010892,
      "tokens_seen": 779550720
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007714944834503511,
      "loss": 1.647,
      "theoretical_loss": 3.7671994035391405,
      "tokens_seen": 779616256
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007714744232698095,
      "loss": 1.6241,
      "theoretical_loss": 3.7671681044451297,
      "tokens_seen": 779681792
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007714543630892678,
      "loss": 1.5851,
      "theoretical_loss": 3.767136808718412,
      "tokens_seen": 779747328
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007714343029087262,
      "loss": 1.6368,
      "theoretical_loss": 3.7671055163583422,
      "tokens_seen": 779812864
    },
    {
      "epoch": 0.24,
      "objective/train/advantage_avg": 0.4958833158016205,
      "objective/train/docs_used": 288522,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4711905717849731,
      "objective/train/original_loss": 1.4711906909942627,
      "objective/train/theoretical_loss": 3.767074227364275,
      "objective/train/tokens_used": 800338400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24610991775989532,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508395433425903,
      "objective/train/weighted_lm_loss": 1.545902967453003,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.767074227364275,
      "tokens_seen": 779878400
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007714142427281847,
      "loss": 1.6525,
      "theoretical_loss": 3.767074227364275,
      "tokens_seen": 779878400
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007713941825476429,
      "loss": 1.7193,
      "theoretical_loss": 3.767042941735566,
      "tokens_seen": 779943936
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007713741223671014,
      "loss": 1.6416,
      "theoretical_loss": 3.76701165947157,
      "tokens_seen": 780009472
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007713540621865596,
      "loss": 1.6083,
      "theoretical_loss": 3.766980380571643,
      "tokens_seen": 780075008
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007713340020060181,
      "loss": 1.5439,
      "theoretical_loss": 3.7669491050351396,
      "tokens_seen": 780140544
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007713139418254765,
      "loss": 1.5689,
      "theoretical_loss": 3.7669178328614166,
      "tokens_seen": 780206080
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007712938816449348,
      "loss": 1.4895,
      "theoretical_loss": 3.76688656404983,
      "tokens_seen": 780271616
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007712738214643932,
      "loss": 1.6185,
      "theoretical_loss": 3.7668552985997357,
      "tokens_seen": 780337152
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007712537612838516,
      "loss": 1.739,
      "theoretical_loss": 3.7668240365104895,
      "tokens_seen": 780402688
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007712337011033099,
      "loss": 1.6098,
      "theoretical_loss": 3.7667927777814487,
      "tokens_seen": 780468224
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007712136409227684,
      "loss": 1.5923,
      "theoretical_loss": 3.7667615224119704,
      "tokens_seen": 780533760
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007711935807422266,
      "loss": 1.5821,
      "theoretical_loss": 3.7667302704014105,
      "tokens_seen": 780599296
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007711735205616851,
      "loss": 1.6597,
      "theoretical_loss": 3.7666990217491265,
      "tokens_seen": 780664832
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007711534603811434,
      "loss": 1.5226,
      "theoretical_loss": 3.766667776454476,
      "tokens_seen": 780730368
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007711334002006018,
      "loss": 1.5693,
      "theoretical_loss": 3.766636534516816,
      "tokens_seen": 780795904
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007711133400200602,
      "loss": 1.6576,
      "theoretical_loss": 3.766605295935504,
      "tokens_seen": 780861440
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007710932798395186,
      "loss": 1.5351,
      "theoretical_loss": 3.7665740607098974,
      "tokens_seen": 780926976
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007710732196589769,
      "loss": 1.5765,
      "theoretical_loss": 3.766542828839355,
      "tokens_seen": 780992512
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007710531594784354,
      "loss": 1.5218,
      "theoretical_loss": 3.766511600323235,
      "tokens_seen": 781058048
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007710330992978937,
      "loss": 1.6331,
      "theoretical_loss": 3.7664803751608957,
      "tokens_seen": 781123584
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007710130391173521,
      "loss": 1.5247,
      "theoretical_loss": 3.7664491533516946,
      "tokens_seen": 781189120
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007709929789368104,
      "loss": 1.5119,
      "theoretical_loss": 3.7664179348949913,
      "tokens_seen": 781254656
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007709729187562688,
      "loss": 1.5394,
      "theoretical_loss": 3.7663867197901437,
      "tokens_seen": 781320192
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007709528585757272,
      "loss": 1.6244,
      "theoretical_loss": 3.7663555080365123,
      "tokens_seen": 781385728
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007709327983951856,
      "loss": 1.5667,
      "theoretical_loss": 3.766324299633455,
      "tokens_seen": 781451264
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007709127382146439,
      "loss": 1.5698,
      "theoretical_loss": 3.7662930945803317,
      "tokens_seen": 781516800
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007708926780341023,
      "loss": 1.4866,
      "theoretical_loss": 3.7662618928765017,
      "tokens_seen": 781582336
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007708726178535607,
      "loss": 1.5478,
      "theoretical_loss": 3.7662306945213255,
      "tokens_seen": 781647872
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007708525576730191,
      "loss": 1.7532,
      "theoretical_loss": 3.766199499514162,
      "tokens_seen": 781713408
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007708324974924774,
      "loss": 1.5569,
      "theoretical_loss": 3.7661683078543717,
      "tokens_seen": 781778944
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007708124373119359,
      "loss": 1.4561,
      "theoretical_loss": 3.7661371195413142,
      "tokens_seen": 781844480
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007707923771313941,
      "loss": 1.6013,
      "theoretical_loss": 3.7661059345743513,
      "tokens_seen": 781910016
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007707723169508526,
      "loss": 1.4959,
      "theoretical_loss": 3.7660747529528424,
      "tokens_seen": 781975552
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007707522567703109,
      "loss": 1.5822,
      "theoretical_loss": 3.766043574676149,
      "tokens_seen": 782041088
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007707321965897693,
      "loss": 1.5494,
      "theoretical_loss": 3.7660123997436314,
      "tokens_seen": 782106624
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007707121364092277,
      "loss": 1.5849,
      "theoretical_loss": 3.7659812281546516,
      "tokens_seen": 782172160
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007706920762286861,
      "loss": 1.6841,
      "theoretical_loss": 3.76595005990857,
      "tokens_seen": 782237696
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007706720160481444,
      "loss": 1.5538,
      "theoretical_loss": 3.765918895004748,
      "tokens_seen": 782303232
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007706519558676029,
      "loss": 1.575,
      "theoretical_loss": 3.7658877334425487,
      "tokens_seen": 782368768
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007706318956870611,
      "loss": 1.4637,
      "theoretical_loss": 3.765856575221332,
      "tokens_seen": 782434304
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007706118355065196,
      "loss": 1.4895,
      "theoretical_loss": 3.7658254203404615,
      "tokens_seen": 782499840
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000770591775325978,
      "loss": 1.6225,
      "theoretical_loss": 3.765794268799298,
      "tokens_seen": 782565376
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007705717151454363,
      "loss": 1.5017,
      "theoretical_loss": 3.7657631205972053,
      "tokens_seen": 782630912
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007705516549648947,
      "loss": 1.5668,
      "theoretical_loss": 3.765731975733545,
      "tokens_seen": 782696448
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000770531594784353,
      "loss": 1.5598,
      "theoretical_loss": 3.7657008342076796,
      "tokens_seen": 782761984
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007705115346038114,
      "loss": 1.4191,
      "theoretical_loss": 3.765669696018973,
      "tokens_seen": 782827520
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007704914744232699,
      "loss": 1.632,
      "theoretical_loss": 3.7656385611667873,
      "tokens_seen": 782893056
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007704714142427281,
      "loss": 1.6471,
      "theoretical_loss": 3.765607429650486,
      "tokens_seen": 782958592
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007704513540621866,
      "loss": 1.472,
      "theoretical_loss": 3.765576301469433,
      "tokens_seen": 783024128
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000770431293881645,
      "loss": 1.5309,
      "theoretical_loss": 3.765545176622991,
      "tokens_seen": 783089664
    },
    {
      "epoch": 0.24,
      "objective/train/advantage_avg": 0.49230560660362244,
      "objective/train/docs_used": 289672,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6103899478912354,
      "objective/train/original_loss": 1.6103899478912354,
      "objective/train/theoretical_loss": 3.7655140551105246,
      "objective/train/tokens_used": 803615200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24277178943157196,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504647493362427,
      "objective/train/weighted_lm_loss": 1.6917227506637573,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7655140551105246,
      "tokens_seen": 783155200
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007704112337011033,
      "loss": 1.5824,
      "theoretical_loss": 3.7655140551105246,
      "tokens_seen": 783155200
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007703911735205617,
      "loss": 1.6093,
      "theoretical_loss": 3.765482936931397,
      "tokens_seen": 783220736
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00077037111334002,
      "loss": 1.5474,
      "theoretical_loss": 3.7654518220849726,
      "tokens_seen": 783286272
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007703510531594784,
      "loss": 1.4728,
      "theoretical_loss": 3.7654207105706154,
      "tokens_seen": 783351808
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007703309929789369,
      "loss": 1.6066,
      "theoretical_loss": 3.765389602387691,
      "tokens_seen": 783417344
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007703109327983951,
      "loss": 1.6021,
      "theoretical_loss": 3.765358497535563,
      "tokens_seen": 783482880
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007702908726178536,
      "loss": 1.5336,
      "theoretical_loss": 3.7653273960135962,
      "tokens_seen": 783548416
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007702708124373119,
      "loss": 1.6364,
      "theoretical_loss": 3.7652962978211555,
      "tokens_seen": 783613952
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007702507522567703,
      "loss": 1.6868,
      "theoretical_loss": 3.7652652029576066,
      "tokens_seen": 783679488
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007702306920762288,
      "loss": 1.7103,
      "theoretical_loss": 3.7652341114223145,
      "tokens_seen": 783745024
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000770210631895687,
      "loss": 1.5303,
      "theoretical_loss": 3.765203023214645,
      "tokens_seen": 783810560
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007701905717151455,
      "loss": 1.5135,
      "theoretical_loss": 3.765171938333963,
      "tokens_seen": 783876096
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007701705115346038,
      "loss": 1.5167,
      "theoretical_loss": 3.7651408567796354,
      "tokens_seen": 783941632
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007701504513540622,
      "loss": 1.5696,
      "theoretical_loss": 3.765109778551027,
      "tokens_seen": 784007168
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007701303911735206,
      "loss": 1.5142,
      "theoretical_loss": 3.7650787036475055,
      "tokens_seen": 784072704
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000770110330992979,
      "loss": 1.3989,
      "theoretical_loss": 3.765047632068436,
      "tokens_seen": 784138240
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007700902708124373,
      "loss": 1.5794,
      "theoretical_loss": 3.765016563813185,
      "tokens_seen": 784203776
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007700702106318958,
      "loss": 1.6232,
      "theoretical_loss": 3.76498549888112,
      "tokens_seen": 784269312
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000770050150451354,
      "loss": 1.5979,
      "theoretical_loss": 3.764954437271608,
      "tokens_seen": 784334848
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007700300902708125,
      "loss": 1.4158,
      "theoretical_loss": 3.7649233789840153,
      "tokens_seen": 784400384
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007700100300902708,
      "loss": 1.4842,
      "theoretical_loss": 3.7648923240177097,
      "tokens_seen": 784465920
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007699899699097292,
      "loss": 1.4458,
      "theoretical_loss": 3.764861272372058,
      "tokens_seen": 784531456
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007699699097291876,
      "loss": 1.4449,
      "theoretical_loss": 3.7648302240464284,
      "tokens_seen": 784596992
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000769949849548646,
      "loss": 1.5584,
      "theoretical_loss": 3.7647991790401885,
      "tokens_seen": 784662528
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007699297893681043,
      "loss": 1.5224,
      "theoretical_loss": 3.764768137352706,
      "tokens_seen": 784728064
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007699097291875627,
      "loss": 1.5574,
      "theoretical_loss": 3.764737098983349,
      "tokens_seen": 784793600
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007698896690070211,
      "loss": 1.5137,
      "theoretical_loss": 3.764706063931486,
      "tokens_seen": 784859136
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007698696088264795,
      "loss": 1.6239,
      "theoretical_loss": 3.7646750321964855,
      "tokens_seen": 784924672
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007698495486459378,
      "loss": 1.5686,
      "theoretical_loss": 3.764644003777716,
      "tokens_seen": 784990208
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007698294884653963,
      "loss": 1.5062,
      "theoretical_loss": 3.7646129786745464,
      "tokens_seen": 785055744
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007698094282848545,
      "loss": 1.6362,
      "theoretical_loss": 3.764581956886345,
      "tokens_seen": 785121280
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000769789368104313,
      "loss": 1.5062,
      "theoretical_loss": 3.764550938412482,
      "tokens_seen": 785186816
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007697693079237713,
      "loss": 1.5079,
      "theoretical_loss": 3.764519923252326,
      "tokens_seen": 785252352
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007697492477432297,
      "loss": 1.5395,
      "theoretical_loss": 3.7644889114052464,
      "tokens_seen": 785317888
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007697291875626881,
      "loss": 1.728,
      "theoretical_loss": 3.7644579028706135,
      "tokens_seen": 785383424
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007697091273821465,
      "loss": 1.5752,
      "theoretical_loss": 3.764426897647797,
      "tokens_seen": 785448960
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007696890672016048,
      "loss": 1.4924,
      "theoretical_loss": 3.764395895736166,
      "tokens_seen": 785514496
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007696690070210633,
      "loss": 1.5323,
      "theoretical_loss": 3.7643648971350916,
      "tokens_seen": 785580032
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007696489468405215,
      "loss": 1.5276,
      "theoretical_loss": 3.764333901843944,
      "tokens_seen": 785645568
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00076962888665998,
      "loss": 1.4896,
      "theoretical_loss": 3.7643029098620935,
      "tokens_seen": 785711104
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007696088264794383,
      "loss": 1.5057,
      "theoretical_loss": 3.764271921188911,
      "tokens_seen": 785776640
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007695887662988967,
      "loss": 1.6802,
      "theoretical_loss": 3.764240935823767,
      "tokens_seen": 785842176
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007695687061183551,
      "loss": 1.5805,
      "theoretical_loss": 3.764209953766033,
      "tokens_seen": 785907712
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007695486459378134,
      "loss": 1.438,
      "theoretical_loss": 3.7641789750150796,
      "tokens_seen": 785973248
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007695285857572718,
      "loss": 1.4642,
      "theoretical_loss": 3.7641479995702793,
      "tokens_seen": 786038784
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007695085255767303,
      "loss": 1.5398,
      "theoretical_loss": 3.764117027431002,
      "tokens_seen": 786104320
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007694884653961885,
      "loss": 1.6588,
      "theoretical_loss": 3.7640860585966207,
      "tokens_seen": 786169856
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000769468405215647,
      "loss": 1.5324,
      "theoretical_loss": 3.7640550930665073,
      "tokens_seen": 786235392
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007694483450351054,
      "loss": 1.5083,
      "theoretical_loss": 3.764024130840033,
      "tokens_seen": 786300928
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007694282848545637,
      "loss": 1.5271,
      "theoretical_loss": 3.763993171916571,
      "tokens_seen": 786366464
    },
    {
      "epoch": 0.24,
      "objective/train/advantage_avg": 0.4925498962402344,
      "objective/train/docs_used": 290810,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4064291715621948,
      "objective/train/original_loss": 1.4064290523529053,
      "objective/train/theoretical_loss": 3.763962216295493,
      "objective/train/tokens_used": 806892000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24311783909797668,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504910945892334,
      "objective/train/weighted_lm_loss": 1.4774826765060425,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.763962216295493,
      "tokens_seen": 786432000
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007694082246740221,
      "loss": 1.5393,
      "theoretical_loss": 3.763962216295493,
      "tokens_seen": 786432000
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007693881644934804,
      "loss": 1.5168,
      "theoretical_loss": 3.763931263976172,
      "tokens_seen": 786497536
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007693681043129388,
      "loss": 1.5409,
      "theoretical_loss": 3.76390031495798,
      "tokens_seen": 786563072
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007693480441323973,
      "loss": 1.5231,
      "theoretical_loss": 3.7638693692402914,
      "tokens_seen": 786628608
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007693279839518555,
      "loss": 1.5602,
      "theoretical_loss": 3.7638384268224776,
      "tokens_seen": 786694144
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000769307923771314,
      "loss": 1.676,
      "theoretical_loss": 3.763807487703913,
      "tokens_seen": 786759680
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007692878635907723,
      "loss": 1.4986,
      "theoretical_loss": 3.7637765518839705,
      "tokens_seen": 786825216
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007692678034102307,
      "loss": 1.5583,
      "theoretical_loss": 3.7637456193620245,
      "tokens_seen": 786890752
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007692477432296891,
      "loss": 1.5269,
      "theoretical_loss": 3.7637146901374474,
      "tokens_seen": 786956288
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007692276830491474,
      "loss": 1.4849,
      "theoretical_loss": 3.763683764209614,
      "tokens_seen": 787021824
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007692076228686058,
      "loss": 1.6048,
      "theoretical_loss": 3.7636528415778994,
      "tokens_seen": 787087360
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007691875626880642,
      "loss": 1.6717,
      "theoretical_loss": 3.7636219222416756,
      "tokens_seen": 787152896
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007691675025075225,
      "loss": 1.6059,
      "theoretical_loss": 3.7635910062003193,
      "tokens_seen": 787218432
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000769147442326981,
      "loss": 1.5944,
      "theoretical_loss": 3.7635600934532034,
      "tokens_seen": 787283968
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007691273821464393,
      "loss": 1.6308,
      "theoretical_loss": 3.7635291839997036,
      "tokens_seen": 787349504
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007691073219658977,
      "loss": 1.4932,
      "theoretical_loss": 3.763498277839195,
      "tokens_seen": 787415040
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007690872617853561,
      "loss": 1.4809,
      "theoretical_loss": 3.7634673749710523,
      "tokens_seen": 787480576
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007690672016048145,
      "loss": 1.5467,
      "theoretical_loss": 3.763436475394651,
      "tokens_seen": 787546112
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007690471414242728,
      "loss": 1.6277,
      "theoretical_loss": 3.7634055791093664,
      "tokens_seen": 787611648
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007690270812437312,
      "loss": 1.4895,
      "theoretical_loss": 3.7633746861145747,
      "tokens_seen": 787677184
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007690070210631895,
      "loss": 1.4323,
      "theoretical_loss": 3.763343796409651,
      "tokens_seen": 787742720
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000768986960882648,
      "loss": 1.5547,
      "theoretical_loss": 3.763312909993972,
      "tokens_seen": 787808256
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007689669007021063,
      "loss": 1.6712,
      "theoretical_loss": 3.7632820268669134,
      "tokens_seen": 787873792
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007689468405215647,
      "loss": 1.3757,
      "theoretical_loss": 3.7632511470278516,
      "tokens_seen": 787939328
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000768926780341023,
      "loss": 1.5393,
      "theoretical_loss": 3.7632202704761637,
      "tokens_seen": 788004864
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007689067201604815,
      "loss": 1.4907,
      "theoretical_loss": 3.763189397211225,
      "tokens_seen": 788070400
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007688866599799398,
      "loss": 1.5503,
      "theoretical_loss": 3.7631585272324135,
      "tokens_seen": 788135936
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007688665997993982,
      "loss": 1.61,
      "theoretical_loss": 3.763127660539106,
      "tokens_seen": 788201472
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007688465396188565,
      "loss": 1.4616,
      "theoretical_loss": 3.7630967971306797,
      "tokens_seen": 788267008
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007688264794383149,
      "loss": 1.4953,
      "theoretical_loss": 3.7630659370065116,
      "tokens_seen": 788332544
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007688064192577733,
      "loss": 1.4957,
      "theoretical_loss": 3.76303508016598,
      "tokens_seen": 788398080
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007687863590772317,
      "loss": 1.6235,
      "theoretical_loss": 3.763004226608462,
      "tokens_seen": 788463616
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007687662988966901,
      "loss": 1.5846,
      "theoretical_loss": 3.762973376333335,
      "tokens_seen": 788529152
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007687462387161485,
      "loss": 1.5668,
      "theoretical_loss": 3.762942529339978,
      "tokens_seen": 788594688
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007687261785356069,
      "loss": 1.4758,
      "theoretical_loss": 3.7629116856277687,
      "tokens_seen": 788660224
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007687061183550652,
      "loss": 1.6209,
      "theoretical_loss": 3.7628808451960856,
      "tokens_seen": 788725760
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007686860581745237,
      "loss": 1.5973,
      "theoretical_loss": 3.7628500080443077,
      "tokens_seen": 788791296
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007686659979939819,
      "loss": 1.5865,
      "theoretical_loss": 3.7628191741718124,
      "tokens_seen": 788856832
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007686459378134404,
      "loss": 1.626,
      "theoretical_loss": 3.76278834357798,
      "tokens_seen": 788922368
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007686258776328987,
      "loss": 1.506,
      "theoretical_loss": 3.762757516262189,
      "tokens_seen": 788987904
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007686058174523571,
      "loss": 1.4974,
      "theoretical_loss": 3.7627266922238185,
      "tokens_seen": 789053440
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007685857572718155,
      "loss": 1.6281,
      "theoretical_loss": 3.762695871462248,
      "tokens_seen": 789118976
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007685656970912738,
      "loss": 1.5123,
      "theoretical_loss": 3.762665053976857,
      "tokens_seen": 789184512
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007685456369107322,
      "loss": 1.5924,
      "theoretical_loss": 3.762634239767025,
      "tokens_seen": 789250048
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007685255767301907,
      "loss": 1.4886,
      "theoretical_loss": 3.762603428832133,
      "tokens_seen": 789315584
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007685055165496489,
      "loss": 1.6031,
      "theoretical_loss": 3.762572621171559,
      "tokens_seen": 789381120
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007684854563691074,
      "loss": 1.4767,
      "theoretical_loss": 3.762541816784686,
      "tokens_seen": 789446656
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007684653961885658,
      "loss": 1.4767,
      "theoretical_loss": 3.7625110156708916,
      "tokens_seen": 789512192
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007684453360080241,
      "loss": 1.5101,
      "theoretical_loss": 3.7624802178295584,
      "tokens_seen": 789577728
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007684252758274825,
      "loss": 1.5328,
      "theoretical_loss": 3.7624494232600663,
      "tokens_seen": 789643264
    },
    {
      "epoch": 0.24,
      "objective/train/advantage_avg": 0.49356845021247864,
      "objective/train/docs_used": 292000,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6631910800933838,
      "objective/train/original_loss": 1.6631911993026733,
      "objective/train/theoretical_loss": 3.762418631961796,
      "objective/train/tokens_used": 810168800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2438865453004837,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505967140197754,
      "objective/train/weighted_lm_loss": 1.7472585439682007,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 3.762418631961796,
      "tokens_seen": 789708800
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007684052156469408,
      "loss": 1.6368,
      "theoretical_loss": 3.762418631961796,
      "tokens_seen": 789708800
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007683851554663992,
      "loss": 1.591,
      "theoretical_loss": 3.762387843934129,
      "tokens_seen": 789774336
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007683650952858577,
      "loss": 1.5279,
      "theoretical_loss": 3.762357059176447,
      "tokens_seen": 789839872
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007683450351053159,
      "loss": 1.6013,
      "theoretical_loss": 3.76232627768813,
      "tokens_seen": 789905408
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007683249749247744,
      "loss": 1.527,
      "theoretical_loss": 3.7622954994685616,
      "tokens_seen": 789970944
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007683049147442327,
      "loss": 1.5638,
      "theoretical_loss": 3.762264724517122,
      "tokens_seen": 790036480
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007682848545636911,
      "loss": 1.5854,
      "theoretical_loss": 3.762233952833193,
      "tokens_seen": 790102016
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007682647943831495,
      "loss": 1.5868,
      "theoretical_loss": 3.7622031844161574,
      "tokens_seen": 790167552
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007682447342026078,
      "loss": 1.6349,
      "theoretical_loss": 3.762172419265397,
      "tokens_seen": 790233088
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007682246740220662,
      "loss": 1.5443,
      "theoretical_loss": 3.7621416573802953,
      "tokens_seen": 790298624
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007682046138415246,
      "loss": 1.7027,
      "theoretical_loss": 3.7621108987602336,
      "tokens_seen": 790364160
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007681845536609829,
      "loss": 1.5699,
      "theoretical_loss": 3.762080143404595,
      "tokens_seen": 790429696
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007681644934804414,
      "loss": 1.5184,
      "theoretical_loss": 3.7620493913127633,
      "tokens_seen": 790495232
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007681444332998997,
      "loss": 1.567,
      "theoretical_loss": 3.7620186424841204,
      "tokens_seen": 790560768
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007681243731193581,
      "loss": 1.5399,
      "theoretical_loss": 3.76198789691805,
      "tokens_seen": 790626304
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007681043129388165,
      "loss": 1.5747,
      "theoretical_loss": 3.7619571546139356,
      "tokens_seen": 790691840
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007680842527582749,
      "loss": 1.7117,
      "theoretical_loss": 3.761926415571161,
      "tokens_seen": 790757376
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007680641925777332,
      "loss": 1.5964,
      "theoretical_loss": 3.7618956797891094,
      "tokens_seen": 790822912
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007680441323971916,
      "loss": 1.5564,
      "theoretical_loss": 3.7618649472671652,
      "tokens_seen": 790888448
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007680240722166499,
      "loss": 1.5697,
      "theoretical_loss": 3.7618342180047124,
      "tokens_seen": 790953984
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007680040120361084,
      "loss": 1.6758,
      "theoretical_loss": 3.7618034920011354,
      "tokens_seen": 791019520
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007679839518555667,
      "loss": 1.522,
      "theoretical_loss": 3.7617727692558187,
      "tokens_seen": 791085056
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007679638916750251,
      "loss": 1.5598,
      "theoretical_loss": 3.761742049768146,
      "tokens_seen": 791150592
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007679438314944834,
      "loss": 1.568,
      "theoretical_loss": 3.7617113335375034,
      "tokens_seen": 791216128
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007679237713139419,
      "loss": 1.6034,
      "theoretical_loss": 3.7616806205632747,
      "tokens_seen": 791281664
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007679037111334002,
      "loss": 1.6726,
      "theoretical_loss": 3.761649910844846,
      "tokens_seen": 791347200
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007678836509528586,
      "loss": 1.5397,
      "theoretical_loss": 3.761619204381602,
      "tokens_seen": 791412736
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000767863590772317,
      "loss": 1.5374,
      "theoretical_loss": 3.7615885011729278,
      "tokens_seen": 791478272
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007678435305917753,
      "loss": 1.5309,
      "theoretical_loss": 3.76155780121821,
      "tokens_seen": 791543808
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007678234704112337,
      "loss": 1.5569,
      "theoretical_loss": 3.7615271045168335,
      "tokens_seen": 791609344
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007678034102306921,
      "loss": 1.4615,
      "theoretical_loss": 3.7614964110681846,
      "tokens_seen": 791674880
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007677833500501504,
      "loss": 1.5546,
      "theoretical_loss": 3.7614657208716493,
      "tokens_seen": 791740416
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007677632898696089,
      "loss": 1.6533,
      "theoretical_loss": 3.761435033926614,
      "tokens_seen": 791805952
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007677432296890672,
      "loss": 1.5623,
      "theoretical_loss": 3.761404350232465,
      "tokens_seen": 791871488
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007677231695085256,
      "loss": 1.4596,
      "theoretical_loss": 3.761373669788589,
      "tokens_seen": 791937024
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000767703109327984,
      "loss": 1.5853,
      "theoretical_loss": 3.761342992594373,
      "tokens_seen": 792002560
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007676830491474423,
      "loss": 1.4637,
      "theoretical_loss": 3.7613123186492032,
      "tokens_seen": 792068096
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007676629889669007,
      "loss": 1.5143,
      "theoretical_loss": 3.7612816479524676,
      "tokens_seen": 792133632
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007676429287863591,
      "loss": 1.6328,
      "theoretical_loss": 3.7612509805035526,
      "tokens_seen": 792199168
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007676228686058174,
      "loss": 1.4724,
      "theoretical_loss": 3.7612203163018467,
      "tokens_seen": 792264704
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007676028084252759,
      "loss": 1.5678,
      "theoretical_loss": 3.7611896553467368,
      "tokens_seen": 792330240
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007675827482447341,
      "loss": 1.5655,
      "theoretical_loss": 3.7611589976376107,
      "tokens_seen": 792395776
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007675626880641926,
      "loss": 1.4869,
      "theoretical_loss": 3.761128343173856,
      "tokens_seen": 792461312
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000767542627883651,
      "loss": 1.5796,
      "theoretical_loss": 3.7610976919548618,
      "tokens_seen": 792526848
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007675225677031093,
      "loss": 1.4877,
      "theoretical_loss": 3.7610670439800153,
      "tokens_seen": 792592384
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007675025075225677,
      "loss": 1.5576,
      "theoretical_loss": 3.761036399248706,
      "tokens_seen": 792657920
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007674824473420262,
      "loss": 1.4931,
      "theoretical_loss": 3.7610057577603215,
      "tokens_seen": 792723456
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007674623871614844,
      "loss": 1.5777,
      "theoretical_loss": 3.760975119514251,
      "tokens_seen": 792788992
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007674423269809429,
      "loss": 1.5912,
      "theoretical_loss": 3.7609444845098836,
      "tokens_seen": 792854528
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007674222668004011,
      "loss": 1.5408,
      "theoretical_loss": 3.760913852746608,
      "tokens_seen": 792920064
    },
    {
      "epoch": 0.24,
      "objective/train/advantage_avg": 0.49038174748420715,
      "objective/train/docs_used": 293128,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.419954776763916,
      "objective/train/original_loss": 1.419954776763916,
      "objective/train/theoretical_loss": 3.7608832242238144,
      "objective/train/tokens_used": 813445600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24094732105731964,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502629280090332,
      "objective/train/weighted_lm_loss": 1.4913206100463867,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7608832242238144,
      "tokens_seen": 792985600
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007674022066198596,
      "loss": 1.6114,
      "theoretical_loss": 3.7608832242238144,
      "tokens_seen": 792985600
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000767382146439318,
      "loss": 1.572,
      "theoretical_loss": 3.7608525989408905,
      "tokens_seen": 793051136
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007673620862587763,
      "loss": 1.5256,
      "theoretical_loss": 3.7608219768972275,
      "tokens_seen": 793116672
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007673420260782347,
      "loss": 1.6743,
      "theoretical_loss": 3.7607913580922148,
      "tokens_seen": 793182208
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000767321965897693,
      "loss": 1.5397,
      "theoretical_loss": 3.7607607425252416,
      "tokens_seen": 793247744
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007673019057171514,
      "loss": 1.5917,
      "theoretical_loss": 3.7607301301956984,
      "tokens_seen": 793313280
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007672818455366099,
      "loss": 1.5805,
      "theoretical_loss": 3.7606995211029757,
      "tokens_seen": 793378816
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007672617853560682,
      "loss": 1.5235,
      "theoretical_loss": 3.7606689152464643,
      "tokens_seen": 793444352
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007672417251755266,
      "loss": 1.6795,
      "theoretical_loss": 3.7606383126255536,
      "tokens_seen": 793509888
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000767221664994985,
      "loss": 1.623,
      "theoretical_loss": 3.760607713239635,
      "tokens_seen": 793575424
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007672016048144433,
      "loss": 1.6713,
      "theoretical_loss": 3.7605771170880997,
      "tokens_seen": 793640960
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007671815446339018,
      "loss": 1.5916,
      "theoretical_loss": 3.760546524170339,
      "tokens_seen": 793706496
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007671614844533601,
      "loss": 1.5529,
      "theoretical_loss": 3.760515934485743,
      "tokens_seen": 793772032
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007671414242728185,
      "loss": 1.605,
      "theoretical_loss": 3.7604853480337033,
      "tokens_seen": 793837568
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007671213640922769,
      "loss": 1.6569,
      "theoretical_loss": 3.7604547648136126,
      "tokens_seen": 793903104
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007671013039117353,
      "loss": 1.5361,
      "theoretical_loss": 3.760424184824862,
      "tokens_seen": 793968640
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007670812437311936,
      "loss": 1.4949,
      "theoretical_loss": 3.760393608066843,
      "tokens_seen": 794034176
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000767061183550652,
      "loss": 1.61,
      "theoretical_loss": 3.7603630345389485,
      "tokens_seen": 794099712
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007670411233701103,
      "loss": 1.4037,
      "theoretical_loss": 3.7603324642405704,
      "tokens_seen": 794165248
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007670210631895688,
      "loss": 1.5885,
      "theoretical_loss": 3.7603018971711,
      "tokens_seen": 794230784
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007670010030090271,
      "loss": 1.5158,
      "theoretical_loss": 3.760271333329932,
      "tokens_seen": 794296320
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007669809428284855,
      "loss": 1.6156,
      "theoretical_loss": 3.7602407727164575,
      "tokens_seen": 794361856
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007669608826479438,
      "loss": 1.5157,
      "theoretical_loss": 3.76021021533007,
      "tokens_seen": 794427392
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007669408224674023,
      "loss": 1.7043,
      "theoretical_loss": 3.760179661170162,
      "tokens_seen": 794492928
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007669207622868606,
      "loss": 1.4807,
      "theoretical_loss": 3.7601491102361275,
      "tokens_seen": 794558464
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000766900702106319,
      "loss": 1.6377,
      "theoretical_loss": 3.760118562527359,
      "tokens_seen": 794624000
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007668806419257773,
      "loss": 1.541,
      "theoretical_loss": 3.760088018043252,
      "tokens_seen": 794689536
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007668605817452357,
      "loss": 1.5254,
      "theoretical_loss": 3.7600574767831976,
      "tokens_seen": 794755072
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007668405215646941,
      "loss": 1.5861,
      "theoretical_loss": 3.7600269387465914,
      "tokens_seen": 794820608
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007668204613841525,
      "loss": 1.5806,
      "theoretical_loss": 3.759996403932826,
      "tokens_seen": 794886144
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007668004012036108,
      "loss": 1.579,
      "theoretical_loss": 3.7599658723412976,
      "tokens_seen": 794951680
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007667803410230693,
      "loss": 1.597,
      "theoretical_loss": 3.759935343971399,
      "tokens_seen": 795017216
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007667602808425276,
      "loss": 1.5516,
      "theoretical_loss": 3.759904818822525,
      "tokens_seen": 795082752
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000766740220661986,
      "loss": 1.5316,
      "theoretical_loss": 3.759874296894071,
      "tokens_seen": 795148288
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007667201604814444,
      "loss": 1.5484,
      "theoretical_loss": 3.759843778185431,
      "tokens_seen": 795213824
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007667001003009027,
      "loss": 1.7136,
      "theoretical_loss": 3.7598132626960004,
      "tokens_seen": 795279360
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007666800401203611,
      "loss": 1.5863,
      "theoretical_loss": 3.759782750425175,
      "tokens_seen": 795344896
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007666599799398195,
      "loss": 1.5898,
      "theoretical_loss": 3.7597522413723485,
      "tokens_seen": 795410432
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007666399197592778,
      "loss": 1.5328,
      "theoretical_loss": 3.7597217355369184,
      "tokens_seen": 795475968
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007666198595787363,
      "loss": 1.549,
      "theoretical_loss": 3.7596912329182786,
      "tokens_seen": 795541504
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007665997993981945,
      "loss": 1.4882,
      "theoretical_loss": 3.759660733515826,
      "tokens_seen": 795607040
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000766579739217653,
      "loss": 1.601,
      "theoretical_loss": 3.759630237328957,
      "tokens_seen": 795672576
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007665596790371114,
      "loss": 1.6076,
      "theoretical_loss": 3.7595997443570663,
      "tokens_seen": 795738112
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007665396188565697,
      "loss": 1.5379,
      "theoretical_loss": 3.759569254599552,
      "tokens_seen": 795803648
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007665195586760281,
      "loss": 1.465,
      "theoretical_loss": 3.7595387680558088,
      "tokens_seen": 795869184
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007664994984954866,
      "loss": 1.5445,
      "theoretical_loss": 3.759508284725235,
      "tokens_seen": 795934720
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007664794383149448,
      "loss": 1.7383,
      "theoretical_loss": 3.7594778046072257,
      "tokens_seen": 796000256
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007664593781344033,
      "loss": 1.5825,
      "theoretical_loss": 3.759447327701179,
      "tokens_seen": 796065792
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007664393179538615,
      "loss": 1.5201,
      "theoretical_loss": 3.759416854006492,
      "tokens_seen": 796131328
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00076641925777332,
      "loss": 1.6036,
      "theoretical_loss": 3.7593863835225623,
      "tokens_seen": 796196864
    },
    {
      "epoch": 0.24,
      "objective/train/advantage_avg": 0.4926673173904419,
      "objective/train/docs_used": 294439,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5741297006607056,
      "objective/train/original_loss": 1.574129581451416,
      "objective/train/theoretical_loss": 3.7593559162487864,
      "objective/train/tokens_used": 816722400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24324657022953033,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505032539367676,
      "objective/train/weighted_lm_loss": 1.6535959243774414,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7593559162487864,
      "tokens_seen": 796262400
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007663991975927784,
      "loss": 1.6695,
      "theoretical_loss": 3.7593559162487864,
      "tokens_seen": 796262400
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007663791374122367,
      "loss": 1.5325,
      "theoretical_loss": 3.759325452184563,
      "tokens_seen": 796327936
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007663590772316951,
      "loss": 1.6109,
      "theoretical_loss": 3.7592949913292886,
      "tokens_seen": 796393472
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007663390170511535,
      "loss": 1.6423,
      "theoretical_loss": 3.7592645336823627,
      "tokens_seen": 796459008
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007663189568706118,
      "loss": 1.5822,
      "theoretical_loss": 3.759234079243182,
      "tokens_seen": 796524544
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007662988966900703,
      "loss": 1.5707,
      "theoretical_loss": 3.7592036280111456,
      "tokens_seen": 796590080
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007662788365095285,
      "loss": 1.5627,
      "theoretical_loss": 3.759173179985652,
      "tokens_seen": 796655616
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000766258776328987,
      "loss": 1.4359,
      "theoretical_loss": 3.7591427351660993,
      "tokens_seen": 796721152
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007662387161484453,
      "loss": 1.4781,
      "theoretical_loss": 3.759112293551887,
      "tokens_seen": 796786688
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007662186559679037,
      "loss": 1.5357,
      "theoretical_loss": 3.759081855142413,
      "tokens_seen": 796852224
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007661985957873621,
      "loss": 1.5908,
      "theoretical_loss": 3.7590514199370775,
      "tokens_seen": 796917760
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007661785356068205,
      "loss": 1.521,
      "theoretical_loss": 3.759020987935279,
      "tokens_seen": 796983296
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007661584754262788,
      "loss": 1.4271,
      "theoretical_loss": 3.758990559136417,
      "tokens_seen": 797048832
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007661384152457373,
      "loss": 1.5137,
      "theoretical_loss": 3.7589601335398912,
      "tokens_seen": 797114368
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007661183550651955,
      "loss": 1.4797,
      "theoretical_loss": 3.758929711145101,
      "tokens_seen": 797179904
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000766098294884654,
      "loss": 1.5345,
      "theoretical_loss": 3.7588992919514475,
      "tokens_seen": 797245440
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007660782347041123,
      "loss": 1.4906,
      "theoretical_loss": 3.7588688759583295,
      "tokens_seen": 797310976
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007660581745235707,
      "loss": 1.5565,
      "theoretical_loss": 3.7588384631651475,
      "tokens_seen": 797376512
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007660381143430291,
      "loss": 1.5377,
      "theoretical_loss": 3.758808053571302,
      "tokens_seen": 797442048
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007660180541624875,
      "loss": 1.6332,
      "theoretical_loss": 3.758777647176194,
      "tokens_seen": 797507584
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007659979939819458,
      "loss": 1.6415,
      "theoretical_loss": 3.7587472439792236,
      "tokens_seen": 797573120
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007659779338014042,
      "loss": 1.658,
      "theoretical_loss": 3.7587168439797916,
      "tokens_seen": 797638656
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007659578736208626,
      "loss": 1.5736,
      "theoretical_loss": 3.7586864471772996,
      "tokens_seen": 797704192
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000765937813440321,
      "loss": 1.6401,
      "theoretical_loss": 3.7586560535711486,
      "tokens_seen": 797769728
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007659177532597793,
      "loss": 1.4734,
      "theoretical_loss": 3.75862566316074,
      "tokens_seen": 797835264
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007658976930792377,
      "loss": 1.5075,
      "theoretical_loss": 3.758595275945475,
      "tokens_seen": 797900800
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000765877632898696,
      "loss": 1.6149,
      "theoretical_loss": 3.758564891924755,
      "tokens_seen": 797966336
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007658575727181545,
      "loss": 1.5226,
      "theoretical_loss": 3.7585345110979835,
      "tokens_seen": 798031872
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007658375125376128,
      "loss": 1.7076,
      "theoretical_loss": 3.75850413346456,
      "tokens_seen": 798097408
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007658174523570712,
      "loss": 1.5649,
      "theoretical_loss": 3.7584737590238886,
      "tokens_seen": 798162944
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007657973921765296,
      "loss": 1.617,
      "theoretical_loss": 3.758443387775371,
      "tokens_seen": 798228480
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000765777331995988,
      "loss": 1.5663,
      "theoretical_loss": 3.75841301971841,
      "tokens_seen": 798294016
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007657572718154464,
      "loss": 1.5212,
      "theoretical_loss": 3.7583826548524075,
      "tokens_seen": 798359552
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007657372116349048,
      "loss": 1.596,
      "theoretical_loss": 3.758352293176767,
      "tokens_seen": 798425088
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007657171514543631,
      "loss": 1.4979,
      "theoretical_loss": 3.7583219346908905,
      "tokens_seen": 798490624
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007656970912738215,
      "loss": 1.6226,
      "theoretical_loss": 3.7582915793941822,
      "tokens_seen": 798556160
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00076567703109328,
      "loss": 1.5024,
      "theoretical_loss": 3.758261227286045,
      "tokens_seen": 798621696
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007656569709127382,
      "loss": 1.4722,
      "theoretical_loss": 3.7582308783658833,
      "tokens_seen": 798687232
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007656369107321967,
      "loss": 1.6263,
      "theoretical_loss": 3.758200532633099,
      "tokens_seen": 798752768
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007656168505516549,
      "loss": 1.6312,
      "theoretical_loss": 3.758170190087096,
      "tokens_seen": 798818304
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007655967903711134,
      "loss": 1.5286,
      "theoretical_loss": 3.7581398507272796,
      "tokens_seen": 798883840
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007655767301905718,
      "loss": 1.5078,
      "theoretical_loss": 3.7581095145530536,
      "tokens_seen": 798949376
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007655566700100301,
      "loss": 1.5478,
      "theoretical_loss": 3.7580791815638213,
      "tokens_seen": 799014912
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007655366098294885,
      "loss": 1.6045,
      "theoretical_loss": 3.7580488517589874,
      "tokens_seen": 799080448
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000765516549648947,
      "loss": 1.522,
      "theoretical_loss": 3.758018525137957,
      "tokens_seen": 799145984
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007654964894684052,
      "loss": 1.562,
      "theoretical_loss": 3.757988201700135,
      "tokens_seen": 799211520
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007654764292878637,
      "loss": 1.6466,
      "theoretical_loss": 3.7579578814449253,
      "tokens_seen": 799277056
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007654563691073219,
      "loss": 1.5571,
      "theoretical_loss": 3.757927564371734,
      "tokens_seen": 799342592
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007654363089267804,
      "loss": 1.5877,
      "theoretical_loss": 3.7578972504799646,
      "tokens_seen": 799408128
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007654162487462388,
      "loss": 1.6059,
      "theoretical_loss": 3.7578669397690248,
      "tokens_seen": 799473664
    },
    {
      "epoch": 0.24,
      "objective/train/advantage_avg": 0.49627405405044556,
      "objective/train/docs_used": 295724,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3914297819137573,
      "objective/train/original_loss": 1.3914296627044678,
      "objective/train/theoretical_loss": 3.7578366322383188,
      "objective/train/tokens_used": 819999200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2465026080608368,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508806705474854,
      "objective/train/weighted_lm_loss": 1.4621886014938354,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7578366322383188,
      "tokens_seen": 799539200
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007653961885656971,
      "loss": 1.5704,
      "theoretical_loss": 3.7578366322383188,
      "tokens_seen": 799539200
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007653761283851555,
      "loss": 1.6183,
      "theoretical_loss": 3.757806327887252,
      "tokens_seen": 799604736
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007653560682046139,
      "loss": 1.5425,
      "theoretical_loss": 3.7577760267152316,
      "tokens_seen": 799670272
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007653360080240722,
      "loss": 1.5439,
      "theoretical_loss": 3.757745728721662,
      "tokens_seen": 799735808
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007653159478435307,
      "loss": 1.4769,
      "theoretical_loss": 3.7577154339059504,
      "tokens_seen": 799801344
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007652958876629889,
      "loss": 1.5043,
      "theoretical_loss": 3.757685142267503,
      "tokens_seen": 799866880
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007652758274824474,
      "loss": 1.6786,
      "theoretical_loss": 3.757654853805726,
      "tokens_seen": 799932416
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007652557673019057,
      "loss": 1.5794,
      "theoretical_loss": 3.7576245685200265,
      "tokens_seen": 799997952
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007652357071213641,
      "loss": 1.4718,
      "theoretical_loss": 3.7575942864098106,
      "tokens_seen": 800063488
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007652156469408225,
      "loss": 1.5616,
      "theoretical_loss": 3.7575640074744863,
      "tokens_seen": 800129024
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007651955867602809,
      "loss": 1.5145,
      "theoretical_loss": 3.757533731713459,
      "tokens_seen": 800194560
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007651755265797392,
      "loss": 1.5734,
      "theoretical_loss": 3.757503459126138,
      "tokens_seen": 800260096
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007651554663991977,
      "loss": 1.4753,
      "theoretical_loss": 3.75747318971193,
      "tokens_seen": 800325632
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000765135406218656,
      "loss": 1.5605,
      "theoretical_loss": 3.757442923470242,
      "tokens_seen": 800391168
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007651153460381144,
      "loss": 1.5342,
      "theoretical_loss": 3.7574126604004823,
      "tokens_seen": 800456704
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007650952858575727,
      "loss": 1.5264,
      "theoretical_loss": 3.757382400502059,
      "tokens_seen": 800522240
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007650752256770311,
      "loss": 1.4892,
      "theoretical_loss": 3.7573521437743795,
      "tokens_seen": 800587776
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007650551654964895,
      "loss": 1.5665,
      "theoretical_loss": 3.7573218902168524,
      "tokens_seen": 800653312
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007650351053159479,
      "loss": 1.54,
      "theoretical_loss": 3.7572916398288867,
      "tokens_seen": 800718848
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007650150451354062,
      "loss": 1.5566,
      "theoretical_loss": 3.75726139260989,
      "tokens_seen": 800784384
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007649949849548646,
      "loss": 1.641,
      "theoretical_loss": 3.7572311485592715,
      "tokens_seen": 800849920
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000764974924774323,
      "loss": 1.4919,
      "theoretical_loss": 3.7572009076764403,
      "tokens_seen": 800915456
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007649548645937814,
      "loss": 1.4822,
      "theoretical_loss": 3.757170669960805,
      "tokens_seen": 800980992
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007649348044132397,
      "loss": 1.5841,
      "theoretical_loss": 3.7571404354117752,
      "tokens_seen": 801046528
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007649147442326981,
      "loss": 1.5572,
      "theoretical_loss": 3.7571102040287596,
      "tokens_seen": 801112064
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007648946840521564,
      "loss": 1.5954,
      "theoretical_loss": 3.7570799758111684,
      "tokens_seen": 801177600
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007648746238716149,
      "loss": 1.5706,
      "theoretical_loss": 3.757049750758411,
      "tokens_seen": 801243136
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007648545636910732,
      "loss": 1.6164,
      "theoretical_loss": 3.757019528869897,
      "tokens_seen": 801308672
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007648345035105316,
      "loss": 1.7327,
      "theoretical_loss": 3.7569893101450367,
      "tokens_seen": 801374208
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00076481444332999,
      "loss": 1.6002,
      "theoretical_loss": 3.75695909458324,
      "tokens_seen": 801439744
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007647943831494484,
      "loss": 1.6242,
      "theoretical_loss": 3.7569288821839173,
      "tokens_seen": 801505280
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007647743229689067,
      "loss": 1.663,
      "theoretical_loss": 3.7568986729464795,
      "tokens_seen": 801570816
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007647542627883652,
      "loss": 1.6457,
      "theoretical_loss": 3.756868466870337,
      "tokens_seen": 801636352
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007647342026078234,
      "loss": 1.6119,
      "theoretical_loss": 3.7568382639549,
      "tokens_seen": 801701888
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007647141424272819,
      "loss": 1.5575,
      "theoretical_loss": 3.75680806419958,
      "tokens_seen": 801767424
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007646940822467402,
      "loss": 1.6001,
      "theoretical_loss": 3.7567778676037875,
      "tokens_seen": 801832960
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007646740220661986,
      "loss": 1.5751,
      "theoretical_loss": 3.7567476741669346,
      "tokens_seen": 801898496
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000764653961885657,
      "loss": 1.5877,
      "theoretical_loss": 3.756717483888432,
      "tokens_seen": 801964032
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007646339017051153,
      "loss": 1.6035,
      "theoretical_loss": 3.7566872967676916,
      "tokens_seen": 802029568
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007646138415245737,
      "loss": 1.6353,
      "theoretical_loss": 3.7566571128041257,
      "tokens_seen": 802095104
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007645937813440322,
      "loss": 1.4843,
      "theoretical_loss": 3.756626931997145,
      "tokens_seen": 802160640
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007645737211634904,
      "loss": 1.4345,
      "theoretical_loss": 3.756596754346162,
      "tokens_seen": 802226176
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007645536609829489,
      "loss": 1.5597,
      "theoretical_loss": 3.756566579850589,
      "tokens_seen": 802291712
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007645336008024071,
      "loss": 1.5042,
      "theoretical_loss": 3.756536408509839,
      "tokens_seen": 802357248
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007645135406218656,
      "loss": 1.5465,
      "theoretical_loss": 3.7565062403233234,
      "tokens_seen": 802422784
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000764493480441324,
      "loss": 1.4479,
      "theoretical_loss": 3.7564760752904554,
      "tokens_seen": 802488320
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007644734202607823,
      "loss": 1.5013,
      "theoretical_loss": 3.756445913410648,
      "tokens_seen": 802553856
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007644533600802407,
      "loss": 1.5074,
      "theoretical_loss": 3.7564157546833132,
      "tokens_seen": 802619392
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007644332998996992,
      "loss": 1.5272,
      "theoretical_loss": 3.7563855991078654,
      "tokens_seen": 802684928
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007644132397191574,
      "loss": 1.4647,
      "theoretical_loss": 3.7563554466837177,
      "tokens_seen": 802750464
    },
    {
      "epoch": 0.24,
      "objective/train/advantage_avg": 0.4896015524864197,
      "objective/train/docs_used": 296852,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4510241746902466,
      "objective/train/original_loss": 1.451024055480957,
      "objective/train/theoretical_loss": 3.7563252974102825,
      "objective/train/tokens_used": 823276000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24044343829154968,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501824617385864,
      "objective/train/weighted_lm_loss": 1.5239347219467163,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7563252974102825,
      "tokens_seen": 802816000
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007643931795386159,
      "loss": 1.4638,
      "theoretical_loss": 3.7563252974102825,
      "tokens_seen": 802816000
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007643731193580741,
      "loss": 1.4675,
      "theoretical_loss": 3.7562951512869747,
      "tokens_seen": 802881536
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007643530591775326,
      "loss": 1.4929,
      "theoretical_loss": 3.7562650083132074,
      "tokens_seen": 802947072
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000764332998996991,
      "loss": 1.5849,
      "theoretical_loss": 3.7562348684883946,
      "tokens_seen": 803012608
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007643129388164493,
      "loss": 1.5851,
      "theoretical_loss": 3.75620473181195,
      "tokens_seen": 803078144
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007642928786359077,
      "loss": 1.4552,
      "theoretical_loss": 3.7561745982832893,
      "tokens_seen": 803143680
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007642728184553661,
      "loss": 1.7135,
      "theoretical_loss": 3.756144467901825,
      "tokens_seen": 803209216
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007642527582748245,
      "loss": 1.5736,
      "theoretical_loss": 3.756114340666973,
      "tokens_seen": 803274752
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007642326980942829,
      "loss": 1.6051,
      "theoretical_loss": 3.7560842165781474,
      "tokens_seen": 803340288
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007642126379137413,
      "loss": 1.5655,
      "theoretical_loss": 3.756054095634763,
      "tokens_seen": 803405824
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007641925777331996,
      "loss": 1.5547,
      "theoretical_loss": 3.756023977836235,
      "tokens_seen": 803471360
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007641725175526581,
      "loss": 1.4984,
      "theoretical_loss": 3.7559938631819794,
      "tokens_seen": 803536896
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007641524573721163,
      "loss": 1.6292,
      "theoretical_loss": 3.7559637516714095,
      "tokens_seen": 803602432
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007641323971915748,
      "loss": 1.5461,
      "theoretical_loss": 3.7559336433039427,
      "tokens_seen": 803667968
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007641123370110331,
      "loss": 1.5325,
      "theoretical_loss": 3.755903538078994,
      "tokens_seen": 803733504
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007640922768304915,
      "loss": 1.5873,
      "theoretical_loss": 3.755873435995979,
      "tokens_seen": 803799040
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007640722166499499,
      "loss": 1.586,
      "theoretical_loss": 3.755843337054314,
      "tokens_seen": 803864576
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007640521564694083,
      "loss": 1.652,
      "theoretical_loss": 3.755813241253415,
      "tokens_seen": 803930112
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007640320962888666,
      "loss": 1.6816,
      "theoretical_loss": 3.7557831485926982,
      "tokens_seen": 803995648
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000764012036108325,
      "loss": 1.6526,
      "theoretical_loss": 3.75575305907158,
      "tokens_seen": 804061184
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007639919759277834,
      "loss": 1.6643,
      "theoretical_loss": 3.755722972689477,
      "tokens_seen": 804126720
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007639719157472418,
      "loss": 1.5262,
      "theoretical_loss": 3.7556928894458057,
      "tokens_seen": 804192256
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007639518555667001,
      "loss": 1.527,
      "theoretical_loss": 3.7556628093399835,
      "tokens_seen": 804257792
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007639317953861585,
      "loss": 1.6366,
      "theoretical_loss": 3.755632732371428,
      "tokens_seen": 804323328
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007639117352056168,
      "loss": 1.6131,
      "theoretical_loss": 3.7556026585395546,
      "tokens_seen": 804388864
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007638916750250753,
      "loss": 1.574,
      "theoretical_loss": 3.7555725878437825,
      "tokens_seen": 804454400
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007638716148445336,
      "loss": 1.6765,
      "theoretical_loss": 3.7555425202835275,
      "tokens_seen": 804519936
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000763851554663992,
      "loss": 1.4982,
      "theoretical_loss": 3.755512455858209,
      "tokens_seen": 804585472
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007638314944834504,
      "loss": 1.4189,
      "theoretical_loss": 3.755482394567244,
      "tokens_seen": 804651008
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007638114343029088,
      "loss": 1.4694,
      "theoretical_loss": 3.75545233641005,
      "tokens_seen": 804716544
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007637913741223671,
      "loss": 1.5377,
      "theoretical_loss": 3.7554222813860463,
      "tokens_seen": 804782080
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007637713139418256,
      "loss": 1.5802,
      "theoretical_loss": 3.75539222949465,
      "tokens_seen": 804847616
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007637512537612838,
      "loss": 1.5712,
      "theoretical_loss": 3.755362180735281,
      "tokens_seen": 804913152
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007637311935807423,
      "loss": 1.5611,
      "theoretical_loss": 3.7553321351073556,
      "tokens_seen": 804978688
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007637111334002006,
      "loss": 1.5236,
      "theoretical_loss": 3.7553020926102954,
      "tokens_seen": 805044224
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000763691073219659,
      "loss": 1.4657,
      "theoretical_loss": 3.7552720532435173,
      "tokens_seen": 805109760
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007636710130391174,
      "loss": 1.6226,
      "theoretical_loss": 3.7552420170064407,
      "tokens_seen": 805175296
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007636509528585757,
      "loss": 1.6492,
      "theoretical_loss": 3.755211983898486,
      "tokens_seen": 805240832
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007636308926780341,
      "loss": 1.611,
      "theoretical_loss": 3.755181953919071,
      "tokens_seen": 805306368
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007636108324974926,
      "loss": 1.4535,
      "theoretical_loss": 3.755151927067616,
      "tokens_seen": 805371904
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007635907723169508,
      "loss": 1.575,
      "theoretical_loss": 3.755121903343541,
      "tokens_seen": 805437440
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007635707121364093,
      "loss": 1.6457,
      "theoretical_loss": 3.755091882746265,
      "tokens_seen": 805502976
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007635506519558675,
      "loss": 1.5593,
      "theoretical_loss": 3.755061865275209,
      "tokens_seen": 805568512
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000763530591775326,
      "loss": 1.6587,
      "theoretical_loss": 3.755031850929792,
      "tokens_seen": 805634048
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007635105315947844,
      "loss": 1.5347,
      "theoretical_loss": 3.7550018397094354,
      "tokens_seen": 805699584
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007634904714142427,
      "loss": 1.5105,
      "theoretical_loss": 3.7549718316135596,
      "tokens_seen": 805765120
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007634704112337011,
      "loss": 1.595,
      "theoretical_loss": 3.754941826641584,
      "tokens_seen": 805830656
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007634503510531596,
      "loss": 1.6457,
      "theoretical_loss": 3.7549118247929307,
      "tokens_seen": 805896192
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007634302908726178,
      "loss": 1.5186,
      "theoretical_loss": 3.7548818260670203,
      "tokens_seen": 805961728
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007634102306920763,
      "loss": 1.459,
      "theoretical_loss": 3.7548518304632736,
      "tokens_seen": 806027264
    },
    {
      "epoch": 0.24,
      "objective/train/advantage_avg": 0.49647656083106995,
      "objective/train/docs_used": 297892,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4915120601654053,
      "objective/train/original_loss": 1.4915121793746948,
      "objective/train/theoretical_loss": 3.754821837981112,
      "objective/train/tokens_used": 826552800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24659492075443268,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0509014129638672,
      "objective/train/weighted_lm_loss": 1.5673531293869019,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 3.754821837981112,
      "tokens_seen": 806092800
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007633901705115345,
      "loss": 1.481,
      "theoretical_loss": 3.754821837981112,
      "tokens_seen": 806092800
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000763370110330993,
      "loss": 1.5951,
      "theoretical_loss": 3.7547918486199565,
      "tokens_seen": 806158336
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007633500501504514,
      "loss": 1.5771,
      "theoretical_loss": 3.7547618623792296,
      "tokens_seen": 806223872
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007633299899699097,
      "loss": 1.4961,
      "theoretical_loss": 3.7547318792583524,
      "tokens_seen": 806289408
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007633099297893681,
      "loss": 1.4961,
      "theoretical_loss": 3.7547018992567462,
      "tokens_seen": 806354944
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007632898696088265,
      "loss": 1.6869,
      "theoretical_loss": 3.7546719223738343,
      "tokens_seen": 806420480
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007632698094282848,
      "loss": 1.6951,
      "theoretical_loss": 3.754641948609038,
      "tokens_seen": 806486016
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007632497492477433,
      "loss": 1.5878,
      "theoretical_loss": 3.7546119779617797,
      "tokens_seen": 806551552
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007632296890672016,
      "loss": 1.4295,
      "theoretical_loss": 3.7545820104314815,
      "tokens_seen": 806617088
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00076320962888666,
      "loss": 1.4463,
      "theoretical_loss": 3.7545520460175674,
      "tokens_seen": 806682624
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007631895687061183,
      "loss": 1.54,
      "theoretical_loss": 3.7545220847194587,
      "tokens_seen": 806748160
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007631695085255767,
      "loss": 1.4753,
      "theoretical_loss": 3.7544921265365794,
      "tokens_seen": 806813696
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007631494483450351,
      "loss": 1.5589,
      "theoretical_loss": 3.7544621714683517,
      "tokens_seen": 806879232
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007631293881644935,
      "loss": 1.5648,
      "theoretical_loss": 3.7544322195141993,
      "tokens_seen": 806944768
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007631093279839518,
      "loss": 1.5855,
      "theoretical_loss": 3.7544022706735456,
      "tokens_seen": 807010304
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007630892678034103,
      "loss": 1.4236,
      "theoretical_loss": 3.7543723249458134,
      "tokens_seen": 807075840
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007630692076228686,
      "loss": 1.589,
      "theoretical_loss": 3.754342382330428,
      "tokens_seen": 807141376
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000763049147442327,
      "loss": 1.505,
      "theoretical_loss": 3.754312442826812,
      "tokens_seen": 807206912
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007630290872617853,
      "loss": 1.5345,
      "theoretical_loss": 3.75428250643439,
      "tokens_seen": 807272448
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007630090270812438,
      "loss": 1.5404,
      "theoretical_loss": 3.754252573152585,
      "tokens_seen": 807337984
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007629889669007021,
      "loss": 1.4598,
      "theoretical_loss": 3.7542226429808236,
      "tokens_seen": 807403520
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007629689067201605,
      "loss": 1.5297,
      "theoretical_loss": 3.754192715918528,
      "tokens_seen": 807469056
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007629488465396188,
      "loss": 1.5178,
      "theoretical_loss": 3.754162791965124,
      "tokens_seen": 807534592
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007629287863590772,
      "loss": 1.4608,
      "theoretical_loss": 3.754132871120036,
      "tokens_seen": 807600128
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007629087261785356,
      "loss": 1.4949,
      "theoretical_loss": 3.7541029533826893,
      "tokens_seen": 807665664
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000762888665997994,
      "loss": 1.6127,
      "theoretical_loss": 3.7540730387525088,
      "tokens_seen": 807731200
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007628686058174523,
      "loss": 1.4522,
      "theoretical_loss": 3.754043127228919,
      "tokens_seen": 807796736
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007628485456369108,
      "loss": 1.4974,
      "theoretical_loss": 3.7540132188113464,
      "tokens_seen": 807862272
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007628284854563691,
      "loss": 1.5383,
      "theoretical_loss": 3.7539833134992158,
      "tokens_seen": 807927808
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007628084252758275,
      "loss": 1.5344,
      "theoretical_loss": 3.7539534112919535,
      "tokens_seen": 807993344
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007627883650952858,
      "loss": 1.6219,
      "theoretical_loss": 3.753923512188985,
      "tokens_seen": 808058880
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007627683049147442,
      "loss": 1.5747,
      "theoretical_loss": 3.7538936161897363,
      "tokens_seen": 808124416
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007627482447342027,
      "loss": 1.6211,
      "theoretical_loss": 3.753863723293634,
      "tokens_seen": 808189952
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000762728184553661,
      "loss": 1.4984,
      "theoretical_loss": 3.753833833500103,
      "tokens_seen": 808255488
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007627081243731194,
      "loss": 1.5326,
      "theoretical_loss": 3.7538039468085715,
      "tokens_seen": 808321024
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007626880641925778,
      "loss": 1.6157,
      "theoretical_loss": 3.753774063218465,
      "tokens_seen": 808386560
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0007626680040120361,
      "loss": 1.5569,
      "theoretical_loss": 3.7537441827292106,
      "tokens_seen": 808452096
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007626479438314945,
      "loss": 1.4794,
      "theoretical_loss": 3.7537143053402353,
      "tokens_seen": 808517632
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000762627883650953,
      "loss": 1.6081,
      "theoretical_loss": 3.7536844310509663,
      "tokens_seen": 808583168
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007626078234704112,
      "loss": 1.6485,
      "theoretical_loss": 3.7536545598608306,
      "tokens_seen": 808648704
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007625877632898697,
      "loss": 1.5082,
      "theoretical_loss": 3.753624691769255,
      "tokens_seen": 808714240
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007625677031093279,
      "loss": 1.5468,
      "theoretical_loss": 3.753594826775668,
      "tokens_seen": 808779776
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007625476429287864,
      "loss": 1.4947,
      "theoretical_loss": 3.753564964879497,
      "tokens_seen": 808845312
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007625275827482448,
      "loss": 1.6366,
      "theoretical_loss": 3.75353510608017,
      "tokens_seen": 808910848
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007625075225677031,
      "loss": 1.7082,
      "theoretical_loss": 3.7535052503771142,
      "tokens_seen": 808976384
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007624874623871615,
      "loss": 1.5346,
      "theoretical_loss": 3.7534753977697584,
      "tokens_seen": 809041920
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00076246740220662,
      "loss": 1.4791,
      "theoretical_loss": 3.7534455482575306,
      "tokens_seen": 809107456
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007624473420260782,
      "loss": 1.5342,
      "theoretical_loss": 3.7534157018398595,
      "tokens_seen": 809172992
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007624272818455367,
      "loss": 1.5631,
      "theoretical_loss": 3.7533858585161735,
      "tokens_seen": 809238528
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000762407221664995,
      "loss": 1.6056,
      "theoretical_loss": 3.7533560182859014,
      "tokens_seen": 809304064
    },
    {
      "epoch": 0.25,
      "objective/train/advantage_avg": 0.49424847960472107,
      "objective/train/docs_used": 298985,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5160239934921265,
      "objective/train/original_loss": 1.516023874282837,
      "objective/train/theoretical_loss": 3.753326181148472,
      "objective/train/tokens_used": 829829600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24453648924827576,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506681203842163,
      "objective/train/weighted_lm_loss": 1.5927289724349976,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.753326181148472,
      "tokens_seen": 809369600
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007623871614844534,
      "loss": 1.6231,
      "theoretical_loss": 3.753326181148472,
      "tokens_seen": 809369600
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007623671013039118,
      "loss": 1.6126,
      "theoretical_loss": 3.7532963471033147,
      "tokens_seen": 809435136
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007623470411233701,
      "loss": 1.5947,
      "theoretical_loss": 3.753266516149858,
      "tokens_seen": 809500672
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007623269809428285,
      "loss": 1.5415,
      "theoretical_loss": 3.753236688287532,
      "tokens_seen": 809566208
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007623069207622869,
      "loss": 1.5001,
      "theoretical_loss": 3.7532068635157656,
      "tokens_seen": 809631744
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007622868605817452,
      "loss": 1.6053,
      "theoretical_loss": 3.7531770418339887,
      "tokens_seen": 809697280
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007622668004012037,
      "loss": 1.5351,
      "theoretical_loss": 3.7531472232416316,
      "tokens_seen": 809762816
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000762246740220662,
      "loss": 1.4707,
      "theoretical_loss": 3.753117407738123,
      "tokens_seen": 809828352
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007622266800401204,
      "loss": 1.4867,
      "theoretical_loss": 3.753087595322894,
      "tokens_seen": 809893888
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007622066198595787,
      "loss": 1.4865,
      "theoretical_loss": 3.7530577859953747,
      "tokens_seen": 809959424
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007621865596790371,
      "loss": 1.5121,
      "theoretical_loss": 3.7530279797549957,
      "tokens_seen": 810024960
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007621664994984955,
      "loss": 1.4604,
      "theoretical_loss": 3.752998176601187,
      "tokens_seen": 810090496
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007621464393179539,
      "loss": 1.4833,
      "theoretical_loss": 3.7529683765333792,
      "tokens_seen": 810156032
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007621263791374122,
      "loss": 1.5999,
      "theoretical_loss": 3.752938579551004,
      "tokens_seen": 810221568
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007621063189568707,
      "loss": 1.4885,
      "theoretical_loss": 3.752908785653492,
      "tokens_seen": 810287104
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000762086258776329,
      "loss": 1.4622,
      "theoretical_loss": 3.752878994840274,
      "tokens_seen": 810352640
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007620661985957874,
      "loss": 1.5815,
      "theoretical_loss": 3.752849207110782,
      "tokens_seen": 810418176
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007620461384152457,
      "loss": 1.5143,
      "theoretical_loss": 3.752819422464447,
      "tokens_seen": 810483712
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007620260782347042,
      "loss": 1.547,
      "theoretical_loss": 3.7527896409007004,
      "tokens_seen": 810549248
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007620060180541625,
      "loss": 1.6519,
      "theoretical_loss": 3.7527598624189746,
      "tokens_seen": 810614784
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007619859578736209,
      "loss": 1.5236,
      "theoretical_loss": 3.7527300870187013,
      "tokens_seen": 810680320
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007619658976930792,
      "loss": 1.6837,
      "theoretical_loss": 3.752700314699312,
      "tokens_seen": 810745856
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007619458375125376,
      "loss": 1.4664,
      "theoretical_loss": 3.7526705454602394,
      "tokens_seen": 810811392
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000761925777331996,
      "loss": 1.5867,
      "theoretical_loss": 3.7526407793009158,
      "tokens_seen": 810876928
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007619057171514544,
      "loss": 1.5876,
      "theoretical_loss": 3.752611016220774,
      "tokens_seen": 810942464
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007618856569709127,
      "loss": 1.4822,
      "theoretical_loss": 3.752581256219247,
      "tokens_seen": 811008000
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007618655967903712,
      "loss": 1.6021,
      "theoretical_loss": 3.752551499295766,
      "tokens_seen": 811073536
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007618455366098295,
      "loss": 1.603,
      "theoretical_loss": 3.7525217454497657,
      "tokens_seen": 811139072
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007618254764292879,
      "loss": 1.5399,
      "theoretical_loss": 3.7524919946806783,
      "tokens_seen": 811204608
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007618054162487462,
      "loss": 1.5038,
      "theoretical_loss": 3.752462246987937,
      "tokens_seen": 811270144
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007617853560682046,
      "loss": 1.5469,
      "theoretical_loss": 3.7524325023709757,
      "tokens_seen": 811335680
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000761765295887663,
      "loss": 1.4333,
      "theoretical_loss": 3.752402760829227,
      "tokens_seen": 811401216
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007617452357071214,
      "loss": 1.5553,
      "theoretical_loss": 3.7523730223621268,
      "tokens_seen": 811466752
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007617251755265797,
      "loss": 1.5283,
      "theoretical_loss": 3.752343286969107,
      "tokens_seen": 811532288
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007617051153460382,
      "loss": 1.4444,
      "theoretical_loss": 3.7523135546496023,
      "tokens_seen": 811597824
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007616850551654964,
      "loss": 1.5561,
      "theoretical_loss": 3.7522838254030466,
      "tokens_seen": 811663360
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007616649949849549,
      "loss": 1.596,
      "theoretical_loss": 3.752254099228874,
      "tokens_seen": 811728896
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007616449348044133,
      "loss": 1.5323,
      "theoretical_loss": 3.75222437612652,
      "tokens_seen": 811794432
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007616248746238716,
      "loss": 1.5573,
      "theoretical_loss": 3.7521946560954182,
      "tokens_seen": 811859968
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00076160481444333,
      "loss": 1.4836,
      "theoretical_loss": 3.7521649391350036,
      "tokens_seen": 811925504
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007615847542627883,
      "loss": 1.531,
      "theoretical_loss": 3.7521352252447118,
      "tokens_seen": 811991040
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007615646940822467,
      "loss": 1.6164,
      "theoretical_loss": 3.7521055144239766,
      "tokens_seen": 812056576
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007615446339017052,
      "loss": 1.5664,
      "theoretical_loss": 3.7520758066722344,
      "tokens_seen": 812122112
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007615245737211634,
      "loss": 1.4759,
      "theoretical_loss": 3.75204610198892,
      "tokens_seen": 812187648
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007615045135406219,
      "loss": 1.4986,
      "theoretical_loss": 3.752016400373469,
      "tokens_seen": 812253184
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007614844533600803,
      "loss": 1.508,
      "theoretical_loss": 3.751986701825317,
      "tokens_seen": 812318720
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007614643931795386,
      "loss": 1.5801,
      "theoretical_loss": 3.7519570063438996,
      "tokens_seen": 812384256
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000761444332998997,
      "loss": 1.6145,
      "theoretical_loss": 3.751927313928654,
      "tokens_seen": 812449792
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007614242728184553,
      "loss": 1.5285,
      "theoretical_loss": 3.7518976245790148,
      "tokens_seen": 812515328
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007614042126379137,
      "loss": 1.4732,
      "theoretical_loss": 3.7518679382944184,
      "tokens_seen": 812580864
    },
    {
      "epoch": 0.25,
      "objective/train/advantage_avg": 0.49399682879447937,
      "objective/train/docs_used": 300102,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4264678955078125,
      "objective/train/original_loss": 1.4264678955078125,
      "objective/train/theoretical_loss": 3.7518382550743024,
      "objective/train/tokens_used": 833106400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24423182010650635,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506412982940674,
      "objective/train/weighted_lm_loss": 1.498720407485962,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7518382550743024,
      "tokens_seen": 812646400
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007613841524573722,
      "loss": 1.5452,
      "theoretical_loss": 3.7518382550743024,
      "tokens_seen": 812646400
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007613640922768304,
      "loss": 1.6556,
      "theoretical_loss": 3.751808574918102,
      "tokens_seen": 812711936
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007613440320962889,
      "loss": 1.5226,
      "theoretical_loss": 3.751778897825255,
      "tokens_seen": 812777472
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007613239719157472,
      "loss": 1.5707,
      "theoretical_loss": 3.7517492237951973,
      "tokens_seen": 812843008
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007613039117352056,
      "loss": 1.5391,
      "theoretical_loss": 3.7517195528273666,
      "tokens_seen": 812908544
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000761283851554664,
      "loss": 1.5547,
      "theoretical_loss": 3.7516898849211993,
      "tokens_seen": 812974080
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007612637913741224,
      "loss": 1.5946,
      "theoretical_loss": 3.751660220076134,
      "tokens_seen": 813039616
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007612437311935808,
      "loss": 1.441,
      "theoretical_loss": 3.751630558291607,
      "tokens_seen": 813105152
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007612236710130391,
      "loss": 1.5622,
      "theoretical_loss": 3.751600899567057,
      "tokens_seen": 813170688
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007612036108324975,
      "loss": 1.3893,
      "theoretical_loss": 3.75157124390192,
      "tokens_seen": 813236224
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007611835506519559,
      "loss": 1.4786,
      "theoretical_loss": 3.7515415912956356,
      "tokens_seen": 813301760
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007611634904714143,
      "loss": 1.557,
      "theoretical_loss": 3.751511941747641,
      "tokens_seen": 813367296
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007611434302908726,
      "loss": 1.4311,
      "theoretical_loss": 3.7514822952573743,
      "tokens_seen": 813432832
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007611233701103311,
      "loss": 1.519,
      "theoretical_loss": 3.7514526518242746,
      "tokens_seen": 813498368
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007611033099297894,
      "loss": 1.62,
      "theoretical_loss": 3.7514230114477796,
      "tokens_seen": 813563904
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007610832497492478,
      "loss": 1.5615,
      "theoretical_loss": 3.7513933741273284,
      "tokens_seen": 813629440
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007610631895687061,
      "loss": 1.5656,
      "theoretical_loss": 3.7513637398623603,
      "tokens_seen": 813694976
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007610431293881646,
      "loss": 1.4785,
      "theoretical_loss": 3.751334108652313,
      "tokens_seen": 813760512
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007610230692076229,
      "loss": 1.5357,
      "theoretical_loss": 3.751304480496626,
      "tokens_seen": 813826048
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007610030090270813,
      "loss": 1.6047,
      "theoretical_loss": 3.7512748553947386,
      "tokens_seen": 813891584
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007609829488465396,
      "loss": 1.4995,
      "theoretical_loss": 3.751245233346091,
      "tokens_seen": 813957120
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000760962888665998,
      "loss": 1.5277,
      "theoretical_loss": 3.751215614350121,
      "tokens_seen": 814022656
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007609428284854564,
      "loss": 1.5886,
      "theoretical_loss": 3.75118599840627,
      "tokens_seen": 814088192
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007609227683049148,
      "loss": 1.5548,
      "theoretical_loss": 3.751156385513977,
      "tokens_seen": 814153728
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007609027081243731,
      "loss": 1.54,
      "theoretical_loss": 3.7511267756726823,
      "tokens_seen": 814219264
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007608826479438316,
      "loss": 1.4919,
      "theoretical_loss": 3.7510971688818255,
      "tokens_seen": 814284800
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007608625877632899,
      "loss": 1.686,
      "theoretical_loss": 3.751067565140848,
      "tokens_seen": 814350336
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007608425275827483,
      "loss": 1.5705,
      "theoretical_loss": 3.7510379644491882,
      "tokens_seen": 814415872
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007608224674022066,
      "loss": 1.6311,
      "theoretical_loss": 3.7510083668062886,
      "tokens_seen": 814481408
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000760802407221665,
      "loss": 1.5955,
      "theoretical_loss": 3.750978772211589,
      "tokens_seen": 814546944
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007607823470411234,
      "loss": 1.5297,
      "theoretical_loss": 3.7509491806645308,
      "tokens_seen": 814612480
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007607622868605818,
      "loss": 1.4309,
      "theoretical_loss": 3.7509195921645544,
      "tokens_seen": 814678016
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007607422266800401,
      "loss": 1.497,
      "theoretical_loss": 3.750890006711101,
      "tokens_seen": 814743552
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007607221664994986,
      "loss": 1.5215,
      "theoretical_loss": 3.750860424303612,
      "tokens_seen": 814809088
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007607021063189568,
      "loss": 1.4158,
      "theoretical_loss": 3.7508308449415297,
      "tokens_seen": 814874624
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007606820461384153,
      "loss": 1.6025,
      "theoretical_loss": 3.750801268624295,
      "tokens_seen": 814940160
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007606619859578737,
      "loss": 1.5218,
      "theoretical_loss": 3.7507716953513492,
      "tokens_seen": 815005696
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000760641925777332,
      "loss": 1.6148,
      "theoretical_loss": 3.7507421251221347,
      "tokens_seen": 815071232
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007606218655967904,
      "loss": 1.6298,
      "theoretical_loss": 3.750712557936093,
      "tokens_seen": 815136768
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007606018054162487,
      "loss": 1.6008,
      "theoretical_loss": 3.7506829937926676,
      "tokens_seen": 815202304
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007605817452357071,
      "loss": 1.513,
      "theoretical_loss": 3.7506534326912995,
      "tokens_seen": 815267840
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007605616850551656,
      "loss": 1.4757,
      "theoretical_loss": 3.7506238746314313,
      "tokens_seen": 815333376
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007605416248746238,
      "loss": 1.6563,
      "theoretical_loss": 3.7505943196125067,
      "tokens_seen": 815398912
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007605215646940823,
      "loss": 1.4674,
      "theoretical_loss": 3.7505647676339677,
      "tokens_seen": 815464448
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007605015045135407,
      "loss": 1.6366,
      "theoretical_loss": 3.7505352186952567,
      "tokens_seen": 815529984
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000760481444332999,
      "loss": 1.5276,
      "theoretical_loss": 3.7505056727958177,
      "tokens_seen": 815595520
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007604613841524574,
      "loss": 1.4657,
      "theoretical_loss": 3.7504761299350937,
      "tokens_seen": 815661056
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007604413239719157,
      "loss": 1.4475,
      "theoretical_loss": 3.750446590112528,
      "tokens_seen": 815726592
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007604212637913741,
      "loss": 1.5481,
      "theoretical_loss": 3.7504170533275634,
      "tokens_seen": 815792128
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007604012036108326,
      "loss": 1.3939,
      "theoretical_loss": 3.750387519579645,
      "tokens_seen": 815857664
    },
    {
      "epoch": 0.25,
      "objective/train/advantage_avg": 0.4913868308067322,
      "objective/train/docs_used": 301260,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4402931928634644,
      "objective/train/original_loss": 1.4402931928634644,
      "objective/train/theoretical_loss": 3.7503579888682155,
      "objective/train/tokens_used": 836383200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24211423099040985,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503695011138916,
      "objective/train/weighted_lm_loss": 1.5127294063568115,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7503579888682155,
      "tokens_seen": 815923200
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007603811434302908,
      "loss": 1.5388,
      "theoretical_loss": 3.7503579888682155,
      "tokens_seen": 815923200
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007603610832497493,
      "loss": 1.4122,
      "theoretical_loss": 3.7503284611927192,
      "tokens_seen": 815988736
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007603410230692076,
      "loss": 1.565,
      "theoretical_loss": 3.7502989365526,
      "tokens_seen": 816054272
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000760320962888666,
      "loss": 1.4208,
      "theoretical_loss": 3.750269414947302,
      "tokens_seen": 816119808
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007603009027081244,
      "loss": 1.5684,
      "theoretical_loss": 3.75023989637627,
      "tokens_seen": 816185344
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007602808425275828,
      "loss": 1.5269,
      "theoretical_loss": 3.750210380838949,
      "tokens_seen": 816250880
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007602607823470411,
      "loss": 1.6459,
      "theoretical_loss": 3.7501808683347826,
      "tokens_seen": 816316416
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007602407221664995,
      "loss": 1.5976,
      "theoretical_loss": 3.7501513588632163,
      "tokens_seen": 816381952
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007602206619859578,
      "loss": 1.5639,
      "theoretical_loss": 3.750121852423695,
      "tokens_seen": 816447488
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007602006018054163,
      "loss": 1.5213,
      "theoretical_loss": 3.750092349015663,
      "tokens_seen": 816513024
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007601805416248746,
      "loss": 1.3796,
      "theoretical_loss": 3.7500628486385668,
      "tokens_seen": 816578560
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000760160481444333,
      "loss": 1.6425,
      "theoretical_loss": 3.750033351291851,
      "tokens_seen": 816644096
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007601404212637914,
      "loss": 1.5244,
      "theoretical_loss": 3.7500038569749616,
      "tokens_seen": 816709632
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007601203610832498,
      "loss": 1.3719,
      "theoretical_loss": 3.7499743656873443,
      "tokens_seen": 816775168
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007601003009027081,
      "loss": 1.4923,
      "theoretical_loss": 3.7499448774284447,
      "tokens_seen": 816840704
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007600802407221665,
      "loss": 1.595,
      "theoretical_loss": 3.7499153921977086,
      "tokens_seen": 816906240
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007600601805416248,
      "loss": 1.5579,
      "theoretical_loss": 3.749885909994583,
      "tokens_seen": 816971776
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007600401203610833,
      "loss": 1.4921,
      "theoretical_loss": 3.7498564308185136,
      "tokens_seen": 817037312
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007600200601805416,
      "loss": 1.6023,
      "theoretical_loss": 3.749826954668946,
      "tokens_seen": 817102848
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00076,
      "loss": 1.5006,
      "theoretical_loss": 3.7497974815453285,
      "tokens_seen": 817168384
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007599799398194583,
      "loss": 1.5527,
      "theoretical_loss": 3.749768011447107,
      "tokens_seen": 817233920
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007599598796389168,
      "loss": 1.6046,
      "theoretical_loss": 3.7497385443737272,
      "tokens_seen": 817299456
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007599398194583751,
      "loss": 1.5966,
      "theoretical_loss": 3.7497090803246387,
      "tokens_seen": 817364992
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007599197592778335,
      "loss": 1.5704,
      "theoretical_loss": 3.749679619299286,
      "tokens_seen": 817430528
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007598996990972919,
      "loss": 1.4806,
      "theoretical_loss": 3.749650161297118,
      "tokens_seen": 817496064
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007598796389167503,
      "loss": 1.5282,
      "theoretical_loss": 3.7496207063175815,
      "tokens_seen": 817561600
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007598595787362086,
      "loss": 1.4806,
      "theoretical_loss": 3.7495912543601246,
      "tokens_seen": 817627136
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000759839518555667,
      "loss": 1.6206,
      "theoretical_loss": 3.749561805424195,
      "tokens_seen": 817692672
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007598194583751253,
      "loss": 1.6038,
      "theoretical_loss": 3.7495323595092405,
      "tokens_seen": 817758208
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007597993981945838,
      "loss": 1.5528,
      "theoretical_loss": 3.7495029166147082,
      "tokens_seen": 817823744
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007597793380140421,
      "loss": 1.5238,
      "theoretical_loss": 3.7494734767400475,
      "tokens_seen": 817889280
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007597592778335005,
      "loss": 1.5717,
      "theoretical_loss": 3.749444039884706,
      "tokens_seen": 817954816
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000759739217652959,
      "loss": 1.6112,
      "theoretical_loss": 3.7494146060481333,
      "tokens_seen": 818020352
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007597191574724172,
      "loss": 1.4891,
      "theoretical_loss": 3.7493851752297767,
      "tokens_seen": 818085888
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007596990972918757,
      "loss": 1.5306,
      "theoretical_loss": 3.7493557474290853,
      "tokens_seen": 818151424
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007596790371113341,
      "loss": 1.5071,
      "theoretical_loss": 3.749326322645508,
      "tokens_seen": 818216960
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007596589769307924,
      "loss": 1.5022,
      "theoretical_loss": 3.7492969008784938,
      "tokens_seen": 818282496
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007596389167502508,
      "loss": 1.4661,
      "theoretical_loss": 3.7492674821274927,
      "tokens_seen": 818348032
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007596188565697091,
      "loss": 1.5408,
      "theoretical_loss": 3.7492380663919533,
      "tokens_seen": 818413568
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007595987963891675,
      "loss": 1.3982,
      "theoretical_loss": 3.7492086536713245,
      "tokens_seen": 818479104
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000759578736208626,
      "loss": 1.6386,
      "theoretical_loss": 3.749179243965057,
      "tokens_seen": 818544640
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007595586760280842,
      "loss": 1.5379,
      "theoretical_loss": 3.7491498372726006,
      "tokens_seen": 818610176
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007595386158475427,
      "loss": 1.5833,
      "theoretical_loss": 3.7491204335934043,
      "tokens_seen": 818675712
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007595185556670011,
      "loss": 1.5469,
      "theoretical_loss": 3.749091032926919,
      "tokens_seen": 818741248
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007594984954864594,
      "loss": 1.5903,
      "theoretical_loss": 3.749061635272594,
      "tokens_seen": 818806784
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007594784353059178,
      "loss": 1.5797,
      "theoretical_loss": 3.74903224062988,
      "tokens_seen": 818872320
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007594583751253761,
      "loss": 1.4472,
      "theoretical_loss": 3.7490028489982286,
      "tokens_seen": 818937856
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007594383149448345,
      "loss": 1.3631,
      "theoretical_loss": 3.748973460377089,
      "tokens_seen": 819003392
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000759418254764293,
      "loss": 1.3764,
      "theoretical_loss": 3.7489440747659124,
      "tokens_seen": 819068928
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007593981945837512,
      "loss": 1.5522,
      "theoretical_loss": 3.74891469216415,
      "tokens_seen": 819134464
    },
    {
      "epoch": 0.25,
      "objective/train/advantage_avg": 0.490060418844223,
      "objective/train/docs_used": 302476,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5254801511764526,
      "objective/train/original_loss": 1.5254799127578735,
      "objective/train/theoretical_loss": 3.7488853125712525,
      "objective/train/tokens_used": 839660000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2409975826740265,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502310991287231,
      "objective/train/weighted_lm_loss": 1.6022536754608154,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7488853125712525,
      "tokens_seen": 819200000
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007593781344032097,
      "loss": 1.5987,
      "theoretical_loss": 3.7488853125712525,
      "tokens_seen": 819200000
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000759358074222668,
      "loss": 1.3987,
      "theoretical_loss": 3.7488559359866716,
      "tokens_seen": 819265536
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007593380140421264,
      "loss": 1.5073,
      "theoretical_loss": 3.7488265624098585,
      "tokens_seen": 819331072
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007593179538615848,
      "loss": 1.5368,
      "theoretical_loss": 3.748797191840264,
      "tokens_seen": 819396608
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007592978936810432,
      "loss": 1.5007,
      "theoretical_loss": 3.7487678242773406,
      "tokens_seen": 819462144
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007592778335005015,
      "loss": 1.6073,
      "theoretical_loss": 3.7487384597205393,
      "tokens_seen": 819527680
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007592577733199599,
      "loss": 1.5138,
      "theoretical_loss": 3.7487090981693125,
      "tokens_seen": 819593216
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007592377131394182,
      "loss": 1.5633,
      "theoretical_loss": 3.748679739623113,
      "tokens_seen": 819658752
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007592176529588767,
      "loss": 1.5686,
      "theoretical_loss": 3.748650384081392,
      "tokens_seen": 819724288
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000759197592778335,
      "loss": 1.487,
      "theoretical_loss": 3.748621031543602,
      "tokens_seen": 819789824
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007591775325977934,
      "loss": 1.6486,
      "theoretical_loss": 3.748591682009196,
      "tokens_seen": 819855360
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007591574724172518,
      "loss": 1.5683,
      "theoretical_loss": 3.748562335477626,
      "tokens_seen": 819920896
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007591374122367102,
      "loss": 1.5026,
      "theoretical_loss": 3.7485329919483448,
      "tokens_seen": 819986432
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007591173520561685,
      "loss": 1.4518,
      "theoretical_loss": 3.748503651420806,
      "tokens_seen": 820051968
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007590972918756269,
      "loss": 1.4273,
      "theoretical_loss": 3.7484743138944623,
      "tokens_seen": 820117504
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007590772316950852,
      "loss": 1.5707,
      "theoretical_loss": 3.748444979368767,
      "tokens_seen": 820183040
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007590571715145437,
      "loss": 1.5804,
      "theoretical_loss": 3.7484156478431734,
      "tokens_seen": 820248576
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000759037111334002,
      "loss": 1.5473,
      "theoretical_loss": 3.748386319317135,
      "tokens_seen": 820314112
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007590170511534604,
      "loss": 1.4986,
      "theoretical_loss": 3.7483569937901056,
      "tokens_seen": 820379648
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007589969909729187,
      "loss": 1.4763,
      "theoretical_loss": 3.7483276712615385,
      "tokens_seen": 820445184
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007589769307923772,
      "loss": 1.5526,
      "theoretical_loss": 3.748298351730888,
      "tokens_seen": 820510720
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007589568706118355,
      "loss": 1.4832,
      "theoretical_loss": 3.748269035197609,
      "tokens_seen": 820576256
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007589368104312939,
      "loss": 1.6902,
      "theoretical_loss": 3.748239721661154,
      "tokens_seen": 820641792
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007589167502507523,
      "loss": 1.3717,
      "theoretical_loss": 3.7482104111209784,
      "tokens_seen": 820707328
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007588966900702107,
      "loss": 1.5717,
      "theoretical_loss": 3.748181103576537,
      "tokens_seen": 820772864
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000758876629889669,
      "loss": 1.507,
      "theoretical_loss": 3.748151799027284,
      "tokens_seen": 820838400
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007588565697091274,
      "loss": 1.4715,
      "theoretical_loss": 3.748122497472674,
      "tokens_seen": 820903936
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007588365095285857,
      "loss": 1.6636,
      "theoretical_loss": 3.748093198912162,
      "tokens_seen": 820969472
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007588164493480442,
      "loss": 1.6213,
      "theoretical_loss": 3.7480639033452032,
      "tokens_seen": 821035008
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007587963891675025,
      "loss": 1.5619,
      "theoretical_loss": 3.7480346107712537,
      "tokens_seen": 821100544
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007587763289869609,
      "loss": 1.485,
      "theoretical_loss": 3.7480053211897673,
      "tokens_seen": 821166080
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007587562688064193,
      "loss": 1.6679,
      "theoretical_loss": 3.7479760346002005,
      "tokens_seen": 821231616
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007587362086258776,
      "loss": 1.5947,
      "theoretical_loss": 3.747946751002009,
      "tokens_seen": 821297152
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000758716148445336,
      "loss": 1.6082,
      "theoretical_loss": 3.7479174703946474,
      "tokens_seen": 821362688
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007586960882647945,
      "loss": 1.516,
      "theoretical_loss": 3.747888192777573,
      "tokens_seen": 821428224
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007586760280842527,
      "loss": 1.6706,
      "theoretical_loss": 3.747858918150242,
      "tokens_seen": 821493760
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007586559679037112,
      "loss": 1.54,
      "theoretical_loss": 3.747829646512109,
      "tokens_seen": 821559296
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007586359077231694,
      "loss": 1.523,
      "theoretical_loss": 3.747800377862632,
      "tokens_seen": 821624832
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007586158475426279,
      "loss": 1.5787,
      "theoretical_loss": 3.747771112201267,
      "tokens_seen": 821690368
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007585957873620863,
      "loss": 1.5463,
      "theoretical_loss": 3.74774184952747,
      "tokens_seen": 821755904
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007585757271815446,
      "loss": 1.5385,
      "theoretical_loss": 3.747712589840699,
      "tokens_seen": 821821440
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000758555667001003,
      "loss": 1.4352,
      "theoretical_loss": 3.74768333314041,
      "tokens_seen": 821886976
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007585356068204615,
      "loss": 1.4425,
      "theoretical_loss": 3.7476540794260607,
      "tokens_seen": 821952512
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007585155466399197,
      "loss": 1.7064,
      "theoretical_loss": 3.7476248286971074,
      "tokens_seen": 822018048
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007584954864593782,
      "loss": 1.4926,
      "theoretical_loss": 3.7475955809530084,
      "tokens_seen": 822083584
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007584754262788364,
      "loss": 1.4311,
      "theoretical_loss": 3.7475663361932208,
      "tokens_seen": 822149120
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007584553660982949,
      "loss": 1.5341,
      "theoretical_loss": 3.747537094417202,
      "tokens_seen": 822214656
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007584353059177533,
      "loss": 1.4512,
      "theoretical_loss": 3.74750785562441,
      "tokens_seen": 822280192
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007584152457372116,
      "loss": 1.5068,
      "theoretical_loss": 3.747478619814303,
      "tokens_seen": 822345728
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00075839518555667,
      "loss": 1.5285,
      "theoretical_loss": 3.747449386986339,
      "tokens_seen": 822411264
    },
    {
      "epoch": 0.25,
      "objective/train/advantage_avg": 0.4885343909263611,
      "objective/train/docs_used": 303734,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4919618368148804,
      "objective/train/original_loss": 1.4919617176055908,
      "objective/train/theoretical_loss": 3.7474201571399757,
      "objective/train/tokens_used": 842936800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23959946632385254,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500714778900146,
      "objective/train/weighted_lm_loss": 1.5671418905258179,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7474201571399757,
      "tokens_seen": 822476800
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007583751253761284,
      "loss": 1.6235,
      "theoretical_loss": 3.7474201571399757,
      "tokens_seen": 822476800
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007583550651955867,
      "loss": 1.6065,
      "theoretical_loss": 3.7473909302746717,
      "tokens_seen": 822542336
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007583350050150452,
      "loss": 1.6384,
      "theoretical_loss": 3.7473617063898863,
      "tokens_seen": 822607872
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007583149448345034,
      "loss": 1.6183,
      "theoretical_loss": 3.747332485485077,
      "tokens_seen": 822673408
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007582948846539619,
      "loss": 1.416,
      "theoretical_loss": 3.747303267559703,
      "tokens_seen": 822738944
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007582748244734202,
      "loss": 1.6361,
      "theoretical_loss": 3.747274052613223,
      "tokens_seen": 822804480
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007582547642928786,
      "loss": 1.5246,
      "theoretical_loss": 3.747244840645097,
      "tokens_seen": 822870016
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007582347041123371,
      "loss": 1.5112,
      "theoretical_loss": 3.7472156316547824,
      "tokens_seen": 822935552
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007582146439317954,
      "loss": 1.5475,
      "theoretical_loss": 3.7471864256417406,
      "tokens_seen": 823001088
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007581945837512538,
      "loss": 1.5057,
      "theoretical_loss": 3.7471572226054297,
      "tokens_seen": 823066624
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007581745235707122,
      "loss": 1.5262,
      "theoretical_loss": 3.7471280225453096,
      "tokens_seen": 823132160
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007581544633901706,
      "loss": 1.6378,
      "theoretical_loss": 3.7470988254608404,
      "tokens_seen": 823197696
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007581344032096289,
      "loss": 1.6198,
      "theoretical_loss": 3.7470696313514815,
      "tokens_seen": 823263232
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007581143430290873,
      "loss": 1.4595,
      "theoretical_loss": 3.7470404402166935,
      "tokens_seen": 823328768
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007580942828485456,
      "loss": 1.4636,
      "theoretical_loss": 3.747011252055936,
      "tokens_seen": 823394304
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007580742226680041,
      "loss": 1.4993,
      "theoretical_loss": 3.74698206686867,
      "tokens_seen": 823459840
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007580541624874624,
      "loss": 1.4852,
      "theoretical_loss": 3.7469528846543554,
      "tokens_seen": 823525376
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007580341023069208,
      "loss": 1.6148,
      "theoretical_loss": 3.746923705412453,
      "tokens_seen": 823590912
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007580140421263791,
      "loss": 1.4813,
      "theoretical_loss": 3.746894529142424,
      "tokens_seen": 823656448
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007579939819458376,
      "loss": 1.5919,
      "theoretical_loss": 3.7468653558437284,
      "tokens_seen": 823721984
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007579739217652959,
      "loss": 1.5782,
      "theoretical_loss": 3.7468361855158276,
      "tokens_seen": 823787520
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007579538615847543,
      "loss": 1.4922,
      "theoretical_loss": 3.7468070181581834,
      "tokens_seen": 823853056
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007579338014042127,
      "loss": 1.4983,
      "theoretical_loss": 3.746777853770256,
      "tokens_seen": 823918592
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007579137412236711,
      "loss": 1.3805,
      "theoretical_loss": 3.746748692351508,
      "tokens_seen": 823984128
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007578936810431294,
      "loss": 1.5614,
      "theoretical_loss": 3.7467195339014,
      "tokens_seen": 824049664
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007578736208625878,
      "loss": 1.5291,
      "theoretical_loss": 3.7466903784193946,
      "tokens_seen": 824115200
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007578535606820461,
      "loss": 1.5984,
      "theoretical_loss": 3.746661225904953,
      "tokens_seen": 824180736
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007578335005015046,
      "loss": 1.5991,
      "theoretical_loss": 3.746632076357537,
      "tokens_seen": 824246272
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007578134403209629,
      "loss": 1.5619,
      "theoretical_loss": 3.7466029297766097,
      "tokens_seen": 824311808
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007577933801404213,
      "loss": 1.5119,
      "theoretical_loss": 3.7465737861616333,
      "tokens_seen": 824377344
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007577733199598797,
      "loss": 1.6109,
      "theoretical_loss": 3.746544645512069,
      "tokens_seen": 824442880
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000757753259779338,
      "loss": 1.6473,
      "theoretical_loss": 3.7465155078273806,
      "tokens_seen": 824508416
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007577331995987964,
      "loss": 1.5046,
      "theoretical_loss": 3.7464863731070306,
      "tokens_seen": 824573952
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007577131394182549,
      "loss": 1.4371,
      "theoretical_loss": 3.7464572413504813,
      "tokens_seen": 824639488
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007576930792377131,
      "loss": 1.5249,
      "theoretical_loss": 3.7464281125571963,
      "tokens_seen": 824705024
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007576730190571716,
      "loss": 1.586,
      "theoretical_loss": 3.7463989867266383,
      "tokens_seen": 824770560
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007576529588766298,
      "loss": 1.6034,
      "theoretical_loss": 3.746369863858271,
      "tokens_seen": 824836096
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007576328986960883,
      "loss": 1.466,
      "theoretical_loss": 3.7463407439515577,
      "tokens_seen": 824901632
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007576128385155467,
      "loss": 1.5183,
      "theoretical_loss": 3.7463116270059618,
      "tokens_seen": 824967168
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000757592778335005,
      "loss": 1.5485,
      "theoretical_loss": 3.746282513020947,
      "tokens_seen": 825032704
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007575727181544634,
      "loss": 1.5943,
      "theoretical_loss": 3.7462534019959772,
      "tokens_seen": 825098240
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007575526579739219,
      "loss": 1.4953,
      "theoretical_loss": 3.746224293930516,
      "tokens_seen": 825163776
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007575325977933801,
      "loss": 1.4286,
      "theoretical_loss": 3.7461951888240286,
      "tokens_seen": 825229312
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007575125376128386,
      "loss": 1.6536,
      "theoretical_loss": 3.746166086675978,
      "tokens_seen": 825294848
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007574924774322968,
      "loss": 1.4948,
      "theoretical_loss": 3.7461369874858295,
      "tokens_seen": 825360384
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007574724172517553,
      "loss": 1.4295,
      "theoretical_loss": 3.7461078912530468,
      "tokens_seen": 825425920
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007574523570712137,
      "loss": 1.6772,
      "theoretical_loss": 3.7460787979770958,
      "tokens_seen": 825491456
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000757432296890672,
      "loss": 1.5057,
      "theoretical_loss": 3.7460497076574395,
      "tokens_seen": 825556992
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007574122367101304,
      "loss": 1.5524,
      "theoretical_loss": 3.7460206202935447,
      "tokens_seen": 825622528
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007573921765295888,
      "loss": 1.6037,
      "theoretical_loss": 3.745991535884875,
      "tokens_seen": 825688064
    },
    {
      "epoch": 0.25,
      "objective/train/advantage_avg": 0.49512171745300293,
      "objective/train/docs_used": 304950,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4001202583312988,
      "objective/train/original_loss": 1.4001200199127197,
      "objective/train/theoretical_loss": 3.745962454430897,
      "objective/train/tokens_used": 846213600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2453310787677765,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507594347000122,
      "objective/train/weighted_lm_loss": 1.471086859703064,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.745962454430897,
      "tokens_seen": 825753600
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007573721163490471,
      "loss": 1.5368,
      "theoretical_loss": 3.745962454430897,
      "tokens_seen": 825753600
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007573520561685056,
      "loss": 1.6164,
      "theoretical_loss": 3.7459333759310756,
      "tokens_seen": 825819136
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007573319959879638,
      "loss": 1.6079,
      "theoretical_loss": 3.7459043003848755,
      "tokens_seen": 825884672
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007573119358074223,
      "loss": 1.5926,
      "theoretical_loss": 3.745875227791763,
      "tokens_seen": 825950208
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007572918756268806,
      "loss": 1.5571,
      "theoretical_loss": 3.745846158151204,
      "tokens_seen": 826015744
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000757271815446339,
      "loss": 1.5218,
      "theoretical_loss": 3.745817091462664,
      "tokens_seen": 826081280
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007572517552657974,
      "loss": 1.5597,
      "theoretical_loss": 3.7457880277256095,
      "tokens_seen": 826146816
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007572316950852558,
      "loss": 1.5633,
      "theoretical_loss": 3.7457589669395066,
      "tokens_seen": 826212352
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007572116349047141,
      "loss": 1.5648,
      "theoretical_loss": 3.7457299091038214,
      "tokens_seen": 826277888
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007571915747241726,
      "loss": 1.4567,
      "theoretical_loss": 3.7457008542180206,
      "tokens_seen": 826343424
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007571715145436309,
      "loss": 1.5318,
      "theoretical_loss": 3.7456718022815707,
      "tokens_seen": 826408960
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007571514543630893,
      "loss": 1.4959,
      "theoretical_loss": 3.7456427532939385,
      "tokens_seen": 826474496
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007571313941825476,
      "loss": 1.5591,
      "theoretical_loss": 3.745613707254591,
      "tokens_seen": 826540032
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000757111334002006,
      "loss": 1.3853,
      "theoretical_loss": 3.745584664162995,
      "tokens_seen": 826605568
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007570912738214644,
      "loss": 1.5945,
      "theoretical_loss": 3.745555624018618,
      "tokens_seen": 826671104
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007570712136409228,
      "loss": 1.5657,
      "theoretical_loss": 3.7455265868209273,
      "tokens_seen": 826736640
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007570511534603811,
      "loss": 1.6343,
      "theoretical_loss": 3.7454975525693897,
      "tokens_seen": 826802176
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007570310932798395,
      "loss": 1.5318,
      "theoretical_loss": 3.745468521263474,
      "tokens_seen": 826867712
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007570110330992979,
      "loss": 1.497,
      "theoretical_loss": 3.7454394929026464,
      "tokens_seen": 826933248
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007569909729187563,
      "loss": 1.4725,
      "theoretical_loss": 3.7454104674863764,
      "tokens_seen": 826998784
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007569709127382146,
      "loss": 1.6375,
      "theoretical_loss": 3.7453814450141305,
      "tokens_seen": 827064320
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000756950852557673,
      "loss": 1.6415,
      "theoretical_loss": 3.745352425485378,
      "tokens_seen": 827129856
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007569307923771313,
      "loss": 1.6891,
      "theoretical_loss": 3.745323408899586,
      "tokens_seen": 827195392
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007569107321965898,
      "loss": 1.5985,
      "theoretical_loss": 3.7452943952562245,
      "tokens_seen": 827260928
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007568906720160481,
      "loss": 1.6089,
      "theoretical_loss": 3.7452653845547603,
      "tokens_seen": 827326464
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007568706118355065,
      "loss": 1.6226,
      "theoretical_loss": 3.7452363767946637,
      "tokens_seen": 827392000
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007568505516549649,
      "loss": 1.721,
      "theoretical_loss": 3.7452073719754027,
      "tokens_seen": 827457536
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007568304914744233,
      "loss": 1.596,
      "theoretical_loss": 3.745178370096446,
      "tokens_seen": 827523072
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007568104312938816,
      "loss": 1.5447,
      "theoretical_loss": 3.745149371157263,
      "tokens_seen": 827588608
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007567903711133401,
      "loss": 1.5337,
      "theoretical_loss": 3.745120375157324,
      "tokens_seen": 827654144
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007567703109327983,
      "loss": 1.5808,
      "theoretical_loss": 3.7450913820960965,
      "tokens_seen": 827719680
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007567502507522568,
      "loss": 1.5922,
      "theoretical_loss": 3.7450623919730512,
      "tokens_seen": 827785216
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007567301905717153,
      "loss": 1.604,
      "theoretical_loss": 3.7450334047876574,
      "tokens_seen": 827850752
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007567101303911735,
      "loss": 1.554,
      "theoretical_loss": 3.7450044205393853,
      "tokens_seen": 827916288
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000756690070210632,
      "loss": 1.604,
      "theoretical_loss": 3.744975439227704,
      "tokens_seen": 827981824
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007566700100300902,
      "loss": 1.5666,
      "theoretical_loss": 3.744946460852085,
      "tokens_seen": 828047360
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007566499498495487,
      "loss": 1.6043,
      "theoretical_loss": 3.744917485411997,
      "tokens_seen": 828112896
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007566298896690071,
      "loss": 1.5168,
      "theoretical_loss": 3.744888512906911,
      "tokens_seen": 828178432
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007566098294884654,
      "loss": 1.6387,
      "theoretical_loss": 3.7448595433362977,
      "tokens_seen": 828243968
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007565897693079238,
      "loss": 1.6656,
      "theoretical_loss": 3.744830576699627,
      "tokens_seen": 828309504
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007565697091273823,
      "loss": 1.7129,
      "theoretical_loss": 3.744801612996371,
      "tokens_seen": 828375040
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007565496489468405,
      "loss": 1.643,
      "theoretical_loss": 3.7447726522259988,
      "tokens_seen": 828440576
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000756529588766299,
      "loss": 1.5873,
      "theoretical_loss": 3.744743694387983,
      "tokens_seen": 828506112
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007565095285857572,
      "loss": 1.4557,
      "theoretical_loss": 3.7447147394817937,
      "tokens_seen": 828571648
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007564894684052157,
      "loss": 1.526,
      "theoretical_loss": 3.744685787506903,
      "tokens_seen": 828637184
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007564694082246741,
      "loss": 1.5859,
      "theoretical_loss": 3.744656838462782,
      "tokens_seen": 828702720
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007564493480441324,
      "loss": 1.5619,
      "theoretical_loss": 3.744627892348902,
      "tokens_seen": 828768256
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007564292878635908,
      "loss": 1.5186,
      "theoretical_loss": 3.744598949164735,
      "tokens_seen": 828833792
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007564092276830492,
      "loss": 1.6357,
      "theoretical_loss": 3.7445700089097533,
      "tokens_seen": 828899328
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007563891675025075,
      "loss": 1.503,
      "theoretical_loss": 3.7445410715834284,
      "tokens_seen": 828964864
    },
    {
      "epoch": 0.25,
      "objective/train/advantage_avg": 0.49200740456581116,
      "objective/train/docs_used": 306096,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3664567470550537,
      "objective/train/original_loss": 1.3664568662643433,
      "objective/train/theoretical_loss": 3.7445121371852323,
      "objective/train/tokens_used": 849490400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24247290194034576,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504332780838013,
      "objective/train/weighted_lm_loss": 1.4354283809661865,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7445121371852323,
      "tokens_seen": 829030400
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000756369107321966,
      "loss": 1.5432,
      "theoretical_loss": 3.7445121371852323,
      "tokens_seen": 829030400
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007563490471414242,
      "loss": 1.4903,
      "theoretical_loss": 3.7444832057146376,
      "tokens_seen": 829095936
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007563289869608827,
      "loss": 1.6645,
      "theoretical_loss": 3.7444542771711165,
      "tokens_seen": 829161472
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000756308926780341,
      "loss": 1.5253,
      "theoretical_loss": 3.744425351554142,
      "tokens_seen": 829227008
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007562888665997994,
      "loss": 1.442,
      "theoretical_loss": 3.744396428863186,
      "tokens_seen": 829292544
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007562688064192578,
      "loss": 1.5027,
      "theoretical_loss": 3.7443675090977218,
      "tokens_seen": 829358080
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007562487462387162,
      "loss": 1.5631,
      "theoretical_loss": 3.744338592257222,
      "tokens_seen": 829423616
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007562286860581745,
      "loss": 1.5855,
      "theoretical_loss": 3.7443096783411605,
      "tokens_seen": 829489152
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000756208625877633,
      "loss": 1.4404,
      "theoretical_loss": 3.744280767349009,
      "tokens_seen": 829554688
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007561885656970913,
      "loss": 1.5844,
      "theoretical_loss": 3.7442518592802427,
      "tokens_seen": 829620224
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007561685055165497,
      "loss": 1.5509,
      "theoretical_loss": 3.744222954134334,
      "tokens_seen": 829685760
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000756148445336008,
      "loss": 1.4044,
      "theoretical_loss": 3.744194051910757,
      "tokens_seen": 829751296
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007561283851554664,
      "loss": 1.4598,
      "theoretical_loss": 3.7441651526089847,
      "tokens_seen": 829816832
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007561083249749248,
      "loss": 1.6492,
      "theoretical_loss": 3.744136256228492,
      "tokens_seen": 829882368
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007560882647943832,
      "loss": 1.5154,
      "theoretical_loss": 3.7441073627687524,
      "tokens_seen": 829947904
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007560682046138415,
      "loss": 1.5396,
      "theoretical_loss": 3.7440784722292406,
      "tokens_seen": 830013440
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007560481444332999,
      "loss": 1.5506,
      "theoretical_loss": 3.7440495846094297,
      "tokens_seen": 830078976
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007560280842527583,
      "loss": 1.5914,
      "theoretical_loss": 3.7440206999087953,
      "tokens_seen": 830144512
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007560080240722167,
      "loss": 1.5604,
      "theoretical_loss": 3.743991818126812,
      "tokens_seen": 830210048
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000755987963891675,
      "loss": 1.443,
      "theoretical_loss": 3.7439629392629534,
      "tokens_seen": 830275584
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007559679037111335,
      "loss": 1.5696,
      "theoretical_loss": 3.7439340633166958,
      "tokens_seen": 830341120
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007559478435305917,
      "loss": 1.5371,
      "theoretical_loss": 3.743905190287513,
      "tokens_seen": 830406656
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007559277833500502,
      "loss": 1.6187,
      "theoretical_loss": 3.7438763201748815,
      "tokens_seen": 830472192
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007559077231695085,
      "loss": 1.5046,
      "theoretical_loss": 3.743847452978275,
      "tokens_seen": 830537728
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007558876629889669,
      "loss": 1.5162,
      "theoretical_loss": 3.7438185886971693,
      "tokens_seen": 830603264
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007558676028084253,
      "loss": 1.5431,
      "theoretical_loss": 3.743789727331041,
      "tokens_seen": 830668800
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007558475426278837,
      "loss": 1.6141,
      "theoretical_loss": 3.743760868879365,
      "tokens_seen": 830734336
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000755827482447342,
      "loss": 1.4622,
      "theoretical_loss": 3.7437320133416168,
      "tokens_seen": 830799872
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007558074222668005,
      "loss": 1.493,
      "theoretical_loss": 3.743703160717273,
      "tokens_seen": 830865408
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007557873620862587,
      "loss": 1.5722,
      "theoretical_loss": 3.743674311005809,
      "tokens_seen": 830930944
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007557673019057172,
      "loss": 1.5258,
      "theoretical_loss": 3.743645464206702,
      "tokens_seen": 830996480
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007557472417251755,
      "loss": 1.5966,
      "theoretical_loss": 3.743616620319427,
      "tokens_seen": 831062016
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007557271815446339,
      "loss": 1.4711,
      "theoretical_loss": 3.7435877793434615,
      "tokens_seen": 831127552
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007557071213640923,
      "loss": 1.469,
      "theoretical_loss": 3.743558941278282,
      "tokens_seen": 831193088
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007556870611835506,
      "loss": 1.6031,
      "theoretical_loss": 3.743530106123365,
      "tokens_seen": 831258624
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000755667001003009,
      "loss": 1.4962,
      "theoretical_loss": 3.743501273878188,
      "tokens_seen": 831324160
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007556469408224675,
      "loss": 1.5808,
      "theoretical_loss": 3.743472444542227,
      "tokens_seen": 831389696
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007556268806419257,
      "loss": 1.4925,
      "theoretical_loss": 3.74344361811496,
      "tokens_seen": 831455232
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007556068204613842,
      "loss": 1.5592,
      "theoretical_loss": 3.7434147945958642,
      "tokens_seen": 831520768
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007555867602808426,
      "loss": 1.4975,
      "theoretical_loss": 3.7433859739844166,
      "tokens_seen": 831586304
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007555667001003009,
      "loss": 1.5831,
      "theoretical_loss": 3.743357156280095,
      "tokens_seen": 831651840
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007555466399197593,
      "loss": 1.6913,
      "theoretical_loss": 3.743328341482377,
      "tokens_seen": 831717376
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007555265797392176,
      "loss": 1.4423,
      "theoretical_loss": 3.7432995295907405,
      "tokens_seen": 831782912
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000755506519558676,
      "loss": 1.4329,
      "theoretical_loss": 3.743270720604664,
      "tokens_seen": 831848448
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007554864593781345,
      "loss": 1.6817,
      "theoretical_loss": 3.743241914523625,
      "tokens_seen": 831913984
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007554663991975927,
      "loss": 1.5762,
      "theoretical_loss": 3.7432131113471017,
      "tokens_seen": 831979520
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007554463390170512,
      "loss": 1.6363,
      "theoretical_loss": 3.7431843110745726,
      "tokens_seen": 832045056
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007554262788365095,
      "loss": 1.5311,
      "theoretical_loss": 3.7431555137055166,
      "tokens_seen": 832110592
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007554062186559679,
      "loss": 1.4961,
      "theoretical_loss": 3.743126719239412,
      "tokens_seen": 832176128
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007553861584754263,
      "loss": 1.5629,
      "theoretical_loss": 3.7430979276757377,
      "tokens_seen": 832241664
    },
    {
      "epoch": 0.25,
      "objective/train/advantage_avg": 0.4912421405315399,
      "objective/train/docs_used": 307306,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4710862636566162,
      "objective/train/original_loss": 1.471086025238037,
      "objective/train/theoretical_loss": 3.743069139013972,
      "objective/train/tokens_used": 852767200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24175754189491272,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503530502319336,
      "objective/train/weighted_lm_loss": 1.545080542564392,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.743069139013972,
      "tokens_seen": 832307200
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007553660982948846,
      "loss": 1.4833,
      "theoretical_loss": 3.743069139013972,
      "tokens_seen": 832307200
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000755346038114343,
      "loss": 1.5636,
      "theoretical_loss": 3.743040353253595,
      "tokens_seen": 832372736
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007553259779338014,
      "loss": 1.5124,
      "theoretical_loss": 3.7430115703940854,
      "tokens_seen": 832438272
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007553059177532597,
      "loss": 1.4776,
      "theoretical_loss": 3.7429827904349224,
      "tokens_seen": 832503808
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007552858575727182,
      "loss": 1.6196,
      "theoretical_loss": 3.742954013375586,
      "tokens_seen": 832569344
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007552657973921765,
      "loss": 1.5383,
      "theoretical_loss": 3.7429252392155545,
      "tokens_seen": 832634880
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007552457372116349,
      "loss": 1.6408,
      "theoretical_loss": 3.742896467954309,
      "tokens_seen": 832700416
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007552256770310934,
      "loss": 1.5152,
      "theoretical_loss": 3.7428676995913293,
      "tokens_seen": 832765952
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007552056168505517,
      "loss": 1.5235,
      "theoretical_loss": 3.742838934126094,
      "tokens_seen": 832831488
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007551855566700101,
      "loss": 1.6595,
      "theoretical_loss": 3.7428101715580846,
      "tokens_seen": 832897024
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007551654964894684,
      "loss": 1.5117,
      "theoretical_loss": 3.7427814118867815,
      "tokens_seen": 832962560
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007551454363089268,
      "loss": 1.443,
      "theoretical_loss": 3.742752655111664,
      "tokens_seen": 833028096
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007551253761283852,
      "loss": 1.5752,
      "theoretical_loss": 3.742723901232213,
      "tokens_seen": 833093632
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007551053159478436,
      "loss": 1.5348,
      "theoretical_loss": 3.7426951502479096,
      "tokens_seen": 833159168
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007550852557673019,
      "loss": 1.5254,
      "theoretical_loss": 3.7426664021582337,
      "tokens_seen": 833224704
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007550651955867603,
      "loss": 1.5587,
      "theoretical_loss": 3.7426376569626676,
      "tokens_seen": 833290240
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007550451354062187,
      "loss": 1.5755,
      "theoretical_loss": 3.742608914660692,
      "tokens_seen": 833355776
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007550250752256771,
      "loss": 1.5999,
      "theoretical_loss": 3.7425801752517867,
      "tokens_seen": 833421312
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007550050150451354,
      "loss": 1.5386,
      "theoretical_loss": 3.7425514387354344,
      "tokens_seen": 833486848
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007549849548645939,
      "loss": 1.5692,
      "theoretical_loss": 3.7425227051111163,
      "tokens_seen": 833552384
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007549648946840521,
      "loss": 1.6324,
      "theoretical_loss": 3.742493974378314,
      "tokens_seen": 833617920
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007549448345035106,
      "loss": 1.5592,
      "theoretical_loss": 3.742465246536509,
      "tokens_seen": 833683456
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007549247743229689,
      "loss": 1.6065,
      "theoretical_loss": 3.7424365215851836,
      "tokens_seen": 833748992
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007549047141424273,
      "loss": 1.487,
      "theoretical_loss": 3.7424077995238196,
      "tokens_seen": 833814528
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007548846539618857,
      "loss": 1.4314,
      "theoretical_loss": 3.742379080351899,
      "tokens_seen": 833880064
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007548645937813441,
      "loss": 1.595,
      "theoretical_loss": 3.742350364068904,
      "tokens_seen": 833945600
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007548445336008024,
      "loss": 1.5951,
      "theoretical_loss": 3.7423216506743167,
      "tokens_seen": 834011136
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007548244734202609,
      "loss": 1.5699,
      "theoretical_loss": 3.7422929401676206,
      "tokens_seen": 834076672
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007548044132397191,
      "loss": 1.6379,
      "theoretical_loss": 3.7422642325482975,
      "tokens_seen": 834142208
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007547843530591776,
      "loss": 1.6456,
      "theoretical_loss": 3.742235527815831,
      "tokens_seen": 834207744
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000754764292878636,
      "loss": 1.6048,
      "theoretical_loss": 3.7422068259697037,
      "tokens_seen": 834273280
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007547442326980943,
      "loss": 1.6854,
      "theoretical_loss": 3.7421781270093986,
      "tokens_seen": 834338816
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007547241725175527,
      "loss": 1.6284,
      "theoretical_loss": 3.742149430934398,
      "tokens_seen": 834404352
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000754704112337011,
      "loss": 1.6139,
      "theoretical_loss": 3.742120737744187,
      "tokens_seen": 834469888
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007546840521564694,
      "loss": 1.6395,
      "theoretical_loss": 3.742092047438248,
      "tokens_seen": 834535424
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007546639919759279,
      "loss": 1.6599,
      "theoretical_loss": 3.7420633600160644,
      "tokens_seen": 834600960
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007546439317953861,
      "loss": 1.6314,
      "theoretical_loss": 3.7420346754771208,
      "tokens_seen": 834666496
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007546238716148446,
      "loss": 1.5749,
      "theoretical_loss": 3.7420059938209,
      "tokens_seen": 834732032
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000754603811434303,
      "loss": 1.5115,
      "theoretical_loss": 3.741977315046887,
      "tokens_seen": 834797568
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007545837512537613,
      "loss": 1.6436,
      "theoretical_loss": 3.7419486391545655,
      "tokens_seen": 834863104
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007545636910732197,
      "loss": 1.4517,
      "theoretical_loss": 3.7419199661434197,
      "tokens_seen": 834928640
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000754543630892678,
      "loss": 1.5856,
      "theoretical_loss": 3.741891296012934,
      "tokens_seen": 834994176
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007545235707121364,
      "loss": 1.6453,
      "theoretical_loss": 3.7418626287625933,
      "tokens_seen": 835059712
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007545035105315949,
      "loss": 1.5438,
      "theoretical_loss": 3.7418339643918817,
      "tokens_seen": 835125248
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007544834503510531,
      "loss": 1.557,
      "theoretical_loss": 3.7418053029002842,
      "tokens_seen": 835190784
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007544633901705116,
      "loss": 1.4746,
      "theoretical_loss": 3.741776644287286,
      "tokens_seen": 835256320
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007544433299899699,
      "loss": 1.5006,
      "theoretical_loss": 3.7417479885523717,
      "tokens_seen": 835321856
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007544232698094283,
      "loss": 1.5148,
      "theoretical_loss": 3.741719335695027,
      "tokens_seen": 835387392
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007544032096288867,
      "loss": 1.4897,
      "theoretical_loss": 3.7416906857147367,
      "tokens_seen": 835452928
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000754383149448345,
      "loss": 1.593,
      "theoretical_loss": 3.741662038610987,
      "tokens_seen": 835518464
    },
    {
      "epoch": 0.25,
      "objective/train/advantage_avg": 0.48869842290878296,
      "objective/train/docs_used": 308483,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.430486798286438,
      "objective/train/original_loss": 1.430486798286438,
      "objective/train/theoretical_loss": 3.741633394383263,
      "objective/train/tokens_used": 856044000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23980596661567688,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500887632369995,
      "objective/train/weighted_lm_loss": 1.501882553100586,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.741633394383263,
      "tokens_seen": 835584000
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007543630892678034,
      "loss": 1.4851,
      "theoretical_loss": 3.741633394383263,
      "tokens_seen": 835584000
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007543430290872618,
      "loss": 1.459,
      "theoretical_loss": 3.7416047530310506,
      "tokens_seen": 835649536
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007543229689067201,
      "loss": 1.592,
      "theoretical_loss": 3.741576114553835,
      "tokens_seen": 835715072
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007543029087261786,
      "loss": 1.6269,
      "theoretical_loss": 3.741547478951103,
      "tokens_seen": 835780608
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007542828485456369,
      "loss": 1.5377,
      "theoretical_loss": 3.7415188462223403,
      "tokens_seen": 835846144
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007542627883650953,
      "loss": 1.6276,
      "theoretical_loss": 3.7414902163670334,
      "tokens_seen": 835911680
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007542427281845537,
      "loss": 1.504,
      "theoretical_loss": 3.7414615893846683,
      "tokens_seen": 835977216
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000754222668004012,
      "loss": 1.543,
      "theoretical_loss": 3.741432965274732,
      "tokens_seen": 836042752
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007542026078234704,
      "loss": 1.5845,
      "theoretical_loss": 3.7414043440367113,
      "tokens_seen": 836108288
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007541825476429288,
      "loss": 1.6035,
      "theoretical_loss": 3.741375725670092,
      "tokens_seen": 836173824
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007541624874623871,
      "loss": 1.4841,
      "theoretical_loss": 3.741347110174362,
      "tokens_seen": 836239360
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007541424272818456,
      "loss": 1.5724,
      "theoretical_loss": 3.741318497549008,
      "tokens_seen": 836304896
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007541223671013039,
      "loss": 1.521,
      "theoretical_loss": 3.741289887793517,
      "tokens_seen": 836370432
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007541023069207623,
      "loss": 1.6116,
      "theoretical_loss": 3.741261280907376,
      "tokens_seen": 836435968
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007540822467402206,
      "loss": 1.5549,
      "theoretical_loss": 3.741232676890074,
      "tokens_seen": 836501504
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007540621865596791,
      "loss": 1.5367,
      "theoretical_loss": 3.7412040757410967,
      "tokens_seen": 836567040
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007540421263791374,
      "loss": 1.5972,
      "theoretical_loss": 3.741175477459933,
      "tokens_seen": 836632576
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007540220661985958,
      "loss": 1.639,
      "theoretical_loss": 3.7411468820460705,
      "tokens_seen": 836698112
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007540020060180541,
      "loss": 1.5806,
      "theoretical_loss": 3.7411182894989965,
      "tokens_seen": 836763648
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007539819458375125,
      "loss": 1.5557,
      "theoretical_loss": 3.7410896998181995,
      "tokens_seen": 836829184
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007539618856569709,
      "loss": 1.7171,
      "theoretical_loss": 3.741061113003168,
      "tokens_seen": 836894720
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007539418254764293,
      "loss": 1.4715,
      "theoretical_loss": 3.74103252905339,
      "tokens_seen": 836960256
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007539217652958876,
      "loss": 1.5927,
      "theoretical_loss": 3.7410039479683546,
      "tokens_seen": 837025792
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007539017051153461,
      "loss": 1.5528,
      "theoretical_loss": 3.7409753697475496,
      "tokens_seen": 837091328
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007538816449348044,
      "loss": 1.6177,
      "theoretical_loss": 3.7409467943904646,
      "tokens_seen": 837156864
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007538615847542628,
      "loss": 1.5009,
      "theoretical_loss": 3.740918221896588,
      "tokens_seen": 837222400
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007538415245737212,
      "loss": 1.4693,
      "theoretical_loss": 3.740889652265408,
      "tokens_seen": 837287936
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007538214643931795,
      "loss": 1.5493,
      "theoretical_loss": 3.740861085496416,
      "tokens_seen": 837353472
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007538014042126379,
      "loss": 1.5485,
      "theoretical_loss": 3.7408325215890987,
      "tokens_seen": 837419008
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007537813440320963,
      "loss": 1.5362,
      "theoretical_loss": 3.740803960542947,
      "tokens_seen": 837484544
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007537612838515546,
      "loss": 1.6278,
      "theoretical_loss": 3.7407754023574507,
      "tokens_seen": 837550080
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007537412236710131,
      "loss": 1.6681,
      "theoretical_loss": 3.7407468470320984,
      "tokens_seen": 837615616
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007537211634904714,
      "loss": 1.5349,
      "theoretical_loss": 3.7407182945663804,
      "tokens_seen": 837681152
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007537011033099298,
      "loss": 1.4744,
      "theoretical_loss": 3.740689744959787,
      "tokens_seen": 837746688
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007536810431293883,
      "loss": 1.5358,
      "theoretical_loss": 3.7406611982118076,
      "tokens_seen": 837812224
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007536609829488465,
      "loss": 1.6256,
      "theoretical_loss": 3.740632654321933,
      "tokens_seen": 837877760
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000753640922768305,
      "loss": 1.5036,
      "theoretical_loss": 3.740604113289653,
      "tokens_seen": 837943296
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007536208625877634,
      "loss": 1.6421,
      "theoretical_loss": 3.7405755751144585,
      "tokens_seen": 838008832
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007536008024072217,
      "loss": 1.4505,
      "theoretical_loss": 3.74054703979584,
      "tokens_seen": 838074368
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007535807422266801,
      "loss": 1.6085,
      "theoretical_loss": 3.740518507333288,
      "tokens_seen": 838139904
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007535606820461384,
      "loss": 1.5295,
      "theoretical_loss": 3.740489977726293,
      "tokens_seen": 838205440
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007535406218655968,
      "loss": 1.5317,
      "theoretical_loss": 3.740461450974347,
      "tokens_seen": 838270976
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007535205616850553,
      "loss": 1.6773,
      "theoretical_loss": 3.7404329270769403,
      "tokens_seen": 838336512
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007535005015045135,
      "loss": 1.533,
      "theoretical_loss": 3.740404406033565,
      "tokens_seen": 838402048
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000753480441323972,
      "loss": 1.561,
      "theoretical_loss": 3.740375887843711,
      "tokens_seen": 838467584
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007534603811434303,
      "loss": 1.5232,
      "theoretical_loss": 3.7403473725068714,
      "tokens_seen": 838533120
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007534403209628887,
      "loss": 1.5341,
      "theoretical_loss": 3.740318860022537,
      "tokens_seen": 838598656
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007534202607823471,
      "loss": 1.5059,
      "theoretical_loss": 3.7402903503901994,
      "tokens_seen": 838664192
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007534002006018054,
      "loss": 1.4549,
      "theoretical_loss": 3.7402618436093507,
      "tokens_seen": 838729728
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007533801404212638,
      "loss": 1.5546,
      "theoretical_loss": 3.7402333396794836,
      "tokens_seen": 838795264
    },
    {
      "epoch": 0.25,
      "objective/train/advantage_avg": 0.48842087388038635,
      "objective/train/docs_used": 309674,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6240943670272827,
      "objective/train/original_loss": 1.6240941286087036,
      "objective/train/theoretical_loss": 3.7402048386000892,
      "objective/train/tokens_used": 859320800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.239604189991951,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500600337982178,
      "objective/train/weighted_lm_loss": 1.7052289247512817,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7402048386000892,
      "tokens_seen": 838860800
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007533600802407222,
      "loss": 1.5896,
      "theoretical_loss": 3.7402048386000892,
      "tokens_seen": 838860800
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007533400200601805,
      "loss": 1.5204,
      "theoretical_loss": 3.7401763403706605,
      "tokens_seen": 838926336
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000753319959879639,
      "loss": 1.5178,
      "theoretical_loss": 3.7401478449906893,
      "tokens_seen": 838991872
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007532998996990973,
      "loss": 1.4565,
      "theoretical_loss": 3.7401193524596685,
      "tokens_seen": 839057408
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007532798395185557,
      "loss": 1.5001,
      "theoretical_loss": 3.740090862777091,
      "tokens_seen": 839122944
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007532597793380141,
      "loss": 1.6334,
      "theoretical_loss": 3.740062375942449,
      "tokens_seen": 839188480
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007532397191574725,
      "loss": 1.5466,
      "theoretical_loss": 3.7400338919552363,
      "tokens_seen": 839254016
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007532196589769308,
      "loss": 1.6565,
      "theoretical_loss": 3.7400054108149456,
      "tokens_seen": 839319552
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007531995987963892,
      "loss": 1.4955,
      "theoretical_loss": 3.7399769325210697,
      "tokens_seen": 839385088
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007531795386158475,
      "loss": 1.5765,
      "theoretical_loss": 3.739948457073102,
      "tokens_seen": 839450624
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000753159478435306,
      "loss": 1.5434,
      "theoretical_loss": 3.7399199844705366,
      "tokens_seen": 839516160
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007531394182547643,
      "loss": 1.4989,
      "theoretical_loss": 3.739891514712866,
      "tokens_seen": 839581696
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007531193580742227,
      "loss": 1.6399,
      "theoretical_loss": 3.7398630477995853,
      "tokens_seen": 839647232
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000753099297893681,
      "loss": 1.5399,
      "theoretical_loss": 3.7398345837301874,
      "tokens_seen": 839712768
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007530792377131395,
      "loss": 1.5446,
      "theoretical_loss": 3.7398061225041666,
      "tokens_seen": 839778304
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007530591775325978,
      "loss": 1.4925,
      "theoretical_loss": 3.739777664121016,
      "tokens_seen": 839843840
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007530391173520562,
      "loss": 1.5057,
      "theoretical_loss": 3.7397492085802315,
      "tokens_seen": 839909376
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007530190571715145,
      "loss": 1.5112,
      "theoretical_loss": 3.7397207558813066,
      "tokens_seen": 839974912
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007529989969909729,
      "loss": 1.51,
      "theoretical_loss": 3.739692306023736,
      "tokens_seen": 840040448
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007529789368104313,
      "loss": 1.5999,
      "theoretical_loss": 3.7396638590070133,
      "tokens_seen": 840105984
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007529588766298897,
      "loss": 1.5856,
      "theoretical_loss": 3.739635414830635,
      "tokens_seen": 840171520
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000752938816449348,
      "loss": 1.4993,
      "theoretical_loss": 3.739606973494095,
      "tokens_seen": 840237056
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007529187562688065,
      "loss": 1.5519,
      "theoretical_loss": 3.7395785349968884,
      "tokens_seen": 840302592
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007528986960882648,
      "loss": 1.5795,
      "theoretical_loss": 3.73955009933851,
      "tokens_seen": 840368128
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007528786359077232,
      "loss": 1.549,
      "theoretical_loss": 3.7395216665184554,
      "tokens_seen": 840433664
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007528585757271816,
      "loss": 1.5394,
      "theoretical_loss": 3.7394932365362203,
      "tokens_seen": 840499200
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007528385155466399,
      "loss": 1.6368,
      "theoretical_loss": 3.7394648093912997,
      "tokens_seen": 840564736
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007528184553660983,
      "loss": 1.6394,
      "theoretical_loss": 3.7394363850831898,
      "tokens_seen": 840630272
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007527983951855567,
      "loss": 1.6021,
      "theoretical_loss": 3.739407963611386,
      "tokens_seen": 840695808
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000752778335005015,
      "loss": 1.5265,
      "theoretical_loss": 3.739379544975384,
      "tokens_seen": 840761344
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007527582748244735,
      "loss": 1.456,
      "theoretical_loss": 3.73935112917468,
      "tokens_seen": 840826880
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007527382146439317,
      "loss": 1.588,
      "theoretical_loss": 3.739322716208771,
      "tokens_seen": 840892416
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007527181544633902,
      "loss": 1.4061,
      "theoretical_loss": 3.739294306077152,
      "tokens_seen": 840957952
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007526980942828486,
      "loss": 1.5727,
      "theoretical_loss": 3.73926589877932,
      "tokens_seen": 841023488
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007526780341023069,
      "loss": 1.6156,
      "theoretical_loss": 3.7392374943147715,
      "tokens_seen": 841089024
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007526579739217653,
      "loss": 1.5839,
      "theoretical_loss": 3.7392090926830037,
      "tokens_seen": 841154560
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007526379137412238,
      "loss": 1.5879,
      "theoretical_loss": 3.7391806938835126,
      "tokens_seen": 841220096
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000752617853560682,
      "loss": 1.5525,
      "theoretical_loss": 3.739152297915796,
      "tokens_seen": 841285632
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007525977933801405,
      "loss": 1.5303,
      "theoretical_loss": 3.7391239047793494,
      "tokens_seen": 841351168
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007525777331995987,
      "loss": 1.5798,
      "theoretical_loss": 3.7390955144736724,
      "tokens_seen": 841416704
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0007525576730190572,
      "loss": 1.4629,
      "theoretical_loss": 3.7390671269982603,
      "tokens_seen": 841482240
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007525376128385156,
      "loss": 1.5027,
      "theoretical_loss": 3.7390387423526112,
      "tokens_seen": 841547776
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007525175526579739,
      "loss": 1.5406,
      "theoretical_loss": 3.7390103605362226,
      "tokens_seen": 841613312
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007524974924774323,
      "loss": 1.564,
      "theoretical_loss": 3.7389819815485925,
      "tokens_seen": 841678848
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007524774322968907,
      "loss": 1.583,
      "theoretical_loss": 3.7389536053892187,
      "tokens_seen": 841744384
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000752457372116349,
      "loss": 1.5158,
      "theoretical_loss": 3.738925232057599,
      "tokens_seen": 841809920
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007524373119358075,
      "loss": 1.4892,
      "theoretical_loss": 3.7388968615532314,
      "tokens_seen": 841875456
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007524172517552657,
      "loss": 1.4491,
      "theoretical_loss": 3.7388684938756143,
      "tokens_seen": 841940992
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007523971915747242,
      "loss": 1.5758,
      "theoretical_loss": 3.738840129024246,
      "tokens_seen": 842006528
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007523771313941825,
      "loss": 1.5729,
      "theoretical_loss": 3.7388117669986256,
      "tokens_seen": 842072064
    },
    {
      "epoch": 0.26,
      "objective/train/advantage_avg": 0.497471421957016,
      "objective/train/docs_used": 310992,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3830622434616089,
      "objective/train/original_loss": 1.3830621242523193,
      "objective/train/theoretical_loss": 3.73878340779825,
      "objective/train/tokens_used": 862597600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24755065143108368,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0510058403015137,
      "objective/train/weighted_lm_loss": 1.45350182056427,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.73878340779825,
      "tokens_seen": 842137600
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007523570712136409,
      "loss": 1.5498,
      "theoretical_loss": 3.73878340779825,
      "tokens_seen": 842137600
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007523370110330993,
      "loss": 1.511,
      "theoretical_loss": 3.73875505142262,
      "tokens_seen": 842203136
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007523169508525577,
      "loss": 1.6271,
      "theoretical_loss": 3.738726697871233,
      "tokens_seen": 842268672
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000752296890672016,
      "loss": 1.5885,
      "theoretical_loss": 3.738698347143589,
      "tokens_seen": 842334208
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007522768304914745,
      "loss": 1.6028,
      "theoretical_loss": 3.7386699992391863,
      "tokens_seen": 842399744
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007522567703109327,
      "loss": 1.6317,
      "theoretical_loss": 3.738641654157524,
      "tokens_seen": 842465280
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007522367101303912,
      "loss": 1.6164,
      "theoretical_loss": 3.738613311898103,
      "tokens_seen": 842530816
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007522166499498496,
      "loss": 1.5632,
      "theoretical_loss": 3.738584972460421,
      "tokens_seen": 842596352
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007521965897693079,
      "loss": 1.5123,
      "theoretical_loss": 3.7385566358439783,
      "tokens_seen": 842661888
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007521765295887664,
      "loss": 1.5289,
      "theoretical_loss": 3.738528302048275,
      "tokens_seen": 842727424
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007521564694082247,
      "loss": 1.5827,
      "theoretical_loss": 3.7384999710728106,
      "tokens_seen": 842792960
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007521364092276831,
      "loss": 1.5191,
      "theoretical_loss": 3.7384716429170854,
      "tokens_seen": 842858496
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007521163490471414,
      "loss": 1.6011,
      "theoretical_loss": 3.7384433175805993,
      "tokens_seen": 842924032
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007520962888665999,
      "loss": 1.5683,
      "theoretical_loss": 3.7384149950628522,
      "tokens_seen": 842989568
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007520762286860582,
      "loss": 1.5471,
      "theoretical_loss": 3.738386675363346,
      "tokens_seen": 843055104
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007520561685055166,
      "loss": 1.4593,
      "theoretical_loss": 3.7383583584815794,
      "tokens_seen": 843120640
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000752036108324975,
      "loss": 1.5507,
      "theoretical_loss": 3.7383300444170535,
      "tokens_seen": 843186176
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007520160481444333,
      "loss": 1.6072,
      "theoretical_loss": 3.73830173316927,
      "tokens_seen": 843251712
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007519959879638917,
      "loss": 1.5783,
      "theoretical_loss": 3.738273424737729,
      "tokens_seen": 843317248
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007519759277833501,
      "loss": 1.6217,
      "theoretical_loss": 3.738245119121932,
      "tokens_seen": 843382784
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007519558676028084,
      "loss": 1.5077,
      "theoretical_loss": 3.7382168163213794,
      "tokens_seen": 843448320
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007519358074222669,
      "loss": 1.643,
      "theoretical_loss": 3.7381885163355735,
      "tokens_seen": 843513856
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007519157472417252,
      "loss": 1.6405,
      "theoretical_loss": 3.7381602191640146,
      "tokens_seen": 843579392
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007518956870611836,
      "loss": 1.5195,
      "theoretical_loss": 3.738131924806205,
      "tokens_seen": 843644928
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000751875626880642,
      "loss": 1.6215,
      "theoretical_loss": 3.7381036332616464,
      "tokens_seen": 843710464
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007518555667001003,
      "loss": 1.5798,
      "theoretical_loss": 3.7380753445298405,
      "tokens_seen": 843776000
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007518355065195587,
      "loss": 1.5071,
      "theoretical_loss": 3.738047058610289,
      "tokens_seen": 843841536
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007518154463390171,
      "loss": 1.5521,
      "theoretical_loss": 3.738018775502494,
      "tokens_seen": 843907072
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007517953861584754,
      "loss": 1.5729,
      "theoretical_loss": 3.737990495205958,
      "tokens_seen": 843972608
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007517753259779339,
      "loss": 1.547,
      "theoretical_loss": 3.737962217720183,
      "tokens_seen": 844038144
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007517552657973921,
      "loss": 1.5505,
      "theoretical_loss": 3.7379339430446707,
      "tokens_seen": 844103680
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007517352056168506,
      "loss": 1.4746,
      "theoretical_loss": 3.7379056711789254,
      "tokens_seen": 844169216
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000751715145436309,
      "loss": 1.4957,
      "theoretical_loss": 3.7378774021224483,
      "tokens_seen": 844234752
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007516950852557673,
      "loss": 1.5137,
      "theoretical_loss": 3.7378491358747428,
      "tokens_seen": 844300288
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007516750250752257,
      "loss": 1.5961,
      "theoretical_loss": 3.7378208724353117,
      "tokens_seen": 844365824
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007516549648946842,
      "loss": 1.4582,
      "theoretical_loss": 3.737792611803658,
      "tokens_seen": 844431360
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007516349047141424,
      "loss": 1.5893,
      "theoretical_loss": 3.7377643539792853,
      "tokens_seen": 844496896
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007516148445336009,
      "loss": 1.7006,
      "theoretical_loss": 3.7377360989616966,
      "tokens_seen": 844562432
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007515947843530591,
      "loss": 1.5392,
      "theoretical_loss": 3.7377078467503955,
      "tokens_seen": 844627968
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007515747241725176,
      "loss": 1.5349,
      "theoretical_loss": 3.7376795973448855,
      "tokens_seen": 844693504
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000751554663991976,
      "loss": 1.6522,
      "theoretical_loss": 3.73765135074467,
      "tokens_seen": 844759040
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007515346038114343,
      "loss": 1.6435,
      "theoretical_loss": 3.7376231069492527,
      "tokens_seen": 844824576
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007515145436308927,
      "loss": 1.4764,
      "theoretical_loss": 3.737594865958138,
      "tokens_seen": 844890112
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000751494483450351,
      "loss": 1.5079,
      "theoretical_loss": 3.73756662777083,
      "tokens_seen": 844955648
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007514744232698094,
      "loss": 1.536,
      "theoretical_loss": 3.737538392386833,
      "tokens_seen": 845021184
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007514543630892679,
      "loss": 1.5965,
      "theoretical_loss": 3.737510159805651,
      "tokens_seen": 845086720
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007514343029087261,
      "loss": 1.5188,
      "theoretical_loss": 3.7374819300267883,
      "tokens_seen": 845152256
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007514142427281846,
      "loss": 1.6061,
      "theoretical_loss": 3.73745370304975,
      "tokens_seen": 845217792
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007513941825476429,
      "loss": 1.5451,
      "theoretical_loss": 3.73742547887404,
      "tokens_seen": 845283328
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007513741223671013,
      "loss": 1.532,
      "theoretical_loss": 3.7373972574991643,
      "tokens_seen": 845348864
    },
    {
      "epoch": 0.26,
      "objective/train/advantage_avg": 0.49380365014076233,
      "objective/train/docs_used": 312167,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4567852020263672,
      "objective/train/original_loss": 1.4567852020263672,
      "objective/train/theoretical_loss": 3.7373690389246272,
      "objective/train/tokens_used": 865874400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24405476450920105,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506211519241333,
      "objective/train/weighted_lm_loss": 1.530515193939209,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0293704271316528,
      "theoretical_loss": 3.7373690389246272,
      "tokens_seen": 845414400
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007513540621865597,
      "loss": 1.5228,
      "theoretical_loss": 3.7373690389246272,
      "tokens_seen": 845414400
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007513340020060181,
      "loss": 1.6306,
      "theoretical_loss": 3.737340823149933,
      "tokens_seen": 845479936
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007513139418254764,
      "loss": 1.6652,
      "theoretical_loss": 3.7373126101745884,
      "tokens_seen": 845545472
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007512938816449349,
      "loss": 1.4816,
      "theoretical_loss": 3.737284399998098,
      "tokens_seen": 845611008
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007512738214643931,
      "loss": 1.5889,
      "theoretical_loss": 3.737256192619967,
      "tokens_seen": 845676544
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007512537612838516,
      "loss": 1.5589,
      "theoretical_loss": 3.7372279880397015,
      "tokens_seen": 845742080
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007512337011033099,
      "loss": 1.598,
      "theoretical_loss": 3.7371997862568067,
      "tokens_seen": 845807616
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007512136409227683,
      "loss": 1.4905,
      "theoretical_loss": 3.7371715872707894,
      "tokens_seen": 845873152
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007511935807422267,
      "loss": 1.4481,
      "theoretical_loss": 3.737143391081154,
      "tokens_seen": 845938688
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007511735205616851,
      "loss": 1.5655,
      "theoretical_loss": 3.737115197687408,
      "tokens_seen": 846004224
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007511534603811434,
      "loss": 1.5256,
      "theoretical_loss": 3.7370870070890567,
      "tokens_seen": 846069760
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007511334002006018,
      "loss": 1.5503,
      "theoretical_loss": 3.7370588192856067,
      "tokens_seen": 846135296
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007511133400200602,
      "loss": 1.5232,
      "theoretical_loss": 3.7370306342765653,
      "tokens_seen": 846200832
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007510932798395186,
      "loss": 1.4504,
      "theoretical_loss": 3.7370024520614376,
      "tokens_seen": 846266368
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007510732196589769,
      "loss": 1.5799,
      "theoretical_loss": 3.7369742726397313,
      "tokens_seen": 846331904
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007510531594784353,
      "loss": 1.4536,
      "theoretical_loss": 3.7369460960109535,
      "tokens_seen": 846397440
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007510330992978936,
      "loss": 1.4992,
      "theoretical_loss": 3.73691792217461,
      "tokens_seen": 846462976
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007510130391173521,
      "loss": 1.5582,
      "theoretical_loss": 3.7368897511302093,
      "tokens_seen": 846528512
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007509929789368104,
      "loss": 1.5726,
      "theoretical_loss": 3.736861582877257,
      "tokens_seen": 846594048
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007509729187562688,
      "loss": 1.6649,
      "theoretical_loss": 3.7368334174152618,
      "tokens_seen": 846659584
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007509528585757272,
      "loss": 1.528,
      "theoretical_loss": 3.7368052547437305,
      "tokens_seen": 846725120
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007509327983951856,
      "loss": 1.5241,
      "theoretical_loss": 3.736777094862171,
      "tokens_seen": 846790656
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007509127382146439,
      "loss": 1.6048,
      "theoretical_loss": 3.736748937770091,
      "tokens_seen": 846856192
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007508926780341024,
      "loss": 1.5251,
      "theoretical_loss": 3.736720783466998,
      "tokens_seen": 846921728
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007508726178535606,
      "loss": 1.5874,
      "theoretical_loss": 3.7366926319524003,
      "tokens_seen": 846987264
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007508525576730191,
      "loss": 1.4589,
      "theoretical_loss": 3.736664483225806,
      "tokens_seen": 847052800
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007508324974924774,
      "loss": 1.59,
      "theoretical_loss": 3.736636337286723,
      "tokens_seen": 847118336
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007508124373119358,
      "loss": 1.5507,
      "theoretical_loss": 3.73660819413466,
      "tokens_seen": 847183872
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007507923771313942,
      "loss": 1.4493,
      "theoretical_loss": 3.736580053769125,
      "tokens_seen": 847249408
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007507723169508525,
      "loss": 1.443,
      "theoretical_loss": 3.736551916189627,
      "tokens_seen": 847314944
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007507522567703109,
      "loss": 1.5318,
      "theoretical_loss": 3.736523781395675,
      "tokens_seen": 847380480
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007507321965897694,
      "loss": 1.4341,
      "theoretical_loss": 3.736495649386777,
      "tokens_seen": 847446016
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007507121364092277,
      "loss": 1.4859,
      "theoretical_loss": 3.736467520162442,
      "tokens_seen": 847511552
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007506920762286861,
      "loss": 1.4939,
      "theoretical_loss": 3.7364393937221805,
      "tokens_seen": 847577088
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007506720160481446,
      "loss": 1.6182,
      "theoretical_loss": 3.7364112700655,
      "tokens_seen": 847642624
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007506519558676028,
      "loss": 1.5679,
      "theoretical_loss": 3.736383149191911,
      "tokens_seen": 847708160
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007506318956870613,
      "loss": 1.5597,
      "theoretical_loss": 3.736355031100922,
      "tokens_seen": 847773696
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007506118355065195,
      "loss": 1.4896,
      "theoretical_loss": 3.7363269157920436,
      "tokens_seen": 847839232
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000750591775325978,
      "loss": 1.6328,
      "theoretical_loss": 3.7362988032647846,
      "tokens_seen": 847904768
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007505717151454364,
      "loss": 1.4984,
      "theoretical_loss": 3.736270693518655,
      "tokens_seen": 847970304
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007505516549648947,
      "loss": 1.5644,
      "theoretical_loss": 3.7362425865531654,
      "tokens_seen": 848035840
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007505315947843531,
      "loss": 1.5268,
      "theoretical_loss": 3.736214482367825,
      "tokens_seen": 848101376
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007505115346038115,
      "loss": 1.621,
      "theoretical_loss": 3.7361863809621454,
      "tokens_seen": 848166912
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007504914744232698,
      "loss": 1.6251,
      "theoretical_loss": 3.736158282335635,
      "tokens_seen": 848232448
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007504714142427283,
      "loss": 1.4976,
      "theoretical_loss": 3.736130186487806,
      "tokens_seen": 848297984
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007504513540621865,
      "loss": 1.6222,
      "theoretical_loss": 3.7361020934181677,
      "tokens_seen": 848363520
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000750431293881645,
      "loss": 1.4902,
      "theoretical_loss": 3.7360740031262316,
      "tokens_seen": 848429056
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007504112337011033,
      "loss": 1.4338,
      "theoretical_loss": 3.7360459156115082,
      "tokens_seen": 848494592
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007503911735205617,
      "loss": 1.5121,
      "theoretical_loss": 3.736017830873508,
      "tokens_seen": 848560128
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007503711133400201,
      "loss": 1.5524,
      "theoretical_loss": 3.7359897489117433,
      "tokens_seen": 848625664
    },
    {
      "epoch": 0.26,
      "objective/train/advantage_avg": 0.49475252628326416,
      "objective/train/docs_used": 313372,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5160740613937378,
      "objective/train/original_loss": 1.5160740613937378,
      "objective/train/theoretical_loss": 3.7359616697257243,
      "objective/train/tokens_used": 869151200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2455303817987442,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507235527038574,
      "objective/train/weighted_lm_loss": 1.5928984880447388,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.7788007855415344,
      "theoretical_loss": 3.7359616697257243,
      "tokens_seen": 848691200
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007503510531594785,
      "loss": 1.5384,
      "theoretical_loss": 3.7359616697257243,
      "tokens_seen": 848691200
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007503309929789368,
      "loss": 1.6608,
      "theoretical_loss": 3.7359335933149627,
      "tokens_seen": 848756736
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007503109327983953,
      "loss": 1.4984,
      "theoretical_loss": 3.7359055196789694,
      "tokens_seen": 848822272
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007502908726178535,
      "loss": 1.4786,
      "theoretical_loss": 3.735877448817257,
      "tokens_seen": 848887808
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000750270812437312,
      "loss": 1.5812,
      "theoretical_loss": 3.735849380729336,
      "tokens_seen": 848953344
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007502507522567703,
      "loss": 1.5561,
      "theoretical_loss": 3.7358213154147193,
      "tokens_seen": 849018880
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007502306920762287,
      "loss": 1.4531,
      "theoretical_loss": 3.7357932528729183,
      "tokens_seen": 849084416
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007502106318956871,
      "loss": 1.5089,
      "theoretical_loss": 3.735765193103445,
      "tokens_seen": 849149952
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007501905717151455,
      "loss": 1.5777,
      "theoretical_loss": 3.735737136105812,
      "tokens_seen": 849215488
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007501705115346038,
      "loss": 1.6997,
      "theoretical_loss": 3.7357090818795307,
      "tokens_seen": 849281024
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007501504513540622,
      "loss": 1.5849,
      "theoretical_loss": 3.7356810304241144,
      "tokens_seen": 849346560
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007501303911735206,
      "loss": 1.7269,
      "theoretical_loss": 3.735652981739076,
      "tokens_seen": 849412096
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000750110330992979,
      "loss": 1.5565,
      "theoretical_loss": 3.735624935823927,
      "tokens_seen": 849477632
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007500902708124373,
      "loss": 1.5821,
      "theoretical_loss": 3.7355968926781804,
      "tokens_seen": 849543168
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007500702106318957,
      "loss": 1.4759,
      "theoretical_loss": 3.73556885230135,
      "tokens_seen": 849608704
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000750050150451354,
      "loss": 1.6343,
      "theoretical_loss": 3.735540814692949,
      "tokens_seen": 849674240
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007500300902708125,
      "loss": 1.472,
      "theoretical_loss": 3.735512779852489,
      "tokens_seen": 849739776
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007500100300902708,
      "loss": 1.4426,
      "theoretical_loss": 3.7354847477794846,
      "tokens_seen": 849805312
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007499899699097292,
      "loss": 1.5739,
      "theoretical_loss": 3.735456718473449,
      "tokens_seen": 849870848
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007499699097291876,
      "loss": 1.4898,
      "theoretical_loss": 3.735428691933895,
      "tokens_seen": 849936384
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000749949849548646,
      "loss": 1.4296,
      "theoretical_loss": 3.735400668160337,
      "tokens_seen": 850001920
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007499297893681043,
      "loss": 1.4479,
      "theoretical_loss": 3.7353726471522894,
      "tokens_seen": 850067456
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007499097291875628,
      "loss": 1.4514,
      "theoretical_loss": 3.7353446289092647,
      "tokens_seen": 850132992
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000749889669007021,
      "loss": 1.5855,
      "theoretical_loss": 3.7353166134307783,
      "tokens_seen": 850198528
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007498696088264795,
      "loss": 1.6601,
      "theoretical_loss": 3.735288600716343,
      "tokens_seen": 850264064
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007498495486459378,
      "loss": 1.4877,
      "theoretical_loss": 3.735260590765474,
      "tokens_seen": 850329600
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007498294884653962,
      "loss": 1.5998,
      "theoretical_loss": 3.7352325835776856,
      "tokens_seen": 850395136
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007498094282848546,
      "loss": 1.6309,
      "theoretical_loss": 3.7352045791524917,
      "tokens_seen": 850460672
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007497893681043129,
      "loss": 1.5276,
      "theoretical_loss": 3.7351765774894075,
      "tokens_seen": 850526208
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007497693079237713,
      "loss": 1.5248,
      "theoretical_loss": 3.735148578587948,
      "tokens_seen": 850591744
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007497492477432298,
      "loss": 1.5823,
      "theoretical_loss": 3.7351205824476277,
      "tokens_seen": 850657280
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000749729187562688,
      "loss": 1.6346,
      "theoretical_loss": 3.7350925890679614,
      "tokens_seen": 850722816
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007497091273821465,
      "loss": 1.4355,
      "theoretical_loss": 3.735064598448465,
      "tokens_seen": 850788352
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007496890672016047,
      "loss": 1.6909,
      "theoretical_loss": 3.735036610588653,
      "tokens_seen": 850853888
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007496690070210632,
      "loss": 1.6574,
      "theoretical_loss": 3.7350086254880415,
      "tokens_seen": 850919424
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007496489468405216,
      "loss": 1.5837,
      "theoretical_loss": 3.734980643146145,
      "tokens_seen": 850984960
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007496288866599799,
      "loss": 1.5587,
      "theoretical_loss": 3.73495266356248,
      "tokens_seen": 851050496
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007496088264794383,
      "loss": 1.3885,
      "theoretical_loss": 3.734924686736562,
      "tokens_seen": 851116032
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007495887662988968,
      "loss": 1.4773,
      "theoretical_loss": 3.734896712667907,
      "tokens_seen": 851181568
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000749568706118355,
      "loss": 1.4704,
      "theoretical_loss": 3.7348687413560304,
      "tokens_seen": 851247104
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007495486459378135,
      "loss": 1.5695,
      "theoretical_loss": 3.7348407728004487,
      "tokens_seen": 851312640
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007495285857572717,
      "loss": 1.5377,
      "theoretical_loss": 3.7348128070006785,
      "tokens_seen": 851378176
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007495085255767302,
      "loss": 1.5308,
      "theoretical_loss": 3.734784843956236,
      "tokens_seen": 851443712
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007494884653961886,
      "loss": 1.3856,
      "theoretical_loss": 3.7347568836666376,
      "tokens_seen": 851509248
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007494684052156469,
      "loss": 1.5181,
      "theoretical_loss": 3.7347289261313996,
      "tokens_seen": 851574784
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007494483450351053,
      "loss": 1.6424,
      "theoretical_loss": 3.7347009713500388,
      "tokens_seen": 851640320
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007494282848545637,
      "loss": 1.5155,
      "theoretical_loss": 3.7346730193220727,
      "tokens_seen": 851705856
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000749408224674022,
      "loss": 1.5545,
      "theoretical_loss": 3.7346450700470175,
      "tokens_seen": 851771392
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007493881644934805,
      "loss": 1.4939,
      "theoretical_loss": 3.734617123524391,
      "tokens_seen": 851836928
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007493681043129388,
      "loss": 1.4956,
      "theoretical_loss": 3.7345891797537094,
      "tokens_seen": 851902464
    },
    {
      "epoch": 0.26,
      "objective/train/advantage_avg": 0.49306437373161316,
      "objective/train/docs_used": 314587,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.494137167930603,
      "objective/train/original_loss": 1.494137167930603,
      "objective/train/theoretical_loss": 3.7345612387344906,
      "objective/train/tokens_used": 872428000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24370484054088593,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050545334815979,
      "objective/train/weighted_lm_loss": 1.5693992376327515,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7345612387344906,
      "tokens_seen": 851968000
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007493480441323972,
      "loss": 1.5542,
      "theoretical_loss": 3.7345612387344906,
      "tokens_seen": 851968000
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007493279839518555,
      "loss": 1.5634,
      "theoretical_loss": 3.7345333004662526,
      "tokens_seen": 852033536
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000749307923771314,
      "loss": 1.6651,
      "theoretical_loss": 3.7345053649485123,
      "tokens_seen": 852099072
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007492878635907723,
      "loss": 1.5765,
      "theoretical_loss": 3.7344774321807876,
      "tokens_seen": 852164608
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007492678034102307,
      "loss": 1.5233,
      "theoretical_loss": 3.734449502162596,
      "tokens_seen": 852230144
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007492477432296891,
      "loss": 1.5709,
      "theoretical_loss": 3.7344215748934566,
      "tokens_seen": 852295680
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007492276830491475,
      "loss": 1.5507,
      "theoretical_loss": 3.7343936503728856,
      "tokens_seen": 852361216
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007492076228686059,
      "loss": 1.621,
      "theoretical_loss": 3.734365728600403,
      "tokens_seen": 852426752
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007491875626880642,
      "loss": 1.5273,
      "theoretical_loss": 3.7343378095755257,
      "tokens_seen": 852492288
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007491675025075226,
      "loss": 1.4868,
      "theoretical_loss": 3.734309893297773,
      "tokens_seen": 852557824
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000749147442326981,
      "loss": 1.5043,
      "theoretical_loss": 3.7342819797666635,
      "tokens_seen": 852623360
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007491273821464394,
      "loss": 1.5537,
      "theoretical_loss": 3.7342540689817154,
      "tokens_seen": 852688896
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007491073219658977,
      "loss": 1.4701,
      "theoretical_loss": 3.7342261609424483,
      "tokens_seen": 852754432
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007490872617853561,
      "loss": 1.669,
      "theoretical_loss": 3.7341982556483795,
      "tokens_seen": 852819968
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007490672016048144,
      "loss": 1.5607,
      "theoretical_loss": 3.73417035309903,
      "tokens_seen": 852885504
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007490471414242729,
      "loss": 1.5886,
      "theoretical_loss": 3.7341424532939174,
      "tokens_seen": 852951040
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007490270812437312,
      "loss": 1.5775,
      "theoretical_loss": 3.7341145562325613,
      "tokens_seen": 853016576
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007490070210631896,
      "loss": 1.5053,
      "theoretical_loss": 3.734086661914482,
      "tokens_seen": 853082112
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000748986960882648,
      "loss": 1.6656,
      "theoretical_loss": 3.734058770339198,
      "tokens_seen": 853147648
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007489669007021064,
      "loss": 1.454,
      "theoretical_loss": 3.7340308815062295,
      "tokens_seen": 853213184
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007489468405215647,
      "loss": 1.5441,
      "theoretical_loss": 3.734002995415096,
      "tokens_seen": 853278720
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007489267803410232,
      "loss": 1.5088,
      "theoretical_loss": 3.7339751120653175,
      "tokens_seen": 853344256
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007489067201604814,
      "loss": 1.5808,
      "theoretical_loss": 3.7339472314564146,
      "tokens_seen": 853409792
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007488866599799399,
      "loss": 1.5623,
      "theoretical_loss": 3.733919353587906,
      "tokens_seen": 853475328
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007488665997993982,
      "loss": 1.5903,
      "theoretical_loss": 3.7338914784593134,
      "tokens_seen": 853540864
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007488465396188566,
      "loss": 1.4726,
      "theoretical_loss": 3.733863606070156,
      "tokens_seen": 853606400
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000748826479438315,
      "loss": 1.5117,
      "theoretical_loss": 3.7338357364199553,
      "tokens_seen": 853671936
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007488064192577733,
      "loss": 1.484,
      "theoretical_loss": 3.7338078695082313,
      "tokens_seen": 853737472
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007487863590772317,
      "loss": 1.6356,
      "theoretical_loss": 3.733780005334505,
      "tokens_seen": 853803008
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007487662988966902,
      "loss": 1.6267,
      "theoretical_loss": 3.7337521438982977,
      "tokens_seen": 853868544
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007487462387161484,
      "loss": 1.5981,
      "theoretical_loss": 3.733724285199129,
      "tokens_seen": 853934080
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007487261785356069,
      "loss": 1.6431,
      "theoretical_loss": 3.733696429236521,
      "tokens_seen": 853999616
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007487061183550651,
      "loss": 1.5195,
      "theoretical_loss": 3.733668576009995,
      "tokens_seen": 854065152
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007486860581745236,
      "loss": 1.6183,
      "theoretical_loss": 3.733640725519072,
      "tokens_seen": 854130688
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000748665997993982,
      "loss": 1.4658,
      "theoretical_loss": 3.7336128777632727,
      "tokens_seen": 854196224
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007486459378134403,
      "loss": 1.6006,
      "theoretical_loss": 3.7335850327421207,
      "tokens_seen": 854261760
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007486258776328987,
      "loss": 1.529,
      "theoretical_loss": 3.733557190455136,
      "tokens_seen": 854327296
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007486058174523572,
      "loss": 1.5564,
      "theoretical_loss": 3.733529350901841,
      "tokens_seen": 854392832
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007485857572718154,
      "loss": 1.4639,
      "theoretical_loss": 3.733501514081757,
      "tokens_seen": 854458368
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007485656970912739,
      "loss": 1.4263,
      "theoretical_loss": 3.7334736799944075,
      "tokens_seen": 854523904
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007485456369107321,
      "loss": 1.5395,
      "theoretical_loss": 3.733445848639313,
      "tokens_seen": 854589440
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007485255767301906,
      "loss": 1.5083,
      "theoretical_loss": 3.733418020015997,
      "tokens_seen": 854654976
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000748505516549649,
      "loss": 1.4033,
      "theoretical_loss": 3.7333901941239818,
      "tokens_seen": 854720512
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007484854563691073,
      "loss": 1.5661,
      "theoretical_loss": 3.733362370962789,
      "tokens_seen": 854786048
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007484653961885657,
      "loss": 1.6272,
      "theoretical_loss": 3.733334550531942,
      "tokens_seen": 854851584
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007484453360080241,
      "loss": 1.584,
      "theoretical_loss": 3.7333067328309637,
      "tokens_seen": 854917120
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007484252758274824,
      "loss": 1.6211,
      "theoretical_loss": 3.7332789178593773,
      "tokens_seen": 854982656
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007484052156469409,
      "loss": 1.5534,
      "theoretical_loss": 3.7332511056167044,
      "tokens_seen": 855048192
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007483851554663992,
      "loss": 1.5258,
      "theoretical_loss": 3.7332232961024694,
      "tokens_seen": 855113728
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007483650952858576,
      "loss": 1.5522,
      "theoretical_loss": 3.733195489316195,
      "tokens_seen": 855179264
    },
    {
      "epoch": 0.26,
      "objective/train/advantage_avg": 0.49476879835128784,
      "objective/train/docs_used": 315806,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3681517839431763,
      "objective/train/original_loss": 1.3681516647338867,
      "objective/train/theoretical_loss": 3.733167685257405,
      "objective/train/tokens_used": 875704800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24504238367080688,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507227182388306,
      "objective/train/weighted_lm_loss": 1.4376466274261475,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.733167685257405,
      "tokens_seen": 855244800
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007483450351053159,
      "loss": 1.4879,
      "theoretical_loss": 3.733167685257405,
      "tokens_seen": 855244800
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007483249749247743,
      "loss": 1.533,
      "theoretical_loss": 3.7331398839256225,
      "tokens_seen": 855310336
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007483049147442327,
      "loss": 1.5525,
      "theoretical_loss": 3.7331120853203714,
      "tokens_seen": 855375872
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007482848545636911,
      "loss": 1.5991,
      "theoretical_loss": 3.733084289441175,
      "tokens_seen": 855441408
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007482647943831494,
      "loss": 1.4132,
      "theoretical_loss": 3.733056496287558,
      "tokens_seen": 855506944
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007482447342026079,
      "loss": 1.4547,
      "theoretical_loss": 3.7330287058590432,
      "tokens_seen": 855572480
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007482246740220662,
      "loss": 1.615,
      "theoretical_loss": 3.733000918155156,
      "tokens_seen": 855638016
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007482046138415246,
      "loss": 1.4971,
      "theoretical_loss": 3.7329731331754195,
      "tokens_seen": 855703552
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007481845536609829,
      "loss": 1.3533,
      "theoretical_loss": 3.7329453509193584,
      "tokens_seen": 855769088
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007481644934804414,
      "loss": 1.5063,
      "theoretical_loss": 3.7329175713864977,
      "tokens_seen": 855834624
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007481444332998997,
      "loss": 1.5179,
      "theoretical_loss": 3.7328897945763617,
      "tokens_seen": 855900160
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007481243731193581,
      "loss": 1.5327,
      "theoretical_loss": 3.7328620204884744,
      "tokens_seen": 855965696
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007481043129388164,
      "loss": 1.6511,
      "theoretical_loss": 3.7328342491223614,
      "tokens_seen": 856031232
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007480842527582748,
      "loss": 1.5962,
      "theoretical_loss": 3.7328064804775476,
      "tokens_seen": 856096768
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007480641925777332,
      "loss": 1.5256,
      "theoretical_loss": 3.7327787145535574,
      "tokens_seen": 856162304
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007480441323971916,
      "loss": 1.5448,
      "theoretical_loss": 3.7327509513499164,
      "tokens_seen": 856227840
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007480240722166499,
      "loss": 1.5216,
      "theoretical_loss": 3.7327231908661496,
      "tokens_seen": 856293376
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007480040120361084,
      "loss": 1.4296,
      "theoretical_loss": 3.7326954331017834,
      "tokens_seen": 856358912
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007479839518555667,
      "loss": 1.4747,
      "theoretical_loss": 3.732667678056342,
      "tokens_seen": 856424448
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007479638916750251,
      "loss": 1.4779,
      "theoretical_loss": 3.7326399257293517,
      "tokens_seen": 856489984
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007479438314944834,
      "loss": 1.4807,
      "theoretical_loss": 3.7326121761203384,
      "tokens_seen": 856555520
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007479237713139418,
      "loss": 1.5245,
      "theoretical_loss": 3.7325844292288277,
      "tokens_seen": 856621056
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007479037111334002,
      "loss": 1.4511,
      "theoretical_loss": 3.732556685054346,
      "tokens_seen": 856686592
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007478836509528586,
      "loss": 1.5498,
      "theoretical_loss": 3.732528943596419,
      "tokens_seen": 856752128
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007478635907723169,
      "loss": 1.5322,
      "theoretical_loss": 3.732501204854573,
      "tokens_seen": 856817664
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007478435305917754,
      "loss": 1.4284,
      "theoretical_loss": 3.732473468828334,
      "tokens_seen": 856883200
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007478234704112336,
      "loss": 1.5383,
      "theoretical_loss": 3.7324457355172296,
      "tokens_seen": 856948736
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007478034102306921,
      "loss": 1.5307,
      "theoretical_loss": 3.7324180049207856,
      "tokens_seen": 857014272
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007477833500501505,
      "loss": 1.5262,
      "theoretical_loss": 3.7323902770385287,
      "tokens_seen": 857079808
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007477632898696088,
      "loss": 1.4918,
      "theoretical_loss": 3.732362551869986,
      "tokens_seen": 857145344
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007477432296890673,
      "loss": 1.512,
      "theoretical_loss": 3.7323348294146843,
      "tokens_seen": 857210880
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007477231695085255,
      "loss": 1.5219,
      "theoretical_loss": 3.73230710967215,
      "tokens_seen": 857276416
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000747703109327984,
      "loss": 1.4569,
      "theoretical_loss": 3.732279392641912,
      "tokens_seen": 857341952
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007476830491474424,
      "loss": 1.6519,
      "theoretical_loss": 3.732251678323496,
      "tokens_seen": 857407488
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007476629889669007,
      "loss": 1.5343,
      "theoretical_loss": 3.73222396671643,
      "tokens_seen": 857473024
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007476429287863591,
      "loss": 1.5369,
      "theoretical_loss": 3.7321962578202417,
      "tokens_seen": 857538560
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007476228686058176,
      "loss": 1.5652,
      "theoretical_loss": 3.7321685516344587,
      "tokens_seen": 857604096
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007476028084252758,
      "loss": 1.4835,
      "theoretical_loss": 3.7321408481586094,
      "tokens_seen": 857669632
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007475827482447343,
      "loss": 1.6292,
      "theoretical_loss": 3.73211314739222,
      "tokens_seen": 857735168
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007475626880641925,
      "loss": 1.5232,
      "theoretical_loss": 3.7320854493348197,
      "tokens_seen": 857800704
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000747542627883651,
      "loss": 1.5538,
      "theoretical_loss": 3.732057753985937,
      "tokens_seen": 857866240
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007475225677031094,
      "loss": 1.4779,
      "theoretical_loss": 3.7320300613450987,
      "tokens_seen": 857931776
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007475025075225677,
      "loss": 1.5218,
      "theoretical_loss": 3.732002371411835,
      "tokens_seen": 857997312
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007474824473420261,
      "loss": 1.5494,
      "theoretical_loss": 3.731974684185673,
      "tokens_seen": 858062848
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007474623871614845,
      "loss": 1.5385,
      "theoretical_loss": 3.731946999666142,
      "tokens_seen": 858128384
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007474423269809428,
      "loss": 1.4299,
      "theoretical_loss": 3.7319193178527708,
      "tokens_seen": 858193920
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007474222668004013,
      "loss": 1.4636,
      "theoretical_loss": 3.7318916387450876,
      "tokens_seen": 858259456
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007474022066198596,
      "loss": 1.5923,
      "theoretical_loss": 3.731863962342622,
      "tokens_seen": 858324992
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000747382146439318,
      "loss": 1.5853,
      "theoretical_loss": 3.731836288644903,
      "tokens_seen": 858390528
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007473620862587763,
      "loss": 1.5247,
      "theoretical_loss": 3.731808617651459,
      "tokens_seen": 858456064
    },
    {
      "epoch": 0.26,
      "objective/train/advantage_avg": 0.492812842130661,
      "objective/train/docs_used": 317142,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.475641131401062,
      "objective/train/original_loss": 1.475641131401062,
      "objective/train/theoretical_loss": 3.7317809493618204,
      "objective/train/tokens_used": 878981600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2434263527393341,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505187511444092,
      "objective/train/weighted_lm_loss": 1.5501574277877808,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7317809493618204,
      "tokens_seen": 858521600
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007473420260782347,
      "loss": 1.5759,
      "theoretical_loss": 3.7317809493618204,
      "tokens_seen": 858521600
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007473219658976931,
      "loss": 1.5229,
      "theoretical_loss": 3.7317532837755163,
      "tokens_seen": 858587136
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007473019057171515,
      "loss": 1.654,
      "theoretical_loss": 3.731725620892076,
      "tokens_seen": 858652672
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007472818455366098,
      "loss": 1.4725,
      "theoretical_loss": 3.7316979607110294,
      "tokens_seen": 858718208
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007472617853560683,
      "loss": 1.4571,
      "theoretical_loss": 3.7316703032319056,
      "tokens_seen": 858783744
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007472417251755266,
      "loss": 1.4652,
      "theoretical_loss": 3.731642648454236,
      "tokens_seen": 858849280
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000747221664994985,
      "loss": 1.6891,
      "theoretical_loss": 3.7316149963775493,
      "tokens_seen": 858914816
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007472016048144433,
      "loss": 1.5391,
      "theoretical_loss": 3.7315873470013763,
      "tokens_seen": 858980352
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007471815446339018,
      "loss": 1.6824,
      "theoretical_loss": 3.7315597003252474,
      "tokens_seen": 859045888
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007471614844533601,
      "loss": 1.5959,
      "theoretical_loss": 3.731532056348692,
      "tokens_seen": 859111424
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007471414242728185,
      "loss": 1.5097,
      "theoretical_loss": 3.7315044150712415,
      "tokens_seen": 859176960
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007471213640922768,
      "loss": 1.6425,
      "theoretical_loss": 3.7314767764924257,
      "tokens_seen": 859242496
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007471013039117352,
      "loss": 1.6004,
      "theoretical_loss": 3.731449140611777,
      "tokens_seen": 859308032
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007470812437311936,
      "loss": 1.5333,
      "theoretical_loss": 3.731421507428824,
      "tokens_seen": 859373568
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000747061183550652,
      "loss": 1.5207,
      "theoretical_loss": 3.731393876943099,
      "tokens_seen": 859439104
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007470411233701103,
      "loss": 1.5064,
      "theoretical_loss": 3.7313662491541333,
      "tokens_seen": 859504640
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007470210631895688,
      "loss": 1.4899,
      "theoretical_loss": 3.7313386240614577,
      "tokens_seen": 859570176
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007470010030090271,
      "loss": 1.4848,
      "theoretical_loss": 3.7313110016646034,
      "tokens_seen": 859635712
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007469809428284855,
      "loss": 1.4819,
      "theoretical_loss": 3.7312833819631015,
      "tokens_seen": 859701248
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007469608826479438,
      "loss": 1.6796,
      "theoretical_loss": 3.731255764956485,
      "tokens_seen": 859766784
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007469408224674022,
      "loss": 1.5387,
      "theoretical_loss": 3.7312281506442835,
      "tokens_seen": 859832320
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007469207622868606,
      "loss": 1.6083,
      "theoretical_loss": 3.73120053902603,
      "tokens_seen": 859897856
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000746900702106319,
      "loss": 1.497,
      "theoretical_loss": 3.7311729301012564,
      "tokens_seen": 859963392
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007468806419257773,
      "loss": 1.5181,
      "theoretical_loss": 3.7311453238694945,
      "tokens_seen": 860028928
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007468605817452358,
      "loss": 1.5372,
      "theoretical_loss": 3.7311177203302766,
      "tokens_seen": 860094464
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000746840521564694,
      "loss": 1.6257,
      "theoretical_loss": 3.7310901194831345,
      "tokens_seen": 860160000
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007468204613841525,
      "loss": 1.6018,
      "theoretical_loss": 3.731062521327601,
      "tokens_seen": 860225536
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007468004012036109,
      "loss": 1.4475,
      "theoretical_loss": 3.7310349258632094,
      "tokens_seen": 860291072
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007467803410230692,
      "loss": 1.5311,
      "theoretical_loss": 3.7310073330894906,
      "tokens_seen": 860356608
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007467602808425276,
      "loss": 1.4463,
      "theoretical_loss": 3.730979743005978,
      "tokens_seen": 860422144
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007467402206619859,
      "loss": 1.615,
      "theoretical_loss": 3.7309521556122047,
      "tokens_seen": 860487680
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007467201604814443,
      "loss": 1.4189,
      "theoretical_loss": 3.7309245709077037,
      "tokens_seen": 860553216
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007467001003009028,
      "loss": 1.5045,
      "theoretical_loss": 3.730896988892008,
      "tokens_seen": 860618752
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000746680040120361,
      "loss": 1.7394,
      "theoretical_loss": 3.73086940956465,
      "tokens_seen": 860684288
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007466599799398195,
      "loss": 1.5463,
      "theoretical_loss": 3.7308418329251642,
      "tokens_seen": 860749824
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007466399197592779,
      "loss": 1.6018,
      "theoretical_loss": 3.7308142589730835,
      "tokens_seen": 860815360
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007466198595787362,
      "loss": 1.6778,
      "theoretical_loss": 3.7307866877079414,
      "tokens_seen": 860880896
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007465997993981946,
      "loss": 1.4577,
      "theoretical_loss": 3.7307591191292717,
      "tokens_seen": 860946432
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000746579739217653,
      "loss": 1.5832,
      "theoretical_loss": 3.730731553236608,
      "tokens_seen": 861011968
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007465596790371113,
      "loss": 1.5506,
      "theoretical_loss": 3.7307039900294843,
      "tokens_seen": 861077504
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007465396188565698,
      "loss": 1.537,
      "theoretical_loss": 3.730676429507435,
      "tokens_seen": 861143040
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000746519558676028,
      "loss": 1.4486,
      "theoretical_loss": 3.7306488716699926,
      "tokens_seen": 861208576
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007464994984954865,
      "loss": 1.5256,
      "theoretical_loss": 3.7306213165166935,
      "tokens_seen": 861274112
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007464794383149448,
      "loss": 1.5713,
      "theoretical_loss": 3.7305937640470708,
      "tokens_seen": 861339648
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007464593781344032,
      "loss": 1.4288,
      "theoretical_loss": 3.730566214260659,
      "tokens_seen": 861405184
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007464393179538616,
      "loss": 1.5484,
      "theoretical_loss": 3.730538667156993,
      "tokens_seen": 861470720
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00074641925777332,
      "loss": 1.581,
      "theoretical_loss": 3.7305111227356074,
      "tokens_seen": 861536256
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007463991975927783,
      "loss": 1.5516,
      "theoretical_loss": 3.730483580996037,
      "tokens_seen": 861601792
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007463791374122367,
      "loss": 1.486,
      "theoretical_loss": 3.730456041937817,
      "tokens_seen": 861667328
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000746359077231695,
      "loss": 1.4889,
      "theoretical_loss": 3.730428505560482,
      "tokens_seen": 861732864
    },
    {
      "epoch": 0.26,
      "objective/train/advantage_avg": 0.4923645853996277,
      "objective/train/docs_used": 318349,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3927994966506958,
      "objective/train/original_loss": 1.3927994966506958,
      "objective/train/theoretical_loss": 3.730400971863568,
      "objective/train/tokens_used": 882258400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24308334290981293,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504721403121948,
      "objective/train/weighted_lm_loss": 1.463008999824524,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.730400971863568,
      "tokens_seen": 861798400
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007463390170511535,
      "loss": 1.5754,
      "theoretical_loss": 3.730400971863568,
      "tokens_seen": 861798400
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007463189568706118,
      "loss": 1.6819,
      "theoretical_loss": 3.7303734408466087,
      "tokens_seen": 861863936
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007462988966900702,
      "loss": 1.4701,
      "theoretical_loss": 3.730345912509141,
      "tokens_seen": 861929472
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007462788365095286,
      "loss": 1.5475,
      "theoretical_loss": 3.7303183868506995,
      "tokens_seen": 861995008
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000746258776328987,
      "loss": 1.5536,
      "theoretical_loss": 3.7302908638708203,
      "tokens_seen": 862060544
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007462387161484454,
      "loss": 1.4643,
      "theoretical_loss": 3.7302633435690398,
      "tokens_seen": 862126080
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007462186559679037,
      "loss": 1.4522,
      "theoretical_loss": 3.7302358259448924,
      "tokens_seen": 862191616
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007461985957873622,
      "loss": 1.5522,
      "theoretical_loss": 3.7302083109979147,
      "tokens_seen": 862257152
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007461785356068205,
      "loss": 1.4988,
      "theoretical_loss": 3.730180798727643,
      "tokens_seen": 862322688
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007461584754262789,
      "loss": 1.6232,
      "theoretical_loss": 3.7301532891336135,
      "tokens_seen": 862388224
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007461384152457372,
      "loss": 1.6418,
      "theoretical_loss": 3.730125782215362,
      "tokens_seen": 862453760
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007461183550651956,
      "loss": 1.5174,
      "theoretical_loss": 3.730098277972426,
      "tokens_seen": 862519296
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000746098294884654,
      "loss": 1.5998,
      "theoretical_loss": 3.730070776404341,
      "tokens_seen": 862584832
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007460782347041124,
      "loss": 1.5393,
      "theoretical_loss": 3.730043277510644,
      "tokens_seen": 862650368
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007460581745235707,
      "loss": 1.5621,
      "theoretical_loss": 3.730015781290872,
      "tokens_seen": 862715904
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007460381143430292,
      "loss": 1.7104,
      "theoretical_loss": 3.729988287744561,
      "tokens_seen": 862781440
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007460180541624875,
      "loss": 1.524,
      "theoretical_loss": 3.7299607968712496,
      "tokens_seen": 862846976
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007459979939819459,
      "loss": 1.5213,
      "theoretical_loss": 3.729933308670474,
      "tokens_seen": 862912512
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007459779338014042,
      "loss": 1.6147,
      "theoretical_loss": 3.729905823141771,
      "tokens_seen": 862978048
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007459578736208626,
      "loss": 1.4692,
      "theoretical_loss": 3.7298783402846793,
      "tokens_seen": 863043584
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000745937813440321,
      "loss": 1.6478,
      "theoretical_loss": 3.7298508600987352,
      "tokens_seen": 863109120
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007459177532597794,
      "loss": 1.4197,
      "theoretical_loss": 3.7298233825834766,
      "tokens_seen": 863174656
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007458976930792377,
      "loss": 1.4827,
      "theoretical_loss": 3.729795907738441,
      "tokens_seen": 863240192
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007458776328986962,
      "loss": 1.5571,
      "theoretical_loss": 3.729768435563167,
      "tokens_seen": 863305728
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007458575727181544,
      "loss": 1.4277,
      "theoretical_loss": 3.729740966057192,
      "tokens_seen": 863371264
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007458375125376129,
      "loss": 1.5176,
      "theoretical_loss": 3.7297134992200536,
      "tokens_seen": 863436800
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007458174523570713,
      "loss": 1.5506,
      "theoretical_loss": 3.729686035051291,
      "tokens_seen": 863502336
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007457973921765296,
      "loss": 1.5648,
      "theoretical_loss": 3.7296585735504415,
      "tokens_seen": 863567872
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000745777331995988,
      "loss": 1.5242,
      "theoretical_loss": 3.7296311147170442,
      "tokens_seen": 863633408
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007457572718154463,
      "loss": 1.5284,
      "theoretical_loss": 3.729603658550637,
      "tokens_seen": 863698944
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007457372116349047,
      "loss": 1.5701,
      "theoretical_loss": 3.7295762050507593,
      "tokens_seen": 863764480
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007457171514543632,
      "loss": 1.5458,
      "theoretical_loss": 3.7295487542169488,
      "tokens_seen": 863830016
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007456970912738214,
      "loss": 1.4887,
      "theoretical_loss": 3.7295213060487455,
      "tokens_seen": 863895552
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007456770310932799,
      "loss": 1.5523,
      "theoretical_loss": 3.7294938605456878,
      "tokens_seen": 863961088
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007456569709127383,
      "loss": 1.5334,
      "theoretical_loss": 3.7294664177073145,
      "tokens_seen": 864026624
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007456369107321966,
      "loss": 1.5113,
      "theoretical_loss": 3.7294389775331647,
      "tokens_seen": 864092160
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000745616850551655,
      "loss": 1.6344,
      "theoretical_loss": 3.729411540022779,
      "tokens_seen": 864157696
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007455967903711133,
      "loss": 1.6064,
      "theoretical_loss": 3.7293841051756953,
      "tokens_seen": 864223232
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007455767301905717,
      "loss": 1.5315,
      "theoretical_loss": 3.7293566729914547,
      "tokens_seen": 864288768
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007455566700100302,
      "loss": 1.5569,
      "theoretical_loss": 3.729329243469595,
      "tokens_seen": 864354304
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007455366098294884,
      "loss": 1.5516,
      "theoretical_loss": 3.729301816609657,
      "tokens_seen": 864419840
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007455165496489469,
      "loss": 1.4926,
      "theoretical_loss": 3.729274392411181,
      "tokens_seen": 864485376
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007454964894684052,
      "loss": 1.5601,
      "theoretical_loss": 3.7292469708737066,
      "tokens_seen": 864550912
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007454764292878636,
      "loss": 1.4952,
      "theoretical_loss": 3.7292195519967732,
      "tokens_seen": 864616448
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000745456369107322,
      "loss": 1.572,
      "theoretical_loss": 3.7291921357799223,
      "tokens_seen": 864681984
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007454363089267804,
      "loss": 1.4908,
      "theoretical_loss": 3.729164722222693,
      "tokens_seen": 864747520
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007454162487462387,
      "loss": 1.6633,
      "theoretical_loss": 3.729137311324627,
      "tokens_seen": 864813056
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007453961885656971,
      "loss": 1.3989,
      "theoretical_loss": 3.729109903085264,
      "tokens_seen": 864878592
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007453761283851554,
      "loss": 1.5293,
      "theoretical_loss": 3.729082497504145,
      "tokens_seen": 864944128
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007453560682046139,
      "loss": 1.4942,
      "theoretical_loss": 3.7290550945808105,
      "tokens_seen": 865009664
    },
    {
      "epoch": 0.26,
      "objective/train/advantage_avg": 0.49091729521751404,
      "objective/train/docs_used": 319421,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3913459777832031,
      "objective/train/original_loss": 1.391345739364624,
      "objective/train/theoretical_loss": 3.7290276943148015,
      "objective/train/tokens_used": 885535200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24194402992725372,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503216981887817,
      "objective/train/weighted_lm_loss": 1.461237907409668,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7290276943148015,
      "tokens_seen": 865075200
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007453360080240722,
      "loss": 1.5038,
      "theoretical_loss": 3.7290276943148015,
      "tokens_seen": 865075200
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007453159478435306,
      "loss": 1.483,
      "theoretical_loss": 3.72900029670566,
      "tokens_seen": 865140736
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000745295887662989,
      "loss": 1.6412,
      "theoretical_loss": 3.7289729017529254,
      "tokens_seen": 865206272
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007452758274824474,
      "loss": 1.6696,
      "theoretical_loss": 3.7289455094561403,
      "tokens_seen": 865271808
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007452557673019057,
      "loss": 1.5456,
      "theoretical_loss": 3.7289181198148458,
      "tokens_seen": 865337344
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007452357071213641,
      "loss": 1.591,
      "theoretical_loss": 3.728890732828583,
      "tokens_seen": 865402880
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007452156469408224,
      "loss": 1.489,
      "theoretical_loss": 3.7288633484968936,
      "tokens_seen": 865468416
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007451955867602809,
      "loss": 1.6283,
      "theoretical_loss": 3.72883596681932,
      "tokens_seen": 865533952
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007451755265797392,
      "loss": 1.5724,
      "theoretical_loss": 3.7288085877954025,
      "tokens_seen": 865599488
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007451554663991976,
      "loss": 1.6065,
      "theoretical_loss": 3.7287812114246845,
      "tokens_seen": 865665024
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007451354062186559,
      "loss": 1.403,
      "theoretical_loss": 3.7287538377067078,
      "tokens_seen": 865730560
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007451153460381144,
      "loss": 1.5819,
      "theoretical_loss": 3.7287264666410147,
      "tokens_seen": 865796096
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007450952858575727,
      "loss": 1.6229,
      "theoretical_loss": 3.728699098227146,
      "tokens_seen": 865861632
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007450752256770311,
      "loss": 1.5506,
      "theoretical_loss": 3.7286717324646466,
      "tokens_seen": 865927168
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007450551654964895,
      "loss": 1.5973,
      "theoretical_loss": 3.7286443693530567,
      "tokens_seen": 865992704
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007450351053159479,
      "loss": 1.4985,
      "theoretical_loss": 3.72861700889192,
      "tokens_seen": 866058240
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007450150451354062,
      "loss": 1.496,
      "theoretical_loss": 3.728589651080779,
      "tokens_seen": 866123776
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007449949849548646,
      "loss": 1.4216,
      "theoretical_loss": 3.7285622959191764,
      "tokens_seen": 866189312
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007449749247743229,
      "loss": 1.5338,
      "theoretical_loss": 3.7285349434066557,
      "tokens_seen": 866254848
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007449548645937814,
      "loss": 1.5605,
      "theoretical_loss": 3.7285075935427603,
      "tokens_seen": 866320384
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007449348044132397,
      "loss": 1.5933,
      "theoretical_loss": 3.728480246327032,
      "tokens_seen": 866385920
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007449147442326981,
      "loss": 1.3883,
      "theoretical_loss": 3.728452901759015,
      "tokens_seen": 866451456
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007448946840521565,
      "loss": 1.5304,
      "theoretical_loss": 3.7284255598382523,
      "tokens_seen": 866516992
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007448746238716148,
      "loss": 1.5612,
      "theoretical_loss": 3.728398220564288,
      "tokens_seen": 866582528
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007448545636910732,
      "loss": 1.5837,
      "theoretical_loss": 3.7283708839366656,
      "tokens_seen": 866648064
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007448345035105317,
      "loss": 1.6079,
      "theoretical_loss": 3.7283435499549284,
      "tokens_seen": 866713600
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007448144433299899,
      "loss": 1.4996,
      "theoretical_loss": 3.7283162186186214,
      "tokens_seen": 866779136
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007447943831494484,
      "loss": 1.5612,
      "theoretical_loss": 3.7282888899272866,
      "tokens_seen": 866844672
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007447743229689066,
      "loss": 1.5492,
      "theoretical_loss": 3.72826156388047,
      "tokens_seen": 866910208
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007447542627883651,
      "loss": 1.4324,
      "theoretical_loss": 3.7282342404777147,
      "tokens_seen": 866975744
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007447342026078236,
      "loss": 1.6231,
      "theoretical_loss": 3.7282069197185654,
      "tokens_seen": 867041280
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007447141424272818,
      "loss": 1.6217,
      "theoretical_loss": 3.728179601602567,
      "tokens_seen": 867106816
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007446940822467403,
      "loss": 1.5246,
      "theoretical_loss": 3.728152286129263,
      "tokens_seen": 867172352
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007446740220661987,
      "loss": 1.4669,
      "theoretical_loss": 3.728124973298199,
      "tokens_seen": 867237888
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000744653961885657,
      "loss": 1.499,
      "theoretical_loss": 3.7280976631089198,
      "tokens_seen": 867303424
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007446339017051154,
      "loss": 1.5951,
      "theoretical_loss": 3.728070355560969,
      "tokens_seen": 867368960
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007446138415245737,
      "loss": 1.4906,
      "theoretical_loss": 3.728043050653893,
      "tokens_seen": 867434496
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007445937813440321,
      "loss": 1.7125,
      "theoretical_loss": 3.7280157483872363,
      "tokens_seen": 867500032
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007445737211634906,
      "loss": 1.5156,
      "theoretical_loss": 3.727988448760544,
      "tokens_seen": 867565568
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007445536609829488,
      "loss": 1.7223,
      "theoretical_loss": 3.7279611517733624,
      "tokens_seen": 867631104
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007445336008024073,
      "loss": 1.5532,
      "theoretical_loss": 3.7279338574252354,
      "tokens_seen": 867696640
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007445135406218656,
      "loss": 1.5686,
      "theoretical_loss": 3.7279065657157098,
      "tokens_seen": 867762176
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000744493480441324,
      "loss": 1.603,
      "theoretical_loss": 3.7278792766443303,
      "tokens_seen": 867827712
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007444734202607824,
      "loss": 1.4876,
      "theoretical_loss": 3.7278519902106435,
      "tokens_seen": 867893248
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007444533600802408,
      "loss": 1.6739,
      "theoretical_loss": 3.7278247064141956,
      "tokens_seen": 867958784
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007444332998996991,
      "loss": 1.6247,
      "theoretical_loss": 3.727797425254532,
      "tokens_seen": 868024320
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007444132397191575,
      "loss": 1.5389,
      "theoretical_loss": 3.7277701467311983,
      "tokens_seen": 868089856
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007443931795386158,
      "loss": 1.4778,
      "theoretical_loss": 3.7277428708437412,
      "tokens_seen": 868155392
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007443731193580743,
      "loss": 1.619,
      "theoretical_loss": 3.7277155975917076,
      "tokens_seen": 868220928
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007443530591775326,
      "loss": 1.5954,
      "theoretical_loss": 3.7276883269746435,
      "tokens_seen": 868286464
    },
    {
      "epoch": 0.26,
      "objective/train/advantage_avg": 0.4922114312648773,
      "objective/train/docs_used": 320668,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.47066330909729,
      "objective/train/original_loss": 1.47066330909729,
      "objective/train/theoretical_loss": 3.727661058992095,
      "objective/train/tokens_used": 888812000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2428159862756729,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504554510116577,
      "objective/train/weighted_lm_loss": 1.544921875,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.727661058992095,
      "tokens_seen": 868352000
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000744332998996991,
      "loss": 1.5547,
      "theoretical_loss": 3.727661058992095,
      "tokens_seen": 868352000
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007443129388164494,
      "loss": 1.5615,
      "theoretical_loss": 3.72763379364361,
      "tokens_seen": 868417536
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007442928786359078,
      "loss": 1.5991,
      "theoretical_loss": 3.7276065309287345,
      "tokens_seen": 868483072
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007442728184553661,
      "loss": 1.492,
      "theoretical_loss": 3.727579270847015,
      "tokens_seen": 868548608
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007442527582748245,
      "loss": 1.6118,
      "theoretical_loss": 3.7275520133979994,
      "tokens_seen": 868614144
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007442326980942828,
      "loss": 1.4911,
      "theoretical_loss": 3.727524758581234,
      "tokens_seen": 868679680
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007442126379137413,
      "loss": 1.6765,
      "theoretical_loss": 3.727497506396267,
      "tokens_seen": 868745216
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007441925777331996,
      "loss": 1.4765,
      "theoretical_loss": 3.727470256842645,
      "tokens_seen": 868810752
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000744172517552658,
      "loss": 1.525,
      "theoretical_loss": 3.727443009919916,
      "tokens_seen": 868876288
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007441524573721163,
      "loss": 1.5811,
      "theoretical_loss": 3.7274157656276277,
      "tokens_seen": 868941824
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007441323971915748,
      "loss": 1.5159,
      "theoretical_loss": 3.7273885239653266,
      "tokens_seen": 869007360
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007441123370110331,
      "loss": 1.5848,
      "theoretical_loss": 3.7273612849325612,
      "tokens_seen": 869072896
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007440922768304915,
      "loss": 1.4096,
      "theoretical_loss": 3.72733404852888,
      "tokens_seen": 869138432
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007440722166499499,
      "loss": 1.537,
      "theoretical_loss": 3.72730681475383,
      "tokens_seen": 869203968
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007440521564694083,
      "loss": 1.5715,
      "theoretical_loss": 3.727279583606961,
      "tokens_seen": 869269504
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007440320962888666,
      "loss": 1.5928,
      "theoretical_loss": 3.7272523550878196,
      "tokens_seen": 869335040
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000744012036108325,
      "loss": 1.475,
      "theoretical_loss": 3.727225129195954,
      "tokens_seen": 869400576
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007439919759277833,
      "loss": 1.6003,
      "theoretical_loss": 3.7271979059309146,
      "tokens_seen": 869466112
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007439719157472418,
      "loss": 1.5475,
      "theoretical_loss": 3.727170685292248,
      "tokens_seen": 869531648
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007439518555667001,
      "loss": 1.6588,
      "theoretical_loss": 3.727143467279504,
      "tokens_seen": 869597184
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007439317953861585,
      "loss": 1.6637,
      "theoretical_loss": 3.727116251892231,
      "tokens_seen": 869662720
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007439117352056169,
      "loss": 1.5132,
      "theoretical_loss": 3.727089039129978,
      "tokens_seen": 869728256
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007438916750250752,
      "loss": 1.412,
      "theoretical_loss": 3.7270618289922943,
      "tokens_seen": 869793792
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007438716148445336,
      "loss": 1.5629,
      "theoretical_loss": 3.727034621478728,
      "tokens_seen": 869859328
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000743851554663992,
      "loss": 1.605,
      "theoretical_loss": 3.72700741658883,
      "tokens_seen": 869924864
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007438314944834503,
      "loss": 1.4099,
      "theoretical_loss": 3.7269802143221487,
      "tokens_seen": 869990400
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007438114343029088,
      "loss": 1.5467,
      "theoretical_loss": 3.7269530146782337,
      "tokens_seen": 870055936
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000743791374122367,
      "loss": 1.6391,
      "theoretical_loss": 3.7269258176566344,
      "tokens_seen": 870121472
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007437713139418255,
      "loss": 1.6246,
      "theoretical_loss": 3.7268986232569006,
      "tokens_seen": 870187008
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007437512537612839,
      "loss": 1.5855,
      "theoretical_loss": 3.7268714314785827,
      "tokens_seen": 870252544
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007437311935807422,
      "loss": 1.7268,
      "theoretical_loss": 3.72684424232123,
      "tokens_seen": 870318080
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007437111334002006,
      "loss": 1.5708,
      "theoretical_loss": 3.7268170557843927,
      "tokens_seen": 870383616
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007436910732196591,
      "loss": 1.5107,
      "theoretical_loss": 3.7267898718676205,
      "tokens_seen": 870449152
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007436710130391173,
      "loss": 1.6331,
      "theoretical_loss": 3.7267626905704647,
      "tokens_seen": 870514688
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007436509528585758,
      "loss": 1.5442,
      "theoretical_loss": 3.7267355118924748,
      "tokens_seen": 870580224
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000743630892678034,
      "loss": 1.5563,
      "theoretical_loss": 3.7267083358332016,
      "tokens_seen": 870645760
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007436108324974925,
      "loss": 1.6982,
      "theoretical_loss": 3.726681162392195,
      "tokens_seen": 870711296
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007435907723169509,
      "loss": 1.3741,
      "theoretical_loss": 3.726653991569007,
      "tokens_seen": 870776832
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007435707121364092,
      "loss": 1.4949,
      "theoretical_loss": 3.726626823363188,
      "tokens_seen": 870842368
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007435506519558676,
      "loss": 1.5869,
      "theoretical_loss": 3.7265996577742877,
      "tokens_seen": 870907904
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000743530591775326,
      "loss": 1.5678,
      "theoretical_loss": 3.726572494801859,
      "tokens_seen": 870973440
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007435105315947843,
      "loss": 1.4351,
      "theoretical_loss": 3.7265453344454516,
      "tokens_seen": 871038976
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007434904714142428,
      "loss": 1.5065,
      "theoretical_loss": 3.7265181767046176,
      "tokens_seen": 871104512
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000743470411233701,
      "loss": 1.6187,
      "theoretical_loss": 3.726491021578908,
      "tokens_seen": 871170048
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007434503510531595,
      "loss": 1.661,
      "theoretical_loss": 3.7264638690678744,
      "tokens_seen": 871235584
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007434302908726178,
      "loss": 1.5116,
      "theoretical_loss": 3.726436719171069,
      "tokens_seen": 871301120
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007434102306920762,
      "loss": 1.4972,
      "theoretical_loss": 3.726409571888042,
      "tokens_seen": 871366656
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007433901705115346,
      "loss": 1.5245,
      "theoretical_loss": 3.726382427218346,
      "tokens_seen": 871432192
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000743370110330993,
      "loss": 1.542,
      "theoretical_loss": 3.7263552851615334,
      "tokens_seen": 871497728
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007433500501504513,
      "loss": 1.6573,
      "theoretical_loss": 3.7263281457171558,
      "tokens_seen": 871563264
    },
    {
      "epoch": 0.26,
      "objective/train/advantage_avg": 0.4908714294433594,
      "objective/train/docs_used": 321822,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6295959949493408,
      "objective/train/original_loss": 1.6295961141586304,
      "objective/train/theoretical_loss": 3.7263010088847652,
      "objective/train/tokens_used": 892088800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.241449236869812,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503145456314087,
      "objective/train/weighted_lm_loss": 1.7115412950515747,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7263010088847652,
      "tokens_seen": 871628800
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007433299899699098,
      "loss": 1.6396,
      "theoretical_loss": 3.7263010088847652,
      "tokens_seen": 871628800
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000743309929789368,
      "loss": 1.6147,
      "theoretical_loss": 3.7262738746639137,
      "tokens_seen": 871694336
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007432898696088265,
      "loss": 1.4501,
      "theoretical_loss": 3.7262467430541544,
      "tokens_seen": 871759872
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007432698094282848,
      "loss": 1.4839,
      "theoretical_loss": 3.726219614055039,
      "tokens_seen": 871825408
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007432497492477432,
      "loss": 1.5548,
      "theoretical_loss": 3.726192487666121,
      "tokens_seen": 871890944
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007432296890672017,
      "loss": 1.6719,
      "theoretical_loss": 3.7261653638869525,
      "tokens_seen": 871956480
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00074320962888666,
      "loss": 1.482,
      "theoretical_loss": 3.7261382427170857,
      "tokens_seen": 872022016
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007431895687061184,
      "loss": 1.5246,
      "theoretical_loss": 3.7261111241560743,
      "tokens_seen": 872087552
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007431695085255767,
      "loss": 1.52,
      "theoretical_loss": 3.7260840082034714,
      "tokens_seen": 872153088
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007431494483450352,
      "loss": 1.6253,
      "theoretical_loss": 3.72605689485883,
      "tokens_seen": 872218624
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007431293881644935,
      "loss": 1.4769,
      "theoretical_loss": 3.726029784121703,
      "tokens_seen": 872284160
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007431093279839519,
      "loss": 1.4806,
      "theoretical_loss": 3.7260026759916443,
      "tokens_seen": 872349696
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007430892678034103,
      "loss": 1.5811,
      "theoretical_loss": 3.7259755704682065,
      "tokens_seen": 872415232
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007430692076228687,
      "loss": 1.513,
      "theoretical_loss": 3.7259484675509444,
      "tokens_seen": 872480768
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000743049147442327,
      "loss": 1.5524,
      "theoretical_loss": 3.7259213672394105,
      "tokens_seen": 872546304
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007430290872617854,
      "loss": 1.5574,
      "theoretical_loss": 3.725894269533159,
      "tokens_seen": 872611840
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007430090270812437,
      "loss": 1.5952,
      "theoretical_loss": 3.7258671744317446,
      "tokens_seen": 872677376
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007429889669007022,
      "loss": 1.5724,
      "theoretical_loss": 3.72584008193472,
      "tokens_seen": 872742912
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007429689067201605,
      "loss": 1.5871,
      "theoretical_loss": 3.7258129920416403,
      "tokens_seen": 872808448
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007429488465396189,
      "loss": 1.5604,
      "theoretical_loss": 3.725785904752059,
      "tokens_seen": 872873984
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007429287863590773,
      "loss": 1.5153,
      "theoretical_loss": 3.725758820065531,
      "tokens_seen": 872939520
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007429087261785356,
      "loss": 1.5189,
      "theoretical_loss": 3.725731737981611,
      "tokens_seen": 873005056
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000742888665997994,
      "loss": 1.6437,
      "theoretical_loss": 3.725704658499853,
      "tokens_seen": 873070592
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007428686058174525,
      "loss": 1.5078,
      "theoretical_loss": 3.7256775816198116,
      "tokens_seen": 873136128
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007428485456369107,
      "loss": 1.4921,
      "theoretical_loss": 3.725650507341042,
      "tokens_seen": 873201664
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007428284854563692,
      "loss": 1.5894,
      "theoretical_loss": 3.7256234356630986,
      "tokens_seen": 873267200
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007428084252758274,
      "loss": 1.6586,
      "theoretical_loss": 3.725596366585537,
      "tokens_seen": 873332736
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007427883650952859,
      "loss": 1.5162,
      "theoretical_loss": 3.725569300107912,
      "tokens_seen": 873398272
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007427683049147443,
      "loss": 1.4969,
      "theoretical_loss": 3.7255422362297788,
      "tokens_seen": 873463808
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007427482447342026,
      "loss": 1.5037,
      "theoretical_loss": 3.7255151749506927,
      "tokens_seen": 873529344
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000742728184553661,
      "loss": 1.6792,
      "theoretical_loss": 3.725488116270209,
      "tokens_seen": 873594880
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007427081243731195,
      "loss": 1.5403,
      "theoretical_loss": 3.725461060187884,
      "tokens_seen": 873660416
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007426880641925777,
      "loss": 1.424,
      "theoretical_loss": 3.7254340067032725,
      "tokens_seen": 873725952
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007426680040120362,
      "loss": 1.5238,
      "theoretical_loss": 3.7254069558159304,
      "tokens_seen": 873791488
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007426479438314944,
      "loss": 1.5869,
      "theoretical_loss": 3.725379907525414,
      "tokens_seen": 873857024
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007426278836509529,
      "loss": 1.5113,
      "theoretical_loss": 3.7253528618312792,
      "tokens_seen": 873922560
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007426078234704113,
      "loss": 1.6029,
      "theoretical_loss": 3.7253258187330816,
      "tokens_seen": 873988096
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007425877632898696,
      "loss": 1.629,
      "theoretical_loss": 3.7252987782303775,
      "tokens_seen": 874053632
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000742567703109328,
      "loss": 1.56,
      "theoretical_loss": 3.7252717403227242,
      "tokens_seen": 874119168
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007425476429287864,
      "loss": 1.5491,
      "theoretical_loss": 3.7252447050096773,
      "tokens_seen": 874184704
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007425275827482447,
      "loss": 1.4252,
      "theoretical_loss": 3.7252176722907926,
      "tokens_seen": 874250240
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007425075225677032,
      "loss": 1.6027,
      "theoretical_loss": 3.7251906421656282,
      "tokens_seen": 874315776
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007424874623871614,
      "loss": 1.6526,
      "theoretical_loss": 3.7251636146337406,
      "tokens_seen": 874381312
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0007424674022066199,
      "loss": 1.5684,
      "theoretical_loss": 3.7251365896946855,
      "tokens_seen": 874446848
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007424473420260782,
      "loss": 1.5727,
      "theoretical_loss": 3.725109567348021,
      "tokens_seen": 874512384
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007424272818455366,
      "loss": 1.5501,
      "theoretical_loss": 3.7250825475933036,
      "tokens_seen": 874577920
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000742407221664995,
      "loss": 1.451,
      "theoretical_loss": 3.7250555304300907,
      "tokens_seen": 874643456
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007423871614844534,
      "loss": 1.6368,
      "theoretical_loss": 3.7250285158579395,
      "tokens_seen": 874708992
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007423671013039117,
      "loss": 1.6774,
      "theoretical_loss": 3.725001503876408,
      "tokens_seen": 874774528
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007423470411233702,
      "loss": 1.5007,
      "theoretical_loss": 3.724974494485053,
      "tokens_seen": 874840064
    },
    {
      "epoch": 0.27,
      "objective/train/advantage_avg": 0.4936264455318451,
      "objective/train/docs_used": 323008,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.471723198890686,
      "objective/train/original_loss": 1.4717230796813965,
      "objective/train/theoretical_loss": 3.7249474876834325,
      "objective/train/tokens_used": 895365600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24419958889484406,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050604224205017,
      "objective/train/weighted_lm_loss": 1.546230435371399,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7249474876834325,
      "tokens_seen": 874905600
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007423269809428285,
      "loss": 1.4843,
      "theoretical_loss": 3.7249474876834325,
      "tokens_seen": 874905600
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007423069207622869,
      "loss": 1.6803,
      "theoretical_loss": 3.724920483471103,
      "tokens_seen": 874971136
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007422868605817452,
      "loss": 1.4638,
      "theoretical_loss": 3.7248934818476247,
      "tokens_seen": 875036672
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007422668004012036,
      "loss": 1.5224,
      "theoretical_loss": 3.7248664828125535,
      "tokens_seen": 875102208
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000742246740220662,
      "loss": 1.5498,
      "theoretical_loss": 3.7248394863654486,
      "tokens_seen": 875167744
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007422266800401204,
      "loss": 1.4866,
      "theoretical_loss": 3.724812492505867,
      "tokens_seen": 875233280
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007422066198595787,
      "loss": 1.5783,
      "theoretical_loss": 3.7247855012333693,
      "tokens_seen": 875298816
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007421865596790371,
      "loss": 1.5748,
      "theoretical_loss": 3.7247585125475107,
      "tokens_seen": 875364352
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007421664994984955,
      "loss": 1.5204,
      "theoretical_loss": 3.724731526447852,
      "tokens_seen": 875429888
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007421464393179539,
      "loss": 1.5049,
      "theoretical_loss": 3.724704542933951,
      "tokens_seen": 875495424
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007421263791374122,
      "loss": 1.5615,
      "theoretical_loss": 3.7246775620053665,
      "tokens_seen": 875560960
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007421063189568707,
      "loss": 1.7131,
      "theoretical_loss": 3.7246505836616577,
      "tokens_seen": 875626496
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007420862587763289,
      "loss": 1.5423,
      "theoretical_loss": 3.724623607902383,
      "tokens_seen": 875692032
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007420661985957874,
      "loss": 1.4527,
      "theoretical_loss": 3.724596634727101,
      "tokens_seen": 875757568
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007420461384152457,
      "loss": 1.5361,
      "theoretical_loss": 3.724569664135372,
      "tokens_seen": 875823104
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007420260782347041,
      "loss": 1.5124,
      "theoretical_loss": 3.724542696126754,
      "tokens_seen": 875888640
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007420060180541625,
      "loss": 1.4584,
      "theoretical_loss": 3.7245157307008077,
      "tokens_seen": 875954176
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007419859578736209,
      "loss": 1.6009,
      "theoretical_loss": 3.7244887678570917,
      "tokens_seen": 876019712
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007419658976930792,
      "loss": 1.6105,
      "theoretical_loss": 3.7244618075951657,
      "tokens_seen": 876085248
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007419458375125377,
      "loss": 1.6366,
      "theoretical_loss": 3.7244348499145894,
      "tokens_seen": 876150784
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007419257773319959,
      "loss": 1.534,
      "theoretical_loss": 3.724407894814923,
      "tokens_seen": 876216320
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007419057171514544,
      "loss": 1.5733,
      "theoretical_loss": 3.7243809422957255,
      "tokens_seen": 876281856
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007418856569709127,
      "loss": 1.5811,
      "theoretical_loss": 3.7243539923565576,
      "tokens_seen": 876347392
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007418655967903711,
      "loss": 1.5236,
      "theoretical_loss": 3.724327044996979,
      "tokens_seen": 876412928
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007418455366098295,
      "loss": 1.6448,
      "theoretical_loss": 3.7243001002165506,
      "tokens_seen": 876478464
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007418254764292878,
      "loss": 1.5224,
      "theoretical_loss": 3.724273158014832,
      "tokens_seen": 876544000
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007418054162487462,
      "loss": 1.5918,
      "theoretical_loss": 3.724246218391384,
      "tokens_seen": 876609536
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007417853560682047,
      "loss": 1.6872,
      "theoretical_loss": 3.724219281345767,
      "tokens_seen": 876675072
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007417652958876629,
      "loss": 1.5276,
      "theoretical_loss": 3.724192346877542,
      "tokens_seen": 876740608
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007417452357071214,
      "loss": 1.6045,
      "theoretical_loss": 3.724165414986269,
      "tokens_seen": 876806144
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007417251755265799,
      "loss": 1.5126,
      "theoretical_loss": 3.7241384856715096,
      "tokens_seen": 876871680
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007417051153460381,
      "loss": 1.4767,
      "theoretical_loss": 3.724111558932824,
      "tokens_seen": 876937216
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007416850551654966,
      "loss": 1.658,
      "theoretical_loss": 3.724084634769775,
      "tokens_seen": 877002752
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007416649949849548,
      "loss": 1.4686,
      "theoretical_loss": 3.7240577131819217,
      "tokens_seen": 877068288
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007416449348044133,
      "loss": 1.5592,
      "theoretical_loss": 3.724030794168826,
      "tokens_seen": 877133824
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007416248746238717,
      "loss": 1.5196,
      "theoretical_loss": 3.72400387773005,
      "tokens_seen": 877199360
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00074160481444333,
      "loss": 1.5377,
      "theoretical_loss": 3.7239769638651548,
      "tokens_seen": 877264896
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007415847542627884,
      "loss": 1.5222,
      "theoretical_loss": 3.723950052573702,
      "tokens_seen": 877330432
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007415646940822468,
      "loss": 1.5344,
      "theoretical_loss": 3.723923143855253,
      "tokens_seen": 877395968
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007415446339017051,
      "loss": 1.5653,
      "theoretical_loss": 3.7238962377093707,
      "tokens_seen": 877461504
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007415245737211636,
      "loss": 1.5572,
      "theoretical_loss": 3.7238693341356157,
      "tokens_seen": 877527040
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007415045135406218,
      "loss": 1.4018,
      "theoretical_loss": 3.7238424331335507,
      "tokens_seen": 877592576
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007414844533600803,
      "loss": 1.5305,
      "theoretical_loss": 3.723815534702738,
      "tokens_seen": 877658112
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007414643931795386,
      "loss": 1.6187,
      "theoretical_loss": 3.72378863884274,
      "tokens_seen": 877723648
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000741444332998997,
      "loss": 1.5816,
      "theoretical_loss": 3.7237617455531185,
      "tokens_seen": 877789184
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007414242728184554,
      "loss": 1.5925,
      "theoretical_loss": 3.723734854833436,
      "tokens_seen": 877854720
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007414042126379138,
      "loss": 1.581,
      "theoretical_loss": 3.7237079666832553,
      "tokens_seen": 877920256
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007413841524573721,
      "loss": 1.5874,
      "theoretical_loss": 3.7236810811021392,
      "tokens_seen": 877985792
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007413640922768306,
      "loss": 1.6223,
      "theoretical_loss": 3.7236541980896503,
      "tokens_seen": 878051328
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007413440320962889,
      "loss": 1.5755,
      "theoretical_loss": 3.7236273176453523,
      "tokens_seen": 878116864
    },
    {
      "epoch": 0.27,
      "objective/train/advantage_avg": 0.48837849497795105,
      "objective/train/docs_used": 324331,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3608418703079224,
      "objective/train/original_loss": 1.3608418703079224,
      "objective/train/theoretical_loss": 3.7236004397688065,
      "objective/train/tokens_used": 898642400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23926524817943573,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500540733337402,
      "objective/train/weighted_lm_loss": 1.4285813570022583,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7236004397688065,
      "tokens_seen": 878182400
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007413239719157473,
      "loss": 1.5554,
      "theoretical_loss": 3.7236004397688065,
      "tokens_seen": 878182400
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007413039117352056,
      "loss": 1.4834,
      "theoretical_loss": 3.7235735644595778,
      "tokens_seen": 878247936
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000741283851554664,
      "loss": 1.5936,
      "theoretical_loss": 3.723546691717228,
      "tokens_seen": 878313472
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007412637913741224,
      "loss": 1.4404,
      "theoretical_loss": 3.7235198215413217,
      "tokens_seen": 878379008
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007412437311935808,
      "loss": 1.6238,
      "theoretical_loss": 3.723492953931421,
      "tokens_seen": 878444544
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007412236710130391,
      "loss": 1.437,
      "theoretical_loss": 3.723466088887091,
      "tokens_seen": 878510080
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007412036108324975,
      "loss": 1.7096,
      "theoretical_loss": 3.723439226407894,
      "tokens_seen": 878575616
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007411835506519559,
      "loss": 1.526,
      "theoretical_loss": 3.7234123664933945,
      "tokens_seen": 878641152
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007411634904714143,
      "loss": 1.4855,
      "theoretical_loss": 3.7233855091431565,
      "tokens_seen": 878706688
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007411434302908726,
      "loss": 1.4808,
      "theoretical_loss": 3.723358654356743,
      "tokens_seen": 878772224
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000741123370110331,
      "loss": 1.5015,
      "theoretical_loss": 3.7233318021337194,
      "tokens_seen": 878837760
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007411033099297893,
      "loss": 1.5737,
      "theoretical_loss": 3.7233049524736486,
      "tokens_seen": 878903296
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007410832497492478,
      "loss": 1.6051,
      "theoretical_loss": 3.723278105376096,
      "tokens_seen": 878968832
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007410631895687061,
      "loss": 1.6191,
      "theoretical_loss": 3.723251260840625,
      "tokens_seen": 879034368
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007410431293881645,
      "loss": 1.4128,
      "theoretical_loss": 3.723224418866801,
      "tokens_seen": 879099904
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007410230692076229,
      "loss": 1.5402,
      "theoretical_loss": 3.723197579454188,
      "tokens_seen": 879165440
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007410030090270813,
      "loss": 1.4888,
      "theoretical_loss": 3.723170742602351,
      "tokens_seen": 879230976
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007409829488465396,
      "loss": 1.5943,
      "theoretical_loss": 3.7231439083108544,
      "tokens_seen": 879296512
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007409628886659981,
      "loss": 1.4868,
      "theoretical_loss": 3.723117076579264,
      "tokens_seen": 879362048
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007409428284854563,
      "loss": 1.6246,
      "theoretical_loss": 3.7230902474071446,
      "tokens_seen": 879427584
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007409227683049148,
      "loss": 1.5868,
      "theoretical_loss": 3.7230634207940607,
      "tokens_seen": 879493120
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007409027081243731,
      "loss": 1.4496,
      "theoretical_loss": 3.723036596739578,
      "tokens_seen": 879558656
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007408826479438315,
      "loss": 1.5951,
      "theoretical_loss": 3.723009775243262,
      "tokens_seen": 879624192
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007408625877632899,
      "loss": 1.6175,
      "theoretical_loss": 3.7229829563046772,
      "tokens_seen": 879689728
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007408425275827482,
      "loss": 1.6096,
      "theoretical_loss": 3.7229561399233906,
      "tokens_seen": 879755264
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007408224674022066,
      "loss": 1.5534,
      "theoretical_loss": 3.722929326098967,
      "tokens_seen": 879820800
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007408024072216651,
      "loss": 1.6555,
      "theoretical_loss": 3.722902514830973,
      "tokens_seen": 879886336
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007407823470411233,
      "loss": 1.6111,
      "theoretical_loss": 3.722875706118973,
      "tokens_seen": 879951872
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007407622868605818,
      "loss": 1.5977,
      "theoretical_loss": 3.7228488999625338,
      "tokens_seen": 880017408
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007407422266800402,
      "loss": 1.6121,
      "theoretical_loss": 3.722822096361222,
      "tokens_seen": 880082944
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007407221664994985,
      "loss": 1.6541,
      "theoretical_loss": 3.7227952953146035,
      "tokens_seen": 880148480
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007407021063189569,
      "loss": 1.4959,
      "theoretical_loss": 3.7227684968222445,
      "tokens_seen": 880214016
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007406820461384152,
      "loss": 1.5202,
      "theoretical_loss": 3.722741700883711,
      "tokens_seen": 880279552
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007406619859578736,
      "loss": 1.5318,
      "theoretical_loss": 3.7227149074985704,
      "tokens_seen": 880345088
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007406419257773321,
      "loss": 1.5597,
      "theoretical_loss": 3.7226881166663883,
      "tokens_seen": 880410624
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007406218655967903,
      "loss": 1.4649,
      "theoretical_loss": 3.7226613283867325,
      "tokens_seen": 880476160
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007406018054162488,
      "loss": 1.5753,
      "theoretical_loss": 3.7226345426591694,
      "tokens_seen": 880541696
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000740581745235707,
      "loss": 1.5851,
      "theoretical_loss": 3.7226077594832656,
      "tokens_seen": 880607232
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007405616850551655,
      "loss": 1.4576,
      "theoretical_loss": 3.7225809788585886,
      "tokens_seen": 880672768
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007405416248746239,
      "loss": 1.4086,
      "theoretical_loss": 3.722554200784705,
      "tokens_seen": 880738304
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007405215646940822,
      "loss": 1.4775,
      "theoretical_loss": 3.722527425261183,
      "tokens_seen": 880803840
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007405015045135406,
      "loss": 1.4509,
      "theoretical_loss": 3.7225006522875894,
      "tokens_seen": 880869376
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000740481444332999,
      "loss": 1.5024,
      "theoretical_loss": 3.7224738818634915,
      "tokens_seen": 880934912
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007404613841524573,
      "loss": 1.6874,
      "theoretical_loss": 3.7224471139884567,
      "tokens_seen": 881000448
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007404413239719158,
      "loss": 1.7144,
      "theoretical_loss": 3.7224203486620535,
      "tokens_seen": 881065984
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007404212637913741,
      "loss": 1.6256,
      "theoretical_loss": 3.7223935858838493,
      "tokens_seen": 881131520
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007404012036108325,
      "loss": 1.5768,
      "theoretical_loss": 3.7223668256534124,
      "tokens_seen": 881197056
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007403811434302909,
      "loss": 1.5712,
      "theoretical_loss": 3.7223400679703094,
      "tokens_seen": 881262592
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007403610832497493,
      "loss": 1.5289,
      "theoretical_loss": 3.7223133128341104,
      "tokens_seen": 881328128
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007403410230692076,
      "loss": 1.5089,
      "theoretical_loss": 3.722286560244382,
      "tokens_seen": 881393664
    },
    {
      "epoch": 0.27,
      "objective/train/advantage_avg": 0.4939919412136078,
      "objective/train/docs_used": 325483,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4167739152908325,
      "objective/train/original_loss": 1.416774034500122,
      "objective/train/theoretical_loss": 3.722259810200693,
      "objective/train/tokens_used": 901919200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2443196326494217,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050641417503357,
      "objective/train/weighted_lm_loss": 1.4885128736495972,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.722259810200693,
      "tokens_seen": 881459200
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000740320962888666,
      "loss": 1.4845,
      "theoretical_loss": 3.722259810200693,
      "tokens_seen": 881459200
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007403009027081243,
      "loss": 1.6287,
      "theoretical_loss": 3.722233062702612,
      "tokens_seen": 881524736
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007402808425275828,
      "loss": 1.4938,
      "theoretical_loss": 3.722206317749708,
      "tokens_seen": 881590272
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007402607823470411,
      "loss": 1.5926,
      "theoretical_loss": 3.7221795753415483,
      "tokens_seen": 881655808
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007402407221664995,
      "loss": 1.5602,
      "theoretical_loss": 3.722152835477703,
      "tokens_seen": 881721344
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007402206619859579,
      "loss": 1.6273,
      "theoretical_loss": 3.72212609815774,
      "tokens_seen": 881786880
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007402006018054163,
      "loss": 1.5327,
      "theoretical_loss": 3.722099363381229,
      "tokens_seen": 881852416
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007401805416248747,
      "loss": 1.5627,
      "theoretical_loss": 3.7220726311477383,
      "tokens_seen": 881917952
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000740160481444333,
      "loss": 1.5315,
      "theoretical_loss": 3.7220459014568377,
      "tokens_seen": 881983488
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007401404212637915,
      "loss": 1.5971,
      "theoretical_loss": 3.7220191743080964,
      "tokens_seen": 882049024
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007401203610832497,
      "loss": 1.5517,
      "theoretical_loss": 3.7219924497010837,
      "tokens_seen": 882114560
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007401003009027082,
      "loss": 1.5536,
      "theoretical_loss": 3.721965727635369,
      "tokens_seen": 882180096
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007400802407221665,
      "loss": 1.5868,
      "theoretical_loss": 3.721939008110521,
      "tokens_seen": 882245632
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007400601805416249,
      "loss": 1.5956,
      "theoretical_loss": 3.721912291126111,
      "tokens_seen": 882311168
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007400401203610833,
      "loss": 1.4812,
      "theoretical_loss": 3.721885576681708,
      "tokens_seen": 882376704
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007400200601805417,
      "loss": 1.6593,
      "theoretical_loss": 3.7218588647768813,
      "tokens_seen": 882442240
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00074,
      "loss": 1.474,
      "theoretical_loss": 3.721832155411202,
      "tokens_seen": 882507776
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007399799398194585,
      "loss": 1.5548,
      "theoretical_loss": 3.7218054485842393,
      "tokens_seen": 882573312
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007399598796389167,
      "loss": 1.6429,
      "theoretical_loss": 3.7217787442955643,
      "tokens_seen": 882638848
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007399398194583752,
      "loss": 1.6546,
      "theoretical_loss": 3.721752042544747,
      "tokens_seen": 882704384
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007399197592778335,
      "loss": 1.576,
      "theoretical_loss": 3.7217253433313573,
      "tokens_seen": 882769920
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007398996990972919,
      "loss": 1.5021,
      "theoretical_loss": 3.721698646654966,
      "tokens_seen": 882835456
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007398796389167503,
      "loss": 1.5078,
      "theoretical_loss": 3.721671952515144,
      "tokens_seen": 882900992
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007398595787362086,
      "loss": 1.4952,
      "theoretical_loss": 3.7216452609114614,
      "tokens_seen": 882966528
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000739839518555667,
      "loss": 1.5972,
      "theoretical_loss": 3.7216185718434898,
      "tokens_seen": 883032064
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007398194583751255,
      "loss": 1.5981,
      "theoretical_loss": 3.7215918853108,
      "tokens_seen": 883097600
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007397993981945837,
      "loss": 1.4982,
      "theoretical_loss": 3.7215652013129628,
      "tokens_seen": 883163136
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007397793380140422,
      "loss": 1.6637,
      "theoretical_loss": 3.721538519849549,
      "tokens_seen": 883228672
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007397592778335006,
      "loss": 1.4993,
      "theoretical_loss": 3.7215118409201304,
      "tokens_seen": 883294208
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007397392176529589,
      "loss": 1.5051,
      "theoretical_loss": 3.7214851645242786,
      "tokens_seen": 883359744
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007397191574724173,
      "loss": 1.481,
      "theoretical_loss": 3.7214584906615644,
      "tokens_seen": 883425280
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007396990972918756,
      "loss": 1.5397,
      "theoretical_loss": 3.72143181933156,
      "tokens_seen": 883490816
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000739679037111334,
      "loss": 1.4114,
      "theoretical_loss": 3.721405150533836,
      "tokens_seen": 883556352
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007396589769307925,
      "loss": 1.4995,
      "theoretical_loss": 3.7213784842679654,
      "tokens_seen": 883621888
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007396389167502507,
      "loss": 1.4948,
      "theoretical_loss": 3.7213518205335196,
      "tokens_seen": 883687424
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007396188565697092,
      "loss": 1.6197,
      "theoretical_loss": 3.7213251593300702,
      "tokens_seen": 883752960
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007395987963891675,
      "loss": 1.5077,
      "theoretical_loss": 3.7212985006571904,
      "tokens_seen": 883818496
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007395787362086259,
      "loss": 1.5782,
      "theoretical_loss": 3.721271844514451,
      "tokens_seen": 883884032
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007395586760280843,
      "loss": 1.5045,
      "theoretical_loss": 3.721245190901425,
      "tokens_seen": 883949568
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007395386158475426,
      "loss": 1.6692,
      "theoretical_loss": 3.7212185398176847,
      "tokens_seen": 884015104
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000739518555667001,
      "loss": 1.5233,
      "theoretical_loss": 3.721191891262803,
      "tokens_seen": 884080640
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007394984954864594,
      "loss": 1.576,
      "theoretical_loss": 3.7211652452363513,
      "tokens_seen": 884146176
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007394784353059177,
      "loss": 1.4988,
      "theoretical_loss": 3.721138601737904,
      "tokens_seen": 884211712
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007394583751253762,
      "loss": 1.548,
      "theoretical_loss": 3.7211119607670327,
      "tokens_seen": 884277248
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007394383149448345,
      "loss": 1.5966,
      "theoretical_loss": 3.7210853223233107,
      "tokens_seen": 884342784
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007394182547642929,
      "loss": 1.3624,
      "theoretical_loss": 3.721058686406311,
      "tokens_seen": 884408320
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007393981945837513,
      "loss": 1.5949,
      "theoretical_loss": 3.721032053015607,
      "tokens_seen": 884473856
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007393781344032097,
      "loss": 1.5336,
      "theoretical_loss": 3.721005422150771,
      "tokens_seen": 884539392
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000739358074222668,
      "loss": 1.4031,
      "theoretical_loss": 3.720978793811378,
      "tokens_seen": 884604928
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007393380140421264,
      "loss": 1.6101,
      "theoretical_loss": 3.720952167997,
      "tokens_seen": 884670464
    },
    {
      "epoch": 0.27,
      "objective/train/advantage_avg": 0.49303656816482544,
      "objective/train/docs_used": 326745,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.664990782737732,
      "objective/train/original_loss": 1.6649906635284424,
      "objective/train/theoretical_loss": 3.720925544707211,
      "objective/train/tokens_used": 905196000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24346923828125,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505414009094238,
      "objective/train/weighted_lm_loss": 1.7489945888519287,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.720925544707211,
      "tokens_seen": 884736000
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007393179538615847,
      "loss": 1.6132,
      "theoretical_loss": 3.720925544707211,
      "tokens_seen": 884736000
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007392978936810432,
      "loss": 1.5637,
      "theoretical_loss": 3.7208989239415846,
      "tokens_seen": 884801536
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007392778335005015,
      "loss": 1.5596,
      "theoretical_loss": 3.7208723056996944,
      "tokens_seen": 884867072
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007392577733199599,
      "loss": 1.5777,
      "theoretical_loss": 3.7208456899811146,
      "tokens_seen": 884932608
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007392377131394182,
      "loss": 1.6254,
      "theoretical_loss": 3.720819076785419,
      "tokens_seen": 884998144
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007392176529588767,
      "loss": 1.5619,
      "theoretical_loss": 3.7207924661121816,
      "tokens_seen": 885063680
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000739197592778335,
      "loss": 1.5783,
      "theoretical_loss": 3.7207658579609766,
      "tokens_seen": 885129216
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007391775325977934,
      "loss": 1.3964,
      "theoretical_loss": 3.720739252331378,
      "tokens_seen": 885194752
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007391574724172517,
      "loss": 1.4862,
      "theoretical_loss": 3.720712649222961,
      "tokens_seen": 885260288
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007391374122367101,
      "loss": 1.4439,
      "theoretical_loss": 3.720686048635299,
      "tokens_seen": 885325824
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007391173520561685,
      "loss": 1.5158,
      "theoretical_loss": 3.7206594505679673,
      "tokens_seen": 885391360
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007390972918756269,
      "loss": 1.557,
      "theoretical_loss": 3.7206328550205408,
      "tokens_seen": 885456896
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007390772316950852,
      "loss": 1.5826,
      "theoretical_loss": 3.720606261992593,
      "tokens_seen": 885522432
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007390571715145437,
      "loss": 1.6015,
      "theoretical_loss": 3.7205796714837005,
      "tokens_seen": 885587968
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000739037111334002,
      "loss": 1.53,
      "theoretical_loss": 3.720553083493437,
      "tokens_seen": 885653504
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007390170511534604,
      "loss": 1.5517,
      "theoretical_loss": 3.720526498021378,
      "tokens_seen": 885719040
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007389969909729188,
      "loss": 1.6243,
      "theoretical_loss": 3.7204999150670988,
      "tokens_seen": 885784576
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007389769307923771,
      "loss": 1.496,
      "theoretical_loss": 3.7204733346301744,
      "tokens_seen": 885850112
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007389568706118355,
      "loss": 1.5409,
      "theoretical_loss": 3.7204467567101807,
      "tokens_seen": 885915648
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000738936810431294,
      "loss": 1.5645,
      "theoretical_loss": 3.720420181306693,
      "tokens_seen": 885981184
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007389167502507522,
      "loss": 1.5123,
      "theoretical_loss": 3.7203936084192866,
      "tokens_seen": 886046720
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007388966900702107,
      "loss": 1.5715,
      "theoretical_loss": 3.7203670380475375,
      "tokens_seen": 886112256
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007388766298896689,
      "loss": 1.4636,
      "theoretical_loss": 3.7203404701910214,
      "tokens_seen": 886177792
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007388565697091274,
      "loss": 1.6178,
      "theoretical_loss": 3.7203139048493146,
      "tokens_seen": 886243328
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007388365095285858,
      "loss": 1.6371,
      "theoretical_loss": 3.720287342021992,
      "tokens_seen": 886308864
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007388164493480441,
      "loss": 1.4193,
      "theoretical_loss": 3.7202607817086313,
      "tokens_seen": 886374400
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007387963891675025,
      "loss": 1.5348,
      "theoretical_loss": 3.7202342239088075,
      "tokens_seen": 886439936
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000738776328986961,
      "loss": 1.5865,
      "theoretical_loss": 3.7202076686220975,
      "tokens_seen": 886505472
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007387562688064192,
      "loss": 1.5465,
      "theoretical_loss": 3.720181115848078,
      "tokens_seen": 886571008
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007387362086258777,
      "loss": 1.4365,
      "theoretical_loss": 3.7201545655863244,
      "tokens_seen": 886636544
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000738716148445336,
      "loss": 1.4955,
      "theoretical_loss": 3.7201280178364144,
      "tokens_seen": 886702080
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007386960882647944,
      "loss": 1.5549,
      "theoretical_loss": 3.720101472597925,
      "tokens_seen": 886767616
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007386760280842529,
      "loss": 1.5306,
      "theoretical_loss": 3.7200749298704316,
      "tokens_seen": 886833152
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007386559679037111,
      "loss": 1.5458,
      "theoretical_loss": 3.720048389653512,
      "tokens_seen": 886898688
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007386359077231696,
      "loss": 1.5625,
      "theoretical_loss": 3.7200218519467434,
      "tokens_seen": 886964224
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007386158475426279,
      "loss": 1.5483,
      "theoretical_loss": 3.719995316749703,
      "tokens_seen": 887029760
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007385957873620863,
      "loss": 1.6459,
      "theoretical_loss": 3.7199687840619675,
      "tokens_seen": 887095296
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007385757271815447,
      "loss": 1.5754,
      "theoretical_loss": 3.719942253883115,
      "tokens_seen": 887160832
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000738555667001003,
      "loss": 1.5327,
      "theoretical_loss": 3.7199157262127223,
      "tokens_seen": 887226368
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007385356068204614,
      "loss": 1.5918,
      "theoretical_loss": 3.719889201050367,
      "tokens_seen": 887291904
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007385155466399198,
      "loss": 1.5391,
      "theoretical_loss": 3.719862678395627,
      "tokens_seen": 887357440
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007384954864593781,
      "loss": 1.5709,
      "theoretical_loss": 3.7198361582480803,
      "tokens_seen": 887422976
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007384754262788366,
      "loss": 1.5289,
      "theoretical_loss": 3.7198096406073047,
      "tokens_seen": 887488512
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007384553660982949,
      "loss": 1.5832,
      "theoretical_loss": 3.719783125472877,
      "tokens_seen": 887554048
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007384353059177533,
      "loss": 1.4561,
      "theoretical_loss": 3.719756612844377,
      "tokens_seen": 887619584
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007384152457372117,
      "loss": 1.4902,
      "theoretical_loss": 3.719730102721382,
      "tokens_seen": 887685120
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00073839518555667,
      "loss": 1.484,
      "theoretical_loss": 3.7197035951034705,
      "tokens_seen": 887750656
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007383751253761284,
      "loss": 1.5796,
      "theoretical_loss": 3.7196770899902205,
      "tokens_seen": 887816192
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007383550651955868,
      "loss": 1.4534,
      "theoretical_loss": 3.7196505873812105,
      "tokens_seen": 887881728
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007383350050150451,
      "loss": 1.6008,
      "theoretical_loss": 3.71962408727602,
      "tokens_seen": 887947264
    },
    {
      "epoch": 0.27,
      "objective/train/advantage_avg": 0.4947243332862854,
      "objective/train/docs_used": 327973,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3370107412338257,
      "objective/train/original_loss": 1.3370107412338257,
      "objective/train/theoretical_loss": 3.719597589674226,
      "objective/train/tokens_used": 908472800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24511003494262695,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507185459136963,
      "objective/train/weighted_lm_loss": 1.404590129852295,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.719597589674226,
      "tokens_seen": 888012800
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007383149448345036,
      "loss": 1.4767,
      "theoretical_loss": 3.719597589674226,
      "tokens_seen": 888012800
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007382948846539619,
      "loss": 1.6183,
      "theoretical_loss": 3.719571094575409,
      "tokens_seen": 888078336
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007382748244734203,
      "loss": 1.5199,
      "theoretical_loss": 3.7195446019791465,
      "tokens_seen": 888143872
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007382547642928786,
      "loss": 1.5538,
      "theoretical_loss": 3.7195181118850185,
      "tokens_seen": 888209408
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007382347041123371,
      "loss": 1.5071,
      "theoretical_loss": 3.719491624292604,
      "tokens_seen": 888274944
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007382146439317954,
      "loss": 1.5108,
      "theoretical_loss": 3.719465139201482,
      "tokens_seen": 888340480
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007381945837512538,
      "loss": 1.5689,
      "theoretical_loss": 3.7194386566112314,
      "tokens_seen": 888406016
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007381745235707121,
      "loss": 1.4405,
      "theoretical_loss": 3.719412176521432,
      "tokens_seen": 888471552
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007381544633901705,
      "loss": 1.5605,
      "theoretical_loss": 3.719385698931663,
      "tokens_seen": 888537088
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007381344032096289,
      "loss": 1.5493,
      "theoretical_loss": 3.7193592238415043,
      "tokens_seen": 888602624
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007381143430290873,
      "loss": 1.4485,
      "theoretical_loss": 3.7193327512505356,
      "tokens_seen": 888668160
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007380942828485456,
      "loss": 1.5138,
      "theoretical_loss": 3.719306281158337,
      "tokens_seen": 888733696
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007380742226680041,
      "loss": 1.5721,
      "theoretical_loss": 3.7192798135644876,
      "tokens_seen": 888799232
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007380541624874624,
      "loss": 1.5248,
      "theoretical_loss": 3.719253348468568,
      "tokens_seen": 888864768
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007380341023069208,
      "loss": 1.676,
      "theoretical_loss": 3.7192268858701576,
      "tokens_seen": 888930304
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007380140421263792,
      "loss": 1.5078,
      "theoretical_loss": 3.7192004257688374,
      "tokens_seen": 888995840
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007379939819458375,
      "loss": 1.5258,
      "theoretical_loss": 3.7191739681641875,
      "tokens_seen": 889061376
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007379739217652959,
      "loss": 1.5171,
      "theoretical_loss": 3.7191475130557885,
      "tokens_seen": 889126912
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007379538615847543,
      "loss": 1.4757,
      "theoretical_loss": 3.7191210604432205,
      "tokens_seen": 889192448
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007379338014042126,
      "loss": 1.5757,
      "theoretical_loss": 3.719094610326064,
      "tokens_seen": 889257984
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007379137412236711,
      "loss": 1.5898,
      "theoretical_loss": 3.7190681627039,
      "tokens_seen": 889323520
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007378936810431293,
      "loss": 1.5945,
      "theoretical_loss": 3.7190417175763097,
      "tokens_seen": 889389056
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007378736208625878,
      "loss": 1.5437,
      "theoretical_loss": 3.7190152749428735,
      "tokens_seen": 889454592
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007378535606820462,
      "loss": 1.6152,
      "theoretical_loss": 3.7189888348031728,
      "tokens_seen": 889520128
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007378335005015045,
      "loss": 1.5619,
      "theoretical_loss": 3.718962397156788,
      "tokens_seen": 889585664
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007378134403209629,
      "loss": 1.4392,
      "theoretical_loss": 3.718935962003301,
      "tokens_seen": 889651200
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007377933801404214,
      "loss": 1.5465,
      "theoretical_loss": 3.7189095293422927,
      "tokens_seen": 889716736
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007377733199598796,
      "loss": 1.5394,
      "theoretical_loss": 3.718883099173345,
      "tokens_seen": 889782272
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007377532597793381,
      "loss": 1.6119,
      "theoretical_loss": 3.7188566714960394,
      "tokens_seen": 889847808
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007377331995987963,
      "loss": 1.5594,
      "theoretical_loss": 3.718830246309957,
      "tokens_seen": 889913344
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007377131394182548,
      "loss": 1.5805,
      "theoretical_loss": 3.71880382361468,
      "tokens_seen": 889978880
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007376930792377132,
      "loss": 1.6005,
      "theoretical_loss": 3.7187774034097894,
      "tokens_seen": 890044416
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007376730190571715,
      "loss": 1.5652,
      "theoretical_loss": 3.7187509856948684,
      "tokens_seen": 890109952
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007376529588766299,
      "loss": 1.5576,
      "theoretical_loss": 3.7187245704694982,
      "tokens_seen": 890175488
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007376328986960883,
      "loss": 1.4933,
      "theoretical_loss": 3.7186981577332614,
      "tokens_seen": 890241024
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007376128385155466,
      "loss": 1.4404,
      "theoretical_loss": 3.7186717474857396,
      "tokens_seen": 890306560
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007375927783350051,
      "loss": 1.4874,
      "theoretical_loss": 3.718645339726516,
      "tokens_seen": 890372096
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007375727181544633,
      "loss": 1.5229,
      "theoretical_loss": 3.718618934455172,
      "tokens_seen": 890437632
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007375526579739218,
      "loss": 1.5385,
      "theoretical_loss": 3.718592531671291,
      "tokens_seen": 890503168
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007375325977933801,
      "loss": 1.5386,
      "theoretical_loss": 3.7185661313744554,
      "tokens_seen": 890568704
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007375125376128385,
      "loss": 1.5603,
      "theoretical_loss": 3.7185397335642474,
      "tokens_seen": 890634240
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007374924774322969,
      "loss": 1.5324,
      "theoretical_loss": 3.7185133382402507,
      "tokens_seen": 890699776
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007374724172517553,
      "loss": 1.5549,
      "theoretical_loss": 3.7184869454020477,
      "tokens_seen": 890765312
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007374523570712136,
      "loss": 1.451,
      "theoretical_loss": 3.718460555049221,
      "tokens_seen": 890830848
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007374322968906721,
      "loss": 1.5199,
      "theoretical_loss": 3.718434167181355,
      "tokens_seen": 890896384
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007374122367101303,
      "loss": 1.5396,
      "theoretical_loss": 3.718407781798032,
      "tokens_seen": 890961920
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007373921765295888,
      "loss": 1.5787,
      "theoretical_loss": 3.7183813988988357,
      "tokens_seen": 891027456
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007373721163490471,
      "loss": 1.4986,
      "theoretical_loss": 3.71835501848335,
      "tokens_seen": 891092992
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007373520561685055,
      "loss": 1.6682,
      "theoretical_loss": 3.7183286405511566,
      "tokens_seen": 891158528
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007373319959879639,
      "loss": 1.669,
      "theoretical_loss": 3.7183022651018414,
      "tokens_seen": 891224064
    },
    {
      "epoch": 0.27,
      "objective/train/advantage_avg": 0.4911864101886749,
      "objective/train/docs_used": 329220,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3891663551330566,
      "objective/train/original_loss": 1.3891664743423462,
      "objective/train/theoretical_loss": 3.7182758921349865,
      "objective/train/tokens_used": 911749600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2422795444726944,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503501892089844,
      "objective/train/weighted_lm_loss": 1.4590027332305908,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7182758921349865,
      "tokens_seen": 891289600
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007373119358074223,
      "loss": 1.497,
      "theoretical_loss": 3.7182758921349865,
      "tokens_seen": 891289600
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007372918756268806,
      "loss": 1.4432,
      "theoretical_loss": 3.7182495216501765,
      "tokens_seen": 891355136
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000737271815446339,
      "loss": 1.5203,
      "theoretical_loss": 3.7182231536469956,
      "tokens_seen": 891420672
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007372517552657974,
      "loss": 1.634,
      "theoretical_loss": 3.7181967881250273,
      "tokens_seen": 891486208
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007372316950852558,
      "loss": 1.5754,
      "theoretical_loss": 3.718170425083856,
      "tokens_seen": 891551744
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007372116349047142,
      "loss": 1.6224,
      "theoretical_loss": 3.7181440645230657,
      "tokens_seen": 891617280
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007371915747241725,
      "loss": 1.4085,
      "theoretical_loss": 3.718117706442241,
      "tokens_seen": 891682816
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007371715145436309,
      "loss": 1.584,
      "theoretical_loss": 3.718091350840966,
      "tokens_seen": 891748352
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007371514543630893,
      "loss": 1.5142,
      "theoretical_loss": 3.718064997718826,
      "tokens_seen": 891813888
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007371313941825477,
      "loss": 1.4979,
      "theoretical_loss": 3.7180386470754048,
      "tokens_seen": 891879424
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000737111334002006,
      "loss": 1.5761,
      "theoretical_loss": 3.7180122989102875,
      "tokens_seen": 891944960
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007370912738214645,
      "loss": 1.5686,
      "theoretical_loss": 3.7179859532230584,
      "tokens_seen": 892010496
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007370712136409228,
      "loss": 1.4928,
      "theoretical_loss": 3.7179596100133034,
      "tokens_seen": 892076032
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007370511534603812,
      "loss": 1.5706,
      "theoretical_loss": 3.7179332692806075,
      "tokens_seen": 892141568
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007370310932798396,
      "loss": 1.6209,
      "theoretical_loss": 3.7179069310245554,
      "tokens_seen": 892207104
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007370110330992979,
      "loss": 1.5485,
      "theoretical_loss": 3.717880595244732,
      "tokens_seen": 892272640
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007369909729187563,
      "loss": 1.6008,
      "theoretical_loss": 3.7178542619407233,
      "tokens_seen": 892338176
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007369709127382147,
      "loss": 1.5462,
      "theoretical_loss": 3.7178279311121143,
      "tokens_seen": 892403712
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000736950852557673,
      "loss": 1.6487,
      "theoretical_loss": 3.717801602758491,
      "tokens_seen": 892469248
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007369307923771315,
      "loss": 1.62,
      "theoretical_loss": 3.717775276879438,
      "tokens_seen": 892534784
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007369107321965897,
      "loss": 1.5631,
      "theoretical_loss": 3.7177489534745427,
      "tokens_seen": 892600320
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007368906720160482,
      "loss": 1.6712,
      "theoretical_loss": 3.7177226325433894,
      "tokens_seen": 892665856
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007368706118355066,
      "loss": 1.5399,
      "theoretical_loss": 3.7176963140855657,
      "tokens_seen": 892731392
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007368505516549649,
      "loss": 1.6741,
      "theoretical_loss": 3.717669998100656,
      "tokens_seen": 892796928
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007368304914744233,
      "loss": 1.5021,
      "theoretical_loss": 3.717643684588247,
      "tokens_seen": 892862464
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007368104312938818,
      "loss": 1.5849,
      "theoretical_loss": 3.7176173735479248,
      "tokens_seen": 892928000
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00073679037111334,
      "loss": 1.6585,
      "theoretical_loss": 3.7175910649792767,
      "tokens_seen": 892993536
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007367703109327985,
      "loss": 1.6115,
      "theoretical_loss": 3.717564758881888,
      "tokens_seen": 893059072
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007367502507522567,
      "loss": 1.3831,
      "theoretical_loss": 3.7175384552553457,
      "tokens_seen": 893124608
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007367301905717152,
      "loss": 1.5941,
      "theoretical_loss": 3.7175121540992366,
      "tokens_seen": 893190144
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007367101303911736,
      "loss": 1.4814,
      "theoretical_loss": 3.7174858554131465,
      "tokens_seen": 893255680
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007366900702106319,
      "loss": 1.375,
      "theoretical_loss": 3.717459559196664,
      "tokens_seen": 893321216
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007366700100300903,
      "loss": 1.5655,
      "theoretical_loss": 3.7174332654493742,
      "tokens_seen": 893386752
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007366499498495487,
      "loss": 1.5439,
      "theoretical_loss": 3.7174069741708653,
      "tokens_seen": 893452288
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000736629889669007,
      "loss": 1.5214,
      "theoretical_loss": 3.717380685360724,
      "tokens_seen": 893517824
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007366098294884655,
      "loss": 1.5316,
      "theoretical_loss": 3.717354399018538,
      "tokens_seen": 893583360
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007365897693079237,
      "loss": 1.631,
      "theoretical_loss": 3.717328115143894,
      "tokens_seen": 893648896
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007365697091273822,
      "loss": 1.6299,
      "theoretical_loss": 3.7173018337363795,
      "tokens_seen": 893714432
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007365496489468405,
      "loss": 1.5717,
      "theoretical_loss": 3.7172755547955827,
      "tokens_seen": 893779968
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007365295887662989,
      "loss": 1.5854,
      "theoretical_loss": 3.7172492783210904,
      "tokens_seen": 893845504
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007365095285857573,
      "loss": 1.4504,
      "theoretical_loss": 3.717223004312491,
      "tokens_seen": 893911040
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007364894684052157,
      "loss": 1.5354,
      "theoretical_loss": 3.717196732769372,
      "tokens_seen": 893976576
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000736469408224674,
      "loss": 1.5052,
      "theoretical_loss": 3.717170463691321,
      "tokens_seen": 894042112
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007364493480441325,
      "loss": 1.5838,
      "theoretical_loss": 3.7171441970779266,
      "tokens_seen": 894107648
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007364292878635907,
      "loss": 1.6257,
      "theoretical_loss": 3.717117932928777,
      "tokens_seen": 894173184
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007364092276830492,
      "loss": 1.62,
      "theoretical_loss": 3.71709167124346,
      "tokens_seen": 894238720
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007363891675025075,
      "loss": 1.5352,
      "theoretical_loss": 3.7170654120215634,
      "tokens_seen": 894304256
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007363691073219659,
      "loss": 1.5453,
      "theoretical_loss": 3.717039155262677,
      "tokens_seen": 894369792
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007363490471414243,
      "loss": 1.439,
      "theoretical_loss": 3.7170129009663886,
      "tokens_seen": 894435328
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007363289869608827,
      "loss": 1.7361,
      "theoretical_loss": 3.716986649132287,
      "tokens_seen": 894500864
    },
    {
      "epoch": 0.27,
      "objective/train/advantage_avg": 0.49507322907447815,
      "objective/train/docs_used": 330336,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3652136325836182,
      "objective/train/original_loss": 1.365213394165039,
      "objective/train/theoretical_loss": 3.7169603997599605,
      "objective/train/tokens_used": 915026400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24529388546943665,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050754427909851,
      "objective/train/weighted_lm_loss": 1.4345537424087524,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7169603997599605,
      "tokens_seen": 894566400
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000736308926780341,
      "loss": 1.504,
      "theoretical_loss": 3.7169603997599605,
      "tokens_seen": 894566400
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007362888665997994,
      "loss": 1.5418,
      "theoretical_loss": 3.716934152848998,
      "tokens_seen": 894631936
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007362688064192578,
      "loss": 1.5402,
      "theoretical_loss": 3.716907908398989,
      "tokens_seen": 894697472
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007362487462387162,
      "loss": 1.4954,
      "theoretical_loss": 3.716881666409522,
      "tokens_seen": 894763008
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007362286860581745,
      "loss": 1.662,
      "theoretical_loss": 3.716855426880186,
      "tokens_seen": 894828544
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000736208625877633,
      "loss": 1.4931,
      "theoretical_loss": 3.716829189810571,
      "tokens_seen": 894894080
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007361885656970912,
      "loss": 1.5073,
      "theoretical_loss": 3.7168029552002655,
      "tokens_seen": 894959616
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007361685055165497,
      "loss": 1.5268,
      "theoretical_loss": 3.7167767230488593,
      "tokens_seen": 895025152
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000736148445336008,
      "loss": 1.604,
      "theoretical_loss": 3.716750493355941,
      "tokens_seen": 895090688
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007361283851554664,
      "loss": 1.5471,
      "theoretical_loss": 3.716724266121102,
      "tokens_seen": 895156224
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007361083249749248,
      "loss": 1.562,
      "theoretical_loss": 3.716698041343931,
      "tokens_seen": 895221760
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007360882647943832,
      "loss": 1.5751,
      "theoretical_loss": 3.7166718190240178,
      "tokens_seen": 895287296
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007360682046138415,
      "loss": 1.6129,
      "theoretical_loss": 3.716645599160952,
      "tokens_seen": 895352832
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007360481444333,
      "loss": 1.6391,
      "theoretical_loss": 3.7166193817543243,
      "tokens_seen": 895418368
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007360280842527582,
      "loss": 1.5138,
      "theoretical_loss": 3.716593166803724,
      "tokens_seen": 895483904
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007360080240722167,
      "loss": 1.6115,
      "theoretical_loss": 3.7165669543087425,
      "tokens_seen": 895549440
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000735987963891675,
      "loss": 1.4009,
      "theoretical_loss": 3.716540744268969,
      "tokens_seen": 895614976
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007359679037111334,
      "loss": 1.4429,
      "theoretical_loss": 3.7165145366839942,
      "tokens_seen": 895680512
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007359478435305918,
      "loss": 1.5126,
      "theoretical_loss": 3.7164883315534087,
      "tokens_seen": 895746048
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007359277833500501,
      "loss": 1.6318,
      "theoretical_loss": 3.7164621288768025,
      "tokens_seen": 895811584
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007359077231695085,
      "loss": 1.6113,
      "theoretical_loss": 3.716435928653768,
      "tokens_seen": 895877120
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000735887662988967,
      "loss": 1.5509,
      "theoretical_loss": 3.7164097308838935,
      "tokens_seen": 895942656
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007358676028084252,
      "loss": 1.5021,
      "theoretical_loss": 3.7163835355667723,
      "tokens_seen": 896008192
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007358475426278837,
      "loss": 1.5639,
      "theoretical_loss": 3.716357342701994,
      "tokens_seen": 896073728
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007358274824473419,
      "loss": 1.5175,
      "theoretical_loss": 3.7163311522891496,
      "tokens_seen": 896139264
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007358074222668004,
      "loss": 1.4844,
      "theoretical_loss": 3.7163049643278314,
      "tokens_seen": 896204800
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007357873620862588,
      "loss": 1.4348,
      "theoretical_loss": 3.7162787788176295,
      "tokens_seen": 896270336
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007357673019057171,
      "loss": 1.5886,
      "theoretical_loss": 3.716252595758136,
      "tokens_seen": 896335872
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007357472417251755,
      "loss": 1.5403,
      "theoretical_loss": 3.7162264151489417,
      "tokens_seen": 896401408
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000735727181544634,
      "loss": 1.4991,
      "theoretical_loss": 3.7162002369896387,
      "tokens_seen": 896466944
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007357071213640923,
      "loss": 1.5659,
      "theoretical_loss": 3.716174061279819,
      "tokens_seen": 896532480
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007356870611835507,
      "loss": 1.4478,
      "theoretical_loss": 3.7161478880190737,
      "tokens_seen": 896598016
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000735667001003009,
      "loss": 1.5761,
      "theoretical_loss": 3.716121717206995,
      "tokens_seen": 896663552
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007356469408224674,
      "loss": 1.4361,
      "theoretical_loss": 3.7160955488431746,
      "tokens_seen": 896729088
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007356268806419259,
      "loss": 1.5569,
      "theoretical_loss": 3.7160693829272047,
      "tokens_seen": 896794624
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007356068204613841,
      "loss": 1.5367,
      "theoretical_loss": 3.716043219458678,
      "tokens_seen": 896860160
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007355867602808426,
      "loss": 1.526,
      "theoretical_loss": 3.7160170584371857,
      "tokens_seen": 896925696
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007355667001003009,
      "loss": 1.5209,
      "theoretical_loss": 3.7159908998623212,
      "tokens_seen": 896991232
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007355466399197593,
      "loss": 1.4502,
      "theoretical_loss": 3.715964743733676,
      "tokens_seen": 897056768
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007355265797392177,
      "loss": 1.5695,
      "theoretical_loss": 3.7159385900508433,
      "tokens_seen": 897122304
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007355065195586761,
      "loss": 1.6992,
      "theoretical_loss": 3.715912438813416,
      "tokens_seen": 897187840
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007354864593781344,
      "loss": 1.5037,
      "theoretical_loss": 3.715886290020986,
      "tokens_seen": 897253376
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007354663991975929,
      "loss": 1.5559,
      "theoretical_loss": 3.7158601436731464,
      "tokens_seen": 897318912
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007354463390170511,
      "loss": 1.6083,
      "theoretical_loss": 3.715833999769491,
      "tokens_seen": 897384448
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007354262788365096,
      "loss": 1.4732,
      "theoretical_loss": 3.7158078583096117,
      "tokens_seen": 897449984
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007354062186559679,
      "loss": 1.4644,
      "theoretical_loss": 3.715781719293102,
      "tokens_seen": 897515520
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007353861584754263,
      "loss": 1.529,
      "theoretical_loss": 3.715755582719556,
      "tokens_seen": 897581056
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007353660982948847,
      "loss": 1.3948,
      "theoretical_loss": 3.715729448588566,
      "tokens_seen": 897646592
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007353460381143431,
      "loss": 1.4205,
      "theoretical_loss": 3.7157033168997256,
      "tokens_seen": 897712128
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007353259779338014,
      "loss": 1.46,
      "theoretical_loss": 3.715677187652628,
      "tokens_seen": 897777664
    },
    {
      "epoch": 0.27,
      "objective/train/advantage_avg": 0.49338528513908386,
      "objective/train/docs_used": 331577,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4540321826934814,
      "objective/train/original_loss": 1.4540319442749023,
      "objective/train/theoretical_loss": 3.7156510608468674,
      "objective/train/tokens_used": 918303200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24429115653038025,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050580382347107,
      "objective/train/weighted_lm_loss": 1.5274916887283325,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.7156510608468674,
      "tokens_seen": 897843200
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007353059177532598,
      "loss": 1.5747,
      "theoretical_loss": 3.7156510608468674,
      "tokens_seen": 897843200
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007352858575727182,
      "loss": 1.5405,
      "theoretical_loss": 3.7156249364820377,
      "tokens_seen": 897908736
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007352657973921766,
      "loss": 1.6722,
      "theoretical_loss": 3.7155988145577323,
      "tokens_seen": 897974272
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007352457372116349,
      "loss": 1.4796,
      "theoretical_loss": 3.7155726950735453,
      "tokens_seen": 898039808
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007352256770310933,
      "loss": 1.5649,
      "theoretical_loss": 3.7155465780290706,
      "tokens_seen": 898105344
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007352056168505516,
      "loss": 1.432,
      "theoretical_loss": 3.7155204634239025,
      "tokens_seen": 898170880
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007351855566700101,
      "loss": 1.4952,
      "theoretical_loss": 3.715494351257635,
      "tokens_seen": 898236416
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007351654964894684,
      "loss": 1.6546,
      "theoretical_loss": 3.7154682415298628,
      "tokens_seen": 898301952
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007351454363089268,
      "loss": 1.5539,
      "theoretical_loss": 3.7154421342401793,
      "tokens_seen": 898367488
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007351253761283852,
      "loss": 1.56,
      "theoretical_loss": 3.7154160293881806,
      "tokens_seen": 898433024
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007351053159478436,
      "loss": 1.5154,
      "theoretical_loss": 3.71538992697346,
      "tokens_seen": 898498560
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007350852557673019,
      "loss": 1.5403,
      "theoretical_loss": 3.715363826995612,
      "tokens_seen": 898564096
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007350651955867604,
      "loss": 1.546,
      "theoretical_loss": 3.7153377294542325,
      "tokens_seen": 898629632
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007350451354062186,
      "loss": 1.4992,
      "theoretical_loss": 3.7153116343489154,
      "tokens_seen": 898695168
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007350250752256771,
      "loss": 1.4948,
      "theoretical_loss": 3.7152855416792567,
      "tokens_seen": 898760704
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007350050150451354,
      "loss": 1.5775,
      "theoretical_loss": 3.715259451444851,
      "tokens_seen": 898826240
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007349849548645938,
      "loss": 1.5772,
      "theoretical_loss": 3.715233363645293,
      "tokens_seen": 898891776
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007349648946840522,
      "loss": 1.5237,
      "theoretical_loss": 3.7152072782801784,
      "tokens_seen": 898957312
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007349448345035105,
      "loss": 1.5208,
      "theoretical_loss": 3.7151811953491025,
      "tokens_seen": 899022848
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007349247743229689,
      "loss": 1.6329,
      "theoretical_loss": 3.715155114851661,
      "tokens_seen": 899088384
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007349047141424274,
      "loss": 1.4821,
      "theoretical_loss": 3.7151290367874497,
      "tokens_seen": 899153920
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007348846539618856,
      "loss": 1.5552,
      "theoretical_loss": 3.715102961156063,
      "tokens_seen": 899219456
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007348645937813441,
      "loss": 1.4916,
      "theoretical_loss": 3.715076887957098,
      "tokens_seen": 899284992
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007348445336008023,
      "loss": 1.508,
      "theoretical_loss": 3.71505081719015,
      "tokens_seen": 899350528
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007348244734202608,
      "loss": 1.51,
      "theoretical_loss": 3.715024748854815,
      "tokens_seen": 899416064
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007348044132397192,
      "loss": 1.4391,
      "theoretical_loss": 3.7149986829506894,
      "tokens_seen": 899481600
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007347843530591775,
      "loss": 1.6433,
      "theoretical_loss": 3.7149726194773685,
      "tokens_seen": 899547136
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007347642928786359,
      "loss": 1.5585,
      "theoretical_loss": 3.714946558434449,
      "tokens_seen": 899612672
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007347442326980944,
      "loss": 1.4747,
      "theoretical_loss": 3.714920499821528,
      "tokens_seen": 899678208
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007347241725175526,
      "loss": 1.5016,
      "theoretical_loss": 3.7148944436382005,
      "tokens_seen": 899743744
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007347041123370111,
      "loss": 1.6175,
      "theoretical_loss": 3.714868389884064,
      "tokens_seen": 899809280
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007346840521564693,
      "loss": 1.4893,
      "theoretical_loss": 3.7148423385587153,
      "tokens_seen": 899874816
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007346639919759278,
      "loss": 1.4388,
      "theoretical_loss": 3.7148162896617505,
      "tokens_seen": 899940352
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007346439317953862,
      "loss": 1.5738,
      "theoretical_loss": 3.714790243192766,
      "tokens_seen": 900005888
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007346238716148445,
      "loss": 1.5373,
      "theoretical_loss": 3.71476419915136,
      "tokens_seen": 900071424
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007346038114343029,
      "loss": 1.4877,
      "theoretical_loss": 3.7147381575371288,
      "tokens_seen": 900136960
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007345837512537613,
      "loss": 1.4306,
      "theoretical_loss": 3.714712118349669,
      "tokens_seen": 900202496
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007345636910732196,
      "loss": 1.6046,
      "theoretical_loss": 3.7146860815885794,
      "tokens_seen": 900268032
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007345436308926781,
      "loss": 1.6605,
      "theoretical_loss": 3.7146600472534557,
      "tokens_seen": 900333568
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007345235707121364,
      "loss": 1.5562,
      "theoretical_loss": 3.714634015343896,
      "tokens_seen": 900399104
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007345035105315948,
      "loss": 1.5952,
      "theoretical_loss": 3.7146079858594976,
      "tokens_seen": 900464640
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007344834503510531,
      "loss": 1.5715,
      "theoretical_loss": 3.714581958799858,
      "tokens_seen": 900530176
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007344633901705115,
      "loss": 1.4754,
      "theoretical_loss": 3.7145559341645753,
      "tokens_seen": 900595712
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007344433299899699,
      "loss": 1.5178,
      "theoretical_loss": 3.714529911953247,
      "tokens_seen": 900661248
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007344232698094283,
      "loss": 1.5697,
      "theoretical_loss": 3.714503892165471,
      "tokens_seen": 900726784
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007344032096288866,
      "loss": 1.488,
      "theoretical_loss": 3.7144778748008456,
      "tokens_seen": 900792320
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007343831494483451,
      "loss": 1.5028,
      "theoretical_loss": 3.714451859858968,
      "tokens_seen": 900857856
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007343630892678034,
      "loss": 1.5317,
      "theoretical_loss": 3.714425847339437,
      "tokens_seen": 900923392
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007343430290872618,
      "loss": 1.6374,
      "theoretical_loss": 3.714399837241851,
      "tokens_seen": 900988928
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007343229689067201,
      "loss": 1.5067,
      "theoretical_loss": 3.714373829565808,
      "tokens_seen": 901054464
    },
    {
      "epoch": 0.27,
      "objective/train/advantage_avg": 0.4941953122615814,
      "objective/train/docs_used": 332819,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5781757831573486,
      "objective/train/original_loss": 1.5781757831573486,
      "objective/train/theoretical_loss": 3.714347824310907,
      "objective/train/tokens_used": 921580000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24453237652778625,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050662636756897,
      "objective/train/weighted_lm_loss": 1.6580100059509277,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.714347824310907,
      "tokens_seen": 901120000
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007343029087261786,
      "loss": 1.6008,
      "theoretical_loss": 3.714347824310907,
      "tokens_seen": 901120000
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007342828485456369,
      "loss": 1.459,
      "theoretical_loss": 3.7143218214767457,
      "tokens_seen": 901185536
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007342627883650953,
      "loss": 1.5398,
      "theoretical_loss": 3.7142958210629233,
      "tokens_seen": 901251072
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007342427281845536,
      "loss": 1.6024,
      "theoretical_loss": 3.7142698230690385,
      "tokens_seen": 901316608
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000734222668004012,
      "loss": 1.5728,
      "theoretical_loss": 3.7142438274946903,
      "tokens_seen": 901382144
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007342026078234705,
      "loss": 1.5302,
      "theoretical_loss": 3.714217834339477,
      "tokens_seen": 901447680
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007341825476429288,
      "loss": 1.5524,
      "theoretical_loss": 3.714191843602998,
      "tokens_seen": 901513216
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007341624874623872,
      "loss": 1.6042,
      "theoretical_loss": 3.7141658552848527,
      "tokens_seen": 901578752
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007341424272818456,
      "loss": 1.5385,
      "theoretical_loss": 3.71413986938464,
      "tokens_seen": 901644288
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000734122367101304,
      "loss": 1.5364,
      "theoretical_loss": 3.7141138859019596,
      "tokens_seen": 901709824
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007341023069207623,
      "loss": 1.5846,
      "theoretical_loss": 3.7140879048364104,
      "tokens_seen": 901775360
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007340822467402208,
      "loss": 1.4888,
      "theoretical_loss": 3.7140619261875916,
      "tokens_seen": 901840896
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000734062186559679,
      "loss": 1.4744,
      "theoretical_loss": 3.714035949955104,
      "tokens_seen": 901906432
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007340421263791375,
      "loss": 1.6392,
      "theoretical_loss": 3.714009976138546,
      "tokens_seen": 901971968
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007340220661985958,
      "loss": 1.5587,
      "theoretical_loss": 3.7139840047375183,
      "tokens_seen": 902037504
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007340020060180542,
      "loss": 1.6001,
      "theoretical_loss": 3.7139580357516206,
      "tokens_seen": 902103040
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007339819458375126,
      "loss": 1.6082,
      "theoretical_loss": 3.7139320691804523,
      "tokens_seen": 902168576
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007339618856569709,
      "loss": 1.5149,
      "theoretical_loss": 3.7139061050236144,
      "tokens_seen": 902234112
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007339418254764293,
      "loss": 1.44,
      "theoretical_loss": 3.713880143280707,
      "tokens_seen": 902299648
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007339217652958878,
      "loss": 1.538,
      "theoretical_loss": 3.7138541839513293,
      "tokens_seen": 902365184
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000733901705115346,
      "loss": 1.498,
      "theoretical_loss": 3.713828227035082,
      "tokens_seen": 902430720
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007338816449348045,
      "loss": 1.4815,
      "theoretical_loss": 3.7138022725315665,
      "tokens_seen": 902496256
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007338615847542627,
      "loss": 1.6513,
      "theoretical_loss": 3.713776320440383,
      "tokens_seen": 902561792
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007338415245737212,
      "loss": 1.5355,
      "theoretical_loss": 3.7137503707611312,
      "tokens_seen": 902627328
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007338214643931796,
      "loss": 1.5123,
      "theoretical_loss": 3.713724423493413,
      "tokens_seen": 902692864
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007338014042126379,
      "loss": 1.5089,
      "theoretical_loss": 3.713698478636829,
      "tokens_seen": 902758400
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007337813440320963,
      "loss": 1.6067,
      "theoretical_loss": 3.7136725361909795,
      "tokens_seen": 902823936
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007337612838515548,
      "loss": 1.5354,
      "theoretical_loss": 3.713646596155466,
      "tokens_seen": 902889472
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000733741223671013,
      "loss": 1.6636,
      "theoretical_loss": 3.71362065852989,
      "tokens_seen": 902955008
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007337211634904715,
      "loss": 1.5023,
      "theoretical_loss": 3.7135947233138515,
      "tokens_seen": 903020544
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007337011033099297,
      "loss": 1.4835,
      "theoretical_loss": 3.713568790506953,
      "tokens_seen": 903086080
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007336810431293882,
      "loss": 1.6501,
      "theoretical_loss": 3.713542860108796,
      "tokens_seen": 903151616
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007336609829488466,
      "loss": 1.3778,
      "theoretical_loss": 3.713516932118981,
      "tokens_seen": 903217152
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007336409227683049,
      "loss": 1.5625,
      "theoretical_loss": 3.7134910065371103,
      "tokens_seen": 903282688
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007336208625877633,
      "loss": 1.4652,
      "theoretical_loss": 3.7134650833627854,
      "tokens_seen": 903348224
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007336008024072217,
      "loss": 1.5245,
      "theoretical_loss": 3.713439162595608,
      "tokens_seen": 903413760
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00073358074222668,
      "loss": 1.4703,
      "theoretical_loss": 3.71341324423518,
      "tokens_seen": 903479296
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007335606820461385,
      "loss": 1.4181,
      "theoretical_loss": 3.7133873282811036,
      "tokens_seen": 903544832
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007335406218655968,
      "loss": 1.5251,
      "theoretical_loss": 3.7133614147329808,
      "tokens_seen": 903610368
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007335205616850552,
      "loss": 1.6838,
      "theoretical_loss": 3.7133355035904136,
      "tokens_seen": 903675904
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007335005015045135,
      "loss": 1.6379,
      "theoretical_loss": 3.713309594853005,
      "tokens_seen": 903741440
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000733480441323972,
      "loss": 1.4379,
      "theoretical_loss": 3.713283688520356,
      "tokens_seen": 903806976
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007334603811434303,
      "loss": 1.5696,
      "theoretical_loss": 3.7132577845920696,
      "tokens_seen": 903872512
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007334403209628887,
      "loss": 1.5813,
      "theoretical_loss": 3.7132318830677487,
      "tokens_seen": 903938048
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000733420260782347,
      "loss": 1.6165,
      "theoretical_loss": 3.713205983946996,
      "tokens_seen": 904003584
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007334002006018055,
      "loss": 1.5224,
      "theoretical_loss": 3.713180087229414,
      "tokens_seen": 904069120
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007333801404212638,
      "loss": 1.4772,
      "theoretical_loss": 3.7131541929146055,
      "tokens_seen": 904134656
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007333600802407222,
      "loss": 1.4436,
      "theoretical_loss": 3.713128301002173,
      "tokens_seen": 904200192
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007333400200601805,
      "loss": 1.4262,
      "theoretical_loss": 3.713102411491721,
      "tokens_seen": 904265728
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000733319959879639,
      "loss": 1.5758,
      "theoretical_loss": 3.71307652438285,
      "tokens_seen": 904331264
    },
    {
      "epoch": 0.27,
      "objective/train/advantage_avg": 0.490201860666275,
      "objective/train/docs_used": 333943,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.402259349822998,
      "objective/train/original_loss": 1.402259349822998,
      "objective/train/theoretical_loss": 3.713050639675166,
      "objective/train/tokens_used": 924856800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24115392565727234,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050246000289917,
      "objective/train/weighted_lm_loss": 1.4726414680480957,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.713050639675166,
      "tokens_seen": 904396800
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007332998996990973,
      "loss": 1.4792,
      "theoretical_loss": 3.713050639675166,
      "tokens_seen": 904396800
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007332798395185557,
      "loss": 1.5239,
      "theoretical_loss": 3.713024757368271,
      "tokens_seen": 904462336
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000733259779338014,
      "loss": 1.4786,
      "theoretical_loss": 3.712998877461768,
      "tokens_seen": 904527872
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007332397191574724,
      "loss": 1.4277,
      "theoretical_loss": 3.7129729999552614,
      "tokens_seen": 904593408
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007332196589769308,
      "loss": 1.513,
      "theoretical_loss": 3.712947124848354,
      "tokens_seen": 904658944
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007331995987963892,
      "loss": 1.5765,
      "theoretical_loss": 3.7129212521406503,
      "tokens_seen": 904724480
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007331795386158475,
      "loss": 1.4948,
      "theoretical_loss": 3.7128953818317534,
      "tokens_seen": 904790016
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000733159478435306,
      "loss": 1.6894,
      "theoretical_loss": 3.712869513921267,
      "tokens_seen": 904855552
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007331394182547643,
      "loss": 1.527,
      "theoretical_loss": 3.7128436484087954,
      "tokens_seen": 904921088
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007331193580742227,
      "loss": 1.5399,
      "theoretical_loss": 3.712817785293943,
      "tokens_seen": 904986624
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000733099297893681,
      "loss": 1.5486,
      "theoretical_loss": 3.712791924576314,
      "tokens_seen": 905052160
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007330792377131394,
      "loss": 1.5266,
      "theoretical_loss": 3.712766066255512,
      "tokens_seen": 905117696
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007330591775325978,
      "loss": 1.4733,
      "theoretical_loss": 3.712740210331141,
      "tokens_seen": 905183232
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007330391173520562,
      "loss": 1.5574,
      "theoretical_loss": 3.712714356802807,
      "tokens_seen": 905248768
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007330190571715145,
      "loss": 1.5035,
      "theoretical_loss": 3.712688505670113,
      "tokens_seen": 905314304
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000732998996990973,
      "loss": 1.5057,
      "theoretical_loss": 3.7126626569326637,
      "tokens_seen": 905379840
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007329789368104312,
      "loss": 1.6768,
      "theoretical_loss": 3.712636810590065,
      "tokens_seen": 905445376
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007329588766298897,
      "loss": 1.5002,
      "theoretical_loss": 3.712610966641921,
      "tokens_seen": 905510912
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000732938816449348,
      "loss": 1.4537,
      "theoretical_loss": 3.712585125087837,
      "tokens_seen": 905576448
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007329187562688064,
      "loss": 1.4468,
      "theoretical_loss": 3.7125592859274166,
      "tokens_seen": 905641984
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007328986960882648,
      "loss": 1.5671,
      "theoretical_loss": 3.7125334491602664,
      "tokens_seen": 905707520
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007328786359077231,
      "loss": 1.4226,
      "theoretical_loss": 3.7125076147859906,
      "tokens_seen": 905773056
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007328585757271815,
      "loss": 1.4046,
      "theoretical_loss": 3.7124817828041956,
      "tokens_seen": 905838592
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00073283851554664,
      "loss": 1.5966,
      "theoretical_loss": 3.712455953214486,
      "tokens_seen": 905904128
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007328184553660982,
      "loss": 1.6447,
      "theoretical_loss": 3.712430126016467,
      "tokens_seen": 905969664
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007327983951855567,
      "loss": 1.5007,
      "theoretical_loss": 3.7124043012097445,
      "tokens_seen": 906035200
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007327783350050151,
      "loss": 1.5595,
      "theoretical_loss": 3.7123784787939247,
      "tokens_seen": 906100736
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007327582748244734,
      "loss": 1.4266,
      "theoretical_loss": 3.712352658768612,
      "tokens_seen": 906166272
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007327382146439318,
      "loss": 1.6237,
      "theoretical_loss": 3.7123268411334136,
      "tokens_seen": 906231808
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007327181544633901,
      "loss": 1.5407,
      "theoretical_loss": 3.712301025887935,
      "tokens_seen": 906297344
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007326980942828486,
      "loss": 1.4856,
      "theoretical_loss": 3.7122752130317815,
      "tokens_seen": 906362880
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.000732678034102307,
      "loss": 1.5986,
      "theoretical_loss": 3.7122494025645607,
      "tokens_seen": 906428416
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007326579739217653,
      "loss": 1.3604,
      "theoretical_loss": 3.7122235944858772,
      "tokens_seen": 906493952
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007326379137412237,
      "loss": 1.4404,
      "theoretical_loss": 3.712197788795338,
      "tokens_seen": 906559488
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007326178535606821,
      "loss": 1.4967,
      "theoretical_loss": 3.7121719854925503,
      "tokens_seen": 906625024
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007325977933801404,
      "loss": 1.5209,
      "theoretical_loss": 3.712146184577119,
      "tokens_seen": 906690560
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007325777331995989,
      "loss": 1.5063,
      "theoretical_loss": 3.712120386048652,
      "tokens_seen": 906756096
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007325576730190572,
      "loss": 1.6111,
      "theoretical_loss": 3.712094589906755,
      "tokens_seen": 906821632
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007325376128385156,
      "loss": 1.4578,
      "theoretical_loss": 3.7120687961510357,
      "tokens_seen": 906887168
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007325175526579739,
      "loss": 1.4891,
      "theoretical_loss": 3.7120430047811004,
      "tokens_seen": 906952704
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007324974924774323,
      "loss": 1.6032,
      "theoretical_loss": 3.712017215796556,
      "tokens_seen": 907018240
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007324774322968907,
      "loss": 1.4976,
      "theoretical_loss": 3.7119914291970098,
      "tokens_seen": 907083776
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007324573721163491,
      "loss": 1.4149,
      "theoretical_loss": 3.711965644982069,
      "tokens_seen": 907149312
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007324373119358074,
      "loss": 1.5783,
      "theoretical_loss": 3.7119398631513403,
      "tokens_seen": 907214848
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007324172517552659,
      "loss": 1.5459,
      "theoretical_loss": 3.7119140837044315,
      "tokens_seen": 907280384
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007323971915747242,
      "loss": 1.4594,
      "theoretical_loss": 3.71188830664095,
      "tokens_seen": 907345920
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007323771313941826,
      "loss": 1.4985,
      "theoretical_loss": 3.711862531960503,
      "tokens_seen": 907411456
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0007323570712136409,
      "loss": 1.5004,
      "theoretical_loss": 3.711836759662699,
      "tokens_seen": 907476992
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007323370110330994,
      "loss": 1.5199,
      "theoretical_loss": 3.7118109897471445,
      "tokens_seen": 907542528
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007323169508525577,
      "loss": 1.4622,
      "theoretical_loss": 3.7117852222134484,
      "tokens_seen": 907608064
    },
    {
      "epoch": 0.28,
      "objective/train/advantage_avg": 0.49309253692626953,
      "objective/train/docs_used": 335091,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.365290880203247,
      "objective/train/original_loss": 1.3652909994125366,
      "objective/train/theoretical_loss": 3.7117594570612176,
      "objective/train/tokens_used": 928133600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2434723675251007,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505470037460327,
      "objective/train/weighted_lm_loss": 1.4342434406280518,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7117594570612176,
      "tokens_seen": 907673600
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007322968906720161,
      "loss": 1.4598,
      "theoretical_loss": 3.7117594570612176,
      "tokens_seen": 907673600
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007322768304914744,
      "loss": 1.5251,
      "theoretical_loss": 3.7117336942900607,
      "tokens_seen": 907739136
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007322567703109328,
      "loss": 1.5069,
      "theoretical_loss": 3.7117079338995858,
      "tokens_seen": 907804672
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007322367101303912,
      "loss": 1.5835,
      "theoretical_loss": 3.7116821758894005,
      "tokens_seen": 907870208
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007322166499498496,
      "loss": 1.4626,
      "theoretical_loss": 3.7116564202591134,
      "tokens_seen": 907935744
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007321965897693079,
      "loss": 1.6062,
      "theoretical_loss": 3.711630667008333,
      "tokens_seen": 908001280
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007321765295887664,
      "loss": 1.6331,
      "theoretical_loss": 3.7116049161366673,
      "tokens_seen": 908066816
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007321564694082247,
      "loss": 1.4469,
      "theoretical_loss": 3.7115791676437255,
      "tokens_seen": 908132352
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007321364092276831,
      "loss": 1.4688,
      "theoretical_loss": 3.711553421529116,
      "tokens_seen": 908197888
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007321163490471414,
      "loss": 1.4304,
      "theoretical_loss": 3.7115276777924473,
      "tokens_seen": 908263424
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007320962888665998,
      "loss": 1.5412,
      "theoretical_loss": 3.7115019364333275,
      "tokens_seen": 908328960
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007320762286860582,
      "loss": 1.6077,
      "theoretical_loss": 3.7114761974513675,
      "tokens_seen": 908394496
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007320561685055166,
      "loss": 1.5779,
      "theoretical_loss": 3.7114504608461747,
      "tokens_seen": 908460032
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007320361083249749,
      "loss": 1.4848,
      "theoretical_loss": 3.7114247266173583,
      "tokens_seen": 908525568
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007320160481444334,
      "loss": 1.5258,
      "theoretical_loss": 3.7113989947645276,
      "tokens_seen": 908591104
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007319959879638916,
      "loss": 1.4771,
      "theoretical_loss": 3.7113732652872926,
      "tokens_seen": 908656640
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007319759277833501,
      "loss": 1.546,
      "theoretical_loss": 3.711347538185262,
      "tokens_seen": 908722176
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007319558676028085,
      "loss": 1.5161,
      "theoretical_loss": 3.7113218134580452,
      "tokens_seen": 908787712
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007319358074222668,
      "loss": 1.5833,
      "theoretical_loss": 3.711296091105252,
      "tokens_seen": 908853248
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007319157472417252,
      "loss": 1.52,
      "theoretical_loss": 3.711270371126492,
      "tokens_seen": 908918784
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007318956870611835,
      "loss": 1.4664,
      "theoretical_loss": 3.7112446535213754,
      "tokens_seen": 908984320
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007318756268806419,
      "loss": 1.5129,
      "theoretical_loss": 3.7112189382895107,
      "tokens_seen": 909049856
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007318555667001004,
      "loss": 1.3925,
      "theoretical_loss": 3.7111932254305096,
      "tokens_seen": 909115392
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007318355065195586,
      "loss": 1.4639,
      "theoretical_loss": 3.7111675149439804,
      "tokens_seen": 909180928
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007318154463390171,
      "loss": 1.578,
      "theoretical_loss": 3.7111418068295343,
      "tokens_seen": 909246464
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007317953861584755,
      "loss": 1.563,
      "theoretical_loss": 3.7111161010867812,
      "tokens_seen": 909312000
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007317753259779338,
      "loss": 1.4527,
      "theoretical_loss": 3.7110903977153313,
      "tokens_seen": 909377536
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007317552657973922,
      "loss": 1.5968,
      "theoretical_loss": 3.7110646967147956,
      "tokens_seen": 909443072
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007317352056168505,
      "loss": 1.4044,
      "theoretical_loss": 3.711038998084783,
      "tokens_seen": 909508608
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007317151454363089,
      "loss": 1.6923,
      "theoretical_loss": 3.711013301824906,
      "tokens_seen": 909574144
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007316950852557674,
      "loss": 1.4548,
      "theoretical_loss": 3.710987607934774,
      "tokens_seen": 909639680
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007316750250752256,
      "loss": 1.5561,
      "theoretical_loss": 3.710961916413998,
      "tokens_seen": 909705216
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007316549648946841,
      "loss": 1.5293,
      "theoretical_loss": 3.7109362272621893,
      "tokens_seen": 909770752
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007316349047141424,
      "loss": 1.5652,
      "theoretical_loss": 3.7109105404789586,
      "tokens_seen": 909836288
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007316148445336008,
      "loss": 1.5264,
      "theoretical_loss": 3.7108848560639167,
      "tokens_seen": 909901824
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007315947843530592,
      "loss": 1.472,
      "theoretical_loss": 3.7108591740166745,
      "tokens_seen": 909967360
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007315747241725176,
      "loss": 1.55,
      "theoretical_loss": 3.7108334943368435,
      "tokens_seen": 910032896
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007315546639919759,
      "loss": 1.5383,
      "theoretical_loss": 3.710807817024036,
      "tokens_seen": 910098432
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007315346038114343,
      "loss": 1.5295,
      "theoretical_loss": 3.7107821420778615,
      "tokens_seen": 910163968
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007315145436308926,
      "loss": 1.5832,
      "theoretical_loss": 3.710756469497933,
      "tokens_seen": 910229504
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007314944834503511,
      "loss": 1.4573,
      "theoretical_loss": 3.710730799283861,
      "tokens_seen": 910295040
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007314744232698094,
      "loss": 1.4642,
      "theoretical_loss": 3.710705131435258,
      "tokens_seen": 910360576
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007314543630892678,
      "loss": 1.5471,
      "theoretical_loss": 3.7106794659517357,
      "tokens_seen": 910426112
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007314343029087262,
      "loss": 1.438,
      "theoretical_loss": 3.7106538028329057,
      "tokens_seen": 910491648
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007314142427281846,
      "loss": 1.5215,
      "theoretical_loss": 3.710628142078379,
      "tokens_seen": 910557184
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007313941825476429,
      "loss": 1.452,
      "theoretical_loss": 3.7106024836877696,
      "tokens_seen": 910622720
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007313741223671013,
      "loss": 1.5105,
      "theoretical_loss": 3.7105768276606885,
      "tokens_seen": 910688256
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007313540621865596,
      "loss": 1.5075,
      "theoretical_loss": 3.710551173996748,
      "tokens_seen": 910753792
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007313340020060181,
      "loss": 1.4472,
      "theoretical_loss": 3.7105255226955602,
      "tokens_seen": 910819328
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007313139418254764,
      "loss": 1.4744,
      "theoretical_loss": 3.7104998737567376,
      "tokens_seen": 910884864
    },
    {
      "epoch": 0.28,
      "objective/train/advantage_avg": 0.49511486291885376,
      "objective/train/docs_used": 336141,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5025779008865356,
      "objective/train/original_loss": 1.5025780200958252,
      "objective/train/theoretical_loss": 3.710474227179893,
      "objective/train/tokens_used": 931410400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2453271597623825,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507588386535645,
      "objective/train/weighted_lm_loss": 1.5787557363510132,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.710474227179893,
      "tokens_seen": 910950400
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007312938816449348,
      "loss": 1.5946,
      "theoretical_loss": 3.710474227179893,
      "tokens_seen": 910950400
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007312738214643931,
      "loss": 1.4776,
      "theoretical_loss": 3.710448582964639,
      "tokens_seen": 911015936
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007312537612838516,
      "loss": 1.4899,
      "theoretical_loss": 3.710422941110588,
      "tokens_seen": 911081472
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007312337011033099,
      "loss": 1.5028,
      "theoretical_loss": 3.710397301617353,
      "tokens_seen": 911147008
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007312136409227683,
      "loss": 1.5371,
      "theoretical_loss": 3.710371664484547,
      "tokens_seen": 911212544
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007311935807422268,
      "loss": 1.4034,
      "theoretical_loss": 3.710346029711782,
      "tokens_seen": 911278080
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007311735205616851,
      "loss": 1.4266,
      "theoretical_loss": 3.710320397298672,
      "tokens_seen": 911343616
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007311534603811435,
      "loss": 1.4864,
      "theoretical_loss": 3.71029476724483,
      "tokens_seen": 911409152
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007311334002006018,
      "loss": 1.469,
      "theoretical_loss": 3.710269139549869,
      "tokens_seen": 911474688
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007311133400200602,
      "loss": 1.4579,
      "theoretical_loss": 3.7102435142134027,
      "tokens_seen": 911540224
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007310932798395186,
      "loss": 1.373,
      "theoretical_loss": 3.710217891235043,
      "tokens_seen": 911605760
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000731073219658977,
      "loss": 1.497,
      "theoretical_loss": 3.710192270614406,
      "tokens_seen": 911671296
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007310531594784353,
      "loss": 1.5498,
      "theoretical_loss": 3.7101666523511034,
      "tokens_seen": 911736832
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007310330992978938,
      "loss": 1.4235,
      "theoretical_loss": 3.7101410364447496,
      "tokens_seen": 911802368
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000731013039117352,
      "loss": 1.6592,
      "theoretical_loss": 3.710115422894958,
      "tokens_seen": 911867904
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007309929789368105,
      "loss": 1.56,
      "theoretical_loss": 3.7100898117013426,
      "tokens_seen": 911933440
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007309729187562689,
      "loss": 1.4365,
      "theoretical_loss": 3.710064202863517,
      "tokens_seen": 911998976
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007309528585757272,
      "loss": 1.547,
      "theoretical_loss": 3.710038596381096,
      "tokens_seen": 912064512
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007309327983951856,
      "loss": 1.6258,
      "theoretical_loss": 3.7100129922536933,
      "tokens_seen": 912130048
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007309127382146439,
      "loss": 1.5512,
      "theoretical_loss": 3.709987390480923,
      "tokens_seen": 912195584
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007308926780341023,
      "loss": 1.5336,
      "theoretical_loss": 3.7099617910623994,
      "tokens_seen": 912261120
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007308726178535608,
      "loss": 1.5221,
      "theoretical_loss": 3.7099361939977378,
      "tokens_seen": 912326656
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000730852557673019,
      "loss": 1.6308,
      "theoretical_loss": 3.7099105992865518,
      "tokens_seen": 912392192
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007308324974924775,
      "loss": 1.4497,
      "theoretical_loss": 3.709885006928456,
      "tokens_seen": 912457728
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007308124373119359,
      "loss": 1.5121,
      "theoretical_loss": 3.7098594169230648,
      "tokens_seen": 912523264
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007307923771313942,
      "loss": 1.5233,
      "theoretical_loss": 3.709833829269994,
      "tokens_seen": 912588800
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007307723169508526,
      "loss": 1.5005,
      "theoretical_loss": 3.709808243968858,
      "tokens_seen": 912654336
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007307522567703109,
      "loss": 1.3822,
      "theoretical_loss": 3.7097826610192715,
      "tokens_seen": 912719872
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007307321965897693,
      "loss": 1.5105,
      "theoretical_loss": 3.7097570804208497,
      "tokens_seen": 912785408
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007307121364092278,
      "loss": 1.5449,
      "theoretical_loss": 3.7097315021732076,
      "tokens_seen": 912850944
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000730692076228686,
      "loss": 1.4522,
      "theoretical_loss": 3.7097059262759604,
      "tokens_seen": 912916480
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007306720160481445,
      "loss": 1.4448,
      "theoretical_loss": 3.709680352728724,
      "tokens_seen": 912982016
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007306519558676028,
      "loss": 1.5511,
      "theoretical_loss": 3.709654781531113,
      "tokens_seen": 913047552
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007306318956870612,
      "loss": 1.53,
      "theoretical_loss": 3.709629212682744,
      "tokens_seen": 913113088
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007306118355065196,
      "loss": 1.5703,
      "theoretical_loss": 3.709603646183231,
      "tokens_seen": 913178624
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000730591775325978,
      "loss": 1.5115,
      "theoretical_loss": 3.7095780820321913,
      "tokens_seen": 913244160
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007305717151454363,
      "loss": 1.5071,
      "theoretical_loss": 3.709552520229239,
      "tokens_seen": 913309696
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007305516549648947,
      "loss": 1.4625,
      "theoretical_loss": 3.7095269607739914,
      "tokens_seen": 913375232
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000730531594784353,
      "loss": 1.4216,
      "theoretical_loss": 3.709501403666064,
      "tokens_seen": 913440768
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007305115346038115,
      "loss": 1.6108,
      "theoretical_loss": 3.7094758489050728,
      "tokens_seen": 913506304
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007304914744232698,
      "loss": 1.5082,
      "theoretical_loss": 3.7094502964906337,
      "tokens_seen": 913571840
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007304714142427282,
      "loss": 1.547,
      "theoretical_loss": 3.7094247464223633,
      "tokens_seen": 913637376
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007304513540621866,
      "loss": 1.5942,
      "theoretical_loss": 3.7093991986998773,
      "tokens_seen": 913702912
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000730431293881645,
      "loss": 1.4681,
      "theoretical_loss": 3.7093736533227926,
      "tokens_seen": 913768448
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007304112337011033,
      "loss": 1.3905,
      "theoretical_loss": 3.709348110290726,
      "tokens_seen": 913833984
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007303911735205617,
      "loss": 1.4601,
      "theoretical_loss": 3.7093225696032937,
      "tokens_seen": 913899520
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00073037111334002,
      "loss": 1.5562,
      "theoretical_loss": 3.7092970312601117,
      "tokens_seen": 913965056
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007303510531594785,
      "loss": 1.5507,
      "theoretical_loss": 3.709271495260798,
      "tokens_seen": 914030592
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007303309929789368,
      "loss": 1.4563,
      "theoretical_loss": 3.7092459616049682,
      "tokens_seen": 914096128
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007303109327983952,
      "loss": 1.4443,
      "theoretical_loss": 3.709220430292241,
      "tokens_seen": 914161664
    },
    {
      "epoch": 0.28,
      "objective/train/advantage_avg": 0.4891274571418762,
      "objective/train/docs_used": 337261,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4483685493469238,
      "objective/train/original_loss": 1.4483685493469238,
      "objective/train/theoretical_loss": 3.709194901322231,
      "objective/train/tokens_used": 934687200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2399061918258667,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501322746276855,
      "objective/train/weighted_lm_loss": 1.520915150642395,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.709194901322231,
      "tokens_seen": 914227200
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007302908726178535,
      "loss": 1.6048,
      "theoretical_loss": 3.709194901322231,
      "tokens_seen": 914227200
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000730270812437312,
      "loss": 1.4676,
      "theoretical_loss": 3.709169374694558,
      "tokens_seen": 914292736
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007302507522567703,
      "loss": 1.5268,
      "theoretical_loss": 3.709143850408837,
      "tokens_seen": 914358272
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007302306920762287,
      "loss": 1.4751,
      "theoretical_loss": 3.7091183284646867,
      "tokens_seen": 914423808
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000730210631895687,
      "loss": 1.5256,
      "theoretical_loss": 3.709092808861724,
      "tokens_seen": 914489344
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007301905717151455,
      "loss": 1.5011,
      "theoretical_loss": 3.709067291599566,
      "tokens_seen": 914554880
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007301705115346038,
      "loss": 1.5044,
      "theoretical_loss": 3.7090417766778305,
      "tokens_seen": 914620416
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007301504513540622,
      "loss": 1.5095,
      "theoretical_loss": 3.7090162640961357,
      "tokens_seen": 914685952
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007301303911735205,
      "loss": 1.481,
      "theoretical_loss": 3.7089907538540987,
      "tokens_seen": 914751488
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000730110330992979,
      "loss": 1.4966,
      "theoretical_loss": 3.7089652459513376,
      "tokens_seen": 914817024
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007300902708124373,
      "loss": 1.484,
      "theoretical_loss": 3.7089397403874704,
      "tokens_seen": 914882560
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007300702106318957,
      "loss": 1.4283,
      "theoretical_loss": 3.708914237162115,
      "tokens_seen": 914948096
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007300501504513541,
      "loss": 1.5042,
      "theoretical_loss": 3.708888736274889,
      "tokens_seen": 915013632
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007300300902708124,
      "loss": 1.5487,
      "theoretical_loss": 3.708863237725412,
      "tokens_seen": 915079168
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007300100300902708,
      "loss": 1.5378,
      "theoretical_loss": 3.708837741513301,
      "tokens_seen": 915144704
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007299899699097293,
      "loss": 1.5432,
      "theoretical_loss": 3.708812247638175,
      "tokens_seen": 915210240
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007299699097291875,
      "loss": 1.5377,
      "theoretical_loss": 3.708786756099652,
      "tokens_seen": 915275776
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000729949849548646,
      "loss": 1.5257,
      "theoretical_loss": 3.7087612668973513,
      "tokens_seen": 915341312
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007299297893681042,
      "loss": 1.494,
      "theoretical_loss": 3.7087357800308904,
      "tokens_seen": 915406848
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007299097291875627,
      "loss": 1.4374,
      "theoretical_loss": 3.7087102954998894,
      "tokens_seen": 915472384
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007298896690070211,
      "loss": 1.5563,
      "theoretical_loss": 3.708684813303966,
      "tokens_seen": 915537920
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007298696088264794,
      "loss": 1.5166,
      "theoretical_loss": 3.7086593334427396,
      "tokens_seen": 915603456
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007298495486459378,
      "loss": 1.4497,
      "theoretical_loss": 3.708633855915829,
      "tokens_seen": 915668992
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007298294884653963,
      "loss": 1.521,
      "theoretical_loss": 3.708608380722854,
      "tokens_seen": 915734528
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007298094282848545,
      "loss": 1.5189,
      "theoretical_loss": 3.708582907863433,
      "tokens_seen": 915800064
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000729789368104313,
      "loss": 1.4966,
      "theoretical_loss": 3.7085574373371855,
      "tokens_seen": 915865600
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007297693079237712,
      "loss": 1.4727,
      "theoretical_loss": 3.708531969143731,
      "tokens_seen": 915931136
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007297492477432297,
      "loss": 1.5122,
      "theoretical_loss": 3.7085065032826883,
      "tokens_seen": 915996672
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007297291875626882,
      "loss": 1.5311,
      "theoretical_loss": 3.7084810397536785,
      "tokens_seen": 916062208
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007297091273821464,
      "loss": 1.5179,
      "theoretical_loss": 3.708455578556319,
      "tokens_seen": 916127744
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007296890672016049,
      "loss": 1.6679,
      "theoretical_loss": 3.708430119690232,
      "tokens_seen": 916193280
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007296690070210632,
      "loss": 1.5441,
      "theoretical_loss": 3.708404663155035,
      "tokens_seen": 916258816
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007296489468405216,
      "loss": 1.5458,
      "theoretical_loss": 3.7083792089503493,
      "tokens_seen": 916324352
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00072962888665998,
      "loss": 1.5719,
      "theoretical_loss": 3.7083537570757947,
      "tokens_seen": 916389888
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007296088264794384,
      "loss": 1.4195,
      "theoretical_loss": 3.708328307530991,
      "tokens_seen": 916455424
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007295887662988967,
      "loss": 1.6615,
      "theoretical_loss": 3.708302860315559,
      "tokens_seen": 916520960
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007295687061183551,
      "loss": 1.3627,
      "theoretical_loss": 3.7082774154291176,
      "tokens_seen": 916586496
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007295486459378134,
      "loss": 1.4461,
      "theoretical_loss": 3.708251972871288,
      "tokens_seen": 916652032
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007295285857572719,
      "loss": 1.5121,
      "theoretical_loss": 3.7082265326416914,
      "tokens_seen": 916717568
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007295085255767302,
      "loss": 1.4843,
      "theoretical_loss": 3.7082010947399473,
      "tokens_seen": 916783104
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007294884653961886,
      "loss": 1.4509,
      "theoretical_loss": 3.708175659165676,
      "tokens_seen": 916848640
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000729468405215647,
      "loss": 1.5201,
      "theoretical_loss": 3.708150225918499,
      "tokens_seen": 916914176
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007294483450351054,
      "loss": 1.4946,
      "theoretical_loss": 3.708124794998037,
      "tokens_seen": 916979712
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007294282848545637,
      "loss": 1.3232,
      "theoretical_loss": 3.7080993664039106,
      "tokens_seen": 917045248
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007294082246740221,
      "loss": 1.4677,
      "theoretical_loss": 3.7080739401357405,
      "tokens_seen": 917110784
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007293881644934804,
      "loss": 1.4122,
      "theoretical_loss": 3.7080485161931493,
      "tokens_seen": 917176320
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007293681043129389,
      "loss": 1.622,
      "theoretical_loss": 3.708023094575756,
      "tokens_seen": 917241856
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007293480441323972,
      "loss": 1.5137,
      "theoretical_loss": 3.707997675283183,
      "tokens_seen": 917307392
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007293279839518556,
      "loss": 1.4818,
      "theoretical_loss": 3.7079722583150514,
      "tokens_seen": 917372928
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007293079237713139,
      "loss": 1.5348,
      "theoretical_loss": 3.7079468436709826,
      "tokens_seen": 917438464
    },
    {
      "epoch": 0.28,
      "objective/train/advantage_avg": 0.49285387992858887,
      "objective/train/docs_used": 338425,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5136154890060425,
      "objective/train/original_loss": 1.5136152505874634,
      "objective/train/theoretical_loss": 3.7079214313505986,
      "objective/train/tokens_used": 937964000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2434709668159485,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505231618881226,
      "objective/train/weighted_lm_loss": 1.5898890495300293,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7079214313505986,
      "tokens_seen": 917504000
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007292878635907724,
      "loss": 1.6218,
      "theoretical_loss": 3.7079214313505986,
      "tokens_seen": 917504000
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007292678034102307,
      "loss": 1.4464,
      "theoretical_loss": 3.70789602135352,
      "tokens_seen": 917569536
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007292477432296891,
      "loss": 1.5588,
      "theoretical_loss": 3.7078706136793693,
      "tokens_seen": 917635072
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007292276830491475,
      "loss": 1.4513,
      "theoretical_loss": 3.707845208327768,
      "tokens_seen": 917700608
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007292076228686059,
      "loss": 1.4758,
      "theoretical_loss": 3.707819805298338,
      "tokens_seen": 917766144
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007291875626880642,
      "loss": 1.5825,
      "theoretical_loss": 3.707794404590701,
      "tokens_seen": 917831680
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007291675025075226,
      "loss": 1.5211,
      "theoretical_loss": 3.7077690062044795,
      "tokens_seen": 917897216
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007291474423269809,
      "loss": 1.5727,
      "theoretical_loss": 3.7077436101392953,
      "tokens_seen": 917962752
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007291273821464394,
      "loss": 1.4509,
      "theoretical_loss": 3.7077182163947704,
      "tokens_seen": 918028288
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007291073219658977,
      "loss": 1.5435,
      "theoretical_loss": 3.707692824970527,
      "tokens_seen": 918093824
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007290872617853561,
      "loss": 1.5576,
      "theoretical_loss": 3.7076674358661883,
      "tokens_seen": 918159360
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007290672016048145,
      "loss": 1.4745,
      "theoretical_loss": 3.707642049081376,
      "tokens_seen": 918224896
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007290471414242728,
      "loss": 1.5114,
      "theoretical_loss": 3.7076166646157134,
      "tokens_seen": 918290432
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007290270812437312,
      "loss": 1.512,
      "theoretical_loss": 3.707591282468823,
      "tokens_seen": 918355968
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007290070210631897,
      "loss": 1.5481,
      "theoretical_loss": 3.7075659026403267,
      "tokens_seen": 918421504
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007289869608826479,
      "loss": 1.5457,
      "theoretical_loss": 3.707540525129848,
      "tokens_seen": 918487040
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007289669007021064,
      "loss": 1.4794,
      "theoretical_loss": 3.7075151499370094,
      "tokens_seen": 918552576
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007289468405215646,
      "loss": 1.6149,
      "theoretical_loss": 3.7074897770614346,
      "tokens_seen": 918618112
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007289267803410231,
      "loss": 1.5017,
      "theoretical_loss": 3.707464406502746,
      "tokens_seen": 918683648
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007289067201604815,
      "loss": 1.6513,
      "theoretical_loss": 3.707439038260567,
      "tokens_seen": 918749184
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007288866599799398,
      "loss": 1.4945,
      "theoretical_loss": 3.7074136723345212,
      "tokens_seen": 918814720
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007288665997993982,
      "loss": 1.6587,
      "theoretical_loss": 3.707388308724232,
      "tokens_seen": 918880256
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007288465396188567,
      "loss": 1.5477,
      "theoretical_loss": 3.7073629474293224,
      "tokens_seen": 918945792
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007288264794383149,
      "loss": 1.406,
      "theoretical_loss": 3.7073375884494157,
      "tokens_seen": 919011328
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007288064192577734,
      "loss": 1.6296,
      "theoretical_loss": 3.707312231784136,
      "tokens_seen": 919076864
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007287863590772316,
      "loss": 1.6131,
      "theoretical_loss": 3.707286877433107,
      "tokens_seen": 919142400
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007287662988966901,
      "loss": 1.4684,
      "theoretical_loss": 3.7072615253959524,
      "tokens_seen": 919207936
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007287462387161485,
      "loss": 1.508,
      "theoretical_loss": 3.707236175672296,
      "tokens_seen": 919273472
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007287261785356068,
      "loss": 1.4736,
      "theoretical_loss": 3.7072108282617617,
      "tokens_seen": 919339008
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007287061183550652,
      "loss": 1.486,
      "theoretical_loss": 3.7071854831639737,
      "tokens_seen": 919404544
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007286860581745236,
      "loss": 1.5752,
      "theoretical_loss": 3.707160140378557,
      "tokens_seen": 919470080
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007286659979939819,
      "loss": 1.5419,
      "theoretical_loss": 3.707134799905134,
      "tokens_seen": 919535616
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007286459378134404,
      "loss": 1.5498,
      "theoretical_loss": 3.7071094617433307,
      "tokens_seen": 919601152
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007286258776328986,
      "loss": 1.4082,
      "theoretical_loss": 3.7070841258927705,
      "tokens_seen": 919666688
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007286058174523571,
      "loss": 1.5653,
      "theoretical_loss": 3.7070587923530782,
      "tokens_seen": 919732224
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007285857572718154,
      "loss": 1.5904,
      "theoretical_loss": 3.7070334611238787,
      "tokens_seen": 919797760
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007285656970912738,
      "loss": 1.5121,
      "theoretical_loss": 3.707008132204796,
      "tokens_seen": 919863296
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007285456369107322,
      "loss": 1.6357,
      "theoretical_loss": 3.7069828055954557,
      "tokens_seen": 919928832
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007285255767301906,
      "loss": 1.4851,
      "theoretical_loss": 3.7069574812954817,
      "tokens_seen": 919994368
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007285055165496489,
      "loss": 1.4461,
      "theoretical_loss": 3.7069321593045,
      "tokens_seen": 920059904
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007284854563691074,
      "loss": 1.3679,
      "theoretical_loss": 3.7069068396221345,
      "tokens_seen": 920125440
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007284653961885657,
      "loss": 1.406,
      "theoretical_loss": 3.7068815222480116,
      "tokens_seen": 920190976
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007284453360080241,
      "loss": 1.5506,
      "theoretical_loss": 3.7068562071817555,
      "tokens_seen": 920256512
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007284252758274824,
      "loss": 1.4538,
      "theoretical_loss": 3.7068308944229917,
      "tokens_seen": 920322048
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007284052156469408,
      "loss": 1.5482,
      "theoretical_loss": 3.706805583971345,
      "tokens_seen": 920387584
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007283851554663992,
      "loss": 1.5772,
      "theoretical_loss": 3.7067802758264428,
      "tokens_seen": 920453120
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007283650952858576,
      "loss": 1.5098,
      "theoretical_loss": 3.7067549699879088,
      "tokens_seen": 920518656
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007283450351053159,
      "loss": 1.5045,
      "theoretical_loss": 3.7067296664553693,
      "tokens_seen": 920584192
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007283249749247743,
      "loss": 1.5579,
      "theoretical_loss": 3.7067043652284495,
      "tokens_seen": 920649728
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007283049147442327,
      "loss": 1.7246,
      "theoretical_loss": 3.7066790663067763,
      "tokens_seen": 920715264
    },
    {
      "epoch": 0.28,
      "objective/train/advantage_avg": 0.49049344658851624,
      "objective/train/docs_used": 339576,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.416655421257019,
      "objective/train/original_loss": 1.4166555404663086,
      "objective/train/theoretical_loss": 3.706653769689974,
      "objective/train/tokens_used": 941240800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24111294746398926,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502750873565674,
      "objective/train/weighted_lm_loss": 1.4878922700881958,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.706653769689974,
      "tokens_seen": 920780800
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007282848545636911,
      "loss": 1.6253,
      "theoretical_loss": 3.706653769689974,
      "tokens_seen": 920780800
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007282647943831494,
      "loss": 1.6711,
      "theoretical_loss": 3.706628475377671,
      "tokens_seen": 920846336
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007282447342026079,
      "loss": 1.5778,
      "theoretical_loss": 3.7066031833694906,
      "tokens_seen": 920911872
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007282246740220663,
      "loss": 1.6557,
      "theoretical_loss": 3.7065778936650604,
      "tokens_seen": 920977408
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007282046138415246,
      "loss": 1.5422,
      "theoretical_loss": 3.7065526062640073,
      "tokens_seen": 921042944
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000728184553660983,
      "loss": 1.4805,
      "theoretical_loss": 3.7065273211659564,
      "tokens_seen": 921108480
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007281644934804413,
      "loss": 1.535,
      "theoretical_loss": 3.7065020383705347,
      "tokens_seen": 921174016
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007281444332998998,
      "loss": 1.4118,
      "theoretical_loss": 3.7064767578773683,
      "tokens_seen": 921239552
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007281243731193581,
      "loss": 1.4001,
      "theoretical_loss": 3.7064514796860846,
      "tokens_seen": 921305088
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007281043129388165,
      "loss": 1.5464,
      "theoretical_loss": 3.706426203796309,
      "tokens_seen": 921370624
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007280842527582749,
      "loss": 1.5001,
      "theoretical_loss": 3.70640093020767,
      "tokens_seen": 921436160
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007280641925777332,
      "loss": 1.4606,
      "theoretical_loss": 3.706375658919793,
      "tokens_seen": 921501696
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007280441323971916,
      "loss": 1.5664,
      "theoretical_loss": 3.7063503899323056,
      "tokens_seen": 921567232
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00072802407221665,
      "loss": 1.5812,
      "theoretical_loss": 3.7063251232448344,
      "tokens_seen": 921632768
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007280040120361083,
      "loss": 1.586,
      "theoretical_loss": 3.7062998588570073,
      "tokens_seen": 921698304
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007279839518555668,
      "loss": 1.5172,
      "theoretical_loss": 3.706274596768451,
      "tokens_seen": 921763840
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000727963891675025,
      "loss": 1.381,
      "theoretical_loss": 3.706249336978792,
      "tokens_seen": 921829376
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007279438314944835,
      "loss": 1.6025,
      "theoretical_loss": 3.7062240794876598,
      "tokens_seen": 921894912
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007279237713139419,
      "loss": 1.5159,
      "theoretical_loss": 3.7061988242946793,
      "tokens_seen": 921960448
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007279037111334002,
      "loss": 1.5373,
      "theoretical_loss": 3.7061735713994803,
      "tokens_seen": 922025984
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007278836509528586,
      "loss": 1.6722,
      "theoretical_loss": 3.706148320801689,
      "tokens_seen": 922091520
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007278635907723171,
      "loss": 1.3894,
      "theoretical_loss": 3.706123072500934,
      "tokens_seen": 922157056
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007278435305917753,
      "loss": 1.4565,
      "theoretical_loss": 3.7060978264968423,
      "tokens_seen": 922222592
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007278234704112338,
      "loss": 1.6769,
      "theoretical_loss": 3.706072582789042,
      "tokens_seen": 922288128
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000727803410230692,
      "loss": 1.5058,
      "theoretical_loss": 3.7060473413771615,
      "tokens_seen": 922353664
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007277833500501505,
      "loss": 1.4791,
      "theoretical_loss": 3.706022102260829,
      "tokens_seen": 922419200
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007277632898696089,
      "loss": 1.4318,
      "theoretical_loss": 3.705996865439672,
      "tokens_seen": 922484736
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007277432296890672,
      "loss": 1.5071,
      "theoretical_loss": 3.7059716309133197,
      "tokens_seen": 922550272
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007277231695085256,
      "loss": 1.4344,
      "theoretical_loss": 3.705946398681399,
      "tokens_seen": 922615808
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000727703109327984,
      "loss": 1.4966,
      "theoretical_loss": 3.7059211687435396,
      "tokens_seen": 922681344
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007276830491474423,
      "loss": 1.4718,
      "theoretical_loss": 3.7058959410993695,
      "tokens_seen": 922746880
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007276629889669008,
      "loss": 1.4606,
      "theoretical_loss": 3.7058707157485173,
      "tokens_seen": 922812416
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000727642928786359,
      "loss": 1.4975,
      "theoretical_loss": 3.705845492690612,
      "tokens_seen": 922877952
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007276228686058175,
      "loss": 1.4944,
      "theoretical_loss": 3.7058202719252815,
      "tokens_seen": 922943488
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007276028084252758,
      "loss": 1.6168,
      "theoretical_loss": 3.7057950534521558,
      "tokens_seen": 923009024
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007275827482447342,
      "loss": 1.5249,
      "theoretical_loss": 3.7057698372708634,
      "tokens_seen": 923074560
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007275626880641926,
      "loss": 1.5771,
      "theoretical_loss": 3.705744623381033,
      "tokens_seen": 923140096
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000727542627883651,
      "loss": 1.6246,
      "theoretical_loss": 3.705719411782294,
      "tokens_seen": 923205632
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007275225677031093,
      "loss": 1.4816,
      "theoretical_loss": 3.705694202474275,
      "tokens_seen": 923271168
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007275025075225678,
      "loss": 1.5396,
      "theoretical_loss": 3.705668995456606,
      "tokens_seen": 923336704
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000727482447342026,
      "loss": 1.5217,
      "theoretical_loss": 3.7056437907289173,
      "tokens_seen": 923402240
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007274623871614845,
      "loss": 1.425,
      "theoretical_loss": 3.705618588290837,
      "tokens_seen": 923467776
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007274423269809428,
      "loss": 1.5429,
      "theoretical_loss": 3.7055933881419936,
      "tokens_seen": 923533312
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007274222668004012,
      "loss": 1.4917,
      "theoretical_loss": 3.705568190282019,
      "tokens_seen": 923598848
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007274022066198596,
      "loss": 1.3722,
      "theoretical_loss": 3.705542994710542,
      "tokens_seen": 923664384
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000727382146439318,
      "loss": 1.4368,
      "theoretical_loss": 3.705517801427192,
      "tokens_seen": 923729920
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007273620862587763,
      "loss": 1.4287,
      "theoretical_loss": 3.7054926104315995,
      "tokens_seen": 923795456
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007273420260782347,
      "loss": 1.4535,
      "theoretical_loss": 3.705467421723394,
      "tokens_seen": 923860992
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007273219658976931,
      "loss": 1.5643,
      "theoretical_loss": 3.705442235302206,
      "tokens_seen": 923926528
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007273019057171515,
      "loss": 1.4092,
      "theoretical_loss": 3.7054170511676654,
      "tokens_seen": 923992064
    },
    {
      "epoch": 0.28,
      "objective/train/advantage_avg": 0.49350494146347046,
      "objective/train/docs_used": 340768,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.575994849205017,
      "objective/train/original_loss": 1.5759947299957275,
      "objective/train/theoretical_loss": 3.705391869319403,
      "objective/train/tokens_used": 944517600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24395433068275452,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505907535552979,
      "objective/train/weighted_lm_loss": 1.6554665565490723,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.705391869319403,
      "tokens_seen": 924057600
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007272818455366098,
      "loss": 1.5538,
      "theoretical_loss": 3.705391869319403,
      "tokens_seen": 924057600
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007272617853560683,
      "loss": 1.5145,
      "theoretical_loss": 3.7053666897570476,
      "tokens_seen": 924123136
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007272417251755265,
      "loss": 1.4648,
      "theoretical_loss": 3.7053415124802314,
      "tokens_seen": 924188672
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000727221664994985,
      "loss": 1.4121,
      "theoretical_loss": 3.7053163374885836,
      "tokens_seen": 924254208
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007272016048144433,
      "loss": 1.4406,
      "theoretical_loss": 3.7052911647817357,
      "tokens_seen": 924319744
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007271815446339017,
      "loss": 1.6954,
      "theoretical_loss": 3.7052659943593174,
      "tokens_seen": 924385280
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007271614844533601,
      "loss": 1.4464,
      "theoretical_loss": 3.7052408262209604,
      "tokens_seen": 924450816
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007271414242728185,
      "loss": 1.4773,
      "theoretical_loss": 3.705215660366295,
      "tokens_seen": 924516352
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007271213640922768,
      "loss": 1.4817,
      "theoretical_loss": 3.7051904967949527,
      "tokens_seen": 924581888
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007271013039117353,
      "loss": 1.4049,
      "theoretical_loss": 3.7051653355065635,
      "tokens_seen": 924647424
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007270812437311935,
      "loss": 1.4757,
      "theoretical_loss": 3.705140176500759,
      "tokens_seen": 924712960
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000727061183550652,
      "loss": 1.4273,
      "theoretical_loss": 3.705115019777171,
      "tokens_seen": 924778496
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007270411233701103,
      "loss": 1.4866,
      "theoretical_loss": 3.7050898653354296,
      "tokens_seen": 924844032
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007270210631895687,
      "loss": 1.6472,
      "theoretical_loss": 3.705064713175167,
      "tokens_seen": 924909568
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007270010030090271,
      "loss": 1.5683,
      "theoretical_loss": 3.7050395632960145,
      "tokens_seen": 924975104
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007269809428284854,
      "loss": 1.4334,
      "theoretical_loss": 3.7050144156976037,
      "tokens_seen": 925040640
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007269608826479438,
      "loss": 1.4257,
      "theoretical_loss": 3.7049892703795653,
      "tokens_seen": 925106176
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007269408224674023,
      "loss": 1.5684,
      "theoretical_loss": 3.7049641273415324,
      "tokens_seen": 925171712
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007269207622868605,
      "loss": 1.4977,
      "theoretical_loss": 3.7049389865831355,
      "tokens_seen": 925237248
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000726900702106319,
      "loss": 1.4187,
      "theoretical_loss": 3.704913848104008,
      "tokens_seen": 925302784
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007268806419257772,
      "loss": 1.53,
      "theoretical_loss": 3.70488871190378,
      "tokens_seen": 925368320
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007268605817452357,
      "loss": 1.4792,
      "theoretical_loss": 3.704863577982085,
      "tokens_seen": 925433856
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007268405215646941,
      "loss": 1.4658,
      "theoretical_loss": 3.7048384463385537,
      "tokens_seen": 925499392
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007268204613841524,
      "loss": 1.508,
      "theoretical_loss": 3.70481331697282,
      "tokens_seen": 925564928
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007268004012036108,
      "loss": 1.5516,
      "theoretical_loss": 3.704788189884515,
      "tokens_seen": 925630464
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007267803410230693,
      "loss": 1.4535,
      "theoretical_loss": 3.7047630650732715,
      "tokens_seen": 925696000
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007267602808425275,
      "loss": 1.5428,
      "theoretical_loss": 3.7047379425387224,
      "tokens_seen": 925761536
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000726740220661986,
      "loss": 1.4619,
      "theoretical_loss": 3.704712822280499,
      "tokens_seen": 925827072
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007267201604814444,
      "loss": 1.6605,
      "theoretical_loss": 3.7046877042982347,
      "tokens_seen": 925892608
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007267001003009027,
      "loss": 1.528,
      "theoretical_loss": 3.704662588591563,
      "tokens_seen": 925958144
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007266800401203612,
      "loss": 1.488,
      "theoretical_loss": 3.7046374751601148,
      "tokens_seen": 926023680
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007266599799398194,
      "loss": 1.5321,
      "theoretical_loss": 3.704612364003525,
      "tokens_seen": 926089216
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007266399197592779,
      "loss": 1.596,
      "theoretical_loss": 3.7045872551214254,
      "tokens_seen": 926154752
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007266198595787362,
      "loss": 1.4666,
      "theoretical_loss": 3.704562148513449,
      "tokens_seen": 926220288
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007265997993981946,
      "loss": 1.4383,
      "theoretical_loss": 3.7045370441792294,
      "tokens_seen": 926285824
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000726579739217653,
      "loss": 1.4365,
      "theoretical_loss": 3.7045119421184003,
      "tokens_seen": 926351360
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007265596790371114,
      "loss": 1.4921,
      "theoretical_loss": 3.704486842330594,
      "tokens_seen": 926416896
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007265396188565697,
      "loss": 1.5425,
      "theoretical_loss": 3.704461744815444,
      "tokens_seen": 926482432
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007265195586760282,
      "loss": 1.5837,
      "theoretical_loss": 3.7044366495725845,
      "tokens_seen": 926547968
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007264994984954865,
      "loss": 1.5043,
      "theoretical_loss": 3.7044115566016487,
      "tokens_seen": 926613504
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007264794383149449,
      "loss": 1.3741,
      "theoretical_loss": 3.7043864659022696,
      "tokens_seen": 926679040
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007264593781344032,
      "loss": 1.4124,
      "theoretical_loss": 3.704361377474082,
      "tokens_seen": 926744576
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007264393179538616,
      "loss": 1.4662,
      "theoretical_loss": 3.7043362913167197,
      "tokens_seen": 926810112
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00072641925777332,
      "loss": 1.5184,
      "theoretical_loss": 3.7043112074298157,
      "tokens_seen": 926875648
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007263991975927784,
      "loss": 1.4321,
      "theoretical_loss": 3.704286125813004,
      "tokens_seen": 926941184
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007263791374122367,
      "loss": 1.5553,
      "theoretical_loss": 3.7042610464659202,
      "tokens_seen": 927006720
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007263590772316951,
      "loss": 1.4337,
      "theoretical_loss": 3.704235969388197,
      "tokens_seen": 927072256
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007263390170511535,
      "loss": 1.4561,
      "theoretical_loss": 3.704210894579469,
      "tokens_seen": 927137792
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007263189568706119,
      "loss": 1.5876,
      "theoretical_loss": 3.7041858220393706,
      "tokens_seen": 927203328
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007262988966900702,
      "loss": 1.4788,
      "theoretical_loss": 3.704160751767537,
      "tokens_seen": 927268864
    },
    {
      "epoch": 0.28,
      "objective/train/advantage_avg": 0.4912850260734558,
      "objective/train/docs_used": 341815,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.337402105331421,
      "objective/train/original_loss": 1.337402105331421,
      "objective/train/theoretical_loss": 3.7041356837636013,
      "objective/train/tokens_used": 947794400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24185810983181,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503580570220947,
      "objective/train/weighted_lm_loss": 1.4047034978866577,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7041356837636013,
      "tokens_seen": 927334400
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007262788365095287,
      "loss": 1.448,
      "theoretical_loss": 3.7041356837636013,
      "tokens_seen": 927334400
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007262587763289869,
      "loss": 1.4664,
      "theoretical_loss": 3.7041106180271983,
      "tokens_seen": 927399936
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007262387161484454,
      "loss": 1.5342,
      "theoretical_loss": 3.704085554557964,
      "tokens_seen": 927465472
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007262186559679037,
      "loss": 1.5683,
      "theoretical_loss": 3.7040604933555317,
      "tokens_seen": 927531008
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007261985957873621,
      "loss": 1.351,
      "theoretical_loss": 3.704035434419537,
      "tokens_seen": 927596544
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007261785356068205,
      "loss": 1.4746,
      "theoretical_loss": 3.704010377749614,
      "tokens_seen": 927662080
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007261584754262789,
      "loss": 1.5159,
      "theoretical_loss": 3.703985323345399,
      "tokens_seen": 927727616
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007261384152457372,
      "loss": 1.3905,
      "theoretical_loss": 3.703960271206527,
      "tokens_seen": 927793152
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007261183550651957,
      "loss": 1.5273,
      "theoretical_loss": 3.7039352213326318,
      "tokens_seen": 927858688
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007260982948846539,
      "loss": 1.4205,
      "theoretical_loss": 3.7039101737233495,
      "tokens_seen": 927924224
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007260782347041124,
      "loss": 1.3979,
      "theoretical_loss": 3.7038851283783156,
      "tokens_seen": 927989760
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007260581745235707,
      "loss": 1.5695,
      "theoretical_loss": 3.7038600852971655,
      "tokens_seen": 928055296
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007260381143430291,
      "loss": 1.4711,
      "theoretical_loss": 3.7038350444795345,
      "tokens_seen": 928120832
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007260180541624875,
      "loss": 1.5732,
      "theoretical_loss": 3.703810005925058,
      "tokens_seen": 928186368
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007259979939819458,
      "loss": 1.4677,
      "theoretical_loss": 3.7037849696333724,
      "tokens_seen": 928251904
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007259779338014042,
      "loss": 1.6445,
      "theoretical_loss": 3.703759935604113,
      "tokens_seen": 928317440
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007259578736208627,
      "loss": 1.5415,
      "theoretical_loss": 3.7037349038369154,
      "tokens_seen": 928382976
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007259378134403209,
      "loss": 1.5068,
      "theoretical_loss": 3.703709874331416,
      "tokens_seen": 928448512
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007259177532597794,
      "loss": 1.5388,
      "theoretical_loss": 3.703684847087251,
      "tokens_seen": 928514048
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007258976930792376,
      "loss": 1.4757,
      "theoretical_loss": 3.7036598221040555,
      "tokens_seen": 928579584
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007258776328986961,
      "loss": 1.4913,
      "theoretical_loss": 3.703634799381467,
      "tokens_seen": 928645120
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007258575727181545,
      "loss": 1.622,
      "theoretical_loss": 3.7036097789191205,
      "tokens_seen": 928710656
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007258375125376128,
      "loss": 1.59,
      "theoretical_loss": 3.7035847607166534,
      "tokens_seen": 928776192
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007258174523570712,
      "loss": 1.5077,
      "theoretical_loss": 3.703559744773702,
      "tokens_seen": 928841728
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007257973921765297,
      "loss": 1.5019,
      "theoretical_loss": 3.7035347310899027,
      "tokens_seen": 928907264
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007257773319959879,
      "loss": 1.5729,
      "theoretical_loss": 3.7035097196648916,
      "tokens_seen": 928972800
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007257572718154464,
      "loss": 1.3882,
      "theoretical_loss": 3.703484710498306,
      "tokens_seen": 929038336
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007257372116349046,
      "loss": 1.6731,
      "theoretical_loss": 3.7034597035897825,
      "tokens_seen": 929103872
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007257171514543631,
      "loss": 1.6606,
      "theoretical_loss": 3.7034346989389575,
      "tokens_seen": 929169408
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007256970912738215,
      "loss": 1.6001,
      "theoretical_loss": 3.703409696545469,
      "tokens_seen": 929234944
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007256770310932798,
      "loss": 1.5943,
      "theoretical_loss": 3.703384696408953,
      "tokens_seen": 929300480
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007256569709127382,
      "loss": 1.544,
      "theoretical_loss": 3.703359698529047,
      "tokens_seen": 929366016
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007256369107321966,
      "loss": 1.5046,
      "theoretical_loss": 3.7033347029053885,
      "tokens_seen": 929431552
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007256168505516549,
      "loss": 1.4519,
      "theoretical_loss": 3.7033097095376144,
      "tokens_seen": 929497088
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007255967903711134,
      "loss": 1.4793,
      "theoretical_loss": 3.7032847184253628,
      "tokens_seen": 929562624
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007255767301905717,
      "loss": 1.3853,
      "theoretical_loss": 3.70325972956827,
      "tokens_seen": 929628160
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007255566700100301,
      "loss": 1.4804,
      "theoretical_loss": 3.7032347429659738,
      "tokens_seen": 929693696
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007255366098294885,
      "loss": 1.5189,
      "theoretical_loss": 3.7032097586181125,
      "tokens_seen": 929759232
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007255165496489469,
      "loss": 1.5705,
      "theoretical_loss": 3.7031847765243233,
      "tokens_seen": 929824768
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007254964894684052,
      "loss": 1.3659,
      "theoretical_loss": 3.703159796684244,
      "tokens_seen": 929890304
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007254764292878636,
      "loss": 1.5044,
      "theoretical_loss": 3.7031348190975133,
      "tokens_seen": 929955840
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007254563691073219,
      "loss": 1.4919,
      "theoretical_loss": 3.7031098437637677,
      "tokens_seen": 930021376
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007254363089267804,
      "loss": 1.4214,
      "theoretical_loss": 3.7030848706826465,
      "tokens_seen": 930086912
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007254162487462387,
      "loss": 1.4183,
      "theoretical_loss": 3.703059899853787,
      "tokens_seen": 930152448
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007253961885656971,
      "loss": 1.4885,
      "theoretical_loss": 3.703034931276828,
      "tokens_seen": 930217984
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007253761283851554,
      "loss": 1.5276,
      "theoretical_loss": 3.703009964951407,
      "tokens_seen": 930283520
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007253560682046139,
      "loss": 1.6602,
      "theoretical_loss": 3.702985000877163,
      "tokens_seen": 930349056
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007253360080240722,
      "loss": 1.5445,
      "theoretical_loss": 3.7029600390537345,
      "tokens_seen": 930414592
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007253159478435306,
      "loss": 1.5066,
      "theoretical_loss": 3.70293507948076,
      "tokens_seen": 930480128
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007252958876629889,
      "loss": 1.3841,
      "theoretical_loss": 3.7029101221578777,
      "tokens_seen": 930545664
    },
    {
      "epoch": 0.28,
      "objective/train/advantage_avg": 0.4946530759334564,
      "objective/train/docs_used": 343003,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5357661247253418,
      "objective/train/original_loss": 1.5357661247253418,
      "objective/train/theoretical_loss": 3.7028851670847267,
      "objective/train/tokens_used": 951071200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2450975626707077,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507113933563232,
      "objective/train/weighted_lm_loss": 1.6135687828063965,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7028851670847267,
      "tokens_seen": 930611200
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007252758274824473,
      "loss": 1.6176,
      "theoretical_loss": 3.7028851670847267,
      "tokens_seen": 930611200
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007252557673019057,
      "loss": 1.4259,
      "theoretical_loss": 3.702860214260946,
      "tokens_seen": 930676736
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007252357071213641,
      "loss": 1.4812,
      "theoretical_loss": 3.702835263686174,
      "tokens_seen": 930742272
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007252156469408225,
      "loss": 1.5097,
      "theoretical_loss": 3.7028103153600496,
      "tokens_seen": 930807808
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007251955867602809,
      "loss": 1.5004,
      "theoretical_loss": 3.7027853692822124,
      "tokens_seen": 930873344
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007251755265797393,
      "loss": 1.5389,
      "theoretical_loss": 3.7027604254523014,
      "tokens_seen": 930938880
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007251554663991976,
      "loss": 1.5845,
      "theoretical_loss": 3.7027354838699553,
      "tokens_seen": 931004416
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007251354062186561,
      "loss": 1.6467,
      "theoretical_loss": 3.7027105445348134,
      "tokens_seen": 931069952
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007251153460381143,
      "loss": 1.4211,
      "theoretical_loss": 3.702685607446516,
      "tokens_seen": 931135488
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007250952858575728,
      "loss": 1.528,
      "theoretical_loss": 3.7026606726047016,
      "tokens_seen": 931201024
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007250752256770311,
      "loss": 1.5263,
      "theoretical_loss": 3.70263574000901,
      "tokens_seen": 931266560
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007250551654964895,
      "loss": 1.5076,
      "theoretical_loss": 3.702610809659081,
      "tokens_seen": 931332096
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007250351053159479,
      "loss": 1.6135,
      "theoretical_loss": 3.7025858815545543,
      "tokens_seen": 931397632
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007250150451354062,
      "loss": 1.6782,
      "theoretical_loss": 3.70256095569507,
      "tokens_seen": 931463168
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007249949849548646,
      "loss": 1.4595,
      "theoretical_loss": 3.7025360320802667,
      "tokens_seen": 931528704
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007249749247743231,
      "loss": 1.4558,
      "theoretical_loss": 3.7025111107097857,
      "tokens_seen": 931594240
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007249548645937813,
      "loss": 1.4582,
      "theoretical_loss": 3.7024861915832665,
      "tokens_seen": 931659776
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007249348044132398,
      "loss": 1.5118,
      "theoretical_loss": 3.702461274700349,
      "tokens_seen": 931725312
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000724914744232698,
      "loss": 1.4941,
      "theoretical_loss": 3.7024363600606742,
      "tokens_seen": 931790848
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007248946840521565,
      "loss": 1.4858,
      "theoretical_loss": 3.702411447663882,
      "tokens_seen": 931856384
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007248746238716149,
      "loss": 1.502,
      "theoretical_loss": 3.7023865375096126,
      "tokens_seen": 931921920
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007248545636910732,
      "loss": 1.4481,
      "theoretical_loss": 3.7023616295975064,
      "tokens_seen": 931987456
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007248345035105316,
      "loss": 1.6111,
      "theoretical_loss": 3.7023367239272034,
      "tokens_seen": 932052992
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007248144433299901,
      "loss": 1.5373,
      "theoretical_loss": 3.7023118204983456,
      "tokens_seen": 932118528
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007247943831494483,
      "loss": 1.562,
      "theoretical_loss": 3.7022869193105734,
      "tokens_seen": 932184064
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007247743229689068,
      "loss": 1.4616,
      "theoretical_loss": 3.702262020363526,
      "tokens_seen": 932249600
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000724754262788365,
      "loss": 1.4238,
      "theoretical_loss": 3.7022371236568468,
      "tokens_seen": 932315136
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007247342026078235,
      "loss": 1.5133,
      "theoretical_loss": 3.7022122291901747,
      "tokens_seen": 932380672
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007247141424272819,
      "loss": 1.517,
      "theoretical_loss": 3.702187336963151,
      "tokens_seen": 932446208
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007246940822467402,
      "loss": 1.4655,
      "theoretical_loss": 3.702162446975418,
      "tokens_seen": 932511744
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007246740220661986,
      "loss": 1.4799,
      "theoretical_loss": 3.7021375592266157,
      "tokens_seen": 932577280
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000724653961885657,
      "loss": 1.5578,
      "theoretical_loss": 3.7021126737163863,
      "tokens_seen": 932642816
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007246339017051153,
      "loss": 1.5245,
      "theoretical_loss": 3.70208779044437,
      "tokens_seen": 932708352
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007246138415245738,
      "loss": 1.4996,
      "theoretical_loss": 3.7020629094102095,
      "tokens_seen": 932773888
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007245937813440321,
      "loss": 1.496,
      "theoretical_loss": 3.7020380306135454,
      "tokens_seen": 932839424
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007245737211634905,
      "loss": 1.3934,
      "theoretical_loss": 3.70201315405402,
      "tokens_seen": 932904960
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007245536609829489,
      "loss": 1.4539,
      "theoretical_loss": 3.7019882797312746,
      "tokens_seen": 932970496
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007245336008024073,
      "loss": 1.6325,
      "theoretical_loss": 3.7019634076449517,
      "tokens_seen": 933036032
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007245135406218656,
      "loss": 1.6974,
      "theoretical_loss": 3.7019385377946916,
      "tokens_seen": 933101568
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000724493480441324,
      "loss": 1.4815,
      "theoretical_loss": 3.7019136701801374,
      "tokens_seen": 933167104
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007244734202607823,
      "loss": 1.6662,
      "theoretical_loss": 3.701888804800931,
      "tokens_seen": 933232640
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007244533600802408,
      "loss": 1.3673,
      "theoretical_loss": 3.7018639416567147,
      "tokens_seen": 933298176
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007244332998996991,
      "loss": 1.4912,
      "theoretical_loss": 3.70183908074713,
      "tokens_seen": 933363712
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007244132397191575,
      "loss": 1.4854,
      "theoretical_loss": 3.70181422207182,
      "tokens_seen": 933429248
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007243931795386158,
      "loss": 1.4158,
      "theoretical_loss": 3.701789365630426,
      "tokens_seen": 933494784
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007243731193580743,
      "loss": 1.5263,
      "theoretical_loss": 3.7017645114225917,
      "tokens_seen": 933560320
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007243530591775326,
      "loss": 1.4787,
      "theoretical_loss": 3.701739659447959,
      "tokens_seen": 933625856
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000724332998996991,
      "loss": 1.4756,
      "theoretical_loss": 3.7017148097061705,
      "tokens_seen": 933691392
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007243129388164493,
      "loss": 1.4186,
      "theoretical_loss": 3.701689962196868,
      "tokens_seen": 933756928
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007242928786359077,
      "loss": 1.5458,
      "theoretical_loss": 3.7016651169196964,
      "tokens_seen": 933822464
    },
    {
      "epoch": 0.28,
      "objective/train/advantage_avg": 0.49441173672676086,
      "objective/train/docs_used": 344002,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5564475059509277,
      "objective/train/original_loss": 1.5564476251602173,
      "objective/train/theoretical_loss": 3.7016402738742964,
      "objective/train/tokens_used": 954348000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24466389417648315,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506850481033325,
      "objective/train/weighted_lm_loss": 1.6352620124816895,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.7016402738742964,
      "tokens_seen": 933888000
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007242728184553661,
      "loss": 1.6135,
      "theoretical_loss": 3.7016402738742964,
      "tokens_seen": 933888000
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007242527582748245,
      "loss": 1.5193,
      "theoretical_loss": 3.7016154330603124,
      "tokens_seen": 933953536
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007242326980942828,
      "loss": 1.5463,
      "theoretical_loss": 3.701590594477387,
      "tokens_seen": 934019072
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007242126379137413,
      "loss": 1.5425,
      "theoretical_loss": 3.701565758125162,
      "tokens_seen": 934084608
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007241925777331996,
      "loss": 1.576,
      "theoretical_loss": 3.701540924003283,
      "tokens_seen": 934150144
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000724172517552658,
      "loss": 1.611,
      "theoretical_loss": 3.701516092111391,
      "tokens_seen": 934215680
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007241524573721164,
      "loss": 1.5177,
      "theoretical_loss": 3.701491262449131,
      "tokens_seen": 934281216
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007241323971915747,
      "loss": 1.6143,
      "theoretical_loss": 3.7014664350161457,
      "tokens_seen": 934346752
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007241123370110331,
      "loss": 1.5082,
      "theoretical_loss": 3.7014416098120786,
      "tokens_seen": 934412288
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007240922768304915,
      "loss": 1.5049,
      "theoretical_loss": 3.7014167868365733,
      "tokens_seen": 934477824
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007240722166499498,
      "loss": 1.5868,
      "theoretical_loss": 3.7013919660892736,
      "tokens_seen": 934543360
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007240521564694083,
      "loss": 1.5462,
      "theoretical_loss": 3.7013671475698233,
      "tokens_seen": 934608896
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007240320962888665,
      "loss": 1.5506,
      "theoretical_loss": 3.701342331277866,
      "tokens_seen": 934674432
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000724012036108325,
      "loss": 1.4113,
      "theoretical_loss": 3.7013175172130452,
      "tokens_seen": 934739968
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007239919759277834,
      "loss": 1.531,
      "theoretical_loss": 3.701292705375006,
      "tokens_seen": 934805504
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007239719157472417,
      "loss": 1.5028,
      "theoretical_loss": 3.701267895763392,
      "tokens_seen": 934871040
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007239518555667001,
      "loss": 1.5578,
      "theoretical_loss": 3.701243088377847,
      "tokens_seen": 934936576
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007239317953861584,
      "loss": 1.5556,
      "theoretical_loss": 3.701218283218016,
      "tokens_seen": 935002112
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007239117352056168,
      "loss": 1.5259,
      "theoretical_loss": 3.701193480283542,
      "tokens_seen": 935067648
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007238916750250753,
      "loss": 1.4193,
      "theoretical_loss": 3.701168679574071,
      "tokens_seen": 935133184
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007238716148445335,
      "loss": 1.5359,
      "theoretical_loss": 3.701143881089246,
      "tokens_seen": 935198720
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000723851554663992,
      "loss": 1.4061,
      "theoretical_loss": 3.701119084828713,
      "tokens_seen": 935264256
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007238314944834504,
      "loss": 1.4745,
      "theoretical_loss": 3.701094290792116,
      "tokens_seen": 935329792
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007238114343029087,
      "loss": 1.517,
      "theoretical_loss": 3.7010694989790993,
      "tokens_seen": 935395328
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007237913741223671,
      "loss": 1.4539,
      "theoretical_loss": 3.701044709389308,
      "tokens_seen": 935460864
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007237713139418255,
      "loss": 1.6794,
      "theoretical_loss": 3.701019922022388,
      "tokens_seen": 935526400
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007237512537612838,
      "loss": 1.4794,
      "theoretical_loss": 3.7009951368779825,
      "tokens_seen": 935591936
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007237311935807423,
      "loss": 1.4547,
      "theoretical_loss": 3.7009703539557375,
      "tokens_seen": 935657472
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007237111334002006,
      "loss": 1.4852,
      "theoretical_loss": 3.7009455732552983,
      "tokens_seen": 935723008
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000723691073219659,
      "loss": 1.517,
      "theoretical_loss": 3.70092079477631,
      "tokens_seen": 935788544
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007236710130391174,
      "loss": 1.4725,
      "theoretical_loss": 3.700896018518418,
      "tokens_seen": 935854080
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007236509528585757,
      "loss": 1.4993,
      "theoretical_loss": 3.7008712444812675,
      "tokens_seen": 935919616
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007236308926780342,
      "loss": 1.5497,
      "theoretical_loss": 3.7008464726645034,
      "tokens_seen": 935985152
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007236108324974925,
      "loss": 1.5012,
      "theoretical_loss": 3.7008217030677724,
      "tokens_seen": 936050688
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007235907723169509,
      "loss": 1.6028,
      "theoretical_loss": 3.700796935690719,
      "tokens_seen": 936116224
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007235707121364093,
      "loss": 1.5361,
      "theoretical_loss": 3.70077217053299,
      "tokens_seen": 936181760
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007235506519558677,
      "loss": 1.4526,
      "theoretical_loss": 3.7007474075942306,
      "tokens_seen": 936247296
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000723530591775326,
      "loss": 1.4516,
      "theoretical_loss": 3.7007226468740866,
      "tokens_seen": 936312832
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007235105315947844,
      "loss": 1.5439,
      "theoretical_loss": 3.700697888372204,
      "tokens_seen": 936378368
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007234904714142427,
      "loss": 1.4962,
      "theoretical_loss": 3.7006731320882293,
      "tokens_seen": 936443904
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007234704112337012,
      "loss": 1.4862,
      "theoretical_loss": 3.700648378021808,
      "tokens_seen": 936509440
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007234503510531595,
      "loss": 1.5216,
      "theoretical_loss": 3.700623626172587,
      "tokens_seen": 936574976
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007234302908726179,
      "loss": 1.5636,
      "theoretical_loss": 3.7005988765402114,
      "tokens_seen": 936640512
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007234102306920762,
      "loss": 1.4919,
      "theoretical_loss": 3.700574129124329,
      "tokens_seen": 936706048
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007233901705115347,
      "loss": 1.5367,
      "theoretical_loss": 3.700549383924585,
      "tokens_seen": 936771584
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000723370110330993,
      "loss": 1.4959,
      "theoretical_loss": 3.700524640940627,
      "tokens_seen": 936837120
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007233500501504514,
      "loss": 1.4473,
      "theoretical_loss": 3.700499900172101,
      "tokens_seen": 936902656
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007233299899699097,
      "loss": 1.4439,
      "theoretical_loss": 3.7004751616186535,
      "tokens_seen": 936968192
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007233099297893681,
      "loss": 1.4495,
      "theoretical_loss": 3.7004504252799313,
      "tokens_seen": 937033728
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007232898696088265,
      "loss": 1.4447,
      "theoretical_loss": 3.700425691155582,
      "tokens_seen": 937099264
    },
    {
      "epoch": 0.28,
      "objective/train/advantage_avg": 0.48927345871925354,
      "objective/train/docs_used": 344763,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3818204402923584,
      "objective/train/original_loss": 1.3818204402923584,
      "objective/train/theoretical_loss": 3.700400959245252,
      "objective/train/tokens_used": 957624800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24084201455116272,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501515865325928,
      "objective/train/weighted_lm_loss": 1.451027512550354,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.700400959245252,
      "tokens_seen": 937164800
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007232698094282849,
      "loss": 1.4665,
      "theoretical_loss": 3.700400959245252,
      "tokens_seen": 937164800
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007232497492477432,
      "loss": 1.3529,
      "theoretical_loss": 3.700376229548588,
      "tokens_seen": 937230336
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007232296890672017,
      "loss": 1.5683,
      "theoretical_loss": 3.7003515020652378,
      "tokens_seen": 937295872
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00072320962888666,
      "loss": 1.5468,
      "theoretical_loss": 3.7003267767948476,
      "tokens_seen": 937361408
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007231895687061184,
      "loss": 1.4723,
      "theoretical_loss": 3.7003020537370657,
      "tokens_seen": 937426944
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007231695085255768,
      "loss": 1.5527,
      "theoretical_loss": 3.700277332891539,
      "tokens_seen": 937492480
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007231494483450351,
      "loss": 1.4233,
      "theoretical_loss": 3.7002526142579146,
      "tokens_seen": 937558016
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007231293881644935,
      "loss": 1.472,
      "theoretical_loss": 3.7002278978358403,
      "tokens_seen": 937623552
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000723109327983952,
      "loss": 1.5364,
      "theoretical_loss": 3.7002031836249643,
      "tokens_seen": 937689088
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007230892678034102,
      "loss": 1.586,
      "theoretical_loss": 3.700178471624933,
      "tokens_seen": 937754624
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007230692076228687,
      "loss": 1.5407,
      "theoretical_loss": 3.700153761835395,
      "tokens_seen": 937820160
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007230491474423269,
      "loss": 1.4722,
      "theoretical_loss": 3.700129054255998,
      "tokens_seen": 937885696
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007230290872617854,
      "loss": 1.5553,
      "theoretical_loss": 3.7001043488863896,
      "tokens_seen": 937951232
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007230090270812438,
      "loss": 1.5595,
      "theoretical_loss": 3.700079645726218,
      "tokens_seen": 938016768
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007229889669007021,
      "loss": 1.4781,
      "theoretical_loss": 3.700054944775131,
      "tokens_seen": 938082304
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007229689067201605,
      "loss": 1.4769,
      "theoretical_loss": 3.700030246032777,
      "tokens_seen": 938147840
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007229488465396188,
      "loss": 1.6117,
      "theoretical_loss": 3.7000055494988047,
      "tokens_seen": 938213376
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007229287863590772,
      "loss": 1.5795,
      "theoretical_loss": 3.699980855172862,
      "tokens_seen": 938278912
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007229087261785357,
      "loss": 1.5097,
      "theoretical_loss": 3.6999561630545967,
      "tokens_seen": 938344448
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007228886659979939,
      "loss": 1.5712,
      "theoretical_loss": 3.699931473143658,
      "tokens_seen": 938409984
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007228686058174524,
      "loss": 1.6014,
      "theoretical_loss": 3.6999067854396936,
      "tokens_seen": 938475520
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007228485456369108,
      "loss": 1.4023,
      "theoretical_loss": 3.6998820999423536,
      "tokens_seen": 938541056
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007228284854563691,
      "loss": 1.439,
      "theoretical_loss": 3.699857416651285,
      "tokens_seen": 938606592
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007228084252758275,
      "loss": 1.4545,
      "theoretical_loss": 3.6998327355661385,
      "tokens_seen": 938672128
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007227883650952859,
      "loss": 1.5029,
      "theoretical_loss": 3.6998080566865608,
      "tokens_seen": 938737664
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007227683049147442,
      "loss": 1.5093,
      "theoretical_loss": 3.699783380012202,
      "tokens_seen": 938803200
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007227482447342027,
      "loss": 1.5164,
      "theoretical_loss": 3.6997587055427115,
      "tokens_seen": 938868736
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007227281845536609,
      "loss": 1.5391,
      "theoretical_loss": 3.699734033277738,
      "tokens_seen": 938934272
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007227081243731194,
      "loss": 1.4238,
      "theoretical_loss": 3.6997093632169307,
      "tokens_seen": 938999808
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007226880641925777,
      "loss": 1.5756,
      "theoretical_loss": 3.699684695359939,
      "tokens_seen": 939065344
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007226680040120361,
      "loss": 1.6001,
      "theoretical_loss": 3.6996600297064113,
      "tokens_seen": 939130880
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007226479438314945,
      "loss": 1.5363,
      "theoretical_loss": 3.6996353662559986,
      "tokens_seen": 939196416
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007226278836509529,
      "loss": 1.5151,
      "theoretical_loss": 3.699610705008349,
      "tokens_seen": 939261952
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007226078234704112,
      "loss": 1.4927,
      "theoretical_loss": 3.699586045963113,
      "tokens_seen": 939327488
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007225877632898697,
      "loss": 1.5265,
      "theoretical_loss": 3.69956138911994,
      "tokens_seen": 939393024
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007225677031093279,
      "loss": 1.3567,
      "theoretical_loss": 3.6995367344784795,
      "tokens_seen": 939458560
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007225476429287864,
      "loss": 1.3797,
      "theoretical_loss": 3.6995120820383818,
      "tokens_seen": 939524096
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007225275827482447,
      "loss": 1.5645,
      "theoretical_loss": 3.6994874317992963,
      "tokens_seen": 939589632
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007225075225677031,
      "loss": 1.4072,
      "theoretical_loss": 3.6994627837608736,
      "tokens_seen": 939655168
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007224874623871615,
      "loss": 1.5406,
      "theoretical_loss": 3.699438137922763,
      "tokens_seen": 939720704
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007224674022066199,
      "loss": 1.5124,
      "theoretical_loss": 3.6994134942846157,
      "tokens_seen": 939786240
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007224473420260782,
      "loss": 1.5612,
      "theoretical_loss": 3.6993888528460808,
      "tokens_seen": 939851776
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007224272818455366,
      "loss": 1.5591,
      "theoretical_loss": 3.6993642136068092,
      "tokens_seen": 939917312
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000722407221664995,
      "loss": 1.6094,
      "theoretical_loss": 3.699339576566451,
      "tokens_seen": 939982848
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007223871614844534,
      "loss": 1.42,
      "theoretical_loss": 3.6993149417246576,
      "tokens_seen": 940048384
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007223671013039117,
      "loss": 1.4978,
      "theoretical_loss": 3.6992903090810785,
      "tokens_seen": 940113920
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007223470411233701,
      "loss": 1.5483,
      "theoretical_loss": 3.6992656786353644,
      "tokens_seen": 940179456
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007223269809428284,
      "loss": 1.5215,
      "theoretical_loss": 3.699241050387166,
      "tokens_seen": 940244992
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007223069207622869,
      "loss": 1.4196,
      "theoretical_loss": 3.699216424336135,
      "tokens_seen": 940310528
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007222868605817452,
      "loss": 1.5163,
      "theoretical_loss": 3.6991918004819215,
      "tokens_seen": 940376064
    },
    {
      "epoch": 0.28,
      "objective/train/advantage_avg": 0.4920152425765991,
      "objective/train/docs_used": 345855,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3839293718338013,
      "objective/train/original_loss": 1.3839294910430908,
      "objective/train/theoretical_loss": 3.6991671788241764,
      "objective/train/tokens_used": 960901600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24251851439476013,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504343509674072,
      "objective/train/weighted_lm_loss": 1.4535373449325562,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6991671788241764,
      "tokens_seen": 940441600
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0007222668004012036,
      "loss": 1.4982,
      "theoretical_loss": 3.6991671788241764,
      "tokens_seen": 940441600
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000722246740220662,
      "loss": 1.6259,
      "theoretical_loss": 3.699142559362551,
      "tokens_seen": 940507136
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007222266800401204,
      "loss": 1.4164,
      "theoretical_loss": 3.6991179420966964,
      "tokens_seen": 940572672
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007222066198595788,
      "loss": 1.5285,
      "theoretical_loss": 3.6990933270262634,
      "tokens_seen": 940638208
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007221865596790372,
      "loss": 1.5041,
      "theoretical_loss": 3.699068714150904,
      "tokens_seen": 940703744
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007221664994984955,
      "loss": 1.3569,
      "theoretical_loss": 3.699044103470269,
      "tokens_seen": 940769280
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007221464393179539,
      "loss": 1.5458,
      "theoretical_loss": 3.69901949498401,
      "tokens_seen": 940834816
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007221263791374123,
      "loss": 1.4807,
      "theoretical_loss": 3.698994888691778,
      "tokens_seen": 940900352
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007221063189568706,
      "loss": 1.5493,
      "theoretical_loss": 3.6989702845932255,
      "tokens_seen": 940965888
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007220862587763291,
      "loss": 1.5664,
      "theoretical_loss": 3.6989456826880036,
      "tokens_seen": 941031424
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007220661985957873,
      "loss": 1.5909,
      "theoretical_loss": 3.6989210829757644,
      "tokens_seen": 941096960
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007220461384152458,
      "loss": 1.6025,
      "theoretical_loss": 3.6988964854561592,
      "tokens_seen": 941162496
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007220260782347042,
      "loss": 1.5441,
      "theoretical_loss": 3.6988718901288404,
      "tokens_seen": 941228032
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007220060180541625,
      "loss": 1.4992,
      "theoretical_loss": 3.69884729699346,
      "tokens_seen": 941293568
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007219859578736209,
      "loss": 1.4971,
      "theoretical_loss": 3.6988227060496692,
      "tokens_seen": 941359104
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007219658976930792,
      "loss": 1.5497,
      "theoretical_loss": 3.6987981172971205,
      "tokens_seen": 941424640
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007219458375125376,
      "loss": 1.5466,
      "theoretical_loss": 3.698773530735467,
      "tokens_seen": 941490176
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007219257773319961,
      "loss": 1.497,
      "theoretical_loss": 3.698748946364361,
      "tokens_seen": 941555712
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007219057171514543,
      "loss": 1.5971,
      "theoretical_loss": 3.6987243641834535,
      "tokens_seen": 941621248
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007218856569709128,
      "loss": 1.5949,
      "theoretical_loss": 3.6986997841923976,
      "tokens_seen": 941686784
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007218655967903712,
      "loss": 1.6695,
      "theoretical_loss": 3.698675206390847,
      "tokens_seen": 941752320
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007218455366098295,
      "loss": 1.3817,
      "theoretical_loss": 3.6986506307784524,
      "tokens_seen": 941817856
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007218254764292879,
      "loss": 1.4212,
      "theoretical_loss": 3.6986260573548675,
      "tokens_seen": 941883392
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007218054162487463,
      "loss": 1.5415,
      "theoretical_loss": 3.6986014861197454,
      "tokens_seen": 941948928
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007217853560682046,
      "loss": 1.558,
      "theoretical_loss": 3.6985769170727383,
      "tokens_seen": 942014464
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007217652958876631,
      "loss": 1.4706,
      "theoretical_loss": 3.6985523502134994,
      "tokens_seen": 942080000
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007217452357071213,
      "loss": 1.4962,
      "theoretical_loss": 3.698527785541682,
      "tokens_seen": 942145536
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007217251755265798,
      "loss": 1.4593,
      "theoretical_loss": 3.698503223056939,
      "tokens_seen": 942211072
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007217051153460381,
      "loss": 1.5144,
      "theoretical_loss": 3.6984786627589226,
      "tokens_seen": 942276608
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007216850551654965,
      "loss": 1.543,
      "theoretical_loss": 3.698454104647287,
      "tokens_seen": 942342144
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007216649949849549,
      "loss": 1.5247,
      "theoretical_loss": 3.6984295487216867,
      "tokens_seen": 942407680
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007216449348044133,
      "loss": 1.6071,
      "theoretical_loss": 3.698404994981773,
      "tokens_seen": 942473216
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007216248746238716,
      "loss": 1.5578,
      "theoretical_loss": 3.6983804434272,
      "tokens_seen": 942538752
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007216048144433301,
      "loss": 1.4984,
      "theoretical_loss": 3.6983558940576224,
      "tokens_seen": 942604288
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007215847542627883,
      "loss": 1.3957,
      "theoretical_loss": 3.6983313468726924,
      "tokens_seen": 942669824
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007215646940822468,
      "loss": 1.5041,
      "theoretical_loss": 3.6983068018720644,
      "tokens_seen": 942735360
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007215446339017051,
      "loss": 1.5987,
      "theoretical_loss": 3.698282259055392,
      "tokens_seen": 942800896
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007215245737211635,
      "loss": 1.6417,
      "theoretical_loss": 3.6982577184223295,
      "tokens_seen": 942866432
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007215045135406219,
      "loss": 1.5365,
      "theoretical_loss": 3.6982331799725303,
      "tokens_seen": 942931968
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007214844533600803,
      "loss": 1.3922,
      "theoretical_loss": 3.698208643705649,
      "tokens_seen": 942997504
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007214643931795386,
      "loss": 1.5624,
      "theoretical_loss": 3.6981841096213395,
      "tokens_seen": 943063040
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000721444332998997,
      "loss": 1.5269,
      "theoretical_loss": 3.698159577719256,
      "tokens_seen": 943128576
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007214242728184554,
      "loss": 1.5016,
      "theoretical_loss": 3.6981350479990525,
      "tokens_seen": 943194112
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007214042126379138,
      "loss": 1.6252,
      "theoretical_loss": 3.698110520460383,
      "tokens_seen": 943259648
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007213841524573721,
      "loss": 1.6469,
      "theoretical_loss": 3.6980859951029035,
      "tokens_seen": 943325184
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007213640922768305,
      "loss": 1.5468,
      "theoretical_loss": 3.698061471926267,
      "tokens_seen": 943390720
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007213440320962888,
      "loss": 1.5313,
      "theoretical_loss": 3.6980369509301285,
      "tokens_seen": 943456256
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007213239719157473,
      "loss": 1.5684,
      "theoretical_loss": 3.6980124321141434,
      "tokens_seen": 943521792
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007213039117352056,
      "loss": 1.5654,
      "theoretical_loss": 3.697987915477966,
      "tokens_seen": 943587328
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000721283851554664,
      "loss": 1.6175,
      "theoretical_loss": 3.6979634010212497,
      "tokens_seen": 943652864
    },
    {
      "epoch": 0.29,
      "objective/train/advantage_avg": 0.4924708604812622,
      "objective/train/docs_used": 346959,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3609083890914917,
      "objective/train/original_loss": 1.3609085083007812,
      "objective/train/theoretical_loss": 3.6979388887436517,
      "objective/train/tokens_used": 964178400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24285486340522766,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504816770553589,
      "objective/train/weighted_lm_loss": 1.4295440912246704,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0293704271316528,
      "theoretical_loss": 3.6979388887436517,
      "tokens_seen": 943718400
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007212637913741224,
      "loss": 1.4148,
      "theoretical_loss": 3.6979388887436517,
      "tokens_seen": 943718400
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007212437311935808,
      "loss": 1.5465,
      "theoretical_loss": 3.6979143786448256,
      "tokens_seen": 943783936
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007212236710130391,
      "loss": 1.5122,
      "theoretical_loss": 3.697889870724427,
      "tokens_seen": 943849472
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007212036108324976,
      "loss": 1.5241,
      "theoretical_loss": 3.6978653649821114,
      "tokens_seen": 943915008
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007211835506519558,
      "loss": 1.5896,
      "theoretical_loss": 3.697840861417533,
      "tokens_seen": 943980544
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007211634904714143,
      "loss": 1.5059,
      "theoretical_loss": 3.6978163600303473,
      "tokens_seen": 944046080
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007211434302908726,
      "loss": 1.5938,
      "theoretical_loss": 3.6977918608202107,
      "tokens_seen": 944111616
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000721123370110331,
      "loss": 1.5233,
      "theoretical_loss": 3.6977673637867774,
      "tokens_seen": 944177152
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007211033099297894,
      "loss": 1.5118,
      "theoretical_loss": 3.697742868929704,
      "tokens_seen": 944242688
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007210832497492477,
      "loss": 1.4842,
      "theoretical_loss": 3.6977183762486456,
      "tokens_seen": 944308224
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007210631895687061,
      "loss": 1.5376,
      "theoretical_loss": 3.6976938857432575,
      "tokens_seen": 944373760
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007210431293881646,
      "loss": 1.5047,
      "theoretical_loss": 3.697669397413196,
      "tokens_seen": 944439296
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007210230692076228,
      "loss": 1.5459,
      "theoretical_loss": 3.6976449112581173,
      "tokens_seen": 944504832
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007210030090270813,
      "loss": 1.5341,
      "theoretical_loss": 3.697620427277676,
      "tokens_seen": 944570368
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007209829488465395,
      "loss": 1.5339,
      "theoretical_loss": 3.6975959454715297,
      "tokens_seen": 944635904
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000720962888665998,
      "loss": 1.4467,
      "theoretical_loss": 3.6975714658393333,
      "tokens_seen": 944701440
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007209428284854564,
      "loss": 1.5497,
      "theoretical_loss": 3.697546988380744,
      "tokens_seen": 944766976
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007209227683049147,
      "loss": 1.5003,
      "theoretical_loss": 3.697522513095417,
      "tokens_seen": 944832512
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007209027081243731,
      "loss": 1.5245,
      "theoretical_loss": 3.6974980399830093,
      "tokens_seen": 944898048
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007208826479438316,
      "loss": 1.5599,
      "theoretical_loss": 3.697473569043177,
      "tokens_seen": 944963584
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007208625877632898,
      "loss": 1.6078,
      "theoretical_loss": 3.697449100275577,
      "tokens_seen": 945029120
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007208425275827483,
      "loss": 1.4111,
      "theoretical_loss": 3.697424633679865,
      "tokens_seen": 945094656
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007208224674022065,
      "loss": 1.5073,
      "theoretical_loss": 3.6974001692556984,
      "tokens_seen": 945160192
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000720802407221665,
      "loss": 1.5062,
      "theoretical_loss": 3.6973757070027338,
      "tokens_seen": 945225728
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007207823470411234,
      "loss": 1.5079,
      "theoretical_loss": 3.6973512469206278,
      "tokens_seen": 945291264
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007207622868605817,
      "loss": 1.5741,
      "theoretical_loss": 3.697326789009037,
      "tokens_seen": 945356800
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007207422266800401,
      "loss": 1.4569,
      "theoretical_loss": 3.697302333267619,
      "tokens_seen": 945422336
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007207221664994985,
      "loss": 1.4118,
      "theoretical_loss": 3.6972778796960304,
      "tokens_seen": 945487872
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007207021063189569,
      "loss": 1.5255,
      "theoretical_loss": 3.6972534282939282,
      "tokens_seen": 945553408
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007206820461384153,
      "loss": 1.4785,
      "theoretical_loss": 3.69722897906097,
      "tokens_seen": 945618944
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007206619859578737,
      "loss": 1.48,
      "theoretical_loss": 3.697204531996813,
      "tokens_seen": 945684480
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000720641925777332,
      "loss": 1.4004,
      "theoretical_loss": 3.6971800871011142,
      "tokens_seen": 945750016
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007206218655967905,
      "loss": 1.5534,
      "theoretical_loss": 3.6971556443735314,
      "tokens_seen": 945815552
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007206018054162487,
      "loss": 1.4791,
      "theoretical_loss": 3.6971312038137216,
      "tokens_seen": 945881088
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007205817452357072,
      "loss": 1.4525,
      "theoretical_loss": 3.697106765421343,
      "tokens_seen": 945946624
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007205616850551655,
      "loss": 1.4768,
      "theoretical_loss": 3.6970823291960526,
      "tokens_seen": 946012160
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007205416248746239,
      "loss": 1.3663,
      "theoretical_loss": 3.697057895137508,
      "tokens_seen": 946077696
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007205215646940823,
      "loss": 1.5944,
      "theoretical_loss": 3.697033463245368,
      "tokens_seen": 946143232
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007205015045135407,
      "loss": 1.5905,
      "theoretical_loss": 3.69700903351929,
      "tokens_seen": 946208768
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000720481444332999,
      "loss": 1.4967,
      "theoretical_loss": 3.6969846059589315,
      "tokens_seen": 946274304
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007204613841524574,
      "loss": 1.5975,
      "theoretical_loss": 3.696960180563951,
      "tokens_seen": 946339840
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007204413239719158,
      "loss": 1.4146,
      "theoretical_loss": 3.696935757334007,
      "tokens_seen": 946405376
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007204212637913742,
      "loss": 1.557,
      "theoretical_loss": 3.696911336268756,
      "tokens_seen": 946470912
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007204012036108325,
      "loss": 1.54,
      "theoretical_loss": 3.6968869173678582,
      "tokens_seen": 946536448
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000720381143430291,
      "loss": 1.6048,
      "theoretical_loss": 3.6968625006309717,
      "tokens_seen": 946601984
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007203610832497492,
      "loss": 1.4707,
      "theoretical_loss": 3.6968380860577534,
      "tokens_seen": 946667520
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007203410230692077,
      "loss": 1.5007,
      "theoretical_loss": 3.696813673647864,
      "tokens_seen": 946733056
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000720320962888666,
      "loss": 1.5863,
      "theoretical_loss": 3.69678926340096,
      "tokens_seen": 946798592
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007203009027081244,
      "loss": 1.5183,
      "theoretical_loss": 3.6967648553167014,
      "tokens_seen": 946864128
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007202808425275828,
      "loss": 1.6508,
      "theoretical_loss": 3.6967404493947464,
      "tokens_seen": 946929664
    },
    {
      "epoch": 0.29,
      "objective/train/advantage_avg": 0.48964735865592957,
      "objective/train/docs_used": 348186,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4477887153625488,
      "objective/train/original_loss": 1.4477888345718384,
      "objective/train/theoretical_loss": 3.696716045634754,
      "objective/train/tokens_used": 967455200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24084705114364624,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501890182495117,
      "objective/train/weighted_lm_loss": 1.5205042362213135,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.696716045634754,
      "tokens_seen": 946995200
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007202607823470412,
      "loss": 1.5156,
      "theoretical_loss": 3.696716045634754,
      "tokens_seen": 946995200
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007202407221664995,
      "loss": 1.4139,
      "theoretical_loss": 3.696691644036383,
      "tokens_seen": 947060736
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000720220661985958,
      "loss": 1.4361,
      "theoretical_loss": 3.696667244599292,
      "tokens_seen": 947126272
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007202006018054162,
      "loss": 1.4624,
      "theoretical_loss": 3.696642847323141,
      "tokens_seen": 947191808
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007201805416248747,
      "loss": 1.5327,
      "theoretical_loss": 3.696618452207588,
      "tokens_seen": 947257344
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000720160481444333,
      "loss": 1.5925,
      "theoretical_loss": 3.696594059252293,
      "tokens_seen": 947322880
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007201404212637914,
      "loss": 1.6156,
      "theoretical_loss": 3.6965696684569154,
      "tokens_seen": 947388416
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007201203610832498,
      "loss": 1.5458,
      "theoretical_loss": 3.6965452798211134,
      "tokens_seen": 947453952
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007201003009027081,
      "loss": 1.5857,
      "theoretical_loss": 3.696520893344548,
      "tokens_seen": 947519488
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007200802407221665,
      "loss": 1.443,
      "theoretical_loss": 3.6964965090268773,
      "tokens_seen": 947585024
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000720060180541625,
      "loss": 1.646,
      "theoretical_loss": 3.6964721268677616,
      "tokens_seen": 947650560
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007200401203610832,
      "loss": 1.4463,
      "theoretical_loss": 3.696447746866861,
      "tokens_seen": 947716096
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007200200601805417,
      "loss": 1.5571,
      "theoretical_loss": 3.696423369023834,
      "tokens_seen": 947781632
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007199999999999999,
      "loss": 1.448,
      "theoretical_loss": 3.696398993338341,
      "tokens_seen": 947847168
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007199799398194584,
      "loss": 1.598,
      "theoretical_loss": 3.696374619810043,
      "tokens_seen": 947912704
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007199598796389168,
      "loss": 1.4015,
      "theoretical_loss": 3.696350248438598,
      "tokens_seen": 947978240
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007199398194583751,
      "loss": 1.5398,
      "theoretical_loss": 3.696325879223667,
      "tokens_seen": 948043776
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007199197592778335,
      "loss": 1.3615,
      "theoretical_loss": 3.6963015121649105,
      "tokens_seen": 948109312
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000719899699097292,
      "loss": 1.5069,
      "theoretical_loss": 3.6962771472619886,
      "tokens_seen": 948174848
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007198796389167502,
      "loss": 1.4771,
      "theoretical_loss": 3.696252784514561,
      "tokens_seen": 948240384
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007198595787362087,
      "loss": 1.477,
      "theoretical_loss": 3.6962284239222885,
      "tokens_seen": 948305920
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007198395185556669,
      "loss": 1.593,
      "theoretical_loss": 3.696204065484831,
      "tokens_seen": 948371456
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007198194583751254,
      "loss": 1.5033,
      "theoretical_loss": 3.69617970920185,
      "tokens_seen": 948436992
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007197993981945838,
      "loss": 1.5096,
      "theoretical_loss": 3.6961553550730053,
      "tokens_seen": 948502528
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007197793380140421,
      "loss": 1.7204,
      "theoretical_loss": 3.6961310030979577,
      "tokens_seen": 948568064
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007197592778335005,
      "loss": 1.4403,
      "theoretical_loss": 3.6961066532763684,
      "tokens_seen": 948633600
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007197392176529589,
      "loss": 1.5816,
      "theoretical_loss": 3.6960823056078973,
      "tokens_seen": 948699136
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007197191574724172,
      "loss": 1.5106,
      "theoretical_loss": 3.696057960092206,
      "tokens_seen": 948764672
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007196990972918757,
      "loss": 1.6124,
      "theoretical_loss": 3.696033616728955,
      "tokens_seen": 948830208
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000719679037111334,
      "loss": 1.5214,
      "theoretical_loss": 3.696009275517806,
      "tokens_seen": 948895744
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007196589769307924,
      "loss": 1.5879,
      "theoretical_loss": 3.6959849364584203,
      "tokens_seen": 948961280
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007196389167502507,
      "loss": 1.519,
      "theoretical_loss": 3.6959605995504576,
      "tokens_seen": 949026816
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007196188565697091,
      "loss": 1.5486,
      "theoretical_loss": 3.6959362647935805,
      "tokens_seen": 949092352
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007195987963891675,
      "loss": 1.5462,
      "theoretical_loss": 3.6959119321874505,
      "tokens_seen": 949157888
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007195787362086259,
      "loss": 1.5094,
      "theoretical_loss": 3.695887601731728,
      "tokens_seen": 949223424
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007195586760280842,
      "loss": 1.5295,
      "theoretical_loss": 3.695863273426075,
      "tokens_seen": 949288960
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007195386158475427,
      "loss": 1.5454,
      "theoretical_loss": 3.695838947270153,
      "tokens_seen": 949354496
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000719518555667001,
      "loss": 1.538,
      "theoretical_loss": 3.6958146232636246,
      "tokens_seen": 949420032
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007194984954864594,
      "loss": 1.6176,
      "theoretical_loss": 3.69579030140615,
      "tokens_seen": 949485568
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007194784353059177,
      "loss": 1.5965,
      "theoretical_loss": 3.695765981697392,
      "tokens_seen": 949551104
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007194583751253762,
      "loss": 1.5014,
      "theoretical_loss": 3.695741664137012,
      "tokens_seen": 949616640
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007194383149448345,
      "loss": 1.5079,
      "theoretical_loss": 3.6957173487246724,
      "tokens_seen": 949682176
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007194182547642929,
      "loss": 1.5123,
      "theoretical_loss": 3.6956930354600352,
      "tokens_seen": 949747712
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007193981945837512,
      "loss": 1.6348,
      "theoretical_loss": 3.695668724342762,
      "tokens_seen": 949813248
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007193781344032096,
      "loss": 1.4692,
      "theoretical_loss": 3.6956444153725156,
      "tokens_seen": 949878784
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000719358074222668,
      "loss": 1.5111,
      "theoretical_loss": 3.695620108548958,
      "tokens_seen": 949944320
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007193380140421264,
      "loss": 1.4789,
      "theoretical_loss": 3.6955958038717522,
      "tokens_seen": 950009856
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007193179538615847,
      "loss": 1.5541,
      "theoretical_loss": 3.6955715013405595,
      "tokens_seen": 950075392
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007192978936810432,
      "loss": 1.5781,
      "theoretical_loss": 3.6955472009550423,
      "tokens_seen": 950140928
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007192778335005015,
      "loss": 1.4989,
      "theoretical_loss": 3.6955229027148646,
      "tokens_seen": 950206464
    },
    {
      "epoch": 0.29,
      "objective/train/advantage_avg": 0.4935845136642456,
      "objective/train/docs_used": 349336,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3703267574310303,
      "objective/train/original_loss": 1.3703267574310303,
      "objective/train/theoretical_loss": 3.695498606619688,
      "objective/train/tokens_used": 970732000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24398204684257507,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505986213684082,
      "objective/train/weighted_lm_loss": 1.4395825862884521,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.695498606619688,
      "tokens_seen": 950272000
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007192577733199599,
      "loss": 1.4604,
      "theoretical_loss": 3.695498606619688,
      "tokens_seen": 950272000
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007192377131394182,
      "loss": 1.4357,
      "theoretical_loss": 3.695474312669176,
      "tokens_seen": 950337536
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007192176529588766,
      "loss": 1.6608,
      "theoretical_loss": 3.6954500208629897,
      "tokens_seen": 950403072
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007191975927783351,
      "loss": 1.3748,
      "theoretical_loss": 3.695425731200794,
      "tokens_seen": 950468608
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007191775325977934,
      "loss": 1.4539,
      "theoretical_loss": 3.6954014436822513,
      "tokens_seen": 950534144
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007191574724172518,
      "loss": 1.4983,
      "theoretical_loss": 3.6953771583070236,
      "tokens_seen": 950599680
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007191374122367102,
      "loss": 1.6647,
      "theoretical_loss": 3.695352875074775,
      "tokens_seen": 950665216
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007191173520561685,
      "loss": 1.4577,
      "theoretical_loss": 3.6953285939851686,
      "tokens_seen": 950730752
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007190972918756269,
      "loss": 1.4672,
      "theoretical_loss": 3.695304315037868,
      "tokens_seen": 950796288
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007190772316950854,
      "loss": 1.5777,
      "theoretical_loss": 3.6952800382325357,
      "tokens_seen": 950861824
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007190571715145436,
      "loss": 1.5638,
      "theoretical_loss": 3.6952557635688352,
      "tokens_seen": 950927360
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007190371113340021,
      "loss": 1.4536,
      "theoretical_loss": 3.695231491046431,
      "tokens_seen": 950992896
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007190170511534603,
      "loss": 1.4863,
      "theoretical_loss": 3.6952072206649857,
      "tokens_seen": 951058432
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007189969909729188,
      "loss": 1.4652,
      "theoretical_loss": 3.6951829524241626,
      "tokens_seen": 951123968
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007189769307923772,
      "loss": 1.5526,
      "theoretical_loss": 3.6951586863236265,
      "tokens_seen": 951189504
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007189568706118355,
      "loss": 1.5713,
      "theoretical_loss": 3.695134422363041,
      "tokens_seen": 951255040
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007189368104312939,
      "loss": 1.5293,
      "theoretical_loss": 3.695110160542069,
      "tokens_seen": 951320576
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007189368104312939,
      "loss": 1.5946,
      "theoretical_loss": 3.695085900860376,
      "tokens_seen": 951386112
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007189167502507524,
      "loss": 1.4885,
      "theoretical_loss": 3.695061643317625,
      "tokens_seen": 951451648
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007188966900702106,
      "loss": 1.5591,
      "theoretical_loss": 3.69503738791348,
      "tokens_seen": 951517184
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007188766298896691,
      "loss": 1.4072,
      "theoretical_loss": 3.6950131346476054,
      "tokens_seen": 951582720
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007188565697091273,
      "loss": 1.5942,
      "theoretical_loss": 3.694988883519666,
      "tokens_seen": 951648256
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007188365095285858,
      "loss": 1.531,
      "theoretical_loss": 3.6949646345293248,
      "tokens_seen": 951713792
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007188164493480442,
      "loss": 1.4934,
      "theoretical_loss": 3.6949403876762474,
      "tokens_seen": 951779328
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007187963891675025,
      "loss": 1.532,
      "theoretical_loss": 3.694916142960098,
      "tokens_seen": 951844864
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007187763289869609,
      "loss": 1.4764,
      "theoretical_loss": 3.6948919003805414,
      "tokens_seen": 951910400
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007187562688064193,
      "loss": 1.3691,
      "theoretical_loss": 3.694867659937242,
      "tokens_seen": 951975936
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007187362086258776,
      "loss": 1.4351,
      "theoretical_loss": 3.694843421629864,
      "tokens_seen": 952041472
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007187161484453361,
      "loss": 1.6208,
      "theoretical_loss": 3.6948191854580728,
      "tokens_seen": 952107008
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007186960882647943,
      "loss": 1.443,
      "theoretical_loss": 3.6947949514215326,
      "tokens_seen": 952172544
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007186760280842528,
      "loss": 1.608,
      "theoretical_loss": 3.6947707195199095,
      "tokens_seen": 952238080
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007186559679037111,
      "loss": 1.4807,
      "theoretical_loss": 3.6947464897528675,
      "tokens_seen": 952303616
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007186359077231695,
      "loss": 1.4595,
      "theoretical_loss": 3.694722262120072,
      "tokens_seen": 952369152
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007186158475426279,
      "loss": 1.5994,
      "theoretical_loss": 3.6946980366211877,
      "tokens_seen": 952434688
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007185957873620863,
      "loss": 1.57,
      "theoretical_loss": 3.694673813255881,
      "tokens_seen": 952500224
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007185757271815446,
      "loss": 1.4659,
      "theoretical_loss": 3.6946495920238167,
      "tokens_seen": 952565760
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007185556670010031,
      "loss": 1.5048,
      "theoretical_loss": 3.6946253729246594,
      "tokens_seen": 952631296
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007185356068204614,
      "loss": 1.5195,
      "theoretical_loss": 3.6946011559580754,
      "tokens_seen": 952696832
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007185155466399198,
      "loss": 1.4108,
      "theoretical_loss": 3.6945769411237297,
      "tokens_seen": 952762368
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007184954864593781,
      "loss": 1.5052,
      "theoretical_loss": 3.6945527284212885,
      "tokens_seen": 952827904
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007184754262788366,
      "loss": 1.5185,
      "theoretical_loss": 3.6945285178504172,
      "tokens_seen": 952893440
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007184553660982949,
      "loss": 1.5093,
      "theoretical_loss": 3.6945043094107817,
      "tokens_seen": 952958976
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007184353059177533,
      "loss": 1.4542,
      "theoretical_loss": 3.6944801031020473,
      "tokens_seen": 953024512
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007184152457372116,
      "loss": 1.4507,
      "theoretical_loss": 3.6944558989238807,
      "tokens_seen": 953090048
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00071839518555667,
      "loss": 1.5067,
      "theoretical_loss": 3.694431696875948,
      "tokens_seen": 953155584
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007183751253761284,
      "loss": 1.5859,
      "theoretical_loss": 3.694407496957914,
      "tokens_seen": 953221120
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007183550651955868,
      "loss": 1.6061,
      "theoretical_loss": 3.694383299169446,
      "tokens_seen": 953286656
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007183350050150451,
      "loss": 1.5297,
      "theoretical_loss": 3.6943591035102097,
      "tokens_seen": 953352192
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007183149448345036,
      "loss": 1.6218,
      "theoretical_loss": 3.6943349099798715,
      "tokens_seen": 953417728
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007182948846539619,
      "loss": 1.3222,
      "theoretical_loss": 3.6943107185780977,
      "tokens_seen": 953483264
    },
    {
      "epoch": 0.29,
      "objective/train/advantage_avg": 0.49555379152297974,
      "objective/train/docs_used": 350619,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3165675401687622,
      "objective/train/original_loss": 1.3165676593780518,
      "objective/train/theoretical_loss": 3.694286529304555,
      "objective/train/tokens_used": 974008800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2457924485206604,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508049726486206,
      "objective/train/weighted_lm_loss": 1.3834824562072754,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.694286529304555,
      "tokens_seen": 953548800
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007182748244734203,
      "loss": 1.4437,
      "theoretical_loss": 3.694286529304555,
      "tokens_seen": 953548800
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007182547642928786,
      "loss": 1.4582,
      "theoretical_loss": 3.6942623421589103,
      "tokens_seen": 953614336
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000718234704112337,
      "loss": 1.5284,
      "theoretical_loss": 3.6942381571408287,
      "tokens_seen": 953679872
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007182146439317954,
      "loss": 1.4424,
      "theoretical_loss": 3.6942139742499784,
      "tokens_seen": 953745408
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007181945837512538,
      "loss": 1.6708,
      "theoretical_loss": 3.694189793486026,
      "tokens_seen": 953810944
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007181745235707121,
      "loss": 1.5503,
      "theoretical_loss": 3.694165614848637,
      "tokens_seen": 953876480
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007181544633901706,
      "loss": 1.5311,
      "theoretical_loss": 3.6941414383374793,
      "tokens_seen": 953942016
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007181344032096288,
      "loss": 1.4911,
      "theoretical_loss": 3.6941172639522204,
      "tokens_seen": 954007552
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007181143430290873,
      "loss": 1.5369,
      "theoretical_loss": 3.6940930916925265,
      "tokens_seen": 954073088
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007180942828485457,
      "loss": 1.4864,
      "theoretical_loss": 3.694068921558065,
      "tokens_seen": 954138624
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000718074222668004,
      "loss": 1.3942,
      "theoretical_loss": 3.6940447535485026,
      "tokens_seen": 954204160
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007180541624874624,
      "loss": 1.5049,
      "theoretical_loss": 3.6940205876635073,
      "tokens_seen": 954269696
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007180341023069207,
      "loss": 1.5732,
      "theoretical_loss": 3.693996423902746,
      "tokens_seen": 954335232
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007180140421263791,
      "loss": 1.4692,
      "theoretical_loss": 3.693972262265887,
      "tokens_seen": 954400768
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007179939819458376,
      "loss": 1.4391,
      "theoretical_loss": 3.6939481027525956,
      "tokens_seen": 954466304
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007179739217652958,
      "loss": 1.4598,
      "theoretical_loss": 3.693923945362542,
      "tokens_seen": 954531840
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007179538615847543,
      "loss": 1.5067,
      "theoretical_loss": 3.6938997900953923,
      "tokens_seen": 954597376
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007179338014042127,
      "loss": 1.5016,
      "theoretical_loss": 3.6938756369508146,
      "tokens_seen": 954662912
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000717913741223671,
      "loss": 1.5436,
      "theoretical_loss": 3.6938514859284766,
      "tokens_seen": 954728448
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007178936810431294,
      "loss": 1.4751,
      "theoretical_loss": 3.693827337028046,
      "tokens_seen": 954793984
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007178736208625877,
      "loss": 1.5415,
      "theoretical_loss": 3.6938031902491915,
      "tokens_seen": 954859520
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007178535606820461,
      "loss": 1.533,
      "theoretical_loss": 3.6937790455915804,
      "tokens_seen": 954925056
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007178335005015046,
      "loss": 1.4386,
      "theoretical_loss": 3.6937549030548813,
      "tokens_seen": 954990592
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007178134403209628,
      "loss": 1.5542,
      "theoretical_loss": 3.6937307626387614,
      "tokens_seen": 955056128
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007177933801404213,
      "loss": 1.374,
      "theoretical_loss": 3.69370662434289,
      "tokens_seen": 955121664
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007177733199598796,
      "loss": 1.5874,
      "theoretical_loss": 3.6936824881669352,
      "tokens_seen": 955187200
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000717753259779338,
      "loss": 1.588,
      "theoretical_loss": 3.693658354110565,
      "tokens_seen": 955252736
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007177331995987964,
      "loss": 1.4508,
      "theoretical_loss": 3.693634222173448,
      "tokens_seen": 955318272
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007177131394182548,
      "loss": 1.5053,
      "theoretical_loss": 3.693610092355253,
      "tokens_seen": 955383808
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007176930792377132,
      "loss": 1.5666,
      "theoretical_loss": 3.6935859646556484,
      "tokens_seen": 955449344
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007176730190571715,
      "loss": 1.4742,
      "theoretical_loss": 3.6935618390743032,
      "tokens_seen": 955514880
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007176529588766299,
      "loss": 1.4658,
      "theoretical_loss": 3.6935377156108853,
      "tokens_seen": 955580416
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007176328986960883,
      "loss": 1.4992,
      "theoretical_loss": 3.6935135942650645,
      "tokens_seen": 955645952
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007176128385155467,
      "loss": 1.5189,
      "theoretical_loss": 3.6934894750365093,
      "tokens_seen": 955711488
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000717592778335005,
      "loss": 1.5632,
      "theoretical_loss": 3.6934653579248886,
      "tokens_seen": 955777024
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007175727181544635,
      "loss": 1.561,
      "theoretical_loss": 3.6934412429298717,
      "tokens_seen": 955842560
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007175526579739218,
      "loss": 1.5443,
      "theoretical_loss": 3.693417130051128,
      "tokens_seen": 955908096
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007175325977933802,
      "loss": 1.3939,
      "theoretical_loss": 3.693393019288326,
      "tokens_seen": 955973632
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007175125376128385,
      "loss": 1.4296,
      "theoretical_loss": 3.693368910641135,
      "tokens_seen": 956039168
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000717492477432297,
      "loss": 1.5783,
      "theoretical_loss": 3.693344804109225,
      "tokens_seen": 956104704
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007174724172517553,
      "loss": 1.5378,
      "theoretical_loss": 3.693320699692265,
      "tokens_seen": 956170240
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007174523570712137,
      "loss": 1.6355,
      "theoretical_loss": 3.6932965973899248,
      "tokens_seen": 956235776
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000717432296890672,
      "loss": 1.453,
      "theoretical_loss": 3.693272497201874,
      "tokens_seen": 956301312
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007174122367101304,
      "loss": 1.5024,
      "theoretical_loss": 3.693248399127781,
      "tokens_seen": 956366848
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007173921765295888,
      "loss": 1.5604,
      "theoretical_loss": 3.6932243031673178,
      "tokens_seen": 956432384
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007173721163490472,
      "loss": 1.5619,
      "theoretical_loss": 3.693200209320153,
      "tokens_seen": 956497920
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007173520561685055,
      "loss": 1.6982,
      "theoretical_loss": 3.6931761175859554,
      "tokens_seen": 956563456
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000717331995987964,
      "loss": 1.5402,
      "theoretical_loss": 3.6931520279643975,
      "tokens_seen": 956628992
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007173119358074223,
      "loss": 1.5647,
      "theoretical_loss": 3.6931279404551467,
      "tokens_seen": 956694528
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007172918756268807,
      "loss": 1.4757,
      "theoretical_loss": 3.693103855057875,
      "tokens_seen": 956760064
    },
    {
      "epoch": 0.29,
      "objective/train/advantage_avg": 0.4956148862838745,
      "objective/train/docs_used": 351696,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4528080224990845,
      "objective/train/original_loss": 1.4528082609176636,
      "objective/train/theoretical_loss": 3.693079771772251,
      "objective/train/tokens_used": 977285600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24582801759243011,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508112907409668,
      "objective/train/weighted_lm_loss": 1.5266010761260986,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.693079771772251,
      "tokens_seen": 956825600
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000717271815446339,
      "loss": 1.501,
      "theoretical_loss": 3.693079771772251,
      "tokens_seen": 956825600
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007172517552657974,
      "loss": 1.5668,
      "theoretical_loss": 3.6930556905979466,
      "tokens_seen": 956891136
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007172316950852558,
      "loss": 1.6318,
      "theoretical_loss": 3.6930316115346304,
      "tokens_seen": 956956672
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007172116349047142,
      "loss": 1.5029,
      "theoretical_loss": 3.693007534581975,
      "tokens_seen": 957022208
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007171915747241725,
      "loss": 1.5111,
      "theoretical_loss": 3.692983459739649,
      "tokens_seen": 957087744
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000717171514543631,
      "loss": 1.5444,
      "theoretical_loss": 3.692959387007323,
      "tokens_seen": 957153280
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007171514543630892,
      "loss": 1.6297,
      "theoretical_loss": 3.6929353163846685,
      "tokens_seen": 957218816
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007171313941825477,
      "loss": 1.6346,
      "theoretical_loss": 3.692911247871356,
      "tokens_seen": 957284352
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000717111334002006,
      "loss": 1.498,
      "theoretical_loss": 3.6928871814670563,
      "tokens_seen": 957349888
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007170912738214644,
      "loss": 1.4934,
      "theoretical_loss": 3.6928631171714397,
      "tokens_seen": 957415424
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007170712136409228,
      "loss": 1.5684,
      "theoretical_loss": 3.6928390549841774,
      "tokens_seen": 957480960
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007170511534603811,
      "loss": 1.5372,
      "theoretical_loss": 3.692814994904941,
      "tokens_seen": 957546496
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007170310932798395,
      "loss": 1.5873,
      "theoretical_loss": 3.6927909369334007,
      "tokens_seen": 957612032
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000717011033099298,
      "loss": 1.5205,
      "theoretical_loss": 3.692766881069228,
      "tokens_seen": 957677568
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007169909729187562,
      "loss": 1.4062,
      "theoretical_loss": 3.692742827312094,
      "tokens_seen": 957743104
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007169709127382147,
      "loss": 1.4998,
      "theoretical_loss": 3.69271877566167,
      "tokens_seen": 957808640
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007169508525576731,
      "loss": 1.4633,
      "theoretical_loss": 3.6926947261176277,
      "tokens_seen": 957874176
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007169307923771314,
      "loss": 1.5998,
      "theoretical_loss": 3.692670678679638,
      "tokens_seen": 957939712
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007169107321965898,
      "loss": 1.4701,
      "theoretical_loss": 3.6926466333473726,
      "tokens_seen": 958005248
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007168906720160481,
      "loss": 1.5296,
      "theoretical_loss": 3.692622590120503,
      "tokens_seen": 958070784
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007168706118355065,
      "loss": 1.5852,
      "theoretical_loss": 3.692598548998702,
      "tokens_seen": 958136320
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000716850551654965,
      "loss": 1.5529,
      "theoretical_loss": 3.692574509981639,
      "tokens_seen": 958201856
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007168304914744232,
      "loss": 1.5819,
      "theoretical_loss": 3.6925504730689878,
      "tokens_seen": 958267392
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007168104312938817,
      "loss": 1.5893,
      "theoretical_loss": 3.6925264382604195,
      "tokens_seen": 958332928
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00071679037111334,
      "loss": 1.5407,
      "theoretical_loss": 3.692502405555606,
      "tokens_seen": 958398464
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007167703109327984,
      "loss": 1.5047,
      "theoretical_loss": 3.69247837495422,
      "tokens_seen": 958464000
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007167502507522568,
      "loss": 1.5692,
      "theoretical_loss": 3.6924543464559325,
      "tokens_seen": 958529536
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007167301905717152,
      "loss": 1.4599,
      "theoretical_loss": 3.6924303200604163,
      "tokens_seen": 958595072
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007167101303911735,
      "loss": 1.7238,
      "theoretical_loss": 3.692406295767344,
      "tokens_seen": 958660608
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007166900702106319,
      "loss": 1.52,
      "theoretical_loss": 3.6923822735763867,
      "tokens_seen": 958726144
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007166700100300902,
      "loss": 1.6038,
      "theoretical_loss": 3.6923582534872184,
      "tokens_seen": 958791680
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007166499498495487,
      "loss": 1.5098,
      "theoretical_loss": 3.6923342354995103,
      "tokens_seen": 958857216
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000716629889669007,
      "loss": 1.507,
      "theoretical_loss": 3.692310219612936,
      "tokens_seen": 958922752
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007166098294884654,
      "loss": 1.5451,
      "theoretical_loss": 3.6922862058271666,
      "tokens_seen": 958988288
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007165897693079238,
      "loss": 1.4867,
      "theoretical_loss": 3.6922621941418763,
      "tokens_seen": 959053824
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007165697091273822,
      "loss": 1.4762,
      "theoretical_loss": 3.6922381845567367,
      "tokens_seen": 959119360
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007165496489468405,
      "loss": 1.506,
      "theoretical_loss": 3.6922141770714214,
      "tokens_seen": 959184896
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007165295887662989,
      "loss": 1.5202,
      "theoretical_loss": 3.692190171685603,
      "tokens_seen": 959250432
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007165095285857572,
      "loss": 1.5625,
      "theoretical_loss": 3.6921661683989546,
      "tokens_seen": 959315968
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007164894684052157,
      "loss": 1.3625,
      "theoretical_loss": 3.6921421672111494,
      "tokens_seen": 959381504
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000716469408224674,
      "loss": 1.5404,
      "theoretical_loss": 3.6921181681218602,
      "tokens_seen": 959447040
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007164493480441324,
      "loss": 1.4399,
      "theoretical_loss": 3.69209417113076,
      "tokens_seen": 959512576
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007164292878635907,
      "loss": 1.6473,
      "theoretical_loss": 3.692070176237523,
      "tokens_seen": 959578112
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007164092276830492,
      "loss": 1.4826,
      "theoretical_loss": 3.6920461834418212,
      "tokens_seen": 959643648
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007163891675025075,
      "loss": 1.5788,
      "theoretical_loss": 3.6920221927433294,
      "tokens_seen": 959709184
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007163691073219659,
      "loss": 1.5833,
      "theoretical_loss": 3.6919982041417203,
      "tokens_seen": 959774720
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007163490471414242,
      "loss": 1.6083,
      "theoretical_loss": 3.691974217636667,
      "tokens_seen": 959840256
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007163289869608827,
      "loss": 1.6187,
      "theoretical_loss": 3.6919502332278444,
      "tokens_seen": 959905792
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000716308926780341,
      "loss": 1.4781,
      "theoretical_loss": 3.691926250914925,
      "tokens_seen": 959971328
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007162888665997994,
      "loss": 1.5793,
      "theoretical_loss": 3.6919022706975833,
      "tokens_seen": 960036864
    },
    {
      "epoch": 0.29,
      "objective/train/advantage_avg": 0.4893192648887634,
      "objective/train/docs_used": 352874,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3887085914611816,
      "objective/train/original_loss": 1.3887087106704712,
      "objective/train/theoretical_loss": 3.6918782925754936,
      "objective/train/tokens_used": 980562400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24047964811325073,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501543283462524,
      "objective/train/weighted_lm_loss": 1.4587208032608032,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6918782925754936,
      "tokens_seen": 960102400
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007162688064192577,
      "loss": 1.497,
      "theoretical_loss": 3.6918782925754936,
      "tokens_seen": 960102400
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007162487462387162,
      "loss": 1.5364,
      "theoretical_loss": 3.691854316548328,
      "tokens_seen": 960167936
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007162286860581745,
      "loss": 1.5076,
      "theoretical_loss": 3.691830342615763,
      "tokens_seen": 960233472
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007162086258776329,
      "loss": 1.6003,
      "theoretical_loss": 3.69180637077747,
      "tokens_seen": 960299008
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007161885656970914,
      "loss": 1.528,
      "theoretical_loss": 3.6917824010331257,
      "tokens_seen": 960364544
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007161685055165496,
      "loss": 1.484,
      "theoretical_loss": 3.6917584333824025,
      "tokens_seen": 960430080
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007161484453360081,
      "loss": 1.4438,
      "theoretical_loss": 3.6917344678249755,
      "tokens_seen": 960495616
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007161283851554665,
      "loss": 1.4055,
      "theoretical_loss": 3.691710504360519,
      "tokens_seen": 960561152
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007161083249749248,
      "loss": 1.4802,
      "theoretical_loss": 3.691686542988707,
      "tokens_seen": 960626688
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007160882647943832,
      "loss": 1.5666,
      "theoretical_loss": 3.6916625837092143,
      "tokens_seen": 960692224
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007160682046138415,
      "loss": 1.5149,
      "theoretical_loss": 3.6916386265217156,
      "tokens_seen": 960757760
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007160481444332999,
      "loss": 1.4258,
      "theoretical_loss": 3.691614671425886,
      "tokens_seen": 960823296
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007160280842527584,
      "loss": 1.5138,
      "theoretical_loss": 3.6915907184213994,
      "tokens_seen": 960888832
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007160080240722166,
      "loss": 1.4792,
      "theoretical_loss": 3.6915667675079304,
      "tokens_seen": 960954368
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007159879638916751,
      "loss": 1.4866,
      "theoretical_loss": 3.6915428186851553,
      "tokens_seen": 961019904
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007159679037111335,
      "loss": 1.5534,
      "theoretical_loss": 3.691518871952747,
      "tokens_seen": 961085440
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007159478435305918,
      "loss": 1.5237,
      "theoretical_loss": 3.6914949273103828,
      "tokens_seen": 961150976
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007159277833500502,
      "loss": 1.4646,
      "theoretical_loss": 3.691470984757736,
      "tokens_seen": 961216512
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007159077231695085,
      "loss": 1.5806,
      "theoretical_loss": 3.6914470442944824,
      "tokens_seen": 961282048
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007158876629889669,
      "loss": 1.5574,
      "theoretical_loss": 3.6914231059202978,
      "tokens_seen": 961347584
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007158676028084254,
      "loss": 1.4704,
      "theoretical_loss": 3.691399169634856,
      "tokens_seen": 961413120
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007158475426278836,
      "loss": 1.4674,
      "theoretical_loss": 3.6913752354378344,
      "tokens_seen": 961478656
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007158274824473421,
      "loss": 1.3912,
      "theoretical_loss": 3.691351303328907,
      "tokens_seen": 961544192
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007158074222668004,
      "loss": 1.5146,
      "theoretical_loss": 3.6913273733077494,
      "tokens_seen": 961609728
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007157873620862588,
      "loss": 1.5437,
      "theoretical_loss": 3.6913034453740377,
      "tokens_seen": 961675264
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007157673019057172,
      "loss": 1.4881,
      "theoretical_loss": 3.6912795195274475,
      "tokens_seen": 961740800
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007157472417251756,
      "loss": 1.5674,
      "theoretical_loss": 3.691255595767654,
      "tokens_seen": 961806336
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007157271815446339,
      "loss": 1.5339,
      "theoretical_loss": 3.691231674094334,
      "tokens_seen": 961871872
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007157071213640923,
      "loss": 1.591,
      "theoretical_loss": 3.691207754507163,
      "tokens_seen": 961937408
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007156870611835506,
      "loss": 1.6487,
      "theoretical_loss": 3.6911838370058163,
      "tokens_seen": 962002944
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007156670010030091,
      "loss": 1.476,
      "theoretical_loss": 3.6911599215899704,
      "tokens_seen": 962068480
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007156469408224674,
      "loss": 1.5186,
      "theoretical_loss": 3.691136008259302,
      "tokens_seen": 962134016
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007156268806419258,
      "loss": 1.5758,
      "theoretical_loss": 3.6911120970134865,
      "tokens_seen": 962199552
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007156068204613842,
      "loss": 1.5549,
      "theoretical_loss": 3.6910881878522,
      "tokens_seen": 962265088
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007155867602808426,
      "loss": 1.5124,
      "theoretical_loss": 3.6910642807751195,
      "tokens_seen": 962330624
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007155667001003009,
      "loss": 1.447,
      "theoretical_loss": 3.6910403757819212,
      "tokens_seen": 962396160
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007155466399197593,
      "loss": 1.5107,
      "theoretical_loss": 3.6910164728722816,
      "tokens_seen": 962461696
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007155265797392176,
      "loss": 1.5653,
      "theoretical_loss": 3.6909925720458774,
      "tokens_seen": 962527232
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007155065195586761,
      "loss": 1.5631,
      "theoretical_loss": 3.6909686733023843,
      "tokens_seen": 962592768
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007154864593781344,
      "loss": 1.4248,
      "theoretical_loss": 3.69094477664148,
      "tokens_seen": 962658304
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007154663991975928,
      "loss": 1.4074,
      "theoretical_loss": 3.6909208820628407,
      "tokens_seen": 962723840
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007154463390170511,
      "loss": 1.4663,
      "theoretical_loss": 3.6908969895661436,
      "tokens_seen": 962789376
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007154262788365096,
      "loss": 1.5528,
      "theoretical_loss": 3.690873099151065,
      "tokens_seen": 962854912
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007154062186559679,
      "loss": 1.5464,
      "theoretical_loss": 3.690849210817283,
      "tokens_seen": 962920448
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007153861584754263,
      "loss": 1.4058,
      "theoretical_loss": 3.6908253245644733,
      "tokens_seen": 962985984
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007153660982948847,
      "loss": 1.5382,
      "theoretical_loss": 3.690801440392314,
      "tokens_seen": 963051520
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007153460381143431,
      "loss": 1.4114,
      "theoretical_loss": 3.690777558300482,
      "tokens_seen": 963117056
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007153259779338014,
      "loss": 1.4413,
      "theoretical_loss": 3.6907536782886545,
      "tokens_seen": 963182592
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007153059177532598,
      "loss": 1.4997,
      "theoretical_loss": 3.6907298003565083,
      "tokens_seen": 963248128
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007152858575727181,
      "loss": 1.589,
      "theoretical_loss": 3.690705924503722,
      "tokens_seen": 963313664
    },
    {
      "epoch": 0.29,
      "objective/train/advantage_avg": 0.49029409885406494,
      "objective/train/docs_used": 353969,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3857896327972412,
      "objective/train/original_loss": 1.3857896327972412,
      "objective/train/theoretical_loss": 3.690682050729972,
      "objective/train/tokens_used": 983839200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24083305895328522,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502536296844482,
      "objective/train/weighted_lm_loss": 1.4553956985473633,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.690682050729972,
      "tokens_seen": 963379200
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007152657973921766,
      "loss": 1.4168,
      "theoretical_loss": 3.690682050729972,
      "tokens_seen": 963379200
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007152457372116349,
      "loss": 1.5192,
      "theoretical_loss": 3.6906581790349366,
      "tokens_seen": 963444736
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007152256770310933,
      "loss": 1.6816,
      "theoretical_loss": 3.690634309418293,
      "tokens_seen": 963510272
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007152056168505517,
      "loss": 1.4731,
      "theoretical_loss": 3.690610441879719,
      "tokens_seen": 963575808
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00071518555667001,
      "loss": 1.478,
      "theoretical_loss": 3.6905865764188923,
      "tokens_seen": 963641344
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007151654964894684,
      "loss": 1.4455,
      "theoretical_loss": 3.690562713035491,
      "tokens_seen": 963706880
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007151454363089269,
      "loss": 1.4353,
      "theoretical_loss": 3.690538851729193,
      "tokens_seen": 963772416
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007151253761283851,
      "loss": 1.4773,
      "theoretical_loss": 3.6905149924996756,
      "tokens_seen": 963837952
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007151053159478436,
      "loss": 1.4866,
      "theoretical_loss": 3.6904911353466177,
      "tokens_seen": 963903488
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007150852557673018,
      "loss": 1.6193,
      "theoretical_loss": 3.690467280269697,
      "tokens_seen": 963969024
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007150651955867603,
      "loss": 1.4598,
      "theoretical_loss": 3.6904434272685918,
      "tokens_seen": 964034560
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007150451354062187,
      "loss": 1.6072,
      "theoretical_loss": 3.6904195763429803,
      "tokens_seen": 964100096
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000715025075225677,
      "loss": 1.4695,
      "theoretical_loss": 3.690395727492541,
      "tokens_seen": 964165632
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007150050150451354,
      "loss": 1.5058,
      "theoretical_loss": 3.690371880716952,
      "tokens_seen": 964231168
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007149849548645939,
      "loss": 1.5286,
      "theoretical_loss": 3.6903480360158922,
      "tokens_seen": 964296704
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007149648946840521,
      "loss": 1.3761,
      "theoretical_loss": 3.69032419338904,
      "tokens_seen": 964362240
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007149448345035106,
      "loss": 1.5425,
      "theoretical_loss": 3.690300352836074,
      "tokens_seen": 964427776
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007149247743229688,
      "loss": 1.6333,
      "theoretical_loss": 3.6902765143566727,
      "tokens_seen": 964493312
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007149047141424273,
      "loss": 1.5271,
      "theoretical_loss": 3.6902526779505145,
      "tokens_seen": 964558848
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007149047141424273,
      "loss": 1.4631,
      "theoretical_loss": 3.6902288436172794,
      "tokens_seen": 964624384
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007148846539618857,
      "loss": 1.6209,
      "theoretical_loss": 3.690205011356646,
      "tokens_seen": 964689920
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000714864593781344,
      "loss": 1.4621,
      "theoretical_loss": 3.6901811811682927,
      "tokens_seen": 964755456
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007148445336008024,
      "loss": 1.6688,
      "theoretical_loss": 3.690157353051899,
      "tokens_seen": 964820992
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007148244734202608,
      "loss": 1.4345,
      "theoretical_loss": 3.690133527007143,
      "tokens_seen": 964886528
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007148044132397191,
      "loss": 1.5604,
      "theoretical_loss": 3.6901097030337056,
      "tokens_seen": 964952064
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007147843530591776,
      "loss": 1.4844,
      "theoretical_loss": 3.6900858811312656,
      "tokens_seen": 965017600
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007147642928786358,
      "loss": 1.4626,
      "theoretical_loss": 3.6900620612995017,
      "tokens_seen": 965083136
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007147442326980943,
      "loss": 1.4057,
      "theoretical_loss": 3.690038243538093,
      "tokens_seen": 965148672
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007147241725175526,
      "loss": 1.5725,
      "theoretical_loss": 3.6900144278467204,
      "tokens_seen": 965214208
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000714704112337011,
      "loss": 1.3801,
      "theoretical_loss": 3.689990614225062,
      "tokens_seen": 965279744
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007146840521564695,
      "loss": 1.6319,
      "theoretical_loss": 3.6899668026727985,
      "tokens_seen": 965345280
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007146639919759278,
      "loss": 1.7258,
      "theoretical_loss": 3.689942993189609,
      "tokens_seen": 965410816
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007146439317953862,
      "loss": 1.5416,
      "theoretical_loss": 3.6899191857751736,
      "tokens_seen": 965476352
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007146238716148446,
      "loss": 1.4699,
      "theoretical_loss": 3.6898953804291716,
      "tokens_seen": 965541888
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000714603811434303,
      "loss": 1.4588,
      "theoretical_loss": 3.6898715771512833,
      "tokens_seen": 965607424
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007145837512537613,
      "loss": 1.557,
      "theoretical_loss": 3.689847775941189,
      "tokens_seen": 965672960
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007145636910732197,
      "loss": 1.5035,
      "theoretical_loss": 3.6898239767985688,
      "tokens_seen": 965738496
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000714543630892678,
      "loss": 1.5023,
      "theoretical_loss": 3.689800179723102,
      "tokens_seen": 965804032
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007145235707121365,
      "loss": 1.6163,
      "theoretical_loss": 3.6897763847144693,
      "tokens_seen": 965869568
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007145035105315948,
      "loss": 1.5976,
      "theoretical_loss": 3.6897525917723515,
      "tokens_seen": 965935104
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007144834503510532,
      "loss": 1.5316,
      "theoretical_loss": 3.689728800896428,
      "tokens_seen": 966000640
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007144633901705115,
      "loss": 1.5586,
      "theoretical_loss": 3.689705012086379,
      "tokens_seen": 966066176
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00071444332998997,
      "loss": 1.3882,
      "theoretical_loss": 3.6896812253418867,
      "tokens_seen": 966131712
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007144232698094283,
      "loss": 1.6061,
      "theoretical_loss": 3.6896574406626295,
      "tokens_seen": 966197248
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007144032096288867,
      "loss": 1.5249,
      "theoretical_loss": 3.68963365804829,
      "tokens_seen": 966262784
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000714383149448345,
      "loss": 1.5455,
      "theoretical_loss": 3.6896098774985475,
      "tokens_seen": 966328320
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007143630892678035,
      "loss": 1.6703,
      "theoretical_loss": 3.6895860990130833,
      "tokens_seen": 966393856
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007143430290872618,
      "loss": 1.5299,
      "theoretical_loss": 3.689562322591578,
      "tokens_seen": 966459392
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007143229689067202,
      "loss": 1.5678,
      "theoretical_loss": 3.689538548233713,
      "tokens_seen": 966524928
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007143029087261785,
      "loss": 1.5246,
      "theoretical_loss": 3.6895147759391684,
      "tokens_seen": 966590464
    },
    {
      "epoch": 0.29,
      "objective/train/advantage_avg": 0.4932740032672882,
      "objective/train/docs_used": 355190,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5405915975570679,
      "objective/train/original_loss": 1.5405914783477783,
      "objective/train/theoretical_loss": 3.6894910057076267,
      "objective/train/tokens_used": 987116000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24391981959342957,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050567388534546,
      "objective/train/weighted_lm_loss": 1.6184053421020508,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6894910057076267,
      "tokens_seen": 966656000
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000714282848545637,
      "loss": 1.6415,
      "theoretical_loss": 3.6894910057076267,
      "tokens_seen": 966656000
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007142627883650953,
      "loss": 1.5786,
      "theoretical_loss": 3.689467237538767,
      "tokens_seen": 966721536
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007142427281845537,
      "loss": 1.6576,
      "theoretical_loss": 3.6894434714322726,
      "tokens_seen": 966787072
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007142226680040121,
      "loss": 1.5481,
      "theoretical_loss": 3.6894197073878234,
      "tokens_seen": 966852608
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007142026078234704,
      "loss": 1.5063,
      "theoretical_loss": 3.689395945405101,
      "tokens_seen": 966918144
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007141825476429288,
      "loss": 1.4914,
      "theoretical_loss": 3.689372185483787,
      "tokens_seen": 966983680
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007141624874623873,
      "loss": 1.4505,
      "theoretical_loss": 3.689348427623563,
      "tokens_seen": 967049216
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007141424272818455,
      "loss": 1.455,
      "theoretical_loss": 3.6893246718241106,
      "tokens_seen": 967114752
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000714122367101304,
      "loss": 1.517,
      "theoretical_loss": 3.689300918085111,
      "tokens_seen": 967180288
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007141023069207622,
      "loss": 1.4716,
      "theoretical_loss": 3.689277166406246,
      "tokens_seen": 967245824
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007140822467402207,
      "loss": 1.5297,
      "theoretical_loss": 3.689253416787197,
      "tokens_seen": 967311360
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007140621865596791,
      "loss": 1.4901,
      "theoretical_loss": 3.6892296692276467,
      "tokens_seen": 967376896
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007140421263791374,
      "loss": 1.4543,
      "theoretical_loss": 3.6892059237272763,
      "tokens_seen": 967442432
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007140220661985958,
      "loss": 1.4914,
      "theoretical_loss": 3.6891821802857687,
      "tokens_seen": 967507968
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007140020060180543,
      "loss": 1.5454,
      "theoretical_loss": 3.6891584389028047,
      "tokens_seen": 967573504
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007139819458375125,
      "loss": 1.4353,
      "theoretical_loss": 3.689134699578067,
      "tokens_seen": 967639040
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000713961885656971,
      "loss": 1.5163,
      "theoretical_loss": 3.689110962311238,
      "tokens_seen": 967704576
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007139418254764292,
      "loss": 1.5782,
      "theoretical_loss": 3.689087227101999,
      "tokens_seen": 967770112
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007139217652958877,
      "loss": 1.579,
      "theoretical_loss": 3.689063493950034,
      "tokens_seen": 967835648
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007139017051153461,
      "loss": 1.523,
      "theoretical_loss": 3.6890397628550247,
      "tokens_seen": 967901184
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007138816449348044,
      "loss": 1.458,
      "theoretical_loss": 3.689016033816652,
      "tokens_seen": 967966720
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007138615847542628,
      "loss": 1.5139,
      "theoretical_loss": 3.688992306834601,
      "tokens_seen": 968032256
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007138415245737212,
      "loss": 1.5914,
      "theoretical_loss": 3.6889685819085525,
      "tokens_seen": 968097792
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007138214643931795,
      "loss": 1.4902,
      "theoretical_loss": 3.6889448590381893,
      "tokens_seen": 968163328
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000713801404212638,
      "loss": 1.6197,
      "theoretical_loss": 3.6889211382231952,
      "tokens_seen": 968228864
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007137813440320962,
      "loss": 1.5345,
      "theoretical_loss": 3.6888974194632524,
      "tokens_seen": 968294400
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007137612838515547,
      "loss": 1.5105,
      "theoretical_loss": 3.688873702758044,
      "tokens_seen": 968359936
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000713741223671013,
      "loss": 1.5818,
      "theoretical_loss": 3.6888499881072523,
      "tokens_seen": 968425472
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007137211634904714,
      "loss": 1.3664,
      "theoretical_loss": 3.6888262755105607,
      "tokens_seen": 968491008
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007137011033099298,
      "loss": 1.5734,
      "theoretical_loss": 3.688802564967653,
      "tokens_seen": 968556544
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007136810431293882,
      "loss": 1.5054,
      "theoretical_loss": 3.688778856478211,
      "tokens_seen": 968622080
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007136609829488465,
      "loss": 1.4976,
      "theoretical_loss": 3.688755150041919,
      "tokens_seen": 968687616
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000713640922768305,
      "loss": 1.4187,
      "theoretical_loss": 3.68873144565846,
      "tokens_seen": 968753152
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007136208625877632,
      "loss": 1.5956,
      "theoretical_loss": 3.6887077433275177,
      "tokens_seen": 968818688
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007136008024072217,
      "loss": 1.5633,
      "theoretical_loss": 3.6886840430487746,
      "tokens_seen": 968884224
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00071358074222668,
      "loss": 1.6159,
      "theoretical_loss": 3.6886603448219155,
      "tokens_seen": 968949760
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007135606820461384,
      "loss": 1.5579,
      "theoretical_loss": 3.688636648646623,
      "tokens_seen": 969015296
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007135406218655968,
      "loss": 1.3964,
      "theoretical_loss": 3.688612954522581,
      "tokens_seen": 969080832
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007135205616850552,
      "loss": 1.3949,
      "theoretical_loss": 3.688589262449474,
      "tokens_seen": 969146368
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007135005015045135,
      "loss": 1.6,
      "theoretical_loss": 3.6885655724269846,
      "tokens_seen": 969211904
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007134804413239719,
      "loss": 1.524,
      "theoretical_loss": 3.6885418844547972,
      "tokens_seen": 969277440
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007134603811434303,
      "loss": 1.5348,
      "theoretical_loss": 3.688518198532596,
      "tokens_seen": 969342976
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007134403209628887,
      "loss": 1.5366,
      "theoretical_loss": 3.6884945146600643,
      "tokens_seen": 969408512
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000713420260782347,
      "loss": 1.5684,
      "theoretical_loss": 3.688470832836887,
      "tokens_seen": 969474048
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007134002006018055,
      "loss": 1.4988,
      "theoretical_loss": 3.6884471530627483,
      "tokens_seen": 969539584
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007133801404212637,
      "loss": 1.4683,
      "theoretical_loss": 3.6884234753373315,
      "tokens_seen": 969605120
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007133600802407222,
      "loss": 1.5252,
      "theoretical_loss": 3.6883997996603215,
      "tokens_seen": 969670656
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007133400200601805,
      "loss": 1.506,
      "theoretical_loss": 3.688376126031403,
      "tokens_seen": 969736192
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007133199598796389,
      "loss": 1.4747,
      "theoretical_loss": 3.6883524544502597,
      "tokens_seen": 969801728
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007132998996990973,
      "loss": 1.6272,
      "theoretical_loss": 3.6883287849165765,
      "tokens_seen": 969867264
    },
    {
      "epoch": 0.29,
      "objective/train/advantage_avg": 0.4915783405303955,
      "objective/train/docs_used": 356486,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4019124507904053,
      "objective/train/original_loss": 1.4019124507904053,
      "objective/train/theoretical_loss": 3.688305117430038,
      "objective/train/tokens_used": 990392800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2424534112215042,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050390362739563,
      "objective/train/weighted_lm_loss": 1.4725009202957153,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.688305117430038,
      "tokens_seen": 969932800
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007132798395185557,
      "loss": 1.4963,
      "theoretical_loss": 3.688305117430038,
      "tokens_seen": 969932800
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000713259779338014,
      "loss": 1.4572,
      "theoretical_loss": 3.688281451990328,
      "tokens_seen": 969998336
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007132397191574725,
      "loss": 1.4617,
      "theoretical_loss": 3.6882577885971326,
      "tokens_seen": 970063872
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007132196589769307,
      "loss": 1.5957,
      "theoretical_loss": 3.688234127250136,
      "tokens_seen": 970129408
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007131995987963892,
      "loss": 1.5722,
      "theoretical_loss": 3.688210467949023,
      "tokens_seen": 970194944
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007131795386158477,
      "loss": 1.44,
      "theoretical_loss": 3.6881868106934785,
      "tokens_seen": 970260480
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007131594784353059,
      "loss": 1.5456,
      "theoretical_loss": 3.688163155483187,
      "tokens_seen": 970326016
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007131394182547644,
      "loss": 1.6113,
      "theoretical_loss": 3.6881395023178345,
      "tokens_seen": 970391552
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007131193580742226,
      "loss": 1.45,
      "theoretical_loss": 3.6881158511971055,
      "tokens_seen": 970457088
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007130992978936811,
      "loss": 1.4657,
      "theoretical_loss": 3.6880922021206857,
      "tokens_seen": 970522624
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007130792377131395,
      "loss": 1.5437,
      "theoretical_loss": 3.68806855508826,
      "tokens_seen": 970588160
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007130591775325978,
      "loss": 1.4395,
      "theoretical_loss": 3.6880449100995136,
      "tokens_seen": 970653696
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007130391173520562,
      "loss": 1.662,
      "theoretical_loss": 3.6880212671541326,
      "tokens_seen": 970719232
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007130190571715147,
      "loss": 1.5585,
      "theoretical_loss": 3.6879976262518017,
      "tokens_seen": 970784768
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007129989969909729,
      "loss": 1.498,
      "theoretical_loss": 3.6879739873922066,
      "tokens_seen": 970850304
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007129789368104314,
      "loss": 1.5178,
      "theoretical_loss": 3.6879503505750333,
      "tokens_seen": 970915840
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007129588766298896,
      "loss": 1.5582,
      "theoretical_loss": 3.687926715799967,
      "tokens_seen": 970981376
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007129388164493481,
      "loss": 1.5256,
      "theoretical_loss": 3.6879030830666943,
      "tokens_seen": 971046912
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007129187562688065,
      "loss": 1.4763,
      "theoretical_loss": 3.6878794523749,
      "tokens_seen": 971112448
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007128986960882648,
      "loss": 1.5922,
      "theoretical_loss": 3.68785582372427,
      "tokens_seen": 971177984
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007128786359077232,
      "loss": 1.4955,
      "theoretical_loss": 3.687832197114491,
      "tokens_seen": 971243520
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007128585757271816,
      "loss": 1.6199,
      "theoretical_loss": 3.687808572545249,
      "tokens_seen": 971309056
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007128385155466399,
      "loss": 1.4695,
      "theoretical_loss": 3.68778495001623,
      "tokens_seen": 971374592
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007128184553660984,
      "loss": 1.5901,
      "theoretical_loss": 3.6877613295271194,
      "tokens_seen": 971440128
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007127983951855566,
      "loss": 1.5788,
      "theoretical_loss": 3.687737711077605,
      "tokens_seen": 971505664
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007127783350050151,
      "loss": 1.5029,
      "theoretical_loss": 3.6877140946673714,
      "tokens_seen": 971571200
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007127582748244734,
      "loss": 1.5879,
      "theoretical_loss": 3.6876904802961064,
      "tokens_seen": 971636736
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007127382146439318,
      "loss": 1.4919,
      "theoretical_loss": 3.687666867963495,
      "tokens_seen": 971702272
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007127181544633902,
      "loss": 1.4733,
      "theoretical_loss": 3.687643257669225,
      "tokens_seen": 971767808
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007126980942828486,
      "loss": 1.5545,
      "theoretical_loss": 3.6876196494129827,
      "tokens_seen": 971833344
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007126780341023069,
      "loss": 1.5352,
      "theoretical_loss": 3.6875960431944543,
      "tokens_seen": 971898880
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007126579739217654,
      "loss": 1.4885,
      "theoretical_loss": 3.6875724390133273,
      "tokens_seen": 971964416
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007126379137412236,
      "loss": 1.5742,
      "theoretical_loss": 3.6875488368692877,
      "tokens_seen": 972029952
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007126178535606821,
      "loss": 1.4737,
      "theoretical_loss": 3.6875252367620224,
      "tokens_seen": 972095488
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007125977933801404,
      "loss": 1.5961,
      "theoretical_loss": 3.687501638691219,
      "tokens_seen": 972161024
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007125777331995988,
      "loss": 1.5005,
      "theoretical_loss": 3.687478042656564,
      "tokens_seen": 972226560
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007125576730190572,
      "loss": 1.4064,
      "theoretical_loss": 3.687454448657745,
      "tokens_seen": 972292096
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007125376128385156,
      "loss": 1.4909,
      "theoretical_loss": 3.6874308566944487,
      "tokens_seen": 972357632
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007125175526579739,
      "loss": 1.5798,
      "theoretical_loss": 3.6874072667663618,
      "tokens_seen": 972423168
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007124974924774323,
      "loss": 1.4058,
      "theoretical_loss": 3.6873836788731724,
      "tokens_seen": 972488704
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007124774322968907,
      "loss": 1.5979,
      "theoretical_loss": 3.687360093014568,
      "tokens_seen": 972554240
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007124573721163491,
      "loss": 1.4624,
      "theoretical_loss": 3.687336509190235,
      "tokens_seen": 972619776
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007124373119358074,
      "loss": 1.5319,
      "theoretical_loss": 3.687312927399862,
      "tokens_seen": 972685312
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007124172517552659,
      "loss": 1.602,
      "theoretical_loss": 3.6872893476431363,
      "tokens_seen": 972750848
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007123971915747241,
      "loss": 1.4596,
      "theoretical_loss": 3.687265769919745,
      "tokens_seen": 972816384
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007123771313941826,
      "loss": 1.6276,
      "theoretical_loss": 3.687242194229376,
      "tokens_seen": 972881920
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007123570712136409,
      "loss": 1.3834,
      "theoretical_loss": 3.6872186205717172,
      "tokens_seen": 972947456
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007123370110330993,
      "loss": 1.4983,
      "theoretical_loss": 3.6871950489464567,
      "tokens_seen": 973012992
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007123169508525577,
      "loss": 1.6425,
      "theoretical_loss": 3.6871714793532826,
      "tokens_seen": 973078528
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007122968906720161,
      "loss": 1.6462,
      "theoretical_loss": 3.6871479117918815,
      "tokens_seen": 973144064
    },
    {
      "epoch": 0.29,
      "objective/train/advantage_avg": 0.493865042924881,
      "objective/train/docs_used": 357717,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5046820640563965,
      "objective/train/original_loss": 1.5046823024749756,
      "objective/train/theoretical_loss": 3.6871243462619425,
      "objective/train/tokens_used": 993669600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24422094225883484,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506280660629272,
      "objective/train/weighted_lm_loss": 1.5807726383209229,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6871243462619425,
      "tokens_seen": 973209600
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007122768304914744,
      "loss": 1.6303,
      "theoretical_loss": 3.6871243462619425,
      "tokens_seen": 973209600
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007122567703109329,
      "loss": 1.5534,
      "theoretical_loss": 3.687100782763154,
      "tokens_seen": 973275136
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007122367101303911,
      "loss": 1.4476,
      "theoretical_loss": 3.687077221295203,
      "tokens_seen": 973340672
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007122166499498496,
      "loss": 1.4619,
      "theoretical_loss": 3.6870536618577794,
      "tokens_seen": 973406208
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0007121965897693079,
      "loss": 1.5644,
      "theoretical_loss": 3.68703010445057,
      "tokens_seen": 973471744
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007121765295887663,
      "loss": 1.4556,
      "theoretical_loss": 3.687006549073264,
      "tokens_seen": 973537280
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007121564694082247,
      "loss": 1.4248,
      "theoretical_loss": 3.6869829957255496,
      "tokens_seen": 973602816
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000712136409227683,
      "loss": 1.5295,
      "theoretical_loss": 3.6869594444071154,
      "tokens_seen": 973668352
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007121163490471414,
      "loss": 1.5446,
      "theoretical_loss": 3.6869358951176503,
      "tokens_seen": 973733888
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007120962888665999,
      "loss": 1.5206,
      "theoretical_loss": 3.686912347856842,
      "tokens_seen": 973799424
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007120762286860581,
      "loss": 1.4104,
      "theoretical_loss": 3.68688880262438,
      "tokens_seen": 973864960
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007120561685055166,
      "loss": 1.5149,
      "theoretical_loss": 3.686865259419953,
      "tokens_seen": 973930496
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007120361083249748,
      "loss": 1.4399,
      "theoretical_loss": 3.68684171824325,
      "tokens_seen": 973996032
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007120160481444333,
      "loss": 1.5794,
      "theoretical_loss": 3.68681817909396,
      "tokens_seen": 974061568
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007119959879638917,
      "loss": 1.3871,
      "theoretical_loss": 3.6867946419717716,
      "tokens_seen": 974127104
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00071197592778335,
      "loss": 1.4623,
      "theoretical_loss": 3.6867711068763738,
      "tokens_seen": 974192640
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007119558676028084,
      "loss": 1.5266,
      "theoretical_loss": 3.686747573807456,
      "tokens_seen": 974258176
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007119358074222669,
      "loss": 1.5587,
      "theoretical_loss": 3.686724042764708,
      "tokens_seen": 974323712
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007119157472417251,
      "loss": 1.3967,
      "theoretical_loss": 3.6867005137478177,
      "tokens_seen": 974389248
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007118956870611836,
      "loss": 1.6201,
      "theoretical_loss": 3.686676986756476,
      "tokens_seen": 974454784
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007118756268806418,
      "loss": 1.4837,
      "theoretical_loss": 3.686653461790371,
      "tokens_seen": 974520320
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007118555667001003,
      "loss": 1.5289,
      "theoretical_loss": 3.686629938849193,
      "tokens_seen": 974585856
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007118355065195587,
      "loss": 1.6583,
      "theoretical_loss": 3.686606417932631,
      "tokens_seen": 974651392
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000711815446339017,
      "loss": 1.4261,
      "theoretical_loss": 3.6865828990403755,
      "tokens_seen": 974716928
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007117953861584754,
      "loss": 1.5295,
      "theoretical_loss": 3.686559382172115,
      "tokens_seen": 974782464
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007117753259779338,
      "loss": 1.4838,
      "theoretical_loss": 3.6865358673275406,
      "tokens_seen": 974848000
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007117552657973921,
      "loss": 1.5026,
      "theoretical_loss": 3.6865123545063403,
      "tokens_seen": 974913536
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007117352056168506,
      "loss": 1.5566,
      "theoretical_loss": 3.686488843708206,
      "tokens_seen": 974979072
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007117151454363089,
      "loss": 1.5646,
      "theoretical_loss": 3.6864653349328265,
      "tokens_seen": 975044608
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007116950852557673,
      "loss": 1.5365,
      "theoretical_loss": 3.686441828179892,
      "tokens_seen": 975110144
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007116750250752258,
      "loss": 1.6184,
      "theoretical_loss": 3.686418323449093,
      "tokens_seen": 975175680
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000711654964894684,
      "loss": 1.5256,
      "theoretical_loss": 3.6863948207401194,
      "tokens_seen": 975241216
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007116349047141425,
      "loss": 1.5608,
      "theoretical_loss": 3.686371320052661,
      "tokens_seen": 975306752
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007116148445336008,
      "loss": 1.5196,
      "theoretical_loss": 3.6863478213864087,
      "tokens_seen": 975372288
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007115947843530592,
      "loss": 1.6116,
      "theoretical_loss": 3.6863243247410526,
      "tokens_seen": 975437824
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007115747241725176,
      "loss": 1.522,
      "theoretical_loss": 3.6863008301162834,
      "tokens_seen": 975503360
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000711554663991976,
      "loss": 1.407,
      "theoretical_loss": 3.6862773375117914,
      "tokens_seen": 975568896
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007115346038114343,
      "loss": 1.4755,
      "theoretical_loss": 3.6862538469272668,
      "tokens_seen": 975634432
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007115145436308927,
      "loss": 1.5954,
      "theoretical_loss": 3.686230358362401,
      "tokens_seen": 975699968
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007114944834503511,
      "loss": 1.5682,
      "theoretical_loss": 3.686206871816885,
      "tokens_seen": 975765504
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007114744232698095,
      "loss": 1.5575,
      "theoretical_loss": 3.686183387290408,
      "tokens_seen": 975831040
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007114543630892678,
      "loss": 1.5302,
      "theoretical_loss": 3.686159904782662,
      "tokens_seen": 975896576
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007114343029087263,
      "loss": 1.3709,
      "theoretical_loss": 3.686136424293338,
      "tokens_seen": 975962112
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007114142427281845,
      "loss": 1.5444,
      "theoretical_loss": 3.686112945822127,
      "tokens_seen": 976027648
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000711394182547643,
      "loss": 1.5209,
      "theoretical_loss": 3.686089469368719,
      "tokens_seen": 976093184
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007113741223671013,
      "loss": 1.4657,
      "theoretical_loss": 3.6860659949328065,
      "tokens_seen": 976158720
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007113540621865597,
      "loss": 1.5079,
      "theoretical_loss": 3.68604252251408,
      "tokens_seen": 976224256
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007113340020060181,
      "loss": 1.498,
      "theoretical_loss": 3.686019052112231,
      "tokens_seen": 976289792
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007113139418254765,
      "loss": 1.4762,
      "theoretical_loss": 3.6859955837269505,
      "tokens_seen": 976355328
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007112938816449348,
      "loss": 1.5314,
      "theoretical_loss": 3.68597211735793,
      "tokens_seen": 976420864
    },
    {
      "epoch": 0.3,
      "objective/train/advantage_avg": 0.49381908774375916,
      "objective/train/docs_used": 358882,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2823368310928345,
      "objective/train/original_loss": 1.282336711883545,
      "objective/train/theoretical_loss": 3.6859486530048615,
      "objective/train/tokens_used": 996946400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24406789243221283,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050622820854187,
      "objective/train/weighted_lm_loss": 1.347160816192627,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.6859486530048615,
      "tokens_seen": 976486400
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007112738214643933,
      "loss": 1.4937,
      "theoretical_loss": 3.6859486530048615,
      "tokens_seen": 976486400
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007112537612838515,
      "loss": 1.556,
      "theoretical_loss": 3.6859251906674357,
      "tokens_seen": 976551936
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00071123370110331,
      "loss": 1.4657,
      "theoretical_loss": 3.6859017303453445,
      "tokens_seen": 976617472
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007112136409227683,
      "loss": 1.4871,
      "theoretical_loss": 3.6858782720382797,
      "tokens_seen": 976683008
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007111935807422267,
      "loss": 1.4716,
      "theoretical_loss": 3.685854815745933,
      "tokens_seen": 976748544
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007111735205616851,
      "loss": 1.5137,
      "theoretical_loss": 3.6858313614679963,
      "tokens_seen": 976814080
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007111534603811434,
      "loss": 1.5277,
      "theoretical_loss": 3.6858079092041613,
      "tokens_seen": 976879616
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007111334002006018,
      "loss": 1.5512,
      "theoretical_loss": 3.68578445895412,
      "tokens_seen": 976945152
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007111133400200603,
      "loss": 1.4875,
      "theoretical_loss": 3.6857610107175645,
      "tokens_seen": 977010688
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007110932798395185,
      "loss": 1.6624,
      "theoretical_loss": 3.6857375644941865,
      "tokens_seen": 977076224
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000711073219658977,
      "loss": 1.604,
      "theoretical_loss": 3.685714120283679,
      "tokens_seen": 977141760
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007110531594784352,
      "loss": 1.4726,
      "theoretical_loss": 3.6856906780857326,
      "tokens_seen": 977207296
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007110330992978937,
      "loss": 1.5451,
      "theoretical_loss": 3.6856672379000415,
      "tokens_seen": 977272832
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007110130391173521,
      "loss": 1.6018,
      "theoretical_loss": 3.6856437997262974,
      "tokens_seen": 977338368
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007109929789368104,
      "loss": 1.6091,
      "theoretical_loss": 3.6856203635641918,
      "tokens_seen": 977403904
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007109729187562688,
      "loss": 1.4354,
      "theoretical_loss": 3.685596929413418,
      "tokens_seen": 977469440
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007109528585757273,
      "loss": 1.59,
      "theoretical_loss": 3.6855734972736682,
      "tokens_seen": 977534976
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007109327983951855,
      "loss": 1.536,
      "theoretical_loss": 3.6855500671446357,
      "tokens_seen": 977600512
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000710912738214644,
      "loss": 1.4602,
      "theoretical_loss": 3.6855266390260124,
      "tokens_seen": 977666048
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007108926780341022,
      "loss": 1.4475,
      "theoretical_loss": 3.6855032129174914,
      "tokens_seen": 977731584
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007108726178535607,
      "loss": 1.4526,
      "theoretical_loss": 3.685479788818766,
      "tokens_seen": 977797120
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007108525576730191,
      "loss": 1.4673,
      "theoretical_loss": 3.685456366729528,
      "tokens_seen": 977862656
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007108324974924774,
      "loss": 1.436,
      "theoretical_loss": 3.6854329466494704,
      "tokens_seen": 977928192
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007108124373119358,
      "loss": 1.6083,
      "theoretical_loss": 3.6854095285782877,
      "tokens_seen": 977993728
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007107923771313942,
      "loss": 1.6763,
      "theoretical_loss": 3.6853861125156717,
      "tokens_seen": 978059264
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007107723169508525,
      "loss": 1.5396,
      "theoretical_loss": 3.685362698461316,
      "tokens_seen": 978124800
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000710752256770311,
      "loss": 1.4352,
      "theoretical_loss": 3.685339286414913,
      "tokens_seen": 978190336
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007107321965897693,
      "loss": 1.5436,
      "theoretical_loss": 3.6853158763761575,
      "tokens_seen": 978255872
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007107121364092277,
      "loss": 1.5501,
      "theoretical_loss": 3.6852924683447412,
      "tokens_seen": 978321408
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007106920762286861,
      "loss": 1.4814,
      "theoretical_loss": 3.6852690623203594,
      "tokens_seen": 978386944
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007106720160481445,
      "loss": 1.4577,
      "theoretical_loss": 3.685245658302704,
      "tokens_seen": 978452480
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007106519558676028,
      "loss": 1.4796,
      "theoretical_loss": 3.6852222562914694,
      "tokens_seen": 978518016
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007106318956870612,
      "loss": 1.5002,
      "theoretical_loss": 3.6851988562863482,
      "tokens_seen": 978583552
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007106118355065195,
      "loss": 1.5412,
      "theoretical_loss": 3.685175458287036,
      "tokens_seen": 978649088
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000710591775325978,
      "loss": 1.4973,
      "theoretical_loss": 3.6851520622932243,
      "tokens_seen": 978714624
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007105717151454363,
      "loss": 1.4367,
      "theoretical_loss": 3.6851286683046087,
      "tokens_seen": 978780160
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007105516549648947,
      "loss": 1.6395,
      "theoretical_loss": 3.6851052763208823,
      "tokens_seen": 978845696
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000710531594784353,
      "loss": 1.51,
      "theoretical_loss": 3.6850818863417394,
      "tokens_seen": 978911232
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007105115346038115,
      "loss": 1.3317,
      "theoretical_loss": 3.6850584983668737,
      "tokens_seen": 978976768
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007104914744232698,
      "loss": 1.5522,
      "theoretical_loss": 3.6850351123959793,
      "tokens_seen": 979042304
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007104714142427282,
      "loss": 1.4807,
      "theoretical_loss": 3.6850117284287505,
      "tokens_seen": 979107840
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007104513540621865,
      "loss": 1.4754,
      "theoretical_loss": 3.684988346464882,
      "tokens_seen": 979173376
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007104312938816449,
      "loss": 1.396,
      "theoretical_loss": 3.684964966504067,
      "tokens_seen": 979238912
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007104112337011033,
      "loss": 1.4126,
      "theoretical_loss": 3.6849415885460006,
      "tokens_seen": 979304448
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007103911735205617,
      "loss": 1.4937,
      "theoretical_loss": 3.6849182125903774,
      "tokens_seen": 979369984
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00071037111334002,
      "loss": 1.4838,
      "theoretical_loss": 3.6848948386368914,
      "tokens_seen": 979435520
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007103510531594785,
      "loss": 1.5637,
      "theoretical_loss": 3.684871466685238,
      "tokens_seen": 979501056
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007103309929789368,
      "loss": 1.6011,
      "theoretical_loss": 3.6848480967351107,
      "tokens_seen": 979566592
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007103109327983952,
      "loss": 1.5899,
      "theoretical_loss": 3.6848247287862046,
      "tokens_seen": 979632128
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007102908726178535,
      "loss": 1.6488,
      "theoretical_loss": 3.684801362838215,
      "tokens_seen": 979697664
    },
    {
      "epoch": 0.3,
      "objective/train/advantage_avg": 0.4935358464717865,
      "objective/train/docs_used": 360128,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5310640335083008,
      "objective/train/original_loss": 1.5310640335083008,
      "objective/train/theoretical_loss": 3.6847779988908362,
      "objective/train/tokens_used": 1000223200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2439456731081009,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050593614578247,
      "objective/train/weighted_lm_loss": 1.6085599660873413,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6847779988908362,
      "tokens_seen": 979763200
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007102708124373119,
      "loss": 1.538,
      "theoretical_loss": 3.6847779988908362,
      "tokens_seen": 979763200
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007102507522567703,
      "loss": 1.52,
      "theoretical_loss": 3.6847546369437634,
      "tokens_seen": 979828736
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007102306920762287,
      "loss": 1.4732,
      "theoretical_loss": 3.684731276996691,
      "tokens_seen": 979894272
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000710210631895687,
      "loss": 1.4793,
      "theoretical_loss": 3.684707919049315,
      "tokens_seen": 979959808
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007101905717151455,
      "loss": 1.5547,
      "theoretical_loss": 3.6846845631013294,
      "tokens_seen": 980025344
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007101705115346038,
      "loss": 1.5436,
      "theoretical_loss": 3.6846612091524302,
      "tokens_seen": 980090880
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007101504513540622,
      "loss": 1.5141,
      "theoretical_loss": 3.684637857202312,
      "tokens_seen": 980156416
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007101303911735207,
      "loss": 1.465,
      "theoretical_loss": 3.6846145072506715,
      "tokens_seen": 980221952
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007101103309929789,
      "loss": 1.5424,
      "theoretical_loss": 3.6845911592972023,
      "tokens_seen": 980287488
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007100902708124374,
      "loss": 1.4714,
      "theoretical_loss": 3.6845678133416007,
      "tokens_seen": 980353024
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007100702106318956,
      "loss": 1.5499,
      "theoretical_loss": 3.684544469383562,
      "tokens_seen": 980418560
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007100501504513541,
      "loss": 1.4614,
      "theoretical_loss": 3.684521127422782,
      "tokens_seen": 980484096
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007100300902708125,
      "loss": 1.4827,
      "theoretical_loss": 3.684497787458956,
      "tokens_seen": 980549632
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007100100300902708,
      "loss": 1.4183,
      "theoretical_loss": 3.68447444949178,
      "tokens_seen": 980615168
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007099899699097292,
      "loss": 1.5433,
      "theoretical_loss": 3.6844511135209497,
      "tokens_seen": 980680704
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007099699097291877,
      "loss": 1.5302,
      "theoretical_loss": 3.6844277795461613,
      "tokens_seen": 980746240
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007099498495486459,
      "loss": 1.4961,
      "theoretical_loss": 3.68440444756711,
      "tokens_seen": 980811776
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007099297893681044,
      "loss": 1.4829,
      "theoretical_loss": 3.6843811175834915,
      "tokens_seen": 980877312
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007099097291875626,
      "loss": 1.47,
      "theoretical_loss": 3.684357789595003,
      "tokens_seen": 980942848
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007098896690070211,
      "loss": 1.5389,
      "theoretical_loss": 3.68433446360134,
      "tokens_seen": 981008384
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007098696088264795,
      "loss": 1.5265,
      "theoretical_loss": 3.6843111396021992,
      "tokens_seen": 981073920
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007098495486459378,
      "loss": 1.5883,
      "theoretical_loss": 3.6842878175972755,
      "tokens_seen": 981139456
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007098294884653962,
      "loss": 1.5037,
      "theoretical_loss": 3.684264497586266,
      "tokens_seen": 981204992
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007098094282848546,
      "loss": 1.5438,
      "theoretical_loss": 3.684241179568868,
      "tokens_seen": 981270528
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007097893681043129,
      "loss": 1.5435,
      "theoretical_loss": 3.6842178635447764,
      "tokens_seen": 981336064
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007097693079237714,
      "loss": 1.6123,
      "theoretical_loss": 3.6841945495136885,
      "tokens_seen": 981401600
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007097492477432297,
      "loss": 1.4413,
      "theoretical_loss": 3.684171237475301,
      "tokens_seen": 981467136
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007097291875626881,
      "loss": 1.466,
      "theoretical_loss": 3.6841479274293096,
      "tokens_seen": 981532672
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007097091273821465,
      "loss": 1.6303,
      "theoretical_loss": 3.6841246193754125,
      "tokens_seen": 981598208
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007096890672016049,
      "loss": 1.5475,
      "theoretical_loss": 3.684101313313305,
      "tokens_seen": 981663744
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007096690070210632,
      "loss": 1.5015,
      "theoretical_loss": 3.6840780092426852,
      "tokens_seen": 981729280
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007096489468405216,
      "loss": 1.5444,
      "theoretical_loss": 3.6840547071632486,
      "tokens_seen": 981794816
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007096288866599799,
      "loss": 1.4455,
      "theoretical_loss": 3.684031407074694,
      "tokens_seen": 981860352
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007096088264794384,
      "loss": 1.5032,
      "theoretical_loss": 3.6840081089767165,
      "tokens_seen": 981925888
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007095887662988967,
      "loss": 1.5297,
      "theoretical_loss": 3.6839848128690145,
      "tokens_seen": 981991424
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007095687061183551,
      "loss": 1.5064,
      "theoretical_loss": 3.6839615187512846,
      "tokens_seen": 982056960
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007095486459378134,
      "loss": 1.5934,
      "theoretical_loss": 3.6839382266232237,
      "tokens_seen": 982122496
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007095285857572719,
      "loss": 1.525,
      "theoretical_loss": 3.6839149364845305,
      "tokens_seen": 982188032
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007095085255767302,
      "loss": 1.5764,
      "theoretical_loss": 3.683891648334901,
      "tokens_seen": 982253568
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007094884653961886,
      "loss": 1.4934,
      "theoretical_loss": 3.6838683621740334,
      "tokens_seen": 982319104
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007094684052156469,
      "loss": 1.4774,
      "theoretical_loss": 3.6838450780016245,
      "tokens_seen": 982384640
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007094483450351053,
      "loss": 1.5853,
      "theoretical_loss": 3.683821795817372,
      "tokens_seen": 982450176
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007094282848545637,
      "loss": 1.5721,
      "theoretical_loss": 3.6837985156209743,
      "tokens_seen": 982515712
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007094082246740221,
      "loss": 1.574,
      "theoretical_loss": 3.6837752374121284,
      "tokens_seen": 982581248
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007093881644934804,
      "loss": 1.6031,
      "theoretical_loss": 3.6837519611905325,
      "tokens_seen": 982646784
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007093681043129389,
      "loss": 1.5645,
      "theoretical_loss": 3.6837286869558836,
      "tokens_seen": 982712320
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007093480441323972,
      "loss": 1.5629,
      "theoretical_loss": 3.683705414707881,
      "tokens_seen": 982777856
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007093279839518556,
      "loss": 1.5532,
      "theoretical_loss": 3.6836821444462213,
      "tokens_seen": 982843392
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000709307923771314,
      "loss": 1.4873,
      "theoretical_loss": 3.683658876170603,
      "tokens_seen": 982908928
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007092878635907723,
      "loss": 1.4468,
      "theoretical_loss": 3.6836356098807244,
      "tokens_seen": 982974464
    },
    {
      "debugging/Compilability": 1.0,
      "debugging/distinct-1-grams": 0.7488293247094114,
      "debugging/entropy-1-grams": 5.349564433264487,
      "debugging/length": 462.46153846153845,
      "debugging/num_segments": 13,
      "debugging/raw_token_scores_avg": 0.00979544036090374,
      "debugging/raw_token_scores_std": 0.022818773984909058,
      "debugging/score": 0.0073770168052465215,
      "debugging/score_std": 0.0054551366535427,
      "epoch": 0.3,
      "objective/train/advantage_avg": 0.49020400643348694,
      "objective/train/docs_used": 361383,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.349044919013977,
      "objective/train/original_loss": 1.3490447998046875,
      "objective/train/theoretical_loss": 3.6836123455762837,
      "objective/train/tokens_used": 1003500000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24082081019878387,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050244688987732,
      "objective/train/weighted_lm_loss": 1.416857123374939,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6836123455762837,
      "tokens_seen": 983040000
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007092678034102307,
      "loss": 1.5057,
      "theoretical_loss": 3.6836123455762837,
      "tokens_seen": 983040000
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007092477432296891,
      "loss": 1.5334,
      "theoretical_loss": 3.6835890832569786,
      "tokens_seen": 983105536
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007092276830491474,
      "loss": 1.5949,
      "theoretical_loss": 3.683565822922508,
      "tokens_seen": 983171072
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007092076228686059,
      "loss": 1.5106,
      "theoretical_loss": 3.68354256457257,
      "tokens_seen": 983236608
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007091875626880641,
      "loss": 1.484,
      "theoretical_loss": 3.683519308206863,
      "tokens_seen": 983302144
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007091675025075226,
      "loss": 1.4923,
      "theoretical_loss": 3.6834960538250856,
      "tokens_seen": 983367680
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000709147442326981,
      "loss": 1.473,
      "theoretical_loss": 3.683472801426936,
      "tokens_seen": 983433216
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007091273821464393,
      "loss": 1.6085,
      "theoretical_loss": 3.6834495510121132,
      "tokens_seen": 983498752
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007091073219658977,
      "loss": 1.537,
      "theoretical_loss": 3.683426302580316,
      "tokens_seen": 983564288
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000709087261785356,
      "loss": 1.5747,
      "theoretical_loss": 3.6834030561312425,
      "tokens_seen": 983629824
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007090672016048144,
      "loss": 1.5598,
      "theoretical_loss": 3.683379811664592,
      "tokens_seen": 983695360
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007090471414242729,
      "loss": 1.5092,
      "theoretical_loss": 3.683356569180064,
      "tokens_seen": 983760896
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007090270812437311,
      "loss": 1.5644,
      "theoretical_loss": 3.683333328677356,
      "tokens_seen": 983826432
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007090070210631896,
      "loss": 1.505,
      "theoretical_loss": 3.6833100901561684,
      "tokens_seen": 983891968
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000708986960882648,
      "loss": 1.3454,
      "theoretical_loss": 3.6832868536161993,
      "tokens_seen": 983957504
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007089669007021063,
      "loss": 1.4537,
      "theoretical_loss": 3.6832636190571484,
      "tokens_seen": 984023040
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007089468405215647,
      "loss": 1.4599,
      "theoretical_loss": 3.6832403864787144,
      "tokens_seen": 984088576
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000708926780341023,
      "loss": 1.5063,
      "theoretical_loss": 3.6832171558805973,
      "tokens_seen": 984154112
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007089067201604814,
      "loss": 1.577,
      "theoretical_loss": 3.6831939272624963,
      "tokens_seen": 984219648
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007088866599799399,
      "loss": 1.451,
      "theoretical_loss": 3.6831707006241103,
      "tokens_seen": 984285184
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007088665997993981,
      "loss": 1.5625,
      "theoretical_loss": 3.683147475965139,
      "tokens_seen": 984350720
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007088465396188566,
      "loss": 1.4765,
      "theoretical_loss": 3.683124253285282,
      "tokens_seen": 984416256
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007088264794383149,
      "loss": 1.5286,
      "theoretical_loss": 3.683101032584239,
      "tokens_seen": 984481792
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007088064192577733,
      "loss": 1.5541,
      "theoretical_loss": 3.6830778138617095,
      "tokens_seen": 984547328
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007087863590772317,
      "loss": 1.6529,
      "theoretical_loss": 3.683054597117393,
      "tokens_seen": 984612864
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007087662988966901,
      "loss": 1.4626,
      "theoretical_loss": 3.68303138235099,
      "tokens_seen": 984678400
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007087462387161484,
      "loss": 1.6117,
      "theoretical_loss": 3.6830081695622003,
      "tokens_seen": 984743936
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007087261785356069,
      "loss": 1.5535,
      "theoretical_loss": 3.6829849587507235,
      "tokens_seen": 984809472
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007087061183550653,
      "loss": 1.5429,
      "theoretical_loss": 3.6829617499162595,
      "tokens_seen": 984875008
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007086860581745236,
      "loss": 1.4523,
      "theoretical_loss": 3.682938543058508,
      "tokens_seen": 984940544
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000708665997993982,
      "loss": 1.5819,
      "theoretical_loss": 3.68291533817717,
      "tokens_seen": 985006080
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007086459378134403,
      "loss": 1.5588,
      "theoretical_loss": 3.6828921352719455,
      "tokens_seen": 985071616
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007086258776328988,
      "loss": 1.5212,
      "theoretical_loss": 3.6828689343425345,
      "tokens_seen": 985137152
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007086058174523571,
      "loss": 1.5678,
      "theoretical_loss": 3.6828457353886375,
      "tokens_seen": 985202688
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007085857572718155,
      "loss": 1.4901,
      "theoretical_loss": 3.6828225384099547,
      "tokens_seen": 985268224
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007085656970912738,
      "loss": 1.5296,
      "theoretical_loss": 3.6827993434061863,
      "tokens_seen": 985333760
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007085456369107323,
      "loss": 1.459,
      "theoretical_loss": 3.682776150377034,
      "tokens_seen": 985399296
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007085255767301906,
      "loss": 1.5759,
      "theoretical_loss": 3.682752959322197,
      "tokens_seen": 985464832
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000708505516549649,
      "loss": 1.5498,
      "theoretical_loss": 3.6827297702413766,
      "tokens_seen": 985530368
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007084854563691073,
      "loss": 1.4932,
      "theoretical_loss": 3.682706583134274,
      "tokens_seen": 985595904
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007084653961885657,
      "loss": 1.6141,
      "theoretical_loss": 3.682683398000589,
      "tokens_seen": 985661440
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007084453360080241,
      "loss": 1.5315,
      "theoretical_loss": 3.6826602148400225,
      "tokens_seen": 985726976
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007084252758274825,
      "loss": 1.5947,
      "theoretical_loss": 3.6826370336522762,
      "tokens_seen": 985792512
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007084052156469408,
      "loss": 1.7226,
      "theoretical_loss": 3.6826138544370512,
      "tokens_seen": 985858048
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007083851554663993,
      "loss": 1.4656,
      "theoretical_loss": 3.6825906771940478,
      "tokens_seen": 985923584
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007083650952858576,
      "loss": 1.5824,
      "theoretical_loss": 3.682567501922967,
      "tokens_seen": 985989120
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000708345035105316,
      "loss": 1.4884,
      "theoretical_loss": 3.6825443286235107,
      "tokens_seen": 986054656
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007083249749247744,
      "loss": 1.453,
      "theoretical_loss": 3.6825211572953793,
      "tokens_seen": 986120192
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007083049147442327,
      "loss": 1.4852,
      "theoretical_loss": 3.682497987938275,
      "tokens_seen": 986185728
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007082848545636911,
      "loss": 1.4171,
      "theoretical_loss": 3.6824748205518985,
      "tokens_seen": 986251264
    },
    {
      "epoch": 0.3,
      "objective/train/advantage_avg": 0.4950016736984253,
      "objective/train/docs_used": 362627,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.460729718208313,
      "objective/train/original_loss": 1.4607298374176025,
      "objective/train/theoretical_loss": 3.682451655135952,
      "objective/train/tokens_used": 1006776800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24525780975818634,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507471561431885,
      "objective/train/weighted_lm_loss": 1.5347301959991455,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.682451655135952,
      "tokens_seen": 986316800
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007082647943831495,
      "loss": 1.5012,
      "theoretical_loss": 3.682451655135952,
      "tokens_seen": 986316800
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007082447342026078,
      "loss": 1.5389,
      "theoretical_loss": 3.682428491690136,
      "tokens_seen": 986382336
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007082246740220663,
      "loss": 1.5163,
      "theoretical_loss": 3.682405330214153,
      "tokens_seen": 986447872
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007082046138415245,
      "loss": 1.5096,
      "theoretical_loss": 3.6823821707077036,
      "tokens_seen": 986513408
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000708184553660983,
      "loss": 1.4356,
      "theoretical_loss": 3.6823590131704904,
      "tokens_seen": 986578944
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007081644934804414,
      "loss": 1.4247,
      "theoretical_loss": 3.682335857602215,
      "tokens_seen": 986644480
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007081444332998997,
      "loss": 1.4645,
      "theoretical_loss": 3.682312704002579,
      "tokens_seen": 986710016
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007081243731193581,
      "loss": 1.5249,
      "theoretical_loss": 3.682289552371284,
      "tokens_seen": 986775552
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007081043129388164,
      "loss": 1.5937,
      "theoretical_loss": 3.682266402708033,
      "tokens_seen": 986841088
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007080842527582748,
      "loss": 1.5335,
      "theoretical_loss": 3.6822432550125272,
      "tokens_seen": 986906624
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007080641925777333,
      "loss": 1.5617,
      "theoretical_loss": 3.6822201092844686,
      "tokens_seen": 986972160
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007080441323971915,
      "loss": 1.5001,
      "theoretical_loss": 3.6821969655235596,
      "tokens_seen": 987037696
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00070802407221665,
      "loss": 1.5898,
      "theoretical_loss": 3.6821738237295025,
      "tokens_seen": 987103232
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007080040120361084,
      "loss": 1.4197,
      "theoretical_loss": 3.6821506839019995,
      "tokens_seen": 987168768
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007079839518555667,
      "loss": 1.4673,
      "theoretical_loss": 3.682127546040753,
      "tokens_seen": 987234304
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007079638916750251,
      "loss": 1.6351,
      "theoretical_loss": 3.682104410145465,
      "tokens_seen": 987299840
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007079438314944834,
      "loss": 1.5208,
      "theoretical_loss": 3.682081276215839,
      "tokens_seen": 987365376
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007079237713139418,
      "loss": 1.4866,
      "theoretical_loss": 3.6820581442515765,
      "tokens_seen": 987430912
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007079037111334003,
      "loss": 1.4982,
      "theoretical_loss": 3.6820350142523806,
      "tokens_seen": 987496448
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007078836509528585,
      "loss": 1.5795,
      "theoretical_loss": 3.6820118862179534,
      "tokens_seen": 987561984
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000707863590772317,
      "loss": 1.5187,
      "theoretical_loss": 3.6819887601479984,
      "tokens_seen": 987627520
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007078435305917753,
      "loss": 1.5159,
      "theoretical_loss": 3.6819656360422184,
      "tokens_seen": 987693056
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007078234704112337,
      "loss": 1.4958,
      "theoretical_loss": 3.6819425139003155,
      "tokens_seen": 987758592
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007078034102306921,
      "loss": 1.6331,
      "theoretical_loss": 3.6819193937219934,
      "tokens_seen": 987824128
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007077833500501505,
      "loss": 1.4976,
      "theoretical_loss": 3.6818962755069546,
      "tokens_seen": 987889664
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007077632898696088,
      "loss": 1.53,
      "theoretical_loss": 3.681873159254902,
      "tokens_seen": 987955200
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007077432296890673,
      "loss": 1.5517,
      "theoretical_loss": 3.6818500449655396,
      "tokens_seen": 988020736
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007077231695085255,
      "loss": 1.5546,
      "theoretical_loss": 3.68182693263857,
      "tokens_seen": 988086272
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000707703109327984,
      "loss": 1.5102,
      "theoretical_loss": 3.681803822273696,
      "tokens_seen": 988151808
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007076830491474423,
      "loss": 1.428,
      "theoretical_loss": 3.681780713870622,
      "tokens_seen": 988217344
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007076629889669007,
      "loss": 1.4207,
      "theoretical_loss": 3.6817576074290503,
      "tokens_seen": 988282880
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007076429287863591,
      "loss": 1.5934,
      "theoretical_loss": 3.681734502948685,
      "tokens_seen": 988348416
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007076228686058175,
      "loss": 1.3751,
      "theoretical_loss": 3.6817114004292297,
      "tokens_seen": 988413952
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007076028084252758,
      "loss": 1.4964,
      "theoretical_loss": 3.6816882998703875,
      "tokens_seen": 988479488
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007075827482447342,
      "loss": 1.4448,
      "theoretical_loss": 3.681665201271862,
      "tokens_seen": 988545024
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007075626880641925,
      "loss": 1.5548,
      "theoretical_loss": 3.681642104633358,
      "tokens_seen": 988610560
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000707542627883651,
      "loss": 1.53,
      "theoretical_loss": 3.681619009954578,
      "tokens_seen": 988676096
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007075225677031093,
      "loss": 1.5974,
      "theoretical_loss": 3.6815959172352257,
      "tokens_seen": 988741632
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007075025075225677,
      "loss": 1.6386,
      "theoretical_loss": 3.681572826475006,
      "tokens_seen": 988807168
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000707482447342026,
      "loss": 1.6574,
      "theoretical_loss": 3.6815497376736226,
      "tokens_seen": 988872704
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007074623871614845,
      "loss": 1.4708,
      "theoretical_loss": 3.6815266508307793,
      "tokens_seen": 988938240
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007074423269809428,
      "loss": 1.4609,
      "theoretical_loss": 3.68150356594618,
      "tokens_seen": 989003776
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007074222668004012,
      "loss": 1.5522,
      "theoretical_loss": 3.681480483019529,
      "tokens_seen": 989069312
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007074022066198596,
      "loss": 1.4002,
      "theoretical_loss": 3.681457402050531,
      "tokens_seen": 989134848
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000707382146439318,
      "loss": 1.5683,
      "theoretical_loss": 3.6814343230388897,
      "tokens_seen": 989200384
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007073620862587763,
      "loss": 1.5354,
      "theoretical_loss": 3.68141124598431,
      "tokens_seen": 989265920
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007073420260782348,
      "loss": 1.6579,
      "theoretical_loss": 3.6813881708864953,
      "tokens_seen": 989331456
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000707321965897693,
      "loss": 1.6021,
      "theoretical_loss": 3.681365097745151,
      "tokens_seen": 989396992
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007073019057171515,
      "loss": 1.5182,
      "theoretical_loss": 3.6813420265599817,
      "tokens_seen": 989462528
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007072818455366098,
      "loss": 1.4794,
      "theoretical_loss": 3.6813189573306913,
      "tokens_seen": 989528064
    },
    {
      "epoch": 0.3,
      "objective/train/advantage_avg": 0.49303433299064636,
      "objective/train/docs_used": 363941,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.381746768951416,
      "objective/train/original_loss": 1.381746530532837,
      "objective/train/theoretical_loss": 3.681295890056985,
      "objective/train/tokens_used": 1010053600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24332168698310852,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505403280258179,
      "objective/train/weighted_lm_loss": 1.4518052339553833,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.681295890056985,
      "tokens_seen": 989593600
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007072617853560682,
      "loss": 1.443,
      "theoretical_loss": 3.681295890056985,
      "tokens_seen": 989593600
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007072417251755266,
      "loss": 1.4521,
      "theoretical_loss": 3.681272824738567,
      "tokens_seen": 989659136
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007072216649949849,
      "loss": 1.3886,
      "theoretical_loss": 3.681249761375143,
      "tokens_seen": 989724672
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007072016048144434,
      "loss": 1.4113,
      "theoretical_loss": 3.681226699966417,
      "tokens_seen": 989790208
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007071815446339018,
      "loss": 1.6131,
      "theoretical_loss": 3.681203640512095,
      "tokens_seen": 989855744
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007071614844533601,
      "loss": 1.3923,
      "theoretical_loss": 3.68118058301188,
      "tokens_seen": 989921280
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007071414242728185,
      "loss": 1.5988,
      "theoretical_loss": 3.681157527465479,
      "tokens_seen": 989986816
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007071213640922768,
      "loss": 1.6322,
      "theoretical_loss": 3.6811344738725964,
      "tokens_seen": 990052352
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007071013039117352,
      "loss": 1.5585,
      "theoretical_loss": 3.681111422232937,
      "tokens_seen": 990117888
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007070812437311937,
      "loss": 1.5384,
      "theoretical_loss": 3.681088372546207,
      "tokens_seen": 990183424
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007070611835506519,
      "loss": 1.5851,
      "theoretical_loss": 3.6810653248121112,
      "tokens_seen": 990248960
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007070411233701104,
      "loss": 1.5624,
      "theoretical_loss": 3.6810422790303545,
      "tokens_seen": 990314496
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007070210631895688,
      "loss": 1.5593,
      "theoretical_loss": 3.681019235200643,
      "tokens_seen": 990380032
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007070010030090271,
      "loss": 1.4977,
      "theoretical_loss": 3.680996193322682,
      "tokens_seen": 990445568
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007069809428284855,
      "loss": 1.5142,
      "theoretical_loss": 3.6809731533961765,
      "tokens_seen": 990511104
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007069608826479439,
      "loss": 1.6199,
      "theoretical_loss": 3.680950115420833,
      "tokens_seen": 990576640
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007069408224674022,
      "loss": 1.4717,
      "theoretical_loss": 3.680927079396357,
      "tokens_seen": 990642176
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007069207622868607,
      "loss": 1.4142,
      "theoretical_loss": 3.680904045322454,
      "tokens_seen": 990707712
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007069007021063189,
      "loss": 1.5144,
      "theoretical_loss": 3.68088101319883,
      "tokens_seen": 990773248
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007068806419257774,
      "loss": 1.4804,
      "theoretical_loss": 3.6808579830251906,
      "tokens_seen": 990838784
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007068605817452357,
      "loss": 1.5645,
      "theoretical_loss": 3.680834954801242,
      "tokens_seen": 990904320
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007068405215646941,
      "loss": 1.4957,
      "theoretical_loss": 3.6808119285266905,
      "tokens_seen": 990969856
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007068204613841525,
      "loss": 1.4432,
      "theoretical_loss": 3.6807889042012416,
      "tokens_seen": 991035392
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007068004012036109,
      "loss": 1.6071,
      "theoretical_loss": 3.6807658818246014,
      "tokens_seen": 991100928
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007067803410230692,
      "loss": 1.453,
      "theoretical_loss": 3.6807428613964763,
      "tokens_seen": 991166464
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007067602808425277,
      "loss": 1.4479,
      "theoretical_loss": 3.6807198429165733,
      "tokens_seen": 991232000
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007067402206619859,
      "loss": 1.8748,
      "theoretical_loss": 3.680696826384598,
      "tokens_seen": 991297536
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007067201604814444,
      "loss": 1.4207,
      "theoretical_loss": 3.680673811800256,
      "tokens_seen": 991363072
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007067001003009027,
      "loss": 1.5885,
      "theoretical_loss": 3.6806507991632555,
      "tokens_seen": 991428608
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007066800401203611,
      "loss": 1.572,
      "theoretical_loss": 3.680627788473302,
      "tokens_seen": 991494144
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007066599799398195,
      "loss": 1.4772,
      "theoretical_loss": 3.680604779730102,
      "tokens_seen": 991559680
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007066399197592779,
      "loss": 1.5203,
      "theoretical_loss": 3.6805817729333627,
      "tokens_seen": 991625216
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007066198595787362,
      "loss": 1.509,
      "theoretical_loss": 3.68055876808279,
      "tokens_seen": 991690752
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007065997993981946,
      "loss": 1.5223,
      "theoretical_loss": 3.6805357651780914,
      "tokens_seen": 991756288
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000706579739217653,
      "loss": 1.6316,
      "theoretical_loss": 3.6805127642189737,
      "tokens_seen": 991821824
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007065596790371114,
      "loss": 1.5876,
      "theoretical_loss": 3.6804897652051434,
      "tokens_seen": 991887360
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007065396188565697,
      "loss": 1.7304,
      "theoretical_loss": 3.680466768136308,
      "tokens_seen": 991952896
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007065195586760281,
      "loss": 1.6105,
      "theoretical_loss": 3.6804437730121737,
      "tokens_seen": 992018432
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007064994984954864,
      "loss": 1.6498,
      "theoretical_loss": 3.6804207798324486,
      "tokens_seen": 992083968
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007064794383149449,
      "loss": 1.4516,
      "theoretical_loss": 3.680397788596839,
      "tokens_seen": 992149504
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007064593781344032,
      "loss": 1.4879,
      "theoretical_loss": 3.680374799305053,
      "tokens_seen": 992215040
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007064393179538616,
      "loss": 1.5733,
      "theoretical_loss": 3.680351811956797,
      "tokens_seen": 992280576
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00070641925777332,
      "loss": 1.4569,
      "theoretical_loss": 3.6803288265517784,
      "tokens_seen": 992346112
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007063991975927784,
      "loss": 1.574,
      "theoretical_loss": 3.680305843089705,
      "tokens_seen": 992411648
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007063791374122367,
      "loss": 1.4519,
      "theoretical_loss": 3.6802828615702845,
      "tokens_seen": 992477184
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007063590772316952,
      "loss": 1.5341,
      "theoretical_loss": 3.6802598819932237,
      "tokens_seen": 992542720
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007063390170511534,
      "loss": 1.5512,
      "theoretical_loss": 3.680236904358231,
      "tokens_seen": 992608256
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007063189568706119,
      "loss": 1.6234,
      "theoretical_loss": 3.6802139286650135,
      "tokens_seen": 992673792
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007062988966900702,
      "loss": 1.5418,
      "theoretical_loss": 3.6801909549132796,
      "tokens_seen": 992739328
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007062788365095286,
      "loss": 1.4679,
      "theoretical_loss": 3.680167983102736,
      "tokens_seen": 992804864
    },
    {
      "epoch": 0.3,
      "objective/train/advantage_avg": 0.49160754680633545,
      "objective/train/docs_used": 365105,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.459633708000183,
      "objective/train/original_loss": 1.4596335887908936,
      "objective/train/theoretical_loss": 3.6801450132330915,
      "objective/train/tokens_used": 1013330400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2423485815525055,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050392746925354,
      "objective/train/weighted_lm_loss": 1.5332401990890503,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6801450132330915,
      "tokens_seen": 992870400
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000706258776328987,
      "loss": 1.4764,
      "theoretical_loss": 3.6801450132330915,
      "tokens_seen": 992870400
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007062387161484453,
      "loss": 1.6195,
      "theoretical_loss": 3.6801220453040537,
      "tokens_seen": 992935936
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007062186559679037,
      "loss": 1.6178,
      "theoretical_loss": 3.6800990793153305,
      "tokens_seen": 993001472
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007061985957873622,
      "loss": 1.6271,
      "theoretical_loss": 3.6800761152666297,
      "tokens_seen": 993067008
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007061785356068204,
      "loss": 1.5392,
      "theoretical_loss": 3.68005315315766,
      "tokens_seen": 993132544
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007061584754262789,
      "loss": 1.4305,
      "theoretical_loss": 3.680030192988129,
      "tokens_seen": 993198080
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007061384152457371,
      "loss": 1.5002,
      "theoretical_loss": 3.6800072347577455,
      "tokens_seen": 993263616
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007061183550651956,
      "loss": 1.555,
      "theoretical_loss": 3.6799842784662173,
      "tokens_seen": 993329152
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000706098294884654,
      "loss": 1.5633,
      "theoretical_loss": 3.679961324113253,
      "tokens_seen": 993394688
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007060782347041123,
      "loss": 1.481,
      "theoretical_loss": 3.6799383716985616,
      "tokens_seen": 993460224
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007060581745235707,
      "loss": 1.4651,
      "theoretical_loss": 3.6799154212218506,
      "tokens_seen": 993525760
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007060381143430292,
      "loss": 1.3451,
      "theoretical_loss": 3.6798924726828286,
      "tokens_seen": 993591296
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007060180541624874,
      "loss": 1.4654,
      "theoretical_loss": 3.6798695260812044,
      "tokens_seen": 993656832
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007059979939819459,
      "loss": 1.5252,
      "theoretical_loss": 3.679846581416687,
      "tokens_seen": 993722368
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007059779338014041,
      "loss": 1.5836,
      "theoretical_loss": 3.679823638688985,
      "tokens_seen": 993787904
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007059578736208626,
      "loss": 1.4332,
      "theoretical_loss": 3.6798006978978073,
      "tokens_seen": 993853440
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000705937813440321,
      "loss": 1.5144,
      "theoretical_loss": 3.679777759042863,
      "tokens_seen": 993918976
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007059177532597793,
      "loss": 1.4803,
      "theoretical_loss": 3.6797548221238596,
      "tokens_seen": 993984512
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007058976930792377,
      "loss": 1.6256,
      "theoretical_loss": 3.679731887140508,
      "tokens_seen": 994050048
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007058776328986961,
      "loss": 1.4678,
      "theoretical_loss": 3.679708954092516,
      "tokens_seen": 994115584
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007058575727181544,
      "loss": 1.5133,
      "theoretical_loss": 3.679686022979593,
      "tokens_seen": 994181120
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007058375125376129,
      "loss": 1.3787,
      "theoretical_loss": 3.679663093801448,
      "tokens_seen": 994246656
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007058174523570711,
      "loss": 1.4351,
      "theoretical_loss": 3.6796401665577916,
      "tokens_seen": 994312192
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007057973921765296,
      "loss": 1.6044,
      "theoretical_loss": 3.679617241248331,
      "tokens_seen": 994377728
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007057773319959879,
      "loss": 1.582,
      "theoretical_loss": 3.679594317872777,
      "tokens_seen": 994443264
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007057572718154463,
      "loss": 1.6171,
      "theoretical_loss": 3.6795713964308385,
      "tokens_seen": 994508800
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007057372116349047,
      "loss": 1.4581,
      "theoretical_loss": 3.679548476922225,
      "tokens_seen": 994574336
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007057171514543631,
      "loss": 1.5534,
      "theoretical_loss": 3.6795255593466463,
      "tokens_seen": 994639872
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007056970912738215,
      "loss": 1.5368,
      "theoretical_loss": 3.6795026437038114,
      "tokens_seen": 994705408
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007056770310932799,
      "loss": 1.4907,
      "theoretical_loss": 3.679479729993431,
      "tokens_seen": 994770944
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007056569709127383,
      "loss": 1.5039,
      "theoretical_loss": 3.6794568182152143,
      "tokens_seen": 994836480
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007056369107321966,
      "loss": 1.5446,
      "theoretical_loss": 3.679433908368871,
      "tokens_seen": 994902016
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000705616850551655,
      "loss": 1.6087,
      "theoretical_loss": 3.679411000454111,
      "tokens_seen": 994967552
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007055967903711133,
      "loss": 1.5004,
      "theoretical_loss": 3.679388094470644,
      "tokens_seen": 995033088
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007055767301905718,
      "loss": 1.5122,
      "theoretical_loss": 3.6793651904181806,
      "tokens_seen": 995098624
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007055566700100301,
      "loss": 1.5619,
      "theoretical_loss": 3.6793422882964313,
      "tokens_seen": 995164160
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007055366098294885,
      "loss": 1.5045,
      "theoretical_loss": 3.6793193881051045,
      "tokens_seen": 995229696
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007055165496489468,
      "loss": 1.5112,
      "theoretical_loss": 3.679296489843911,
      "tokens_seen": 995295232
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007054964894684053,
      "loss": 1.5126,
      "theoretical_loss": 3.679273593512563,
      "tokens_seen": 995360768
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007054764292878636,
      "loss": 1.4722,
      "theoretical_loss": 3.679250699110768,
      "tokens_seen": 995426304
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000705456369107322,
      "loss": 1.5626,
      "theoretical_loss": 3.6792278066382376,
      "tokens_seen": 995491840
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007054363089267804,
      "loss": 1.6477,
      "theoretical_loss": 3.6792049160946823,
      "tokens_seen": 995557376
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007054162487462388,
      "loss": 1.5863,
      "theoretical_loss": 3.679182027479812,
      "tokens_seen": 995622912
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007053961885656971,
      "loss": 1.5572,
      "theoretical_loss": 3.679159140793339,
      "tokens_seen": 995688448
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007053761283851556,
      "loss": 1.548,
      "theoretical_loss": 3.6791362560349716,
      "tokens_seen": 995753984
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007053560682046138,
      "loss": 1.6719,
      "theoretical_loss": 3.679113373204422,
      "tokens_seen": 995819520
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007053360080240723,
      "loss": 1.4669,
      "theoretical_loss": 3.6790904923014005,
      "tokens_seen": 995885056
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007053159478435306,
      "loss": 1.4069,
      "theoretical_loss": 3.6790676133256177,
      "tokens_seen": 995950592
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000705295887662989,
      "loss": 1.5524,
      "theoretical_loss": 3.679044736276785,
      "tokens_seen": 996016128
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007052758274824474,
      "loss": 1.5954,
      "theoretical_loss": 3.679021861154613,
      "tokens_seen": 996081664
    },
    {
      "epoch": 0.3,
      "objective/train/advantage_avg": 0.49079787731170654,
      "objective/train/docs_used": 366329,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.549461007118225,
      "objective/train/original_loss": 1.549461007118225,
      "objective/train/theoretical_loss": 3.6789989879588125,
      "objective/train/tokens_used": 1016607200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2415924221277237,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503077507019043,
      "objective/train/weighted_lm_loss": 1.6274405717849731,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.6789989879588125,
      "tokens_seen": 996147200
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007052557673019057,
      "loss": 1.4983,
      "theoretical_loss": 3.6789989879588125,
      "tokens_seen": 996147200
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007052357071213641,
      "loss": 1.5578,
      "theoretical_loss": 3.678976116689095,
      "tokens_seen": 996212736
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007052156469408226,
      "loss": 1.6397,
      "theoretical_loss": 3.678953247345171,
      "tokens_seen": 996278272
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007051955867602808,
      "loss": 1.3327,
      "theoretical_loss": 3.6789303799267525,
      "tokens_seen": 996343808
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007051755265797393,
      "loss": 1.5565,
      "theoretical_loss": 3.6789075144335497,
      "tokens_seen": 996409344
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007051554663991975,
      "loss": 1.3962,
      "theoretical_loss": 3.6788846508652746,
      "tokens_seen": 996474880
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000705135406218656,
      "loss": 1.4486,
      "theoretical_loss": 3.678861789221639,
      "tokens_seen": 996540416
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007051153460381144,
      "loss": 1.5258,
      "theoretical_loss": 3.6788389295023536,
      "tokens_seen": 996605952
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007050952858575727,
      "loss": 1.592,
      "theoretical_loss": 3.6788160717071303,
      "tokens_seen": 996671488
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007050752256770311,
      "loss": 1.632,
      "theoretical_loss": 3.6787932158356798,
      "tokens_seen": 996737024
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007050551654964896,
      "loss": 1.5806,
      "theoretical_loss": 3.678770361887715,
      "tokens_seen": 996802560
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007050351053159478,
      "loss": 1.4443,
      "theoretical_loss": 3.678747509862947,
      "tokens_seen": 996868096
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007050150451354063,
      "loss": 1.6311,
      "theoretical_loss": 3.678724659761087,
      "tokens_seen": 996933632
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007049949849548645,
      "loss": 1.5001,
      "theoretical_loss": 3.6787018115818473,
      "tokens_seen": 996999168
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000704974924774323,
      "loss": 1.5623,
      "theoretical_loss": 3.67867896532494,
      "tokens_seen": 997064704
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007049548645937814,
      "loss": 1.4993,
      "theoretical_loss": 3.6786561209900768,
      "tokens_seen": 997130240
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007049348044132397,
      "loss": 1.4157,
      "theoretical_loss": 3.6786332785769695,
      "tokens_seen": 997195776
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007049147442326981,
      "loss": 1.5275,
      "theoretical_loss": 3.6786104380853306,
      "tokens_seen": 997261312
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007048946840521565,
      "loss": 1.527,
      "theoretical_loss": 3.678587599514872,
      "tokens_seen": 997326848
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007048746238716148,
      "loss": 1.4997,
      "theoretical_loss": 3.6785647628653058,
      "tokens_seen": 997392384
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007048545636910733,
      "loss": 1.4203,
      "theoretical_loss": 3.678541928136344,
      "tokens_seen": 997457920
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007048345035105315,
      "loss": 1.5309,
      "theoretical_loss": 3.678519095327699,
      "tokens_seen": 997523456
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00070481444332999,
      "loss": 1.5585,
      "theoretical_loss": 3.6784962644390835,
      "tokens_seen": 997588992
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007047943831494483,
      "loss": 1.4765,
      "theoretical_loss": 3.6784734354702096,
      "tokens_seen": 997654528
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007047743229689067,
      "loss": 1.5925,
      "theoretical_loss": 3.6784506084207904,
      "tokens_seen": 997720064
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007047542627883651,
      "loss": 1.6224,
      "theoretical_loss": 3.6784277832905374,
      "tokens_seen": 997785600
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007047342026078235,
      "loss": 1.5528,
      "theoretical_loss": 3.678404960079164,
      "tokens_seen": 997851136
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007047141424272818,
      "loss": 1.6329,
      "theoretical_loss": 3.6783821387863833,
      "tokens_seen": 997916672
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007046940822467403,
      "loss": 1.5381,
      "theoretical_loss": 3.6783593194119066,
      "tokens_seen": 997982208
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007046740220661986,
      "loss": 1.6666,
      "theoretical_loss": 3.6783365019554477,
      "tokens_seen": 998047744
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000704653961885657,
      "loss": 1.3746,
      "theoretical_loss": 3.678313686416719,
      "tokens_seen": 998113280
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007046339017051153,
      "loss": 1.5315,
      "theoretical_loss": 3.678290872795434,
      "tokens_seen": 998178816
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007046138415245738,
      "loss": 1.6162,
      "theoretical_loss": 3.6782680610913054,
      "tokens_seen": 998244352
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007045937813440321,
      "loss": 1.6161,
      "theoretical_loss": 3.6782452513040464,
      "tokens_seen": 998309888
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007045737211634905,
      "loss": 1.601,
      "theoretical_loss": 3.678222443433369,
      "tokens_seen": 998375424
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007045536609829488,
      "loss": 1.6595,
      "theoretical_loss": 3.678199637478988,
      "tokens_seen": 998440960
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007045336008024072,
      "loss": 1.5734,
      "theoretical_loss": 3.6781768334406157,
      "tokens_seen": 998506496
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007045135406218656,
      "loss": 1.5491,
      "theoretical_loss": 3.6781540313179657,
      "tokens_seen": 998572032
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000704493480441324,
      "loss": 1.4846,
      "theoretical_loss": 3.6781312311107515,
      "tokens_seen": 998637568
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007044734202607823,
      "loss": 1.4137,
      "theoretical_loss": 3.6781084328186857,
      "tokens_seen": 998703104
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007044533600802408,
      "loss": 1.5448,
      "theoretical_loss": 3.678085636441482,
      "tokens_seen": 998768640
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007044332998996991,
      "loss": 1.4991,
      "theoretical_loss": 3.6780628419788552,
      "tokens_seen": 998834176
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007044132397191575,
      "loss": 1.6951,
      "theoretical_loss": 3.678040049430517,
      "tokens_seen": 998899712
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007043931795386158,
      "loss": 1.5587,
      "theoretical_loss": 3.678017258796183,
      "tokens_seen": 998965248
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007043731193580742,
      "loss": 1.6383,
      "theoretical_loss": 3.677994470075565,
      "tokens_seen": 999030784
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007043530591775326,
      "loss": 1.5078,
      "theoretical_loss": 3.677971683268378,
      "tokens_seen": 999096320
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000704332998996991,
      "loss": 1.5588,
      "theoretical_loss": 3.6779488983743356,
      "tokens_seen": 999161856
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007043129388164493,
      "loss": 1.502,
      "theoretical_loss": 3.677926115393152,
      "tokens_seen": 999227392
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007042928786359078,
      "loss": 1.5302,
      "theoretical_loss": 3.6779033343245406,
      "tokens_seen": 999292928
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000704272818455366,
      "loss": 1.5391,
      "theoretical_loss": 3.6778805551682154,
      "tokens_seen": 999358464
    },
    {
      "epoch": 0.3,
      "objective/train/advantage_avg": 0.49473220109939575,
      "objective/train/docs_used": 367579,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6029266119003296,
      "objective/train/original_loss": 1.6029267311096191,
      "objective/train/theoretical_loss": 3.677857777923891,
      "objective/train/tokens_used": 1019884000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24494512379169464,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507185459136963,
      "objective/train/weighted_lm_loss": 1.6841078996658325,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.677857777923891,
      "tokens_seen": 999424000
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007042527582748245,
      "loss": 1.5888,
      "theoretical_loss": 3.677857777923891,
      "tokens_seen": 999424000
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007042326980942828,
      "loss": 1.4645,
      "theoretical_loss": 3.677835002591282,
      "tokens_seen": 999489536
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007042126379137412,
      "loss": 1.55,
      "theoretical_loss": 3.677812229170101,
      "tokens_seen": 999555072
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007041925777331997,
      "loss": 1.5088,
      "theoretical_loss": 3.6777894576600634,
      "tokens_seen": 999620608
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007041725175526579,
      "loss": 1.4962,
      "theoretical_loss": 3.677766688060883,
      "tokens_seen": 999686144
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007041524573721164,
      "loss": 1.4662,
      "theoretical_loss": 3.677743920372275,
      "tokens_seen": 999751680
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007041323971915748,
      "loss": 1.4898,
      "theoretical_loss": 3.677721154593953,
      "tokens_seen": 999817216
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007041123370110331,
      "loss": 1.4759,
      "theoretical_loss": 3.6776983907256326,
      "tokens_seen": 999882752
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007040922768304915,
      "loss": 1.6207,
      "theoretical_loss": 3.677675628767027,
      "tokens_seen": 999948288
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00070407221664995,
      "loss": 1.4998,
      "theoretical_loss": 3.677652868717852,
      "tokens_seen": 1000013824
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007040521564694082,
      "loss": 1.5423,
      "theoretical_loss": 3.6776301105778213,
      "tokens_seen": 1000079360
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007040320962888667,
      "loss": 1.5253,
      "theoretical_loss": 3.6776073543466508,
      "tokens_seen": 1000144896
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007040120361083249,
      "loss": 1.5194,
      "theoretical_loss": 3.6775846000240544,
      "tokens_seen": 1000210432
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007039919759277834,
      "loss": 1.6286,
      "theoretical_loss": 3.6775618476097476,
      "tokens_seen": 1000275968
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007039719157472418,
      "loss": 1.4893,
      "theoretical_loss": 3.6775390971034447,
      "tokens_seen": 1000341504
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007039518555667001,
      "loss": 1.349,
      "theoretical_loss": 3.677516348504861,
      "tokens_seen": 1000407040
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007039317953861585,
      "loss": 1.598,
      "theoretical_loss": 3.6774936018137123,
      "tokens_seen": 1000472576
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007039117352056169,
      "loss": 1.4966,
      "theoretical_loss": 3.6774708570297125,
      "tokens_seen": 1000538112
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007038916750250752,
      "loss": 1.5815,
      "theoretical_loss": 3.6774481141525777,
      "tokens_seen": 1000603648
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007038716148445337,
      "loss": 1.6008,
      "theoretical_loss": 3.6774253731820226,
      "tokens_seen": 1000669184
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000703851554663992,
      "loss": 1.5522,
      "theoretical_loss": 3.6774026341177626,
      "tokens_seen": 1000734720
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007038314944834504,
      "loss": 1.4445,
      "theoretical_loss": 3.677379896959514,
      "tokens_seen": 1000800256
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007038114343029087,
      "loss": 1.5163,
      "theoretical_loss": 3.6773571617069907,
      "tokens_seen": 1000865792
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007037913741223671,
      "loss": 1.4514,
      "theoretical_loss": 3.6773344283599094,
      "tokens_seen": 1000931328
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007037713139418255,
      "loss": 1.5759,
      "theoretical_loss": 3.677311696917985,
      "tokens_seen": 1000996864
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007037512537612839,
      "loss": 1.4993,
      "theoretical_loss": 3.6772889673809335,
      "tokens_seen": 1001062400
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007037311935807422,
      "loss": 1.4143,
      "theoretical_loss": 3.6772662397484703,
      "tokens_seen": 1001127936
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007037111334002007,
      "loss": 1.4002,
      "theoretical_loss": 3.6772435140203115,
      "tokens_seen": 1001193472
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000703691073219659,
      "loss": 1.3682,
      "theoretical_loss": 3.6772207901961727,
      "tokens_seen": 1001259008
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007036710130391174,
      "loss": 1.4863,
      "theoretical_loss": 3.6771980682757697,
      "tokens_seen": 1001324544
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007036509528585757,
      "loss": 1.4758,
      "theoretical_loss": 3.6771753482588183,
      "tokens_seen": 1001390080
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007036308926780342,
      "loss": 1.5028,
      "theoretical_loss": 3.6771526301450352,
      "tokens_seen": 1001455616
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007036108324974925,
      "loss": 1.6226,
      "theoretical_loss": 3.6771299139341354,
      "tokens_seen": 1001521152
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007035907723169509,
      "loss": 1.4782,
      "theoretical_loss": 3.6771071996258358,
      "tokens_seen": 1001586688
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007035707121364092,
      "loss": 1.4387,
      "theoretical_loss": 3.6770844872198523,
      "tokens_seen": 1001652224
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007035506519558676,
      "loss": 1.5053,
      "theoretical_loss": 3.6770617767159006,
      "tokens_seen": 1001717760
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000703530591775326,
      "loss": 1.5385,
      "theoretical_loss": 3.6770390681136984,
      "tokens_seen": 1001783296
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007035105315947844,
      "loss": 1.4445,
      "theoretical_loss": 3.677016361412961,
      "tokens_seen": 1001848832
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007034904714142427,
      "loss": 1.5001,
      "theoretical_loss": 3.6769936566134045,
      "tokens_seen": 1001914368
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007034704112337012,
      "loss": 1.5338,
      "theoretical_loss": 3.676970953714746,
      "tokens_seen": 1001979904
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007034503510531595,
      "loss": 1.4642,
      "theoretical_loss": 3.676948252716702,
      "tokens_seen": 1002045440
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007034302908726179,
      "loss": 1.5481,
      "theoretical_loss": 3.6769255536189895,
      "tokens_seen": 1002110976
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007034102306920762,
      "loss": 1.6291,
      "theoretical_loss": 3.676902856421324,
      "tokens_seen": 1002176512
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007033901705115346,
      "loss": 1.4995,
      "theoretical_loss": 3.6768801611234236,
      "tokens_seen": 1002242048
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000703370110330993,
      "loss": 1.5448,
      "theoretical_loss": 3.6768574677250037,
      "tokens_seen": 1002307584
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007033500501504514,
      "loss": 1.6575,
      "theoretical_loss": 3.6768347762257827,
      "tokens_seen": 1002373120
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007033299899699097,
      "loss": 1.5103,
      "theoretical_loss": 3.6768120866254757,
      "tokens_seen": 1002438656
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007033099297893682,
      "loss": 1.4405,
      "theoretical_loss": 3.676789398923801,
      "tokens_seen": 1002504192
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007032898696088264,
      "loss": 1.6058,
      "theoretical_loss": 3.6767667131204753,
      "tokens_seen": 1002569728
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007032698094282849,
      "loss": 1.4563,
      "theoretical_loss": 3.6767440292152154,
      "tokens_seen": 1002635264
    },
    {
      "epoch": 0.3,
      "objective/train/advantage_avg": 0.4905562102794647,
      "objective/train/docs_used": 368716,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.53972327709198,
      "objective/train/original_loss": 1.5397231578826904,
      "objective/train/theoretical_loss": 3.6767213472077387,
      "objective/train/tokens_used": 1023160800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2414381355047226,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502829551696777,
      "objective/train/weighted_lm_loss": 1.6172175407409668,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6767213472077387,
      "tokens_seen": 1002700800
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007032497492477432,
      "loss": 1.5689,
      "theoretical_loss": 3.6767213472077387,
      "tokens_seen": 1002700800
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007032296890672016,
      "loss": 1.5244,
      "theoretical_loss": 3.676698667097762,
      "tokens_seen": 1002766336
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00070320962888666,
      "loss": 1.3898,
      "theoretical_loss": 3.6766759888850036,
      "tokens_seen": 1002831872
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007031895687061183,
      "loss": 1.511,
      "theoretical_loss": 3.6766533125691794,
      "tokens_seen": 1002897408
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007031695085255767,
      "loss": 1.5228,
      "theoretical_loss": 3.676630638150008,
      "tokens_seen": 1002962944
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007031494483450352,
      "loss": 1.4965,
      "theoretical_loss": 3.676607965627206,
      "tokens_seen": 1003028480
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007031293881644934,
      "loss": 1.4587,
      "theoretical_loss": 3.6765852950004914,
      "tokens_seen": 1003094016
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007031093279839519,
      "loss": 1.4355,
      "theoretical_loss": 3.6765626262695816,
      "tokens_seen": 1003159552
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007030892678034103,
      "loss": 1.5039,
      "theoretical_loss": 3.6765399594341943,
      "tokens_seen": 1003225088
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007030692076228686,
      "loss": 1.5981,
      "theoretical_loss": 3.6765172944940474,
      "tokens_seen": 1003290624
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000703049147442327,
      "loss": 1.5812,
      "theoretical_loss": 3.6764946314488585,
      "tokens_seen": 1003356160
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007030290872617853,
      "loss": 1.5572,
      "theoretical_loss": 3.676471970298345,
      "tokens_seen": 1003421696
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007030090270812437,
      "loss": 1.4582,
      "theoretical_loss": 3.676449311042225,
      "tokens_seen": 1003487232
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007029889669007022,
      "loss": 1.5977,
      "theoretical_loss": 3.676426653680217,
      "tokens_seen": 1003552768
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007029689067201604,
      "loss": 1.5765,
      "theoretical_loss": 3.676403998212038,
      "tokens_seen": 1003618304
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007029488465396189,
      "loss": 1.6281,
      "theoretical_loss": 3.6763813446374067,
      "tokens_seen": 1003683840
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007029287863590772,
      "loss": 1.6223,
      "theoretical_loss": 3.6763586929560415,
      "tokens_seen": 1003749376
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007029087261785356,
      "loss": 1.524,
      "theoretical_loss": 3.6763360431676597,
      "tokens_seen": 1003814912
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000702888665997994,
      "loss": 1.5624,
      "theoretical_loss": 3.67631339527198,
      "tokens_seen": 1003880448
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007028686058174523,
      "loss": 1.5928,
      "theoretical_loss": 3.6762907492687207,
      "tokens_seen": 1003945984
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007028485456369107,
      "loss": 1.5105,
      "theoretical_loss": 3.6762681051576003,
      "tokens_seen": 1004011520
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007028284854563691,
      "loss": 1.4234,
      "theoretical_loss": 3.6762454629383368,
      "tokens_seen": 1004077056
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007028084252758274,
      "loss": 1.4384,
      "theoretical_loss": 3.676222822610649,
      "tokens_seen": 1004142592
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007027883650952859,
      "loss": 1.5835,
      "theoretical_loss": 3.6762001841742555,
      "tokens_seen": 1004208128
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007027683049147442,
      "loss": 1.6196,
      "theoretical_loss": 3.6761775476288747,
      "tokens_seen": 1004273664
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007027482447342026,
      "loss": 1.3853,
      "theoretical_loss": 3.676154912974225,
      "tokens_seen": 1004339200
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000702728184553661,
      "loss": 1.5688,
      "theoretical_loss": 3.676132280210026,
      "tokens_seen": 1004404736
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007027081243731194,
      "loss": 1.5274,
      "theoretical_loss": 3.6761096493359955,
      "tokens_seen": 1004470272
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007026880641925778,
      "loss": 1.6464,
      "theoretical_loss": 3.6760870203518525,
      "tokens_seen": 1004535808
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007026680040120361,
      "loss": 1.5164,
      "theoretical_loss": 3.6760643932573167,
      "tokens_seen": 1004601344
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007026479438314946,
      "loss": 1.4398,
      "theoretical_loss": 3.676041768052106,
      "tokens_seen": 1004666880
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007026278836509529,
      "loss": 1.4357,
      "theoretical_loss": 3.6760191447359403,
      "tokens_seen": 1004732416
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007026078234704113,
      "loss": 1.4326,
      "theoretical_loss": 3.6759965233085383,
      "tokens_seen": 1004797952
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007025877632898696,
      "loss": 1.4622,
      "theoretical_loss": 3.675973903769619,
      "tokens_seen": 1004863488
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000702567703109328,
      "loss": 1.6672,
      "theoretical_loss": 3.6759512861189014,
      "tokens_seen": 1004929024
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007025476429287864,
      "loss": 1.4631,
      "theoretical_loss": 3.675928670356105,
      "tokens_seen": 1004994560
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007025275827482448,
      "loss": 1.5252,
      "theoretical_loss": 3.6759060564809496,
      "tokens_seen": 1005060096
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007025075225677031,
      "loss": 1.5008,
      "theoretical_loss": 3.675883444493154,
      "tokens_seen": 1005125632
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007024874623871616,
      "loss": 1.5267,
      "theoretical_loss": 3.675860834392438,
      "tokens_seen": 1005191168
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007024674022066199,
      "loss": 1.5296,
      "theoretical_loss": 3.6758382261785205,
      "tokens_seen": 1005256704
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007024473420260783,
      "loss": 1.5513,
      "theoretical_loss": 3.6758156198511216,
      "tokens_seen": 1005322240
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007024272818455366,
      "loss": 1.6424,
      "theoretical_loss": 3.675793015409961,
      "tokens_seen": 1005387776
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000702407221664995,
      "loss": 1.5271,
      "theoretical_loss": 3.675770412854758,
      "tokens_seen": 1005453312
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007023871614844534,
      "loss": 1.5116,
      "theoretical_loss": 3.6757478121852323,
      "tokens_seen": 1005518848
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007023671013039118,
      "loss": 1.5236,
      "theoretical_loss": 3.675725213401104,
      "tokens_seen": 1005584384
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007023470411233701,
      "loss": 1.689,
      "theoretical_loss": 3.6757026165020927,
      "tokens_seen": 1005649920
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007023269809428286,
      "loss": 1.4822,
      "theoretical_loss": 3.675680021487919,
      "tokens_seen": 1005715456
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007023069207622868,
      "loss": 1.603,
      "theoretical_loss": 3.675657428358302,
      "tokens_seen": 1005780992
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007022868605817453,
      "loss": 1.4438,
      "theoretical_loss": 3.6756348371129617,
      "tokens_seen": 1005846528
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007022668004012037,
      "loss": 1.5488,
      "theoretical_loss": 3.6756122477516184,
      "tokens_seen": 1005912064
    },
    {
      "epoch": 0.3,
      "objective/train/advantage_avg": 0.49357861280441284,
      "objective/train/docs_used": 369909,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5831763744354248,
      "objective/train/original_loss": 1.5831763744354248,
      "objective/train/theoretical_loss": 3.6755896602739933,
      "objective/train/tokens_used": 1026437600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24397243559360504,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05059814453125,
      "objective/train/weighted_lm_loss": 1.6631537675857544,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.6755896602739933,
      "tokens_seen": 1005977600
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000702246740220662,
      "loss": 1.6501,
      "theoretical_loss": 3.6755896602739933,
      "tokens_seen": 1005977600
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007022266800401204,
      "loss": 1.5701,
      "theoretical_loss": 3.675567074679805,
      "tokens_seen": 1006043136
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007022066198595787,
      "loss": 1.5611,
      "theoretical_loss": 3.6755444909687744,
      "tokens_seen": 1006108672
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007021865596790371,
      "loss": 1.5434,
      "theoretical_loss": 3.6755219091406226,
      "tokens_seen": 1006174208
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007021664994984956,
      "loss": 1.4719,
      "theoretical_loss": 3.675499329195069,
      "tokens_seen": 1006239744
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007021464393179538,
      "loss": 1.4943,
      "theoretical_loss": 3.6754767511318347,
      "tokens_seen": 1006305280
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007021263791374123,
      "loss": 1.5805,
      "theoretical_loss": 3.675454174950639,
      "tokens_seen": 1006370816
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0007021063189568707,
      "loss": 1.5032,
      "theoretical_loss": 3.675431600651204,
      "tokens_seen": 1006436352
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000702086258776329,
      "loss": 1.4828,
      "theoretical_loss": 3.6754090282332497,
      "tokens_seen": 1006501888
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007020661985957874,
      "loss": 1.4529,
      "theoretical_loss": 3.6753864576964967,
      "tokens_seen": 1006567424
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007020461384152457,
      "loss": 1.5744,
      "theoretical_loss": 3.675363889040666,
      "tokens_seen": 1006632960
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007020260782347041,
      "loss": 1.4183,
      "theoretical_loss": 3.675341322265478,
      "tokens_seen": 1006698496
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007020060180541626,
      "loss": 1.5818,
      "theoretical_loss": 3.675318757370654,
      "tokens_seen": 1006764032
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007019859578736208,
      "loss": 1.601,
      "theoretical_loss": 3.675296194355915,
      "tokens_seen": 1006829568
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007019658976930793,
      "loss": 1.524,
      "theoretical_loss": 3.675273633220981,
      "tokens_seen": 1006895104
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007019458375125376,
      "loss": 1.4939,
      "theoretical_loss": 3.6752510739655744,
      "tokens_seen": 1006960640
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000701925777331996,
      "loss": 1.5708,
      "theoretical_loss": 3.675228516589416,
      "tokens_seen": 1007026176
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007019057171514544,
      "loss": 1.4934,
      "theoretical_loss": 3.6752059610922263,
      "tokens_seen": 1007091712
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007018856569709127,
      "loss": 1.5399,
      "theoretical_loss": 3.6751834074737264,
      "tokens_seen": 1007157248
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007018655967903711,
      "loss": 1.4695,
      "theoretical_loss": 3.6751608557336386,
      "tokens_seen": 1007222784
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007018455366098295,
      "loss": 1.6317,
      "theoretical_loss": 3.675138305871684,
      "tokens_seen": 1007288320
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007018254764292878,
      "loss": 1.5978,
      "theoretical_loss": 3.6751157578875837,
      "tokens_seen": 1007353856
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007018054162487463,
      "loss": 1.5243,
      "theoretical_loss": 3.675093211781059,
      "tokens_seen": 1007419392
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007017853560682046,
      "loss": 1.4755,
      "theoretical_loss": 3.6750706675518314,
      "tokens_seen": 1007484928
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000701765295887663,
      "loss": 1.6956,
      "theoretical_loss": 3.6750481251996234,
      "tokens_seen": 1007550464
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007017452357071214,
      "loss": 1.5832,
      "theoretical_loss": 3.6750255847241555,
      "tokens_seen": 1007616000
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007017251755265798,
      "loss": 1.5781,
      "theoretical_loss": 3.67500304612515,
      "tokens_seen": 1007681536
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007017051153460381,
      "loss": 1.6108,
      "theoretical_loss": 3.674980509402329,
      "tokens_seen": 1007747072
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007016850551654965,
      "loss": 1.5358,
      "theoretical_loss": 3.674957974555413,
      "tokens_seen": 1007812608
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007016649949849548,
      "loss": 1.5525,
      "theoretical_loss": 3.6749354415841253,
      "tokens_seen": 1007878144
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007016449348044133,
      "loss": 1.4956,
      "theoretical_loss": 3.674912910488187,
      "tokens_seen": 1007943680
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007016248746238716,
      "loss": 1.4714,
      "theoretical_loss": 3.674890381267321,
      "tokens_seen": 1008009216
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00070160481444333,
      "loss": 1.656,
      "theoretical_loss": 3.6748678539212483,
      "tokens_seen": 1008074752
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007015847542627883,
      "loss": 1.6167,
      "theoretical_loss": 3.674845328449691,
      "tokens_seen": 1008140288
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007015646940822468,
      "loss": 1.4724,
      "theoretical_loss": 3.6748228048523726,
      "tokens_seen": 1008205824
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007015446339017051,
      "loss": 1.5701,
      "theoretical_loss": 3.6748002831290143,
      "tokens_seen": 1008271360
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007015245737211635,
      "loss": 1.651,
      "theoretical_loss": 3.674777763279338,
      "tokens_seen": 1008336896
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007015045135406218,
      "loss": 1.5809,
      "theoretical_loss": 3.674755245303068,
      "tokens_seen": 1008402432
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007014844533600803,
      "loss": 1.5914,
      "theoretical_loss": 3.674732729199924,
      "tokens_seen": 1008467968
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007014643931795386,
      "loss": 1.4676,
      "theoretical_loss": 3.6747102149696307,
      "tokens_seen": 1008533504
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000701444332998997,
      "loss": 1.4777,
      "theoretical_loss": 3.6746877026119096,
      "tokens_seen": 1008599040
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007014242728184553,
      "loss": 1.503,
      "theoretical_loss": 3.674665192126483,
      "tokens_seen": 1008664576
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007014042126379138,
      "loss": 1.6068,
      "theoretical_loss": 3.674642683513074,
      "tokens_seen": 1008730112
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007013841524573721,
      "loss": 1.4877,
      "theoretical_loss": 3.674620176771406,
      "tokens_seen": 1008795648
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007013640922768305,
      "loss": 1.6477,
      "theoretical_loss": 3.674597671901201,
      "tokens_seen": 1008861184
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007013440320962889,
      "loss": 1.5033,
      "theoretical_loss": 3.6745751689021815,
      "tokens_seen": 1008926720
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007013239719157472,
      "loss": 1.6025,
      "theoretical_loss": 3.674552667774071,
      "tokens_seen": 1008992256
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007013039117352056,
      "loss": 1.4483,
      "theoretical_loss": 3.6745301685165925,
      "tokens_seen": 1009057792
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000701283851554664,
      "loss": 1.4369,
      "theoretical_loss": 3.6745076711294686,
      "tokens_seen": 1009123328
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007012637913741223,
      "loss": 1.5422,
      "theoretical_loss": 3.6744851756124226,
      "tokens_seen": 1009188864
    },
    {
      "epoch": 0.31,
      "objective/train/advantage_avg": 0.49197760224342346,
      "objective/train/docs_used": 370943,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5528076887130737,
      "objective/train/original_loss": 1.5528078079223633,
      "objective/train/theoretical_loss": 3.6744626819651773,
      "objective/train/tokens_used": 1029714400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24257339537143707,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504308938980103,
      "objective/train/weighted_lm_loss": 1.6308166980743408,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6744626819651773,
      "tokens_seen": 1009254400
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007012437311935808,
      "loss": 1.546,
      "theoretical_loss": 3.6744626819651773,
      "tokens_seen": 1009254400
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000701223671013039,
      "loss": 1.5858,
      "theoretical_loss": 3.6744401901874566,
      "tokens_seen": 1009319936
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007012036108324975,
      "loss": 1.5433,
      "theoretical_loss": 3.674417700278983,
      "tokens_seen": 1009385472
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000701183550651956,
      "loss": 1.5089,
      "theoretical_loss": 3.67439521223948,
      "tokens_seen": 1009451008
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007011634904714142,
      "loss": 1.4545,
      "theoretical_loss": 3.674372726068671,
      "tokens_seen": 1009516544
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007011434302908727,
      "loss": 1.5082,
      "theoretical_loss": 3.6743502417662803,
      "tokens_seen": 1009582080
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007011233701103311,
      "loss": 1.591,
      "theoretical_loss": 3.67432775933203,
      "tokens_seen": 1009647616
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007011033099297894,
      "loss": 1.419,
      "theoretical_loss": 3.674305278765644,
      "tokens_seen": 1009713152
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007010832497492478,
      "loss": 1.5132,
      "theoretical_loss": 3.6742828000668464,
      "tokens_seen": 1009778688
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007010631895687061,
      "loss": 1.5282,
      "theoretical_loss": 3.6742603232353606,
      "tokens_seen": 1009844224
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007010431293881645,
      "loss": 1.4752,
      "theoretical_loss": 3.6742378482709093,
      "tokens_seen": 1009909760
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000701023069207623,
      "loss": 1.5273,
      "theoretical_loss": 3.674215375173218,
      "tokens_seen": 1009975296
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007010030090270812,
      "loss": 1.5562,
      "theoretical_loss": 3.6741929039420103,
      "tokens_seen": 1010040832
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007009829488465397,
      "loss": 1.5619,
      "theoretical_loss": 3.6741704345770088,
      "tokens_seen": 1010106368
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000700962888665998,
      "loss": 1.475,
      "theoretical_loss": 3.6741479670779387,
      "tokens_seen": 1010171904
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007009428284854564,
      "loss": 1.5155,
      "theoretical_loss": 3.6741255014445233,
      "tokens_seen": 1010237440
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007009227683049148,
      "loss": 1.4929,
      "theoretical_loss": 3.6741030376764865,
      "tokens_seen": 1010302976
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007009027081243731,
      "loss": 1.5483,
      "theoretical_loss": 3.6740805757735533,
      "tokens_seen": 1010368512
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007008826479438315,
      "loss": 1.5221,
      "theoretical_loss": 3.6740581157354475,
      "tokens_seen": 1010434048
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007008625877632899,
      "loss": 1.6347,
      "theoretical_loss": 3.674035657561893,
      "tokens_seen": 1010499584
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007008425275827482,
      "loss": 1.5191,
      "theoretical_loss": 3.674013201252614,
      "tokens_seen": 1010565120
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007008224674022067,
      "loss": 1.588,
      "theoretical_loss": 3.6739907468073354,
      "tokens_seen": 1010630656
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000700802407221665,
      "loss": 1.545,
      "theoretical_loss": 3.6739682942257814,
      "tokens_seen": 1010696192
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007007823470411234,
      "loss": 1.4837,
      "theoretical_loss": 3.673945843507677,
      "tokens_seen": 1010761728
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007007622868605818,
      "loss": 1.5084,
      "theoretical_loss": 3.6739233946527454,
      "tokens_seen": 1010827264
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007007422266800402,
      "loss": 1.4455,
      "theoretical_loss": 3.6739009476607123,
      "tokens_seen": 1010892800
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007007221664994985,
      "loss": 1.4828,
      "theoretical_loss": 3.673878502531302,
      "tokens_seen": 1010958336
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007007021063189569,
      "loss": 1.4814,
      "theoretical_loss": 3.6738560592642395,
      "tokens_seen": 1011023872
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007006820461384152,
      "loss": 1.53,
      "theoretical_loss": 3.6738336178592492,
      "tokens_seen": 1011089408
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007006619859578737,
      "loss": 1.667,
      "theoretical_loss": 3.6738111783160567,
      "tokens_seen": 1011154944
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000700641925777332,
      "loss": 1.5704,
      "theoretical_loss": 3.6737887406343854,
      "tokens_seen": 1011220480
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007006218655967904,
      "loss": 1.6325,
      "theoretical_loss": 3.6737663048139613,
      "tokens_seen": 1011286016
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007006018054162487,
      "loss": 1.5977,
      "theoretical_loss": 3.6737438708545094,
      "tokens_seen": 1011351552
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007005817452357072,
      "loss": 1.5822,
      "theoretical_loss": 3.673721438755755,
      "tokens_seen": 1011417088
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007005616850551655,
      "loss": 1.4895,
      "theoretical_loss": 3.673699008517422,
      "tokens_seen": 1011482624
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007005416248746239,
      "loss": 1.5886,
      "theoretical_loss": 3.6736765801392366,
      "tokens_seen": 1011548160
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007005215646940822,
      "loss": 1.4549,
      "theoretical_loss": 3.673654153620924,
      "tokens_seen": 1011613696
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007005015045135407,
      "loss": 1.6169,
      "theoretical_loss": 3.673631728962209,
      "tokens_seen": 1011679232
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000700481444332999,
      "loss": 1.5835,
      "theoretical_loss": 3.6736093061628177,
      "tokens_seen": 1011744768
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007004613841524574,
      "loss": 1.5883,
      "theoretical_loss": 3.673586885222475,
      "tokens_seen": 1011810304
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007004413239719157,
      "loss": 1.6057,
      "theoretical_loss": 3.673564466140906,
      "tokens_seen": 1011875840
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007004212637913742,
      "loss": 1.5115,
      "theoretical_loss": 3.673542048917837,
      "tokens_seen": 1011941376
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007004012036108325,
      "loss": 1.4975,
      "theoretical_loss": 3.6735196335529934,
      "tokens_seen": 1012006912
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007003811434302909,
      "loss": 1.5642,
      "theoretical_loss": 3.6734972200461007,
      "tokens_seen": 1012072448
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007003610832497493,
      "loss": 1.5226,
      "theoretical_loss": 3.6734748083968842,
      "tokens_seen": 1012137984
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007003410230692076,
      "loss": 1.4878,
      "theoretical_loss": 3.6734523986050704,
      "tokens_seen": 1012203520
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000700320962888666,
      "loss": 1.5178,
      "theoretical_loss": 3.673429990670385,
      "tokens_seen": 1012269056
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007003009027081245,
      "loss": 1.4499,
      "theoretical_loss": 3.673407584592553,
      "tokens_seen": 1012334592
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007002808425275827,
      "loss": 1.7169,
      "theoretical_loss": 3.6733851803713016,
      "tokens_seen": 1012400128
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007002607823470412,
      "loss": 1.5403,
      "theoretical_loss": 3.6733627780063562,
      "tokens_seen": 1012465664
    },
    {
      "epoch": 0.31,
      "objective/train/advantage_avg": 0.48277798295021057,
      "objective/train/docs_used": 372166,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.484204649925232,
      "objective/train/original_loss": 1.4842047691345215,
      "objective/train/theoretical_loss": 3.6733403774974427,
      "objective/train/tokens_used": 1032991200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23505361378192902,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.049472451210022,
      "objective/train/weighted_lm_loss": 1.5566273927688599,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.6733403774974427,
      "tokens_seen": 1012531200
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007002407221664994,
      "loss": 1.5901,
      "theoretical_loss": 3.6733403774974427,
      "tokens_seen": 1012531200
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007002206619859579,
      "loss": 1.5397,
      "theoretical_loss": 3.6733179788442873,
      "tokens_seen": 1012596736
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007002006018054163,
      "loss": 1.5415,
      "theoretical_loss": 3.673295582046616,
      "tokens_seen": 1012662272
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007001805416248746,
      "loss": 1.5434,
      "theoretical_loss": 3.6732731871041557,
      "tokens_seen": 1012727808
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000700160481444333,
      "loss": 1.6314,
      "theoretical_loss": 3.6732507940166323,
      "tokens_seen": 1012793344
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007001404212637915,
      "loss": 1.5295,
      "theoretical_loss": 3.6732284027837716,
      "tokens_seen": 1012858880
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007001203610832497,
      "loss": 1.4725,
      "theoretical_loss": 3.6732060134053013,
      "tokens_seen": 1012924416
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007001003009027082,
      "loss": 1.5402,
      "theoretical_loss": 3.673183625880946,
      "tokens_seen": 1012989952
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007000802407221664,
      "loss": 1.5498,
      "theoretical_loss": 3.673161240210434,
      "tokens_seen": 1013055488
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007000601805416249,
      "loss": 1.5106,
      "theoretical_loss": 3.673138856393491,
      "tokens_seen": 1013121024
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007000401203610833,
      "loss": 1.5873,
      "theoretical_loss": 3.673116474429844,
      "tokens_seen": 1013186560
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007000200601805416,
      "loss": 1.627,
      "theoretical_loss": 3.6730940943192194,
      "tokens_seen": 1013252096
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0007,
      "loss": 1.4893,
      "theoretical_loss": 3.6730717160613437,
      "tokens_seen": 1013317632
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006999799398194584,
      "loss": 1.6312,
      "theoretical_loss": 3.6730493396559445,
      "tokens_seen": 1013383168
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006999598796389167,
      "loss": 1.602,
      "theoretical_loss": 3.673026965102748,
      "tokens_seen": 1013448704
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006999398194583752,
      "loss": 1.5063,
      "theoretical_loss": 3.6730045924014814,
      "tokens_seen": 1013514240
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006999197592778334,
      "loss": 1.5686,
      "theoretical_loss": 3.6729822215518713,
      "tokens_seen": 1013579776
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006998996990972919,
      "loss": 1.4278,
      "theoretical_loss": 3.672959852553645,
      "tokens_seen": 1013645312
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006998796389167502,
      "loss": 1.5401,
      "theoretical_loss": 3.67293748540653,
      "tokens_seen": 1013710848
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006998595787362086,
      "loss": 1.4658,
      "theoretical_loss": 3.672915120110253,
      "tokens_seen": 1013776384
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000699839518555667,
      "loss": 1.4574,
      "theoretical_loss": 3.6728927566645413,
      "tokens_seen": 1013841920
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006998194583751254,
      "loss": 1.4926,
      "theoretical_loss": 3.6728703950691224,
      "tokens_seen": 1013907456
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006997993981945837,
      "loss": 1.6022,
      "theoretical_loss": 3.672848035323723,
      "tokens_seen": 1013972992
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006997793380140422,
      "loss": 1.5332,
      "theoretical_loss": 3.6728256774280714,
      "tokens_seen": 1014038528
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006997592778335004,
      "loss": 1.5277,
      "theoretical_loss": 3.672803321381894,
      "tokens_seen": 1014104064
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006997392176529589,
      "loss": 1.5636,
      "theoretical_loss": 3.672780967184919,
      "tokens_seen": 1014169600
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006997191574724172,
      "loss": 1.4783,
      "theoretical_loss": 3.6727586148368743,
      "tokens_seen": 1014235136
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006996990972918756,
      "loss": 1.4591,
      "theoretical_loss": 3.6727362643374866,
      "tokens_seen": 1014300672
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006996790371113341,
      "loss": 1.5983,
      "theoretical_loss": 3.672713915686484,
      "tokens_seen": 1014366208
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006996589769307924,
      "loss": 1.6144,
      "theoretical_loss": 3.672691568883594,
      "tokens_seen": 1014431744
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006996389167502508,
      "loss": 1.4741,
      "theoretical_loss": 3.672669223928545,
      "tokens_seen": 1014497280
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006996188565697091,
      "loss": 1.5191,
      "theoretical_loss": 3.6726468808210644,
      "tokens_seen": 1014562816
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006995987963891676,
      "loss": 1.4376,
      "theoretical_loss": 3.6726245395608803,
      "tokens_seen": 1014628352
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006995787362086259,
      "loss": 1.4703,
      "theoretical_loss": 3.6726022001477205,
      "tokens_seen": 1014693888
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006995586760280843,
      "loss": 1.6371,
      "theoretical_loss": 3.672579862581313,
      "tokens_seen": 1014759424
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006995386158475426,
      "loss": 1.4595,
      "theoretical_loss": 3.672557526861386,
      "tokens_seen": 1014824960
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006995185556670011,
      "loss": 1.6451,
      "theoretical_loss": 3.6725351929876675,
      "tokens_seen": 1014890496
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006994984954864594,
      "loss": 1.4544,
      "theoretical_loss": 3.672512860959886,
      "tokens_seen": 1014956032
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006994784353059178,
      "loss": 1.5339,
      "theoretical_loss": 3.672490530777769,
      "tokens_seen": 1015021568
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006994583751253761,
      "loss": 1.5035,
      "theoretical_loss": 3.672468202441046,
      "tokens_seen": 1015087104
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006994383149448346,
      "loss": 1.5649,
      "theoretical_loss": 3.6724458759494443,
      "tokens_seen": 1015152640
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006994182547642929,
      "loss": 1.5112,
      "theoretical_loss": 3.672423551302693,
      "tokens_seen": 1015218176
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006993981945837513,
      "loss": 1.5527,
      "theoretical_loss": 3.6724012285005196,
      "tokens_seen": 1015283712
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006993781344032097,
      "loss": 1.5171,
      "theoretical_loss": 3.6723789075426545,
      "tokens_seen": 1015349248
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000699358074222668,
      "loss": 1.5402,
      "theoretical_loss": 3.6723565884288245,
      "tokens_seen": 1015414784
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006993380140421264,
      "loss": 1.5737,
      "theoretical_loss": 3.6723342711587588,
      "tokens_seen": 1015480320
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006993179538615849,
      "loss": 1.578,
      "theoretical_loss": 3.6723119557321864,
      "tokens_seen": 1015545856
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006992978936810431,
      "loss": 1.4938,
      "theoretical_loss": 3.672289642148836,
      "tokens_seen": 1015611392
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006992778335005016,
      "loss": 1.5196,
      "theoretical_loss": 3.6722673304084363,
      "tokens_seen": 1015676928
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006992577733199598,
      "loss": 1.6388,
      "theoretical_loss": 3.672245020510716,
      "tokens_seen": 1015742464
    },
    {
      "epoch": 0.31,
      "objective/train/advantage_avg": 0.49098140001296997,
      "objective/train/docs_used": 372984,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4237624406814575,
      "objective/train/original_loss": 1.423762559890747,
      "objective/train/theoretical_loss": 3.6722227124554045,
      "objective/train/tokens_used": 1036268000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24208305776119232,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503287315368652,
      "objective/train/weighted_lm_loss": 1.4956014156341553,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6722227124554045,
      "tokens_seen": 1015808000
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006992377131394183,
      "loss": 1.5225,
      "theoretical_loss": 3.6722227124554045,
      "tokens_seen": 1015808000
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006992176529588767,
      "loss": 1.4619,
      "theoretical_loss": 3.67220040624223,
      "tokens_seen": 1015873536
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000699197592778335,
      "loss": 1.5031,
      "theoretical_loss": 3.6721781018709225,
      "tokens_seen": 1015939072
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006991775325977934,
      "loss": 1.5608,
      "theoretical_loss": 3.6721557993412106,
      "tokens_seen": 1016004608
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006991574724172519,
      "loss": 1.4036,
      "theoretical_loss": 3.6721334986528236,
      "tokens_seen": 1016070144
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006991374122367101,
      "loss": 1.5446,
      "theoretical_loss": 3.6721111998054914,
      "tokens_seen": 1016135680
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006991173520561686,
      "loss": 1.6334,
      "theoretical_loss": 3.672088902798942,
      "tokens_seen": 1016201216
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006990972918756268,
      "loss": 1.6013,
      "theoretical_loss": 3.672066607632905,
      "tokens_seen": 1016266752
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006990772316950853,
      "loss": 1.4772,
      "theoretical_loss": 3.6720443143071106,
      "tokens_seen": 1016332288
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006990571715145437,
      "loss": 1.586,
      "theoretical_loss": 3.672022022821287,
      "tokens_seen": 1016397824
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000699037111334002,
      "loss": 1.5222,
      "theoretical_loss": 3.6719997331751655,
      "tokens_seen": 1016463360
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006990170511534604,
      "loss": 1.4716,
      "theoretical_loss": 3.671977445368474,
      "tokens_seen": 1016528896
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006989969909729188,
      "loss": 1.4877,
      "theoretical_loss": 3.671955159400943,
      "tokens_seen": 1016594432
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006989769307923771,
      "loss": 1.5579,
      "theoretical_loss": 3.6719328752723026,
      "tokens_seen": 1016659968
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006989568706118356,
      "loss": 1.5866,
      "theoretical_loss": 3.6719105929822815,
      "tokens_seen": 1016725504
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006989368104312938,
      "loss": 1.639,
      "theoretical_loss": 3.6718883125306094,
      "tokens_seen": 1016791040
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006989167502507523,
      "loss": 1.4476,
      "theoretical_loss": 3.6718660339170173,
      "tokens_seen": 1016856576
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006988966900702106,
      "loss": 1.5643,
      "theoretical_loss": 3.671843757141234,
      "tokens_seen": 1016922112
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000698876629889669,
      "loss": 1.6106,
      "theoretical_loss": 3.671821482202991,
      "tokens_seen": 1016987648
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006988565697091274,
      "loss": 1.4743,
      "theoretical_loss": 3.6717992091020166,
      "tokens_seen": 1017053184
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006988365095285858,
      "loss": 1.548,
      "theoretical_loss": 3.6717769378380414,
      "tokens_seen": 1017118720
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006988164493480441,
      "loss": 1.5369,
      "theoretical_loss": 3.671754668410796,
      "tokens_seen": 1017184256
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006987963891675026,
      "loss": 1.5091,
      "theoretical_loss": 3.6717324008200105,
      "tokens_seen": 1017249792
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006987763289869608,
      "loss": 1.5735,
      "theoretical_loss": 3.671710135065415,
      "tokens_seen": 1017315328
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006987562688064193,
      "loss": 1.4887,
      "theoretical_loss": 3.671687871146739,
      "tokens_seen": 1017380864
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006987362086258776,
      "loss": 1.4828,
      "theoretical_loss": 3.671665609063715,
      "tokens_seen": 1017446400
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000698716148445336,
      "loss": 1.4312,
      "theoretical_loss": 3.671643348816071,
      "tokens_seen": 1017511936
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006986960882647944,
      "loss": 1.5637,
      "theoretical_loss": 3.671621090403539,
      "tokens_seen": 1017577472
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006986760280842528,
      "loss": 1.4951,
      "theoretical_loss": 3.6715988338258487,
      "tokens_seen": 1017643008
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006986559679037111,
      "loss": 1.5938,
      "theoretical_loss": 3.671576579082732,
      "tokens_seen": 1017708544
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006986359077231695,
      "loss": 1.4573,
      "theoretical_loss": 3.6715543261739185,
      "tokens_seen": 1017774080
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006986158475426279,
      "loss": 1.5356,
      "theoretical_loss": 3.6715320750991385,
      "tokens_seen": 1017839616
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006985957873620863,
      "loss": 1.497,
      "theoretical_loss": 3.6715098258581236,
      "tokens_seen": 1017905152
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006985757271815446,
      "loss": 1.5553,
      "theoretical_loss": 3.6714875784506047,
      "tokens_seen": 1017970688
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000698555667001003,
      "loss": 1.5813,
      "theoretical_loss": 3.671465332876312,
      "tokens_seen": 1018036224
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006985356068204613,
      "loss": 1.5651,
      "theoretical_loss": 3.6714430891349767,
      "tokens_seen": 1018101760
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006985155466399198,
      "loss": 1.5191,
      "theoretical_loss": 3.6714208472263303,
      "tokens_seen": 1018167296
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006984954864593781,
      "loss": 1.5452,
      "theoretical_loss": 3.6713986071501035,
      "tokens_seen": 1018232832
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006984754262788365,
      "loss": 1.4505,
      "theoretical_loss": 3.671376368906027,
      "tokens_seen": 1018298368
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006984553660982949,
      "loss": 1.574,
      "theoretical_loss": 3.671354132493833,
      "tokens_seen": 1018363904
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006984353059177533,
      "loss": 1.462,
      "theoretical_loss": 3.6713318979132517,
      "tokens_seen": 1018429440
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006984152457372116,
      "loss": 1.6114,
      "theoretical_loss": 3.6713096651640145,
      "tokens_seen": 1018494976
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006983951855566701,
      "loss": 1.4734,
      "theoretical_loss": 3.671287434245853,
      "tokens_seen": 1018560512
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006983751253761283,
      "loss": 1.5596,
      "theoretical_loss": 3.6712652051584986,
      "tokens_seen": 1018626048
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006983550651955868,
      "loss": 1.5193,
      "theoretical_loss": 3.671242977901683,
      "tokens_seen": 1018691584
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006983350050150451,
      "loss": 1.5997,
      "theoretical_loss": 3.6712207524751372,
      "tokens_seen": 1018757120
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006983149448345035,
      "loss": 1.6617,
      "theoretical_loss": 3.6711985288785933,
      "tokens_seen": 1018822656
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006982948846539619,
      "loss": 1.4414,
      "theoretical_loss": 3.6711763071117822,
      "tokens_seen": 1018888192
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006982748244734202,
      "loss": 1.5252,
      "theoretical_loss": 3.671154087174436,
      "tokens_seen": 1018953728
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006982547642928786,
      "loss": 1.5831,
      "theoretical_loss": 3.671131869066287,
      "tokens_seen": 1019019264
    },
    {
      "epoch": 0.31,
      "objective/train/advantage_avg": 0.49563610553741455,
      "objective/train/docs_used": 374190,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4085946083068848,
      "objective/train/original_loss": 1.4085943698883057,
      "objective/train/theoretical_loss": 3.6711096527870657,
      "objective/train/tokens_used": 1039544800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24605487287044525,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508145093917847,
      "objective/train/weighted_lm_loss": 1.4800573587417603,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6711096527870657,
      "tokens_seen": 1019084800
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006982347041123371,
      "loss": 1.5449,
      "theoretical_loss": 3.6711096527870657,
      "tokens_seen": 1019084800
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006982146439317953,
      "loss": 1.6128,
      "theoretical_loss": 3.671087438336505,
      "tokens_seen": 1019150336
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006981945837512538,
      "loss": 1.4896,
      "theoretical_loss": 3.6710652257143366,
      "tokens_seen": 1019215872
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006981745235707123,
      "loss": 1.45,
      "theoretical_loss": 3.671043014920292,
      "tokens_seen": 1019281408
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006981544633901705,
      "loss": 1.5979,
      "theoretical_loss": 3.671020805954104,
      "tokens_seen": 1019346944
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000698134403209629,
      "loss": 1.5011,
      "theoretical_loss": 3.670998598815504,
      "tokens_seen": 1019412480
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006981143430290872,
      "loss": 1.6435,
      "theoretical_loss": 3.6709763935042243,
      "tokens_seen": 1019478016
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006980942828485457,
      "loss": 1.5446,
      "theoretical_loss": 3.670954190019997,
      "tokens_seen": 1019543552
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006980742226680041,
      "loss": 1.4661,
      "theoretical_loss": 3.6709319883625544,
      "tokens_seen": 1019609088
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006980541624874624,
      "loss": 1.5744,
      "theoretical_loss": 3.6709097885316293,
      "tokens_seen": 1019674624
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006980341023069208,
      "loss": 1.6177,
      "theoretical_loss": 3.670887590526953,
      "tokens_seen": 1019740160
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006980140421263792,
      "loss": 1.5491,
      "theoretical_loss": 3.670865394348259,
      "tokens_seen": 1019805696
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006979939819458375,
      "loss": 1.395,
      "theoretical_loss": 3.6708431999952795,
      "tokens_seen": 1019871232
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000697973921765296,
      "loss": 1.615,
      "theoretical_loss": 3.6708210074677465,
      "tokens_seen": 1019936768
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006979538615847542,
      "loss": 1.5967,
      "theoretical_loss": 3.6707988167653927,
      "tokens_seen": 1020002304
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006979338014042127,
      "loss": 1.5992,
      "theoretical_loss": 3.6707766278879514,
      "tokens_seen": 1020067840
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000697913741223671,
      "loss": 1.5598,
      "theoretical_loss": 3.6707544408351547,
      "tokens_seen": 1020133376
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006978936810431294,
      "loss": 1.3694,
      "theoretical_loss": 3.670732255606735,
      "tokens_seen": 1020198912
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006978736208625878,
      "loss": 1.5971,
      "theoretical_loss": 3.670710072202426,
      "tokens_seen": 1020264448
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006978535606820462,
      "loss": 1.5847,
      "theoretical_loss": 3.67068789062196,
      "tokens_seen": 1020329984
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006978335005015045,
      "loss": 1.5695,
      "theoretical_loss": 3.6706657108650704,
      "tokens_seen": 1020395520
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000697813440320963,
      "loss": 1.5801,
      "theoretical_loss": 3.67064353293149,
      "tokens_seen": 1020461056
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006977933801404212,
      "loss": 1.4984,
      "theoretical_loss": 3.670621356820951,
      "tokens_seen": 1020526592
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006977733199598797,
      "loss": 1.4888,
      "theoretical_loss": 3.670599182533188,
      "tokens_seen": 1020592128
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000697753259779338,
      "loss": 1.53,
      "theoretical_loss": 3.670577010067933,
      "tokens_seen": 1020657664
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006977331995987964,
      "loss": 1.5135,
      "theoretical_loss": 3.670554839424919,
      "tokens_seen": 1020723200
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006977131394182548,
      "loss": 1.4723,
      "theoretical_loss": 3.67053267060388,
      "tokens_seen": 1020788736
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006976930792377132,
      "loss": 1.3963,
      "theoretical_loss": 3.6705105036045493,
      "tokens_seen": 1020854272
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006976730190571715,
      "loss": 1.6417,
      "theoretical_loss": 3.6704883384266602,
      "tokens_seen": 1020919808
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006976529588766299,
      "loss": 1.671,
      "theoretical_loss": 3.6704661750699454,
      "tokens_seen": 1020985344
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006976328986960883,
      "loss": 1.5025,
      "theoretical_loss": 3.6704440135341394,
      "tokens_seen": 1021050880
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006976128385155467,
      "loss": 1.6067,
      "theoretical_loss": 3.6704218538189752,
      "tokens_seen": 1021116416
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000697592778335005,
      "loss": 1.6466,
      "theoretical_loss": 3.6703996959241865,
      "tokens_seen": 1021181952
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006975727181544635,
      "loss": 1.4805,
      "theoretical_loss": 3.670377539849507,
      "tokens_seen": 1021247488
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006975526579739217,
      "loss": 1.44,
      "theoretical_loss": 3.6703553855946702,
      "tokens_seen": 1021313024
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006975325977933802,
      "loss": 1.5744,
      "theoretical_loss": 3.6703332331594103,
      "tokens_seen": 1021378560
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006975125376128385,
      "loss": 1.5961,
      "theoretical_loss": 3.67031108254346,
      "tokens_seen": 1021444096
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006974924774322969,
      "loss": 1.4313,
      "theoretical_loss": 3.670288933746555,
      "tokens_seen": 1021509632
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006974724172517553,
      "loss": 1.3896,
      "theoretical_loss": 3.6702667867684275,
      "tokens_seen": 1021575168
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006974523570712137,
      "loss": 1.548,
      "theoretical_loss": 3.6702446416088126,
      "tokens_seen": 1021640704
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000697432296890672,
      "loss": 1.4771,
      "theoretical_loss": 3.6702224982674436,
      "tokens_seen": 1021706240
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006974122367101305,
      "loss": 1.5355,
      "theoretical_loss": 3.670200356744055,
      "tokens_seen": 1021771776
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006973921765295887,
      "loss": 1.3769,
      "theoretical_loss": 3.670178217038381,
      "tokens_seen": 1021837312
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006973721163490472,
      "loss": 1.4602,
      "theoretical_loss": 3.670156079150156,
      "tokens_seen": 1021902848
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006973520561685055,
      "loss": 1.5693,
      "theoretical_loss": 3.6701339430791133,
      "tokens_seen": 1021968384
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006973319959879639,
      "loss": 1.5102,
      "theoretical_loss": 3.6701118088249887,
      "tokens_seen": 1022033920
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006973119358074223,
      "loss": 1.4935,
      "theoretical_loss": 3.670089676387515,
      "tokens_seen": 1022099456
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006972918756268806,
      "loss": 1.5788,
      "theoretical_loss": 3.670067545766428,
      "tokens_seen": 1022164992
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000697271815446339,
      "loss": 1.5978,
      "theoretical_loss": 3.670045416961461,
      "tokens_seen": 1022230528
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006972517552657975,
      "loss": 1.5841,
      "theoretical_loss": 3.6700232899723497,
      "tokens_seen": 1022296064
    },
    {
      "epoch": 0.31,
      "objective/train/advantage_avg": 0.48553693294525146,
      "objective/train/docs_used": 375283,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4208401441574097,
      "objective/train/original_loss": 1.4208400249481201,
      "objective/train/theoretical_loss": 3.6700011647988275,
      "objective/train/tokens_used": 1042821600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23767893016338348,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0497620105743408,
      "objective/train/weighted_lm_loss": 1.4903978109359741,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0216598510742188,
      "theoretical_loss": 3.6700011647988275,
      "tokens_seen": 1022361600
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006972316950852557,
      "loss": 1.5457,
      "theoretical_loss": 3.6700011647988275,
      "tokens_seen": 1022361600
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006972116349047142,
      "loss": 1.4759,
      "theoretical_loss": 3.66997904144063,
      "tokens_seen": 1022427136
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006971915747241724,
      "loss": 1.5102,
      "theoretical_loss": 3.6699569198974915,
      "tokens_seen": 1022492672
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006971715145436309,
      "loss": 1.426,
      "theoretical_loss": 3.6699348001691474,
      "tokens_seen": 1022558208
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006971514543630893,
      "loss": 1.5048,
      "theoretical_loss": 3.6699126822553314,
      "tokens_seen": 1022623744
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006971313941825476,
      "loss": 1.5639,
      "theoretical_loss": 3.669890566155779,
      "tokens_seen": 1022689280
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000697111334002006,
      "loss": 1.4912,
      "theoretical_loss": 3.6698684518702254,
      "tokens_seen": 1022754816
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006970912738214645,
      "loss": 1.4106,
      "theoretical_loss": 3.669846339398405,
      "tokens_seen": 1022820352
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006970712136409227,
      "loss": 1.5321,
      "theoretical_loss": 3.669824228740053,
      "tokens_seen": 1022885888
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006970511534603812,
      "loss": 1.6275,
      "theoretical_loss": 3.669802119894905,
      "tokens_seen": 1022951424
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006970310932798394,
      "loss": 1.6881,
      "theoretical_loss": 3.6697800128626956,
      "tokens_seen": 1023016960
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006970110330992979,
      "loss": 1.5404,
      "theoretical_loss": 3.6697579076431603,
      "tokens_seen": 1023082496
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006969909729187563,
      "loss": 1.4658,
      "theoretical_loss": 3.6697358042360344,
      "tokens_seen": 1023148032
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006969709127382146,
      "loss": 1.5277,
      "theoretical_loss": 3.6697137026410527,
      "tokens_seen": 1023213568
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000696950852557673,
      "loss": 1.5224,
      "theoretical_loss": 3.669691602857951,
      "tokens_seen": 1023279104
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006969307923771314,
      "loss": 1.5233,
      "theoretical_loss": 3.669669504886465,
      "tokens_seen": 1023344640
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006969107321965897,
      "loss": 1.6065,
      "theoretical_loss": 3.6696474087263296,
      "tokens_seen": 1023410176
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006968906720160482,
      "loss": 1.6758,
      "theoretical_loss": 3.6696253143772806,
      "tokens_seen": 1023475712
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006968706118355065,
      "loss": 1.5934,
      "theoretical_loss": 3.669603221839054,
      "tokens_seen": 1023541248
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006968505516549649,
      "loss": 1.5026,
      "theoretical_loss": 3.669581131111385,
      "tokens_seen": 1023606784
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006968304914744233,
      "loss": 1.5499,
      "theoretical_loss": 3.6695590421940096,
      "tokens_seen": 1023672320
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006968104312938816,
      "loss": 1.6431,
      "theoretical_loss": 3.669536955086663,
      "tokens_seen": 1023737856
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00069679037111334,
      "loss": 1.6058,
      "theoretical_loss": 3.669514869789081,
      "tokens_seen": 1023803392
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006967703109327984,
      "loss": 1.4438,
      "theoretical_loss": 3.669492786301001,
      "tokens_seen": 1023868928
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006967502507522567,
      "loss": 1.5332,
      "theoretical_loss": 3.6694707046221575,
      "tokens_seen": 1023934464
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006967301905717152,
      "loss": 1.373,
      "theoretical_loss": 3.6694486247522864,
      "tokens_seen": 1024000000
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006967101303911735,
      "loss": 1.4663,
      "theoretical_loss": 3.669426546691124,
      "tokens_seen": 1024065536
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006966900702106319,
      "loss": 1.6063,
      "theoretical_loss": 3.669404470438407,
      "tokens_seen": 1024131072
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006966700100300903,
      "loss": 1.6052,
      "theoretical_loss": 3.669382395993871,
      "tokens_seen": 1024196608
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006966499498495487,
      "loss": 1.5377,
      "theoretical_loss": 3.6693603233572523,
      "tokens_seen": 1024262144
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006966298896690071,
      "loss": 1.5753,
      "theoretical_loss": 3.669338252528287,
      "tokens_seen": 1024327680
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006966098294884654,
      "loss": 1.4518,
      "theoretical_loss": 3.6693161835067123,
      "tokens_seen": 1024393216
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006965897693079239,
      "loss": 1.6018,
      "theoretical_loss": 3.669294116292263,
      "tokens_seen": 1024458752
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006965697091273821,
      "loss": 1.4396,
      "theoretical_loss": 3.6692720508846772,
      "tokens_seen": 1024524288
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006965496489468406,
      "loss": 1.5829,
      "theoretical_loss": 3.66924998728369,
      "tokens_seen": 1024589824
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006965295887662989,
      "loss": 1.5961,
      "theoretical_loss": 3.669227925489039,
      "tokens_seen": 1024655360
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006965095285857573,
      "loss": 1.5119,
      "theoretical_loss": 3.6692058655004605,
      "tokens_seen": 1024720896
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006964894684052157,
      "loss": 1.5184,
      "theoretical_loss": 3.6691838073176903,
      "tokens_seen": 1024786432
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006964694082246741,
      "loss": 1.5399,
      "theoretical_loss": 3.669161750940466,
      "tokens_seen": 1024851968
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006964493480441324,
      "loss": 1.5893,
      "theoretical_loss": 3.6691396963685246,
      "tokens_seen": 1024917504
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006964292878635909,
      "loss": 1.5988,
      "theoretical_loss": 3.669117643601602,
      "tokens_seen": 1024983040
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006964092276830491,
      "loss": 1.4575,
      "theoretical_loss": 3.669095592639436,
      "tokens_seen": 1025048576
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006963891675025076,
      "loss": 1.5193,
      "theoretical_loss": 3.669073543481763,
      "tokens_seen": 1025114112
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006963691073219659,
      "loss": 1.4335,
      "theoretical_loss": 3.6690514961283203,
      "tokens_seen": 1025179648
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006963490471414243,
      "loss": 1.4508,
      "theoretical_loss": 3.6690294505788446,
      "tokens_seen": 1025245184
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006963289869608827,
      "loss": 1.574,
      "theoretical_loss": 3.6690074068330727,
      "tokens_seen": 1025310720
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000696308926780341,
      "loss": 1.4897,
      "theoretical_loss": 3.6689853648907427,
      "tokens_seen": 1025376256
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006962888665997994,
      "loss": 1.5241,
      "theoretical_loss": 3.6689633247515907,
      "tokens_seen": 1025441792
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006962688064192579,
      "loss": 1.4215,
      "theoretical_loss": 3.668941286415355,
      "tokens_seen": 1025507328
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006962487462387161,
      "loss": 1.5687,
      "theoretical_loss": 3.668919249881773,
      "tokens_seen": 1025572864
    },
    {
      "epoch": 0.31,
      "objective/train/advantage_avg": 0.4947572350502014,
      "objective/train/docs_used": 376516,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2606171369552612,
      "objective/train/original_loss": 1.2606171369552612,
      "objective/train/theoretical_loss": 3.6688972151505803,
      "objective/train/tokens_used": 1046098400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24507856369018555,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507217645645142,
      "objective/train/weighted_lm_loss": 1.3245314359664917,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6688972151505803,
      "tokens_seen": 1025638400
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006962286860581746,
      "loss": 1.4128,
      "theoretical_loss": 3.6688972151505803,
      "tokens_seen": 1025638400
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006962086258776328,
      "loss": 1.3891,
      "theoretical_loss": 3.6688751822215164,
      "tokens_seen": 1025703936
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006961885656970913,
      "loss": 1.6199,
      "theoretical_loss": 3.668853151094318,
      "tokens_seen": 1025769472
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006961685055165497,
      "loss": 1.3942,
      "theoretical_loss": 3.668831121768722,
      "tokens_seen": 1025835008
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000696148445336008,
      "loss": 1.5596,
      "theoretical_loss": 3.668809094244467,
      "tokens_seen": 1025900544
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006961283851554664,
      "loss": 1.496,
      "theoretical_loss": 3.66878706852129,
      "tokens_seen": 1025966080
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006961083249749249,
      "loss": 1.4995,
      "theoretical_loss": 3.6687650445989295,
      "tokens_seen": 1026031616
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006960882647943831,
      "loss": 1.5418,
      "theoretical_loss": 3.668743022477122,
      "tokens_seen": 1026097152
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006960682046138416,
      "loss": 1.455,
      "theoretical_loss": 3.668721002155606,
      "tokens_seen": 1026162688
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006960481444332998,
      "loss": 1.3996,
      "theoretical_loss": 3.66869898363412,
      "tokens_seen": 1026228224
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006960280842527583,
      "loss": 1.5729,
      "theoretical_loss": 3.6686769669124004,
      "tokens_seen": 1026293760
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006960080240722167,
      "loss": 1.4377,
      "theoretical_loss": 3.6686549519901863,
      "tokens_seen": 1026359296
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000695987963891675,
      "loss": 1.554,
      "theoretical_loss": 3.6686329388672156,
      "tokens_seen": 1026424832
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006959679037111334,
      "loss": 1.5478,
      "theoretical_loss": 3.6686109275432264,
      "tokens_seen": 1026490368
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006959478435305918,
      "loss": 1.5226,
      "theoretical_loss": 3.6685889180179565,
      "tokens_seen": 1026555904
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006959277833500501,
      "loss": 1.5285,
      "theoretical_loss": 3.6685669102911445,
      "tokens_seen": 1026621440
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006959077231695086,
      "loss": 1.4501,
      "theoretical_loss": 3.6685449043625282,
      "tokens_seen": 1026686976
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006958876629889669,
      "loss": 1.5076,
      "theoretical_loss": 3.6685229002318467,
      "tokens_seen": 1026752512
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006958676028084253,
      "loss": 1.5463,
      "theoretical_loss": 3.668500897898837,
      "tokens_seen": 1026818048
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006958475426278837,
      "loss": 1.5311,
      "theoretical_loss": 3.6684788973632387,
      "tokens_seen": 1026883584
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000695827482447342,
      "loss": 1.5525,
      "theoretical_loss": 3.6684568986247896,
      "tokens_seen": 1026949120
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006958074222668004,
      "loss": 1.4535,
      "theoretical_loss": 3.668434901683229,
      "tokens_seen": 1027014656
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006957873620862588,
      "loss": 1.6237,
      "theoretical_loss": 3.668412906538295,
      "tokens_seen": 1027080192
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006957673019057171,
      "loss": 1.5045,
      "theoretical_loss": 3.668390913189726,
      "tokens_seen": 1027145728
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006957472417251756,
      "loss": 1.6171,
      "theoretical_loss": 3.668368921637261,
      "tokens_seen": 1027211264
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006957271815446339,
      "loss": 1.5564,
      "theoretical_loss": 3.6683469318806385,
      "tokens_seen": 1027276800
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006957071213640923,
      "loss": 1.5057,
      "theoretical_loss": 3.6683249439195977,
      "tokens_seen": 1027342336
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006956870611835506,
      "loss": 1.5977,
      "theoretical_loss": 3.668302957753877,
      "tokens_seen": 1027407872
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006956670010030091,
      "loss": 1.4442,
      "theoretical_loss": 3.6682809733832156,
      "tokens_seen": 1027473408
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006956469408224674,
      "loss": 1.5973,
      "theoretical_loss": 3.6682589908073524,
      "tokens_seen": 1027538944
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006956268806419258,
      "loss": 1.4961,
      "theoretical_loss": 3.668237010026026,
      "tokens_seen": 1027604480
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006956068204613841,
      "loss": 1.6102,
      "theoretical_loss": 3.6682150310389767,
      "tokens_seen": 1027670016
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006955867602808425,
      "loss": 1.4923,
      "theoretical_loss": 3.668193053845942,
      "tokens_seen": 1027735552
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006955667001003009,
      "loss": 1.489,
      "theoretical_loss": 3.6681710784466626,
      "tokens_seen": 1027801088
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006955466399197593,
      "loss": 1.5567,
      "theoretical_loss": 3.668149104840876,
      "tokens_seen": 1027866624
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006955265797392176,
      "loss": 1.381,
      "theoretical_loss": 3.668127133028323,
      "tokens_seen": 1027932160
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006955065195586761,
      "loss": 1.5914,
      "theoretical_loss": 3.6681051630087427,
      "tokens_seen": 1027997696
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006954864593781344,
      "loss": 1.6012,
      "theoretical_loss": 3.668083194781874,
      "tokens_seen": 1028063232
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006954663991975928,
      "loss": 1.5422,
      "theoretical_loss": 3.6680612283474567,
      "tokens_seen": 1028128768
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006954463390170511,
      "loss": 1.4804,
      "theoretical_loss": 3.6680392637052304,
      "tokens_seen": 1028194304
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006954262788365095,
      "loss": 1.4697,
      "theoretical_loss": 3.6680173008549337,
      "tokens_seen": 1028259840
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006954062186559679,
      "loss": 1.512,
      "theoretical_loss": 3.6679953397963074,
      "tokens_seen": 1028325376
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006953861584754263,
      "loss": 1.5995,
      "theoretical_loss": 3.667973380529091,
      "tokens_seen": 1028390912
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006953660982948846,
      "loss": 1.4681,
      "theoretical_loss": 3.6679514230530232,
      "tokens_seen": 1028456448
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006953460381143431,
      "loss": 1.6098,
      "theoretical_loss": 3.6679294673678453,
      "tokens_seen": 1028521984
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006953259779338013,
      "loss": 1.4871,
      "theoretical_loss": 3.6679075134732964,
      "tokens_seen": 1028587520
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006953059177532598,
      "loss": 1.4701,
      "theoretical_loss": 3.6678855613691157,
      "tokens_seen": 1028653056
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006952858575727182,
      "loss": 1.5864,
      "theoretical_loss": 3.6678636110550444,
      "tokens_seen": 1028718592
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006952657973921765,
      "loss": 1.4193,
      "theoretical_loss": 3.6678416625308214,
      "tokens_seen": 1028784128
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006952457372116349,
      "loss": 1.4981,
      "theoretical_loss": 3.6678197157961874,
      "tokens_seen": 1028849664
    },
    {
      "epoch": 0.31,
      "objective/train/advantage_avg": 0.4935339093208313,
      "objective/train/docs_used": 377682,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3692549467086792,
      "objective/train/original_loss": 1.3692550659179688,
      "objective/train/theoretical_loss": 3.667797770850883,
      "objective/train/tokens_used": 1049375200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2438536435365677,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505930185317993,
      "objective/train/weighted_lm_loss": 1.4384739398956299,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.667797770850883,
      "tokens_seen": 1028915200
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006952256770310932,
      "loss": 1.4737,
      "theoretical_loss": 3.667797770850883,
      "tokens_seen": 1028915200
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006952056168505516,
      "loss": 1.5728,
      "theoretical_loss": 3.667775827694647,
      "tokens_seen": 1028980736
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006951855566700101,
      "loss": 1.6369,
      "theoretical_loss": 3.66775388632722,
      "tokens_seen": 1029046272
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006951654964894684,
      "loss": 1.4211,
      "theoretical_loss": 3.667731946748343,
      "tokens_seen": 1029111808
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006951454363089268,
      "loss": 1.4968,
      "theoretical_loss": 3.667710008957756,
      "tokens_seen": 1029177344
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006951253761283853,
      "loss": 1.5617,
      "theoretical_loss": 3.6676880729551993,
      "tokens_seen": 1029242880
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006951053159478435,
      "loss": 1.5881,
      "theoretical_loss": 3.6676661387404135,
      "tokens_seen": 1029308416
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000695085255767302,
      "loss": 1.5964,
      "theoretical_loss": 3.6676442063131383,
      "tokens_seen": 1029373952
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006950651955867602,
      "loss": 1.5791,
      "theoretical_loss": 3.667622275673115,
      "tokens_seen": 1029439488
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006950451354062187,
      "loss": 1.5751,
      "theoretical_loss": 3.6676003468200844,
      "tokens_seen": 1029505024
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006950250752256771,
      "loss": 1.449,
      "theoretical_loss": 3.6675784197537866,
      "tokens_seen": 1029570560
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006950050150451354,
      "loss": 1.5098,
      "theoretical_loss": 3.6675564944739625,
      "tokens_seen": 1029636096
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006949849548645938,
      "loss": 1.5752,
      "theoretical_loss": 3.667534570980353,
      "tokens_seen": 1029701632
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006949648946840522,
      "loss": 1.5866,
      "theoretical_loss": 3.6675126492726986,
      "tokens_seen": 1029767168
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006949448345035105,
      "loss": 1.494,
      "theoretical_loss": 3.66749072935074,
      "tokens_seen": 1029832704
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000694924774322969,
      "loss": 1.6304,
      "theoretical_loss": 3.667468811214219,
      "tokens_seen": 1029898240
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006949047141424273,
      "loss": 1.5002,
      "theoretical_loss": 3.667446894862876,
      "tokens_seen": 1029963776
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006948846539618857,
      "loss": 1.6248,
      "theoretical_loss": 3.6674249802964516,
      "tokens_seen": 1030029312
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006948645937813441,
      "loss": 1.4524,
      "theoretical_loss": 3.6674030675146874,
      "tokens_seen": 1030094848
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006948445336008024,
      "loss": 1.4352,
      "theoretical_loss": 3.667381156517324,
      "tokens_seen": 1030160384
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006948244734202608,
      "loss": 1.4664,
      "theoretical_loss": 3.667359247304104,
      "tokens_seen": 1030225920
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006948044132397192,
      "loss": 1.5265,
      "theoretical_loss": 3.6673373398747673,
      "tokens_seen": 1030291456
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006947843530591775,
      "loss": 1.5339,
      "theoretical_loss": 3.6673154342290553,
      "tokens_seen": 1030356992
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000694764292878636,
      "loss": 1.4536,
      "theoretical_loss": 3.6672935303667096,
      "tokens_seen": 1030422528
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006947442326980943,
      "loss": 1.5219,
      "theoretical_loss": 3.667271628287472,
      "tokens_seen": 1030488064
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006947241725175527,
      "loss": 1.4269,
      "theoretical_loss": 3.6672497279910834,
      "tokens_seen": 1030553600
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000694704112337011,
      "loss": 1.533,
      "theoretical_loss": 3.6672278294772855,
      "tokens_seen": 1030619136
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006946840521564695,
      "loss": 1.5148,
      "theoretical_loss": 3.6672059327458197,
      "tokens_seen": 1030684672
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006946639919759278,
      "loss": 1.6704,
      "theoretical_loss": 3.6671840377964275,
      "tokens_seen": 1030750208
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006946439317953862,
      "loss": 1.5096,
      "theoretical_loss": 3.6671621446288514,
      "tokens_seen": 1030815744
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006946238716148445,
      "loss": 1.509,
      "theoretical_loss": 3.6671402532428323,
      "tokens_seen": 1030881280
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006946038114343029,
      "loss": 1.5996,
      "theoretical_loss": 3.6671183636381124,
      "tokens_seen": 1030946816
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006945837512537613,
      "loss": 1.5349,
      "theoretical_loss": 3.667096475814433,
      "tokens_seen": 1031012352
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006945636910732197,
      "loss": 1.4181,
      "theoretical_loss": 3.667074589771537,
      "tokens_seen": 1031077888
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000694543630892678,
      "loss": 1.5379,
      "theoretical_loss": 3.6670527055091653,
      "tokens_seen": 1031143424
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006945235707121365,
      "loss": 1.6459,
      "theoretical_loss": 3.6670308230270603,
      "tokens_seen": 1031208960
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006945035105315948,
      "loss": 1.5086,
      "theoretical_loss": 3.6670089423249643,
      "tokens_seen": 1031274496
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006944834503510532,
      "loss": 1.451,
      "theoretical_loss": 3.666987063402619,
      "tokens_seen": 1031340032
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006944633901705115,
      "loss": 1.6479,
      "theoretical_loss": 3.666965186259767,
      "tokens_seen": 1031405568
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006944433299899699,
      "loss": 1.6159,
      "theoretical_loss": 3.66694331089615,
      "tokens_seen": 1031471104
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006944232698094283,
      "loss": 1.695,
      "theoretical_loss": 3.6669214373115104,
      "tokens_seen": 1031536640
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006944032096288867,
      "loss": 1.5426,
      "theoretical_loss": 3.6668995655055907,
      "tokens_seen": 1031602176
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000694383149448345,
      "loss": 1.5446,
      "theoretical_loss": 3.6668776954781332,
      "tokens_seen": 1031667712
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006943630892678035,
      "loss": 1.4588,
      "theoretical_loss": 3.6668558272288805,
      "tokens_seen": 1031733248
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006943430290872617,
      "loss": 1.5417,
      "theoretical_loss": 3.6668339607575744,
      "tokens_seen": 1031798784
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006943229689067202,
      "loss": 1.4512,
      "theoretical_loss": 3.6668120960639587,
      "tokens_seen": 1031864320
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006943029087261786,
      "loss": 1.5442,
      "theoretical_loss": 3.6667902331477746,
      "tokens_seen": 1031929856
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006942828485456369,
      "loss": 1.5034,
      "theoretical_loss": 3.6667683720087654,
      "tokens_seen": 1031995392
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006942627883650953,
      "loss": 1.435,
      "theoretical_loss": 3.6667465126466743,
      "tokens_seen": 1032060928
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006942427281845536,
      "loss": 1.4448,
      "theoretical_loss": 3.666724655061243,
      "tokens_seen": 1032126464
    },
    {
      "epoch": 0.31,
      "objective/train/advantage_avg": 0.4941002428531647,
      "objective/train/docs_used": 378946,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4372228384017944,
      "objective/train/original_loss": 1.4372225999832153,
      "objective/train/theoretical_loss": 3.6667027992522145,
      "objective/train/tokens_used": 1052652000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2443668097257614,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506523847579956,
      "objective/train/weighted_lm_loss": 1.5099949836730957,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.6667027992522145,
      "tokens_seen": 1032192000
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000694222668004012,
      "loss": 1.4445,
      "theoretical_loss": 3.6667027992522145,
      "tokens_seen": 1032192000
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006942026078234705,
      "loss": 1.3915,
      "theoretical_loss": 3.6666809452193325,
      "tokens_seen": 1032257536
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006941825476429287,
      "loss": 1.4879,
      "theoretical_loss": 3.6666590929623393,
      "tokens_seen": 1032323072
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006941624874623872,
      "loss": 1.3953,
      "theoretical_loss": 3.666637242480978,
      "tokens_seen": 1032388608
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006941424272818456,
      "loss": 1.4905,
      "theoretical_loss": 3.6666153937749915,
      "tokens_seen": 1032454144
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006941223671013039,
      "loss": 1.4401,
      "theoretical_loss": 3.6665935468441226,
      "tokens_seen": 1032519680
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006941023069207623,
      "loss": 1.5415,
      "theoretical_loss": 3.666571701688115,
      "tokens_seen": 1032585216
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006940822467402206,
      "loss": 1.4851,
      "theoretical_loss": 3.6665498583067118,
      "tokens_seen": 1032650752
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000694062186559679,
      "loss": 1.6131,
      "theoretical_loss": 3.6665280166996554,
      "tokens_seen": 1032716288
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006940421263791375,
      "loss": 1.6306,
      "theoretical_loss": 3.6665061768666902,
      "tokens_seen": 1032781824
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006940220661985957,
      "loss": 1.5486,
      "theoretical_loss": 3.6664843388075594,
      "tokens_seen": 1032847360
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006940020060180542,
      "loss": 1.5892,
      "theoretical_loss": 3.6664625025220055,
      "tokens_seen": 1032912896
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006939819458375125,
      "loss": 1.4528,
      "theoretical_loss": 3.6664406680097725,
      "tokens_seen": 1032978432
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006939618856569709,
      "loss": 1.3382,
      "theoretical_loss": 3.666418835270604,
      "tokens_seen": 1033043968
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006939418254764293,
      "loss": 1.3838,
      "theoretical_loss": 3.6663970043042435,
      "tokens_seen": 1033109504
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006939217652958877,
      "loss": 1.3317,
      "theoretical_loss": 3.6663751751104345,
      "tokens_seen": 1033175040
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000693901705115346,
      "loss": 1.4528,
      "theoretical_loss": 3.6663533476889207,
      "tokens_seen": 1033240576
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006938816449348045,
      "loss": 1.5836,
      "theoretical_loss": 3.666331522039446,
      "tokens_seen": 1033306112
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006938615847542627,
      "loss": 1.4862,
      "theoretical_loss": 3.6663096981617533,
      "tokens_seen": 1033371648
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006938415245737212,
      "loss": 1.5666,
      "theoretical_loss": 3.6662878760555877,
      "tokens_seen": 1033437184
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006938214643931795,
      "loss": 1.536,
      "theoretical_loss": 3.6662660557206923,
      "tokens_seen": 1033502720
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006938014042126379,
      "loss": 1.3343,
      "theoretical_loss": 3.6662442371568114,
      "tokens_seen": 1033568256
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006937813440320963,
      "loss": 1.4216,
      "theoretical_loss": 3.6662224203636886,
      "tokens_seen": 1033633792
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006937612838515547,
      "loss": 1.4441,
      "theoretical_loss": 3.6662006053410683,
      "tokens_seen": 1033699328
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000693741223671013,
      "loss": 1.4402,
      "theoretical_loss": 3.6661787920886937,
      "tokens_seen": 1033764864
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006937211634904714,
      "loss": 1.5703,
      "theoretical_loss": 3.6661569806063103,
      "tokens_seen": 1033830400
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006937011033099297,
      "loss": 1.6391,
      "theoretical_loss": 3.6661351708936616,
      "tokens_seen": 1033895936
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006936810431293882,
      "loss": 1.4615,
      "theoretical_loss": 3.666113362950491,
      "tokens_seen": 1033961472
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006936609829488466,
      "loss": 1.4972,
      "theoretical_loss": 3.6660915567765446,
      "tokens_seen": 1034027008
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006936409227683049,
      "loss": 1.3258,
      "theoretical_loss": 3.6660697523715653,
      "tokens_seen": 1034092544
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006936208625877633,
      "loss": 1.3885,
      "theoretical_loss": 3.6660479497352982,
      "tokens_seen": 1034158080
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006936008024072217,
      "loss": 1.5246,
      "theoretical_loss": 3.6660261488674877,
      "tokens_seen": 1034223616
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006935807422266801,
      "loss": 1.5458,
      "theoretical_loss": 3.666004349767878,
      "tokens_seen": 1034289152
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006935606820461384,
      "loss": 1.5972,
      "theoretical_loss": 3.665982552436213,
      "tokens_seen": 1034354688
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006935406218655969,
      "loss": 1.5,
      "theoretical_loss": 3.665960756872239,
      "tokens_seen": 1034420224
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006935205616850552,
      "loss": 1.5129,
      "theoretical_loss": 3.6659389630756998,
      "tokens_seen": 1034485760
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006935005015045136,
      "loss": 1.6271,
      "theoretical_loss": 3.6659171710463396,
      "tokens_seen": 1034551296
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000693480441323972,
      "loss": 1.5398,
      "theoretical_loss": 3.665895380783904,
      "tokens_seen": 1034616832
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006934603811434303,
      "loss": 1.4097,
      "theoretical_loss": 3.6658735922881376,
      "tokens_seen": 1034682368
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006934403209628887,
      "loss": 1.4741,
      "theoretical_loss": 3.665851805558785,
      "tokens_seen": 1034747904
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006934202607823471,
      "loss": 1.5488,
      "theoretical_loss": 3.6658300205955916,
      "tokens_seen": 1034813440
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006934002006018054,
      "loss": 1.5359,
      "theoretical_loss": 3.665808237398302,
      "tokens_seen": 1034878976
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006933801404212639,
      "loss": 1.5247,
      "theoretical_loss": 3.665786455966661,
      "tokens_seen": 1034944512
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006933600802407221,
      "loss": 1.5592,
      "theoretical_loss": 3.6657646763004146,
      "tokens_seen": 1035010048
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006933400200601806,
      "loss": 1.5035,
      "theoretical_loss": 3.665742898399307,
      "tokens_seen": 1035075584
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000693319959879639,
      "loss": 1.5414,
      "theoretical_loss": 3.6657211222630837,
      "tokens_seen": 1035141120
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006932998996990973,
      "loss": 1.5223,
      "theoretical_loss": 3.6656993478914903,
      "tokens_seen": 1035206656
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006932798395185557,
      "loss": 1.4395,
      "theoretical_loss": 3.665677575284272,
      "tokens_seen": 1035272192
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000693259779338014,
      "loss": 1.7009,
      "theoretical_loss": 3.6656558044411733,
      "tokens_seen": 1035337728
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006932397191574724,
      "loss": 1.5575,
      "theoretical_loss": 3.6656340353619408,
      "tokens_seen": 1035403264
    },
    {
      "epoch": 0.31,
      "objective/train/advantage_avg": 0.49438637495040894,
      "objective/train/docs_used": 379908,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4613293409347534,
      "objective/train/original_loss": 1.4613293409347534,
      "objective/train/theoretical_loss": 3.6656122680463197,
      "objective/train/tokens_used": 1055928800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24461601674556732,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506821870803833,
      "objective/train/weighted_lm_loss": 1.535212516784668,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 3.6656122680463197,
      "tokens_seen": 1035468800
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006932196589769309,
      "loss": 1.4841,
      "theoretical_loss": 3.6656122680463197,
      "tokens_seen": 1035468800
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006931995987963891,
      "loss": 1.4826,
      "theoretical_loss": 3.665590502494055,
      "tokens_seen": 1035534336
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006931795386158476,
      "loss": 1.3936,
      "theoretical_loss": 3.6655687387048923,
      "tokens_seen": 1035599872
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000693159478435306,
      "loss": 1.5152,
      "theoretical_loss": 3.665546976678578,
      "tokens_seen": 1035665408
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006931394182547643,
      "loss": 1.5638,
      "theoretical_loss": 3.6655252164148564,
      "tokens_seen": 1035730944
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006931193580742227,
      "loss": 1.5539,
      "theoretical_loss": 3.665503457913475,
      "tokens_seen": 1035796480
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000693099297893681,
      "loss": 1.5595,
      "theoretical_loss": 3.665481701174178,
      "tokens_seen": 1035862016
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006930792377131394,
      "loss": 1.4939,
      "theoretical_loss": 3.6654599461967123,
      "tokens_seen": 1035927552
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006930591775325979,
      "loss": 1.5837,
      "theoretical_loss": 3.6654381929808233,
      "tokens_seen": 1035993088
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006930391173520561,
      "loss": 1.5373,
      "theoretical_loss": 3.665416441526257,
      "tokens_seen": 1036058624
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006930190571715146,
      "loss": 1.4947,
      "theoretical_loss": 3.6653946918327587,
      "tokens_seen": 1036124160
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006929989969909729,
      "loss": 1.3719,
      "theoretical_loss": 3.665372943900076,
      "tokens_seen": 1036189696
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006929789368104313,
      "loss": 1.6009,
      "theoretical_loss": 3.6653511977279534,
      "tokens_seen": 1036255232
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006929588766298897,
      "loss": 1.4857,
      "theoretical_loss": 3.6653294533161382,
      "tokens_seen": 1036320768
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006929388164493481,
      "loss": 1.4381,
      "theoretical_loss": 3.665307710664376,
      "tokens_seen": 1036386304
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006929187562688064,
      "loss": 1.575,
      "theoretical_loss": 3.6652859697724134,
      "tokens_seen": 1036451840
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006928986960882649,
      "loss": 1.5911,
      "theoretical_loss": 3.6652642306399965,
      "tokens_seen": 1036517376
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006928786359077231,
      "loss": 1.4685,
      "theoretical_loss": 3.6652424932668715,
      "tokens_seen": 1036582912
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006928585757271816,
      "loss": 1.5437,
      "theoretical_loss": 3.6652207576527847,
      "tokens_seen": 1036648448
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006928385155466399,
      "loss": 1.6076,
      "theoretical_loss": 3.6651990237974834,
      "tokens_seen": 1036713984
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006928184553660983,
      "loss": 1.4979,
      "theoretical_loss": 3.6651772917007137,
      "tokens_seen": 1036779520
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006927983951855567,
      "loss": 1.5986,
      "theoretical_loss": 3.6651555613622215,
      "tokens_seen": 1036845056
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006927783350050151,
      "loss": 1.4765,
      "theoretical_loss": 3.665133832781754,
      "tokens_seen": 1036910592
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006927582748244734,
      "loss": 1.5582,
      "theoretical_loss": 3.665112105959058,
      "tokens_seen": 1036976128
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006927382146439318,
      "loss": 1.5526,
      "theoretical_loss": 3.6650903808938806,
      "tokens_seen": 1037041664
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006927181544633901,
      "loss": 1.473,
      "theoretical_loss": 3.6650686575859677,
      "tokens_seen": 1037107200
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006926980942828486,
      "loss": 1.6163,
      "theoretical_loss": 3.6650469360350657,
      "tokens_seen": 1037172736
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006926780341023069,
      "loss": 1.6309,
      "theoretical_loss": 3.6650252162409234,
      "tokens_seen": 1037238272
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006926579739217653,
      "loss": 1.4393,
      "theoretical_loss": 3.6650034982032857,
      "tokens_seen": 1037303808
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006926379137412236,
      "loss": 1.5891,
      "theoretical_loss": 3.664981781921901,
      "tokens_seen": 1037369344
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006926178535606821,
      "loss": 1.578,
      "theoretical_loss": 3.6649600673965157,
      "tokens_seen": 1037434880
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006925977933801404,
      "loss": 1.3479,
      "theoretical_loss": 3.664938354626877,
      "tokens_seen": 1037500416
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006925777331995988,
      "loss": 1.4924,
      "theoretical_loss": 3.664916643612732,
      "tokens_seen": 1037565952
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006925576730190572,
      "loss": 1.4363,
      "theoretical_loss": 3.664894934353828,
      "tokens_seen": 1037631488
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006925376128385156,
      "loss": 1.5474,
      "theoretical_loss": 3.6648732268499122,
      "tokens_seen": 1037697024
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006925175526579739,
      "loss": 1.5068,
      "theoretical_loss": 3.664851521100732,
      "tokens_seen": 1037762560
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006924974924774323,
      "loss": 1.4601,
      "theoretical_loss": 3.6648298171060345,
      "tokens_seen": 1037828096
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006924774322968906,
      "loss": 1.4618,
      "theoretical_loss": 3.664808114865567,
      "tokens_seen": 1037893632
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006924573721163491,
      "loss": 1.514,
      "theoretical_loss": 3.6647864143790776,
      "tokens_seen": 1037959168
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006924373119358074,
      "loss": 1.4549,
      "theoretical_loss": 3.6647647156463137,
      "tokens_seen": 1038024704
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006924172517552658,
      "loss": 1.505,
      "theoretical_loss": 3.6647430186670222,
      "tokens_seen": 1038090240
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006923971915747242,
      "loss": 1.4985,
      "theoretical_loss": 3.664721323440951,
      "tokens_seen": 1038155776
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006923771313941825,
      "loss": 1.406,
      "theoretical_loss": 3.6646996299678483,
      "tokens_seen": 1038221312
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006923570712136409,
      "loss": 1.5651,
      "theoretical_loss": 3.664677938247461,
      "tokens_seen": 1038286848
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006923370110330994,
      "loss": 1.5304,
      "theoretical_loss": 3.6646562482795373,
      "tokens_seen": 1038352384
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006923169508525576,
      "loss": 1.4997,
      "theoretical_loss": 3.664634560063825,
      "tokens_seen": 1038417920
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006922968906720161,
      "loss": 1.4111,
      "theoretical_loss": 3.6646128736000723,
      "tokens_seen": 1038483456
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006922768304914743,
      "loss": 1.3714,
      "theoretical_loss": 3.6645911888880267,
      "tokens_seen": 1038548992
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006922567703109328,
      "loss": 1.61,
      "theoretical_loss": 3.664569505927436,
      "tokens_seen": 1038614528
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006922367101303912,
      "loss": 1.5279,
      "theoretical_loss": 3.664547824718049,
      "tokens_seen": 1038680064
    },
    {
      "epoch": 0.31,
      "objective/train/advantage_avg": 0.4911058843135834,
      "objective/train/docs_used": 381080,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2691367864608765,
      "objective/train/original_loss": 1.2691364288330078,
      "objective/train/theoretical_loss": 3.6645261452596136,
      "objective/train/tokens_used": 1059205600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24170862138271332,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503392219543457,
      "objective/train/weighted_lm_loss": 1.332932949066162,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6645261452596136,
      "tokens_seen": 1038745600
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006922166499498495,
      "loss": 1.4449,
      "theoretical_loss": 3.6645261452596136,
      "tokens_seen": 1038745600
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006921965897693079,
      "loss": 1.5268,
      "theoretical_loss": 3.664504467551877,
      "tokens_seen": 1038811136
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006921765295887664,
      "loss": 1.5245,
      "theoretical_loss": 3.664482791594588,
      "tokens_seen": 1038876672
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006921564694082247,
      "loss": 1.5734,
      "theoretical_loss": 3.6644611173874955,
      "tokens_seen": 1038942208
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006921364092276831,
      "loss": 1.4903,
      "theoretical_loss": 3.6644394449303466,
      "tokens_seen": 1039007744
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006921163490471414,
      "loss": 1.4408,
      "theoretical_loss": 3.6644177742228905,
      "tokens_seen": 1039073280
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006920962888665998,
      "loss": 1.521,
      "theoretical_loss": 3.664396105264875,
      "tokens_seen": 1039138816
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006920762286860583,
      "loss": 1.4553,
      "theoretical_loss": 3.6643744380560497,
      "tokens_seen": 1039204352
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006920561685055165,
      "loss": 1.5364,
      "theoretical_loss": 3.6643527725961613,
      "tokens_seen": 1039269888
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000692036108324975,
      "loss": 1.5526,
      "theoretical_loss": 3.66433110888496,
      "tokens_seen": 1039335424
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006920160481444333,
      "loss": 1.4727,
      "theoretical_loss": 3.6643094469221933,
      "tokens_seen": 1039400960
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0006919959879638917,
      "loss": 1.478,
      "theoretical_loss": 3.6642877867076105,
      "tokens_seen": 1039466496
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006919759277833501,
      "loss": 1.5678,
      "theoretical_loss": 3.66426612824096,
      "tokens_seen": 1039532032
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006919558676028085,
      "loss": 1.5347,
      "theoretical_loss": 3.6642444715219913,
      "tokens_seen": 1039597568
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006919358074222668,
      "loss": 1.4136,
      "theoretical_loss": 3.664222816550452,
      "tokens_seen": 1039663104
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006919157472417253,
      "loss": 1.4444,
      "theoretical_loss": 3.6642011633260916,
      "tokens_seen": 1039728640
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006918956870611835,
      "loss": 1.51,
      "theoretical_loss": 3.664179511848659,
      "tokens_seen": 1039794176
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000691875626880642,
      "loss": 1.4757,
      "theoretical_loss": 3.6641578621179027,
      "tokens_seen": 1039859712
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006918555667001003,
      "loss": 1.466,
      "theoretical_loss": 3.6641362141335727,
      "tokens_seen": 1039925248
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006918355065195587,
      "loss": 1.4769,
      "theoretical_loss": 3.6641145678954175,
      "tokens_seen": 1039990784
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006918154463390171,
      "loss": 1.5201,
      "theoretical_loss": 3.664092923403186,
      "tokens_seen": 1040056320
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006917953861584755,
      "loss": 1.5064,
      "theoretical_loss": 3.6640712806566285,
      "tokens_seen": 1040121856
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006917753259779338,
      "loss": 1.477,
      "theoretical_loss": 3.6640496396554925,
      "tokens_seen": 1040187392
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006917552657973922,
      "loss": 1.567,
      "theoretical_loss": 3.6640280003995285,
      "tokens_seen": 1040252928
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006917352056168505,
      "loss": 1.4003,
      "theoretical_loss": 3.664006362888485,
      "tokens_seen": 1040318464
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000691715145436309,
      "loss": 1.6344,
      "theoretical_loss": 3.663984727122113,
      "tokens_seen": 1040384000
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006916950852557673,
      "loss": 1.5614,
      "theoretical_loss": 3.6639630931001594,
      "tokens_seen": 1040449536
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006916750250752257,
      "loss": 1.4959,
      "theoretical_loss": 3.663941460822376,
      "tokens_seen": 1040515072
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000691654964894684,
      "loss": 1.4472,
      "theoretical_loss": 3.6639198302885116,
      "tokens_seen": 1040580608
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006916349047141425,
      "loss": 1.5169,
      "theoretical_loss": 3.663898201498315,
      "tokens_seen": 1040646144
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006916148445336008,
      "loss": 1.5612,
      "theoretical_loss": 3.6638765744515367,
      "tokens_seen": 1040711680
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006915947843530592,
      "loss": 1.5472,
      "theoretical_loss": 3.663854949147926,
      "tokens_seen": 1040777216
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006915747241725176,
      "loss": 1.5858,
      "theoretical_loss": 3.6638333255872335,
      "tokens_seen": 1040842752
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000691554663991976,
      "loss": 1.6321,
      "theoretical_loss": 3.663811703769208,
      "tokens_seen": 1040908288
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006915346038114343,
      "loss": 1.4524,
      "theoretical_loss": 3.6637900836935993,
      "tokens_seen": 1040973824
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006915145436308928,
      "loss": 1.5056,
      "theoretical_loss": 3.663768465360158,
      "tokens_seen": 1041039360
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000691494483450351,
      "loss": 1.5092,
      "theoretical_loss": 3.6637468487686338,
      "tokens_seen": 1041104896
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006914744232698095,
      "loss": 1.537,
      "theoretical_loss": 3.6637252339187762,
      "tokens_seen": 1041170432
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006914543630892678,
      "loss": 1.4597,
      "theoretical_loss": 3.6637036208103364,
      "tokens_seen": 1041235968
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006914343029087262,
      "loss": 1.4876,
      "theoretical_loss": 3.663682009443063,
      "tokens_seen": 1041301504
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006914142427281846,
      "loss": 1.4317,
      "theoretical_loss": 3.6636603998167074,
      "tokens_seen": 1041367040
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006913941825476429,
      "loss": 1.4403,
      "theoretical_loss": 3.663638791931019,
      "tokens_seen": 1041432576
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006913741223671013,
      "loss": 1.4392,
      "theoretical_loss": 3.663617185785749,
      "tokens_seen": 1041498112
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006913540621865598,
      "loss": 1.4629,
      "theoretical_loss": 3.6635955813806467,
      "tokens_seen": 1041563648
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000691334002006018,
      "loss": 1.3967,
      "theoretical_loss": 3.6635739787154638,
      "tokens_seen": 1041629184
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006913139418254765,
      "loss": 1.4609,
      "theoretical_loss": 3.663552377789949,
      "tokens_seen": 1041694720
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006912938816449347,
      "loss": 1.5869,
      "theoretical_loss": 3.6635307786038536,
      "tokens_seen": 1041760256
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006912738214643932,
      "loss": 1.4555,
      "theoretical_loss": 3.663509181156928,
      "tokens_seen": 1041825792
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006912537612838516,
      "loss": 1.5126,
      "theoretical_loss": 3.6634875854489235,
      "tokens_seen": 1041891328
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006912337011033099,
      "loss": 1.4469,
      "theoretical_loss": 3.6634659914795895,
      "tokens_seen": 1041956864
    },
    {
      "epoch": 0.32,
      "objective/train/advantage_avg": 0.4946609437465668,
      "objective/train/docs_used": 382133,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.403578758239746,
      "objective/train/original_loss": 1.403578519821167,
      "objective/train/theoretical_loss": 3.663444399248678,
      "objective/train/tokens_used": 1062482400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24501168727874756,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507116317749023,
      "objective/train/weighted_lm_loss": 1.4746416807174683,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.663444399248678,
      "tokens_seen": 1042022400
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006912136409227683,
      "loss": 1.4551,
      "theoretical_loss": 3.663444399248678,
      "tokens_seen": 1042022400
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006911935807422268,
      "loss": 1.575,
      "theoretical_loss": 3.6634228087559384,
      "tokens_seen": 1042087936
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000691173520561685,
      "loss": 1.5237,
      "theoretical_loss": 3.6634012200011226,
      "tokens_seen": 1042153472
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006911534603811435,
      "loss": 1.5775,
      "theoretical_loss": 3.6633796329839807,
      "tokens_seen": 1042219008
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006911334002006017,
      "loss": 1.4742,
      "theoretical_loss": 3.6633580477042633,
      "tokens_seen": 1042284544
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006911133400200602,
      "loss": 1.5192,
      "theoretical_loss": 3.6633364641617225,
      "tokens_seen": 1042350080
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006910932798395186,
      "loss": 1.5035,
      "theoretical_loss": 3.663314882356109,
      "tokens_seen": 1042415616
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006910732196589769,
      "loss": 1.4187,
      "theoretical_loss": 3.6632933022871725,
      "tokens_seen": 1042481152
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006910531594784353,
      "loss": 1.5511,
      "theoretical_loss": 3.663271723954665,
      "tokens_seen": 1042546688
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006910330992978937,
      "loss": 1.5187,
      "theoretical_loss": 3.6632501473583385,
      "tokens_seen": 1042612224
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000691013039117352,
      "loss": 1.4936,
      "theoretical_loss": 3.6632285724979434,
      "tokens_seen": 1042677760
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006909929789368105,
      "loss": 1.5508,
      "theoretical_loss": 3.66320699937323,
      "tokens_seen": 1042743296
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006909729187562687,
      "loss": 1.4719,
      "theoretical_loss": 3.6631854279839513,
      "tokens_seen": 1042808832
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006909528585757272,
      "loss": 1.4332,
      "theoretical_loss": 3.663163858329858,
      "tokens_seen": 1042874368
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006909327983951855,
      "loss": 1.4028,
      "theoretical_loss": 3.6631422904107005,
      "tokens_seen": 1042939904
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006909127382146439,
      "loss": 1.4229,
      "theoretical_loss": 3.663120724226232,
      "tokens_seen": 1043005440
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006908926780341023,
      "loss": 1.4905,
      "theoretical_loss": 3.6630991597762024,
      "tokens_seen": 1043070976
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006908726178535607,
      "loss": 1.4702,
      "theoretical_loss": 3.663077597060364,
      "tokens_seen": 1043136512
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000690852557673019,
      "loss": 1.5159,
      "theoretical_loss": 3.6630560360784687,
      "tokens_seen": 1043202048
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006908324974924775,
      "loss": 1.4538,
      "theoretical_loss": 3.6630344768302674,
      "tokens_seen": 1043267584
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006908124373119358,
      "loss": 1.6309,
      "theoretical_loss": 3.6630129193155128,
      "tokens_seen": 1043333120
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006907923771313942,
      "loss": 1.5992,
      "theoretical_loss": 3.6629913635339557,
      "tokens_seen": 1043398656
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006907723169508525,
      "loss": 1.4011,
      "theoretical_loss": 3.6629698094853484,
      "tokens_seen": 1043464192
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000690752256770311,
      "loss": 1.61,
      "theoretical_loss": 3.6629482571694423,
      "tokens_seen": 1043529728
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006907321965897693,
      "loss": 1.3941,
      "theoretical_loss": 3.6629267065859894,
      "tokens_seen": 1043595264
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006907121364092277,
      "loss": 1.5494,
      "theoretical_loss": 3.662905157734742,
      "tokens_seen": 1043660800
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000690692076228686,
      "loss": 1.6367,
      "theoretical_loss": 3.6628836106154523,
      "tokens_seen": 1043726336
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006906720160481444,
      "loss": 1.4736,
      "theoretical_loss": 3.662862065227872,
      "tokens_seen": 1043791872
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006906519558676029,
      "loss": 1.5113,
      "theoretical_loss": 3.662840521571753,
      "tokens_seen": 1043857408
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006906318956870612,
      "loss": 1.5063,
      "theoretical_loss": 3.662818979646848,
      "tokens_seen": 1043922944
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006906118355065196,
      "loss": 1.5592,
      "theoretical_loss": 3.662797439452909,
      "tokens_seen": 1043988480
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000690591775325978,
      "loss": 1.5276,
      "theoretical_loss": 3.6627759009896876,
      "tokens_seen": 1044054016
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006905717151454364,
      "loss": 1.4683,
      "theoretical_loss": 3.662754364256937,
      "tokens_seen": 1044119552
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006905516549648947,
      "loss": 1.61,
      "theoretical_loss": 3.6627328292544092,
      "tokens_seen": 1044185088
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006905315947843532,
      "loss": 1.6678,
      "theoretical_loss": 3.662711295981857,
      "tokens_seen": 1044250624
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006905115346038114,
      "loss": 1.4696,
      "theoretical_loss": 3.6626897644390324,
      "tokens_seen": 1044316160
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006904914744232699,
      "loss": 1.5639,
      "theoretical_loss": 3.662668234625688,
      "tokens_seen": 1044381696
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006904714142427282,
      "loss": 1.5568,
      "theoretical_loss": 3.6626467065415764,
      "tokens_seen": 1044447232
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006904513540621866,
      "loss": 1.5001,
      "theoretical_loss": 3.66262518018645,
      "tokens_seen": 1044512768
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000690431293881645,
      "loss": 1.4957,
      "theoretical_loss": 3.662603655560062,
      "tokens_seen": 1044578304
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006904112337011033,
      "loss": 1.344,
      "theoretical_loss": 3.6625821326621653,
      "tokens_seen": 1044643840
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006903911735205617,
      "loss": 1.4376,
      "theoretical_loss": 3.6625606114925113,
      "tokens_seen": 1044709376
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006903711133400202,
      "loss": 1.6842,
      "theoretical_loss": 3.6625390920508543,
      "tokens_seen": 1044774912
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006903510531594784,
      "loss": 1.575,
      "theoretical_loss": 3.6625175743369462,
      "tokens_seen": 1044840448
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006903309929789369,
      "loss": 1.5313,
      "theoretical_loss": 3.6624960583505404,
      "tokens_seen": 1044905984
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006903109327983951,
      "loss": 1.4198,
      "theoretical_loss": 3.66247454409139,
      "tokens_seen": 1044971520
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006902908726178536,
      "loss": 1.4174,
      "theoretical_loss": 3.6624530315592483,
      "tokens_seen": 1045037056
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000690270812437312,
      "loss": 1.5229,
      "theoretical_loss": 3.6624315207538674,
      "tokens_seen": 1045102592
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006902507522567703,
      "loss": 1.4939,
      "theoretical_loss": 3.662410011675001,
      "tokens_seen": 1045168128
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006902306920762287,
      "loss": 1.5741,
      "theoretical_loss": 3.6623885043224025,
      "tokens_seen": 1045233664
    },
    {
      "epoch": 0.32,
      "objective/train/advantage_avg": 0.49236148595809937,
      "objective/train/docs_used": 383342,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4417933225631714,
      "objective/train/original_loss": 1.4417933225631714,
      "objective/train/theoretical_loss": 3.6623669986958247,
      "objective/train/tokens_used": 1065759200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24337539076805115,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504733324050903,
      "objective/train/weighted_lm_loss": 1.5144271850585938,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6623669986958247,
      "tokens_seen": 1045299200
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006902106318956872,
      "loss": 1.5511,
      "theoretical_loss": 3.6623669986958247,
      "tokens_seen": 1045299200
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006901905717151454,
      "loss": 1.5429,
      "theoretical_loss": 3.662345494795021,
      "tokens_seen": 1045364736
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006901705115346039,
      "loss": 1.4595,
      "theoretical_loss": 3.6623239926197444,
      "tokens_seen": 1045430272
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006901504513540621,
      "loss": 1.5152,
      "theoretical_loss": 3.662302492169749,
      "tokens_seen": 1045495808
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006901303911735206,
      "loss": 1.4229,
      "theoretical_loss": 3.662280993444788,
      "tokens_seen": 1045561344
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000690110330992979,
      "loss": 1.5675,
      "theoretical_loss": 3.662259496444615,
      "tokens_seen": 1045626880
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006900902708124373,
      "loss": 1.5579,
      "theoretical_loss": 3.6622380011689826,
      "tokens_seen": 1045692416
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006900702106318957,
      "loss": 1.4674,
      "theoretical_loss": 3.6622165076176456,
      "tokens_seen": 1045757952
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006900501504513541,
      "loss": 1.5394,
      "theoretical_loss": 3.662195015790357,
      "tokens_seen": 1045823488
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006900300902708124,
      "loss": 1.489,
      "theoretical_loss": 3.6621735256868706,
      "tokens_seen": 1045889024
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006900100300902709,
      "loss": 1.5196,
      "theoretical_loss": 3.66215203730694,
      "tokens_seen": 1045954560
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006899899699097291,
      "loss": 1.4826,
      "theoretical_loss": 3.6621305506503194,
      "tokens_seen": 1046020096
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006899699097291876,
      "loss": 1.5183,
      "theoretical_loss": 3.6621090657167623,
      "tokens_seen": 1046085632
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006899498495486459,
      "loss": 1.3797,
      "theoretical_loss": 3.6620875825060226,
      "tokens_seen": 1046151168
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006899297893681043,
      "loss": 1.5852,
      "theoretical_loss": 3.6620661010178543,
      "tokens_seen": 1046216704
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006899097291875627,
      "loss": 1.4522,
      "theoretical_loss": 3.6620446212520115,
      "tokens_seen": 1046282240
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006898896690070211,
      "loss": 1.4966,
      "theoretical_loss": 3.6620231432082484,
      "tokens_seen": 1046347776
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006898696088264794,
      "loss": 1.517,
      "theoretical_loss": 3.662001666886318,
      "tokens_seen": 1046413312
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006898495486459379,
      "loss": 1.5683,
      "theoretical_loss": 3.6619801922859763,
      "tokens_seen": 1046478848
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006898294884653962,
      "loss": 1.4073,
      "theoretical_loss": 3.6619587194069756,
      "tokens_seen": 1046544384
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006898094282848546,
      "loss": 1.419,
      "theoretical_loss": 3.661937248249071,
      "tokens_seen": 1046609920
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006897893681043129,
      "loss": 1.5707,
      "theoretical_loss": 3.661915778812017,
      "tokens_seen": 1046675456
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006897693079237713,
      "loss": 1.5442,
      "theoretical_loss": 3.661894311095568,
      "tokens_seen": 1046740992
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006897492477432297,
      "loss": 1.6185,
      "theoretical_loss": 3.661872845099478,
      "tokens_seen": 1046806528
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006897291875626881,
      "loss": 1.4944,
      "theoretical_loss": 3.661851380823502,
      "tokens_seen": 1046872064
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006897091273821464,
      "loss": 1.4485,
      "theoretical_loss": 3.6618299182673932,
      "tokens_seen": 1046937600
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006896890672016048,
      "loss": 1.5839,
      "theoretical_loss": 3.6618084574309075,
      "tokens_seen": 1047003136
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006896690070210632,
      "loss": 1.4672,
      "theoretical_loss": 3.661786998313799,
      "tokens_seen": 1047068672
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006896489468405216,
      "loss": 1.4796,
      "theoretical_loss": 3.6617655409158223,
      "tokens_seen": 1047134208
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006896288866599799,
      "loss": 1.5932,
      "theoretical_loss": 3.661744085236732,
      "tokens_seen": 1047199744
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006896088264794384,
      "loss": 1.3986,
      "theoretical_loss": 3.6617226312762834,
      "tokens_seen": 1047265280
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006895887662988967,
      "loss": 1.442,
      "theoretical_loss": 3.6617011790342304,
      "tokens_seen": 1047330816
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006895687061183551,
      "loss": 1.4367,
      "theoretical_loss": 3.661679728510329,
      "tokens_seen": 1047396352
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006895486459378134,
      "loss": 1.5376,
      "theoretical_loss": 3.6616582797043327,
      "tokens_seen": 1047461888
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006895285857572718,
      "loss": 1.4909,
      "theoretical_loss": 3.6616368326159976,
      "tokens_seen": 1047527424
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006895085255767302,
      "loss": 1.4646,
      "theoretical_loss": 3.661615387245078,
      "tokens_seen": 1047592960
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006894884653961886,
      "loss": 1.6508,
      "theoretical_loss": 3.6615939435913294,
      "tokens_seen": 1047658496
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006894684052156469,
      "loss": 1.5179,
      "theoretical_loss": 3.661572501654507,
      "tokens_seen": 1047724032
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006894483450351054,
      "loss": 1.4801,
      "theoretical_loss": 3.6615510614343654,
      "tokens_seen": 1047789568
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006894282848545636,
      "loss": 1.4365,
      "theoretical_loss": 3.66152962293066,
      "tokens_seen": 1047855104
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006894082246740221,
      "loss": 1.4704,
      "theoretical_loss": 3.6615081861431467,
      "tokens_seen": 1047920640
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006893881644934804,
      "loss": 1.3453,
      "theoretical_loss": 3.6614867510715796,
      "tokens_seen": 1047986176
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006893681043129388,
      "loss": 1.5975,
      "theoretical_loss": 3.661465317715715,
      "tokens_seen": 1048051712
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006893480441323972,
      "loss": 1.4847,
      "theoretical_loss": 3.6614438860753076,
      "tokens_seen": 1048117248
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006893279839518555,
      "loss": 1.4927,
      "theoretical_loss": 3.6614224561501136,
      "tokens_seen": 1048182784
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006893079237713139,
      "loss": 1.5698,
      "theoretical_loss": 3.6614010279398883,
      "tokens_seen": 1048248320
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006892878635907724,
      "loss": 1.4156,
      "theoretical_loss": 3.6613796014443865,
      "tokens_seen": 1048313856
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006892678034102306,
      "loss": 1.4809,
      "theoretical_loss": 3.6613581766633647,
      "tokens_seen": 1048379392
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006892477432296891,
      "loss": 1.4511,
      "theoretical_loss": 3.6613367535965784,
      "tokens_seen": 1048444928
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006892276830491475,
      "loss": 1.5007,
      "theoretical_loss": 3.661315332243783,
      "tokens_seen": 1048510464
    },
    {
      "epoch": 0.32,
      "objective/train/advantage_avg": 0.4936440587043762,
      "objective/train/docs_used": 384505,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.528617024421692,
      "objective/train/original_loss": 1.5286171436309814,
      "objective/train/theoretical_loss": 3.661293912604734,
      "objective/train/tokens_used": 1069036000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24397923052310944,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506048202514648,
      "objective/train/weighted_lm_loss": 1.6059162616729736,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.661293912604734,
      "tokens_seen": 1048576000
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006892076228686058,
      "loss": 1.4886,
      "theoretical_loss": 3.661293912604734,
      "tokens_seen": 1048576000
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006891875626880643,
      "loss": 1.3985,
      "theoretical_loss": 3.661272494679188,
      "tokens_seen": 1048641536
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006891675025075225,
      "loss": 1.4851,
      "theoretical_loss": 3.6612510784669006,
      "tokens_seen": 1048707072
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000689147442326981,
      "loss": 1.577,
      "theoretical_loss": 3.661229663967627,
      "tokens_seen": 1048772608
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006891273821464394,
      "loss": 1.5647,
      "theoretical_loss": 3.661208251181124,
      "tokens_seen": 1048838144
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006891073219658977,
      "loss": 1.3858,
      "theoretical_loss": 3.6611868401071472,
      "tokens_seen": 1048903680
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006890872617853561,
      "loss": 1.545,
      "theoretical_loss": 3.661165430745453,
      "tokens_seen": 1048969216
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006890672016048145,
      "loss": 1.4531,
      "theoretical_loss": 3.661144023095797,
      "tokens_seen": 1049034752
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006890471414242728,
      "loss": 1.4823,
      "theoretical_loss": 3.6611226171579356,
      "tokens_seen": 1049100288
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006890270812437313,
      "loss": 1.4765,
      "theoretical_loss": 3.661101212931625,
      "tokens_seen": 1049165824
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006890070210631895,
      "loss": 1.4781,
      "theoretical_loss": 3.661079810416622,
      "tokens_seen": 1049231360
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000688986960882648,
      "loss": 1.4927,
      "theoretical_loss": 3.6610584096126813,
      "tokens_seen": 1049296896
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006889669007021063,
      "loss": 1.5936,
      "theoretical_loss": 3.6610370105195607,
      "tokens_seen": 1049362432
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006889468405215647,
      "loss": 1.5236,
      "theoretical_loss": 3.6610156131370166,
      "tokens_seen": 1049427968
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006889267803410231,
      "loss": 1.5539,
      "theoretical_loss": 3.660994217464805,
      "tokens_seen": 1049493504
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006889067201604815,
      "loss": 1.5861,
      "theoretical_loss": 3.660972823502682,
      "tokens_seen": 1049559040
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006888866599799398,
      "loss": 1.502,
      "theoretical_loss": 3.660951431250405,
      "tokens_seen": 1049624576
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006888665997993983,
      "loss": 1.4376,
      "theoretical_loss": 3.66093004070773,
      "tokens_seen": 1049690112
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006888465396188566,
      "loss": 1.5028,
      "theoretical_loss": 3.6609086518744136,
      "tokens_seen": 1049755648
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000688826479438315,
      "loss": 1.3377,
      "theoretical_loss": 3.6608872647502126,
      "tokens_seen": 1049821184
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006888064192577733,
      "loss": 1.5266,
      "theoretical_loss": 3.6608658793348847,
      "tokens_seen": 1049886720
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006887863590772317,
      "loss": 1.5236,
      "theoretical_loss": 3.6608444956281856,
      "tokens_seen": 1049952256
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006887662988966901,
      "loss": 1.5458,
      "theoretical_loss": 3.6608231136298723,
      "tokens_seen": 1050017792
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006887462387161485,
      "loss": 1.4751,
      "theoretical_loss": 3.660801733339702,
      "tokens_seen": 1050083328
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006887261785356068,
      "loss": 1.5122,
      "theoretical_loss": 3.6607803547574314,
      "tokens_seen": 1050148864
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006887061183550652,
      "loss": 1.4909,
      "theoretical_loss": 3.6607589778828173,
      "tokens_seen": 1050214400
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006886860581745236,
      "loss": 1.6472,
      "theoretical_loss": 3.660737602715617,
      "tokens_seen": 1050279936
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000688665997993982,
      "loss": 1.5791,
      "theoretical_loss": 3.660716229255588,
      "tokens_seen": 1050345472
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006886459378134403,
      "loss": 1.4591,
      "theoretical_loss": 3.660694857502487,
      "tokens_seen": 1050411008
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006886258776328988,
      "loss": 1.4805,
      "theoretical_loss": 3.6606734874560716,
      "tokens_seen": 1050476544
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006886058174523571,
      "loss": 1.5175,
      "theoretical_loss": 3.660652119116098,
      "tokens_seen": 1050542080
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006885857572718155,
      "loss": 1.481,
      "theoretical_loss": 3.660630752482325,
      "tokens_seen": 1050607616
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006885656970912738,
      "loss": 1.4879,
      "theoretical_loss": 3.660609387554509,
      "tokens_seen": 1050673152
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006885456369107322,
      "loss": 1.4184,
      "theoretical_loss": 3.6605880243324074,
      "tokens_seen": 1050738688
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006885255767301906,
      "loss": 1.5794,
      "theoretical_loss": 3.6605666628157776,
      "tokens_seen": 1050804224
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000688505516549649,
      "loss": 1.6354,
      "theoretical_loss": 3.660545303004378,
      "tokens_seen": 1050869760
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006884854563691073,
      "loss": 1.6028,
      "theoretical_loss": 3.6605239448979647,
      "tokens_seen": 1050935296
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006884653961885658,
      "loss": 1.5436,
      "theoretical_loss": 3.6605025884962963,
      "tokens_seen": 1051000832
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000688445336008024,
      "loss": 1.454,
      "theoretical_loss": 3.66048123379913,
      "tokens_seen": 1051066368
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006884252758274825,
      "loss": 1.5388,
      "theoretical_loss": 3.660459880806224,
      "tokens_seen": 1051131904
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006884052156469408,
      "loss": 1.5936,
      "theoretical_loss": 3.660438529517336,
      "tokens_seen": 1051197440
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006883851554663992,
      "loss": 1.5103,
      "theoretical_loss": 3.6604171799322227,
      "tokens_seen": 1051262976
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006883650952858576,
      "loss": 1.5187,
      "theoretical_loss": 3.6603958320506433,
      "tokens_seen": 1051328512
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006883450351053159,
      "loss": 1.5919,
      "theoretical_loss": 3.660374485872355,
      "tokens_seen": 1051394048
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006883249749247743,
      "loss": 1.6146,
      "theoretical_loss": 3.660353141397116,
      "tokens_seen": 1051459584
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006883049147442328,
      "loss": 1.5499,
      "theoretical_loss": 3.6603317986246844,
      "tokens_seen": 1051525120
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000688284854563691,
      "loss": 1.6005,
      "theoretical_loss": 3.660310457554818,
      "tokens_seen": 1051590656
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006882647943831495,
      "loss": 1.4685,
      "theoretical_loss": 3.6602891181872748,
      "tokens_seen": 1051656192
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006882447342026079,
      "loss": 1.6144,
      "theoretical_loss": 3.6602677805218127,
      "tokens_seen": 1051721728
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006882246740220662,
      "loss": 1.5808,
      "theoretical_loss": 3.66024644455819,
      "tokens_seen": 1051787264
    },
    {
      "epoch": 0.32,
      "objective/train/advantage_avg": 0.49428632855415344,
      "objective/train/docs_used": 385698,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.416479229927063,
      "objective/train/original_loss": 1.416479229927063,
      "objective/train/theoretical_loss": 3.660225110296166,
      "objective/train/tokens_used": 1072312800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24495668709278107,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506739616394043,
      "objective/train/weighted_lm_loss": 1.488330602645874,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.660225110296166,
      "tokens_seen": 1051852800
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006882046138415246,
      "loss": 1.4712,
      "theoretical_loss": 3.660225110296166,
      "tokens_seen": 1051852800
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006881845536609829,
      "loss": 1.5784,
      "theoretical_loss": 3.660203777735498,
      "tokens_seen": 1051918336
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006881644934804413,
      "loss": 1.5351,
      "theoretical_loss": 3.660182446875944,
      "tokens_seen": 1051983872
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006881444332998998,
      "loss": 1.4084,
      "theoretical_loss": 3.6601611177172635,
      "tokens_seen": 1052049408
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000688124373119358,
      "loss": 1.48,
      "theoretical_loss": 3.660139790259214,
      "tokens_seen": 1052114944
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006881043129388165,
      "loss": 1.4048,
      "theoretical_loss": 3.6601184645015543,
      "tokens_seen": 1052180480
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006880842527582748,
      "loss": 1.4018,
      "theoretical_loss": 3.6600971404440434,
      "tokens_seen": 1052246016
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006880641925777332,
      "loss": 1.4766,
      "theoretical_loss": 3.660075818086439,
      "tokens_seen": 1052311552
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006880441323971916,
      "loss": 1.343,
      "theoretical_loss": 3.6600544974285,
      "tokens_seen": 1052377088
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00068802407221665,
      "loss": 1.5598,
      "theoretical_loss": 3.6600331784699858,
      "tokens_seen": 1052442624
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006880040120361083,
      "loss": 1.4517,
      "theoretical_loss": 3.660011861210654,
      "tokens_seen": 1052508160
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006879839518555667,
      "loss": 1.4747,
      "theoretical_loss": 3.659990545650264,
      "tokens_seen": 1052573696
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000687963891675025,
      "loss": 1.4082,
      "theoretical_loss": 3.6599692317885753,
      "tokens_seen": 1052639232
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006879438314944835,
      "loss": 1.4592,
      "theoretical_loss": 3.6599479196253455,
      "tokens_seen": 1052704768
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006879237713139418,
      "loss": 1.4763,
      "theoretical_loss": 3.659926609160334,
      "tokens_seen": 1052770304
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006879037111334002,
      "loss": 1.4256,
      "theoretical_loss": 3.6599053003933006,
      "tokens_seen": 1052835840
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006878836509528586,
      "loss": 1.4493,
      "theoretical_loss": 3.6598839933240024,
      "tokens_seen": 1052901376
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000687863590772317,
      "loss": 1.3854,
      "theoretical_loss": 3.6598626879522005,
      "tokens_seen": 1052966912
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006878435305917753,
      "loss": 1.4463,
      "theoretical_loss": 3.6598413842776534,
      "tokens_seen": 1053032448
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006878234704112337,
      "loss": 1.5215,
      "theoretical_loss": 3.6598200823001195,
      "tokens_seen": 1053097984
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000687803410230692,
      "loss": 1.4254,
      "theoretical_loss": 3.659798782019359,
      "tokens_seen": 1053163520
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006877833500501505,
      "loss": 1.4137,
      "theoretical_loss": 3.65977748343513,
      "tokens_seen": 1053229056
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006877632898696088,
      "loss": 1.5841,
      "theoretical_loss": 3.6597561865471935,
      "tokens_seen": 1053294592
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006877432296890672,
      "loss": 1.566,
      "theoretical_loss": 3.6597348913553076,
      "tokens_seen": 1053360128
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006877231695085255,
      "loss": 1.7329,
      "theoretical_loss": 3.6597135978592323,
      "tokens_seen": 1053425664
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000687703109327984,
      "loss": 1.5323,
      "theoretical_loss": 3.6596923060587265,
      "tokens_seen": 1053491200
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006876830491474424,
      "loss": 1.4816,
      "theoretical_loss": 3.6596710159535504,
      "tokens_seen": 1053556736
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006876629889669007,
      "loss": 1.5308,
      "theoretical_loss": 3.6596497275434627,
      "tokens_seen": 1053622272
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006876429287863592,
      "loss": 1.4878,
      "theoretical_loss": 3.6596284408282242,
      "tokens_seen": 1053687808
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006876228686058175,
      "loss": 1.5398,
      "theoretical_loss": 3.6596071558075938,
      "tokens_seen": 1053753344
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006876028084252759,
      "loss": 1.5751,
      "theoretical_loss": 3.659585872481331,
      "tokens_seen": 1053818880
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006875827482447342,
      "loss": 1.5503,
      "theoretical_loss": 3.6595645908491963,
      "tokens_seen": 1053884416
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006875626880641926,
      "loss": 1.5497,
      "theoretical_loss": 3.6595433109109488,
      "tokens_seen": 1053949952
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000687542627883651,
      "loss": 1.56,
      "theoretical_loss": 3.6595220326663487,
      "tokens_seen": 1054015488
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006875225677031094,
      "loss": 1.4684,
      "theoretical_loss": 3.659500756115156,
      "tokens_seen": 1054081024
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006875025075225677,
      "loss": 1.3854,
      "theoretical_loss": 3.6594794812571303,
      "tokens_seen": 1054146560
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006874824473420262,
      "loss": 1.5226,
      "theoretical_loss": 3.6594582080920324,
      "tokens_seen": 1054212096
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006874623871614844,
      "loss": 1.5548,
      "theoretical_loss": 3.6594369366196213,
      "tokens_seen": 1054277632
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006874423269809429,
      "loss": 1.5202,
      "theoretical_loss": 3.659415666839658,
      "tokens_seen": 1054343168
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006874222668004012,
      "loss": 1.4764,
      "theoretical_loss": 3.6593943987519015,
      "tokens_seen": 1054408704
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006874022066198596,
      "loss": 1.472,
      "theoretical_loss": 3.6593731323561136,
      "tokens_seen": 1054474240
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000687382146439318,
      "loss": 1.4419,
      "theoretical_loss": 3.6593518676520533,
      "tokens_seen": 1054539776
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006873620862587763,
      "loss": 1.5018,
      "theoretical_loss": 3.6593306046394813,
      "tokens_seen": 1054605312
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006873420260782347,
      "loss": 1.5212,
      "theoretical_loss": 3.6593093433181583,
      "tokens_seen": 1054670848
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006873219658976932,
      "loss": 1.5284,
      "theoretical_loss": 3.6592880836878443,
      "tokens_seen": 1054736384
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006873019057171514,
      "loss": 1.4848,
      "theoretical_loss": 3.6592668257482996,
      "tokens_seen": 1054801920
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006872818455366099,
      "loss": 1.5609,
      "theoretical_loss": 3.6592455694992854,
      "tokens_seen": 1054867456
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006872617853560683,
      "loss": 1.3198,
      "theoretical_loss": 3.659224314940561,
      "tokens_seen": 1054932992
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006872417251755266,
      "loss": 1.5111,
      "theoretical_loss": 3.6592030620718887,
      "tokens_seen": 1054998528
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000687221664994985,
      "loss": 1.506,
      "theoretical_loss": 3.6591818108930276,
      "tokens_seen": 1055064064
    },
    {
      "epoch": 0.32,
      "objective/train/advantage_avg": 0.49314162135124207,
      "objective/train/docs_used": 386814,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3797991275787354,
      "objective/train/original_loss": 1.3797991275787354,
      "objective/train/theoretical_loss": 3.659160561403739,
      "objective/train/tokens_used": 1075589600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24348647892475128,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505520105361938,
      "objective/train/weighted_lm_loss": 1.4494986534118652,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.659160561403739,
      "tokens_seen": 1055129600
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006872016048144433,
      "loss": 1.3574,
      "theoretical_loss": 3.659160561403739,
      "tokens_seen": 1055129600
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006871815446339017,
      "loss": 1.4263,
      "theoretical_loss": 3.6591393136037835,
      "tokens_seen": 1055195136
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006871614844533602,
      "loss": 1.5183,
      "theoretical_loss": 3.6591180674929227,
      "tokens_seen": 1055260672
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006871414242728184,
      "loss": 1.4252,
      "theoretical_loss": 3.659096823070916,
      "tokens_seen": 1055326208
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006871213640922769,
      "loss": 1.4701,
      "theoretical_loss": 3.6590755803375252,
      "tokens_seen": 1055391744
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006871013039117352,
      "loss": 1.5129,
      "theoretical_loss": 3.6590543392925117,
      "tokens_seen": 1055457280
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006870812437311936,
      "loss": 1.4862,
      "theoretical_loss": 3.6590330999356353,
      "tokens_seen": 1055522816
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000687061183550652,
      "loss": 1.5446,
      "theoretical_loss": 3.6590118622666576,
      "tokens_seen": 1055588352
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006870411233701103,
      "loss": 1.4284,
      "theoretical_loss": 3.65899062628534,
      "tokens_seen": 1055653888
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006870210631895687,
      "loss": 1.4342,
      "theoretical_loss": 3.658969391991443,
      "tokens_seen": 1055719424
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006870010030090271,
      "loss": 1.5281,
      "theoretical_loss": 3.658948159384728,
      "tokens_seen": 1055784960
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006869809428284854,
      "loss": 1.4855,
      "theoretical_loss": 3.6589269284649566,
      "tokens_seen": 1055850496
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006869608826479439,
      "loss": 1.4876,
      "theoretical_loss": 3.65890569923189,
      "tokens_seen": 1055916032
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006869408224674022,
      "loss": 1.4669,
      "theoretical_loss": 3.658884471685289,
      "tokens_seen": 1055981568
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006869207622868606,
      "loss": 1.4647,
      "theoretical_loss": 3.6588632458249153,
      "tokens_seen": 1056047104
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000686900702106319,
      "loss": 1.5049,
      "theoretical_loss": 3.6588420216505306,
      "tokens_seen": 1056112640
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006868806419257774,
      "loss": 1.6379,
      "theoretical_loss": 3.658820799161896,
      "tokens_seen": 1056178176
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006868605817452357,
      "loss": 1.5341,
      "theoretical_loss": 3.658799578358773,
      "tokens_seen": 1056243712
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006868405215646941,
      "loss": 1.455,
      "theoretical_loss": 3.658778359240923,
      "tokens_seen": 1056309248
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006868204613841524,
      "loss": 1.4932,
      "theoretical_loss": 3.658757141808109,
      "tokens_seen": 1056374784
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006868004012036109,
      "loss": 1.6349,
      "theoretical_loss": 3.65873592606009,
      "tokens_seen": 1056440320
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006867803410230692,
      "loss": 1.4395,
      "theoretical_loss": 3.6587147119966303,
      "tokens_seen": 1056505856
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006867602808425276,
      "loss": 1.4584,
      "theoretical_loss": 3.6586934996174905,
      "tokens_seen": 1056571392
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006867402206619859,
      "loss": 1.4452,
      "theoretical_loss": 3.658672288922432,
      "tokens_seen": 1056636928
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006867201604814444,
      "loss": 1.4153,
      "theoretical_loss": 3.658651079911218,
      "tokens_seen": 1056702464
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006867001003009027,
      "loss": 1.4248,
      "theoretical_loss": 3.6586298725836093,
      "tokens_seen": 1056768000
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006866800401203611,
      "loss": 1.5541,
      "theoretical_loss": 3.658608666939368,
      "tokens_seen": 1056833536
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006866599799398194,
      "loss": 1.4877,
      "theoretical_loss": 3.6585874629782564,
      "tokens_seen": 1056899072
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006866399197592778,
      "loss": 1.5577,
      "theoretical_loss": 3.658566260700036,
      "tokens_seen": 1056964608
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006866198595787362,
      "loss": 1.4538,
      "theoretical_loss": 3.65854506010447,
      "tokens_seen": 1057030144
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006865997993981946,
      "loss": 1.4298,
      "theoretical_loss": 3.65852386119132,
      "tokens_seen": 1057095680
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006865797392176529,
      "loss": 1.5163,
      "theoretical_loss": 3.658502663960347,
      "tokens_seen": 1057161216
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006865596790371114,
      "loss": 1.5903,
      "theoretical_loss": 3.658481468411315,
      "tokens_seen": 1057226752
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006865396188565697,
      "loss": 1.5589,
      "theoretical_loss": 3.6584602745439856,
      "tokens_seen": 1057292288
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006865195586760281,
      "loss": 1.4396,
      "theoretical_loss": 3.658439082358121,
      "tokens_seen": 1057357824
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006864994984954865,
      "loss": 1.3816,
      "theoretical_loss": 3.6584178918534844,
      "tokens_seen": 1057423360
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006864794383149448,
      "loss": 1.4507,
      "theoretical_loss": 3.6583967030298368,
      "tokens_seen": 1057488896
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006864593781344032,
      "loss": 1.5478,
      "theoretical_loss": 3.6583755158869415,
      "tokens_seen": 1057554432
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006864393179538616,
      "loss": 1.552,
      "theoretical_loss": 3.6583543304245616,
      "tokens_seen": 1057619968
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006864192577733199,
      "loss": 1.5148,
      "theoretical_loss": 3.6583331466424585,
      "tokens_seen": 1057685504
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006863991975927784,
      "loss": 1.4941,
      "theoretical_loss": 3.6583119645403954,
      "tokens_seen": 1057751040
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006863791374122366,
      "loss": 1.516,
      "theoretical_loss": 3.6582907841181354,
      "tokens_seen": 1057816576
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006863590772316951,
      "loss": 1.4102,
      "theoretical_loss": 3.6582696053754407,
      "tokens_seen": 1057882112
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006863390170511535,
      "loss": 1.4113,
      "theoretical_loss": 3.658248428312074,
      "tokens_seen": 1057947648
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006863189568706118,
      "loss": 1.4742,
      "theoretical_loss": 3.658227252927799,
      "tokens_seen": 1058013184
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006862988966900702,
      "loss": 1.3787,
      "theoretical_loss": 3.658206079222377,
      "tokens_seen": 1058078720
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006862788365095287,
      "loss": 1.501,
      "theoretical_loss": 3.6581849071955723,
      "tokens_seen": 1058144256
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006862587763289869,
      "loss": 1.4006,
      "theoretical_loss": 3.658163736847147,
      "tokens_seen": 1058209792
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006862387161484454,
      "loss": 1.4734,
      "theoretical_loss": 3.6581425681768653,
      "tokens_seen": 1058275328
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006862186559679036,
      "loss": 1.4434,
      "theoretical_loss": 3.6581214011844887,
      "tokens_seen": 1058340864
    },
    {
      "epoch": 0.32,
      "objective/train/advantage_avg": 0.49355921149253845,
      "objective/train/docs_used": 388023,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3658466339111328,
      "objective/train/original_loss": 1.365846872329712,
      "objective/train/theoretical_loss": 3.6581002358697816,
      "objective/train/tokens_used": 1078866400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24390853941440582,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050595760345459,
      "objective/train/weighted_lm_loss": 1.4350497722625732,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6581002358697816,
      "tokens_seen": 1058406400
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006861985957873621,
      "loss": 1.3981,
      "theoretical_loss": 3.6581002358697816,
      "tokens_seen": 1058406400
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006861785356068206,
      "loss": 1.465,
      "theoretical_loss": 3.658079072232506,
      "tokens_seen": 1058471936
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006861584754262788,
      "loss": 1.5676,
      "theoretical_loss": 3.6580579102724267,
      "tokens_seen": 1058537472
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006861384152457373,
      "loss": 1.5118,
      "theoretical_loss": 3.6580367499893054,
      "tokens_seen": 1058603008
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006861183550651956,
      "loss": 1.571,
      "theoretical_loss": 3.658015591382906,
      "tokens_seen": 1058668544
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000686098294884654,
      "loss": 1.4159,
      "theoretical_loss": 3.657994434452992,
      "tokens_seen": 1058734080
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006860782347041124,
      "loss": 1.5059,
      "theoretical_loss": 3.657973279199327,
      "tokens_seen": 1058799616
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006860581745235707,
      "loss": 1.491,
      "theoretical_loss": 3.657952125621674,
      "tokens_seen": 1058865152
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006860381143430291,
      "loss": 1.4379,
      "theoretical_loss": 3.6579309737197963,
      "tokens_seen": 1058930688
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006860180541624875,
      "loss": 1.5565,
      "theoretical_loss": 3.6579098234934575,
      "tokens_seen": 1058996224
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006859979939819458,
      "loss": 1.5067,
      "theoretical_loss": 3.6578886749424226,
      "tokens_seen": 1059061760
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006859779338014043,
      "loss": 1.6862,
      "theoretical_loss": 3.657867528066453,
      "tokens_seen": 1059127296
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006859578736208626,
      "loss": 1.45,
      "theoretical_loss": 3.6578463828653143,
      "tokens_seen": 1059192832
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000685937813440321,
      "loss": 1.5609,
      "theoretical_loss": 3.657825239338769,
      "tokens_seen": 1059258368
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006859177532597794,
      "loss": 1.5919,
      "theoretical_loss": 3.657804097486581,
      "tokens_seen": 1059323904
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006858976930792378,
      "loss": 1.4925,
      "theoretical_loss": 3.657782957308515,
      "tokens_seen": 1059389440
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006858776328986961,
      "loss": 1.4902,
      "theoretical_loss": 3.6577618188043344,
      "tokens_seen": 1059454976
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006858575727181545,
      "loss": 1.4122,
      "theoretical_loss": 3.657740681973803,
      "tokens_seen": 1059520512
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006858375125376128,
      "loss": 1.334,
      "theoretical_loss": 3.657719546816685,
      "tokens_seen": 1059586048
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006858174523570713,
      "loss": 1.356,
      "theoretical_loss": 3.657698413332744,
      "tokens_seen": 1059651584
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006857973921765296,
      "loss": 1.5293,
      "theoretical_loss": 3.657677281521744,
      "tokens_seen": 1059717120
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000685777331995988,
      "loss": 1.4387,
      "theoretical_loss": 3.6576561513834505,
      "tokens_seen": 1059782656
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006857572718154463,
      "loss": 1.462,
      "theoretical_loss": 3.657635022917626,
      "tokens_seen": 1059848192
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006857372116349048,
      "loss": 1.5532,
      "theoretical_loss": 3.6576138961240354,
      "tokens_seen": 1059913728
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006857171514543631,
      "loss": 1.4648,
      "theoretical_loss": 3.657592771002443,
      "tokens_seen": 1059979264
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006856970912738215,
      "loss": 1.3048,
      "theoretical_loss": 3.657571647552613,
      "tokens_seen": 1060044800
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006856770310932798,
      "loss": 1.5058,
      "theoretical_loss": 3.65755052577431,
      "tokens_seen": 1060110336
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006856569709127382,
      "loss": 1.4072,
      "theoretical_loss": 3.6575294056672973,
      "tokens_seen": 1060175872
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006856369107321966,
      "loss": 1.4178,
      "theoretical_loss": 3.6575082872313414,
      "tokens_seen": 1060241408
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000685616850551655,
      "loss": 1.4991,
      "theoretical_loss": 3.6574871704662053,
      "tokens_seen": 1060306944
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006855967903711133,
      "loss": 1.4622,
      "theoretical_loss": 3.657466055371654,
      "tokens_seen": 1060372480
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006855767301905718,
      "loss": 1.5484,
      "theoretical_loss": 3.6574449419474515,
      "tokens_seen": 1060438016
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006855566700100301,
      "loss": 1.4342,
      "theoretical_loss": 3.6574238301933635,
      "tokens_seen": 1060503552
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006855366098294885,
      "loss": 1.5313,
      "theoretical_loss": 3.657402720109154,
      "tokens_seen": 1060569088
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006855165496489469,
      "loss": 1.3867,
      "theoretical_loss": 3.657381611694588,
      "tokens_seen": 1060634624
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006854964894684052,
      "loss": 1.4837,
      "theoretical_loss": 3.65736050494943,
      "tokens_seen": 1060700160
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006854764292878636,
      "loss": 1.4516,
      "theoretical_loss": 3.6573393998734454,
      "tokens_seen": 1060765696
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000685456369107322,
      "loss": 1.5309,
      "theoretical_loss": 3.657318296466398,
      "tokens_seen": 1060831232
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006854363089267803,
      "loss": 1.4986,
      "theoretical_loss": 3.6572971947280544,
      "tokens_seen": 1060896768
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006854162487462388,
      "loss": 1.5457,
      "theoretical_loss": 3.6572760946581777,
      "tokens_seen": 1060962304
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000685396188565697,
      "loss": 1.6211,
      "theoretical_loss": 3.6572549962565346,
      "tokens_seen": 1061027840
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006853761283851555,
      "loss": 1.4428,
      "theoretical_loss": 3.6572338995228892,
      "tokens_seen": 1061093376
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006853560682046139,
      "loss": 1.6298,
      "theoretical_loss": 3.6572128044570067,
      "tokens_seen": 1061158912
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006853360080240722,
      "loss": 1.545,
      "theoretical_loss": 3.6571917110586525,
      "tokens_seen": 1061224448
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006853159478435306,
      "loss": 1.4266,
      "theoretical_loss": 3.6571706193275917,
      "tokens_seen": 1061289984
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006852958876629891,
      "loss": 1.5886,
      "theoretical_loss": 3.65714952926359,
      "tokens_seen": 1061355520
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006852758274824473,
      "loss": 1.3588,
      "theoretical_loss": 3.657128440866412,
      "tokens_seen": 1061421056
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006852557673019058,
      "loss": 1.4657,
      "theoretical_loss": 3.6571073541358237,
      "tokens_seen": 1061486592
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000685235707121364,
      "loss": 1.5114,
      "theoretical_loss": 3.6570862690715895,
      "tokens_seen": 1061552128
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006852156469408225,
      "loss": 1.4371,
      "theoretical_loss": 3.657065185673476,
      "tokens_seen": 1061617664
    },
    {
      "epoch": 0.32,
      "objective/train/advantage_avg": 0.4925467371940613,
      "objective/train/docs_used": 389244,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.481990933418274,
      "objective/train/original_loss": 1.481990933418274,
      "objective/train/theoretical_loss": 3.6570441039412485,
      "objective/train/tokens_used": 1082143200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24319204688072205,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504910945892334,
      "objective/train/weighted_lm_loss": 1.5568042993545532,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6570441039412485,
      "tokens_seen": 1061683200
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006851955867602809,
      "loss": 1.5446,
      "theoretical_loss": 3.6570441039412485,
      "tokens_seen": 1061683200
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006851755265797392,
      "loss": 1.5326,
      "theoretical_loss": 3.6570230238746717,
      "tokens_seen": 1061748736
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006851554663991976,
      "loss": 1.5159,
      "theoretical_loss": 3.657001945473512,
      "tokens_seen": 1061814272
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000685135406218656,
      "loss": 1.4583,
      "theoretical_loss": 3.656980868737535,
      "tokens_seen": 1061879808
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006851153460381143,
      "loss": 1.4862,
      "theoretical_loss": 3.6569597936665064,
      "tokens_seen": 1061945344
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006850952858575728,
      "loss": 1.5909,
      "theoretical_loss": 3.656938720260192,
      "tokens_seen": 1062010880
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000685075225677031,
      "loss": 1.5195,
      "theoretical_loss": 3.656917648518357,
      "tokens_seen": 1062076416
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006850551654964895,
      "loss": 1.5754,
      "theoretical_loss": 3.656896578440768,
      "tokens_seen": 1062141952
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006850351053159478,
      "loss": 1.4672,
      "theoretical_loss": 3.6568755100271897,
      "tokens_seen": 1062207488
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006850150451354062,
      "loss": 1.4793,
      "theoretical_loss": 3.656854443277389,
      "tokens_seen": 1062273024
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006849949849548646,
      "loss": 1.4074,
      "theoretical_loss": 3.6568333781911324,
      "tokens_seen": 1062338560
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000684974924774323,
      "loss": 1.5361,
      "theoretical_loss": 3.656812314768185,
      "tokens_seen": 1062404096
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006849548645937813,
      "loss": 1.5185,
      "theoretical_loss": 3.656791253008313,
      "tokens_seen": 1062469632
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006849348044132398,
      "loss": 1.4431,
      "theoretical_loss": 3.6567701929112824,
      "tokens_seen": 1062535168
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000684914744232698,
      "loss": 1.4135,
      "theoretical_loss": 3.6567491344768595,
      "tokens_seen": 1062600704
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006848946840521565,
      "loss": 1.5701,
      "theoretical_loss": 3.656728077704811,
      "tokens_seen": 1062666240
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006848746238716148,
      "loss": 1.5982,
      "theoretical_loss": 3.6567070225949028,
      "tokens_seen": 1062731776
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006848545636910732,
      "loss": 1.4142,
      "theoretical_loss": 3.656685969146901,
      "tokens_seen": 1062797312
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006848345035105316,
      "loss": 1.4323,
      "theoretical_loss": 3.656664917360572,
      "tokens_seen": 1062862848
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00068481444332999,
      "loss": 1.6165,
      "theoretical_loss": 3.6566438672356827,
      "tokens_seen": 1062928384
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006847943831494483,
      "loss": 1.4872,
      "theoretical_loss": 3.656622818771999,
      "tokens_seen": 1062993920
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006847743229689067,
      "loss": 1.4309,
      "theoretical_loss": 3.6566017719692874,
      "tokens_seen": 1063059456
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000684754262788365,
      "loss": 1.4851,
      "theoretical_loss": 3.6565807268273147,
      "tokens_seen": 1063124992
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006847342026078235,
      "loss": 1.5061,
      "theoretical_loss": 3.656559683345848,
      "tokens_seen": 1063190528
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006847141424272818,
      "loss": 1.3672,
      "theoretical_loss": 3.6565386415246524,
      "tokens_seen": 1063256064
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006846940822467402,
      "loss": 1.4877,
      "theoretical_loss": 3.656517601363496,
      "tokens_seen": 1063321600
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006846740220661986,
      "loss": 1.3376,
      "theoretical_loss": 3.656496562862145,
      "tokens_seen": 1063387136
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000684653961885657,
      "loss": 1.5182,
      "theoretical_loss": 3.6564755260203667,
      "tokens_seen": 1063452672
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006846339017051154,
      "loss": 1.5358,
      "theoretical_loss": 3.6564544908379273,
      "tokens_seen": 1063518208
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006846138415245737,
      "loss": 1.5375,
      "theoretical_loss": 3.6564334573145936,
      "tokens_seen": 1063583744
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006845937813440322,
      "loss": 1.538,
      "theoretical_loss": 3.656412425450133,
      "tokens_seen": 1063649280
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006845737211634905,
      "loss": 1.3224,
      "theoretical_loss": 3.656391395244312,
      "tokens_seen": 1063714816
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006845536609829489,
      "loss": 1.5038,
      "theoretical_loss": 3.6563703666968985,
      "tokens_seen": 1063780352
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006845336008024073,
      "loss": 1.472,
      "theoretical_loss": 3.656349339807658,
      "tokens_seen": 1063845888
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006845135406218656,
      "loss": 1.3465,
      "theoretical_loss": 3.6563283145763594,
      "tokens_seen": 1063911424
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000684493480441324,
      "loss": 1.4208,
      "theoretical_loss": 3.6563072910027685,
      "tokens_seen": 1063976960
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006844734202607825,
      "loss": 1.5458,
      "theoretical_loss": 3.656286269086653,
      "tokens_seen": 1064042496
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006844533600802407,
      "loss": 1.6352,
      "theoretical_loss": 3.656265248827781,
      "tokens_seen": 1064108032
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006844332998996992,
      "loss": 1.3112,
      "theoretical_loss": 3.6562442302259184,
      "tokens_seen": 1064173568
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006844132397191574,
      "loss": 1.6017,
      "theoretical_loss": 3.6562232132808328,
      "tokens_seen": 1064239104
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006843931795386159,
      "loss": 1.4475,
      "theoretical_loss": 3.6562021979922923,
      "tokens_seen": 1064304640
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006843731193580743,
      "loss": 1.5451,
      "theoretical_loss": 3.6561811843600642,
      "tokens_seen": 1064370176
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006843530591775326,
      "loss": 1.6365,
      "theoretical_loss": 3.656160172383915,
      "tokens_seen": 1064435712
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000684332998996991,
      "loss": 1.5228,
      "theoretical_loss": 3.6561391620636137,
      "tokens_seen": 1064501248
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006843129388164495,
      "loss": 1.464,
      "theoretical_loss": 3.6561181533989267,
      "tokens_seen": 1064566784
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006842928786359077,
      "loss": 1.4358,
      "theoretical_loss": 3.656097146389622,
      "tokens_seen": 1064632320
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006842728184553662,
      "loss": 1.4443,
      "theoretical_loss": 3.6560761410354674,
      "tokens_seen": 1064697856
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006842527582748244,
      "loss": 1.5943,
      "theoretical_loss": 3.6560551373362307,
      "tokens_seen": 1064763392
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006842326980942829,
      "loss": 1.4913,
      "theoretical_loss": 3.6560341352916796,
      "tokens_seen": 1064828928
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006842126379137413,
      "loss": 1.499,
      "theoretical_loss": 3.656013134901582,
      "tokens_seen": 1064894464
    },
    {
      "epoch": 0.32,
      "objective/train/advantage_avg": 0.4909442365169525,
      "objective/train/docs_used": 390462,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.342321753501892,
      "objective/train/original_loss": 1.342321515083313,
      "objective/train/theoretical_loss": 3.6559921361657057,
      "objective/train/tokens_used": 1085420000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24155539274215698,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503222942352295,
      "objective/train/weighted_lm_loss": 1.4100650548934937,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6559921361657057,
      "tokens_seen": 1064960000
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006841925777331996,
      "loss": 1.4922,
      "theoretical_loss": 3.6559921361657057,
      "tokens_seen": 1064960000
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000684172517552658,
      "loss": 1.4999,
      "theoretical_loss": 3.655971139083818,
      "tokens_seen": 1065025536
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006841524573721164,
      "loss": 1.4754,
      "theoretical_loss": 3.655950143655688,
      "tokens_seen": 1065091072
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006841323971915747,
      "loss": 1.559,
      "theoretical_loss": 3.655929149881082,
      "tokens_seen": 1065156608
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006841123370110332,
      "loss": 1.5765,
      "theoretical_loss": 3.6559081577597703,
      "tokens_seen": 1065222144
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006840922768304914,
      "loss": 1.4436,
      "theoretical_loss": 3.655887167291519,
      "tokens_seen": 1065287680
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006840722166499499,
      "loss": 1.4426,
      "theoretical_loss": 3.655866178476098,
      "tokens_seen": 1065353216
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006840521564694082,
      "loss": 1.5514,
      "theoretical_loss": 3.655845191313274,
      "tokens_seen": 1065418752
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006840320962888666,
      "loss": 1.4629,
      "theoretical_loss": 3.6558242058028156,
      "tokens_seen": 1065484288
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000684012036108325,
      "loss": 1.4648,
      "theoretical_loss": 3.655803221944492,
      "tokens_seen": 1065549824
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006839919759277834,
      "loss": 1.4795,
      "theoretical_loss": 3.65578223973807,
      "tokens_seen": 1065615360
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006839719157472417,
      "loss": 1.5421,
      "theoretical_loss": 3.6557612591833193,
      "tokens_seen": 1065680896
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006839518555667002,
      "loss": 1.6601,
      "theoretical_loss": 3.655740280280008,
      "tokens_seen": 1065746432
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006839317953861584,
      "loss": 1.5497,
      "theoretical_loss": 3.655719303027904,
      "tokens_seen": 1065811968
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006839117352056169,
      "loss": 1.5435,
      "theoretical_loss": 3.6556983274267765,
      "tokens_seen": 1065877504
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006838916750250752,
      "loss": 1.4548,
      "theoretical_loss": 3.6556773534763933,
      "tokens_seen": 1065943040
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006838716148445336,
      "loss": 1.5531,
      "theoretical_loss": 3.6556563811765237,
      "tokens_seen": 1066008576
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000683851554663992,
      "loss": 1.4558,
      "theoretical_loss": 3.655635410526936,
      "tokens_seen": 1066074112
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006838314944834504,
      "loss": 1.5034,
      "theoretical_loss": 3.6556144415273994,
      "tokens_seen": 1066139648
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006838114343029087,
      "loss": 1.5214,
      "theoretical_loss": 3.6555934741776825,
      "tokens_seen": 1066205184
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006837913741223671,
      "loss": 1.3325,
      "theoretical_loss": 3.6555725084775537,
      "tokens_seen": 1066270720
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006837713139418255,
      "loss": 1.5426,
      "theoretical_loss": 3.6555515444267814,
      "tokens_seen": 1066336256
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006837512537612839,
      "loss": 1.6108,
      "theoretical_loss": 3.655530582025136,
      "tokens_seen": 1066401792
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006837311935807422,
      "loss": 1.3991,
      "theoretical_loss": 3.655509621272385,
      "tokens_seen": 1066467328
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006837111334002006,
      "loss": 1.4697,
      "theoretical_loss": 3.655488662168298,
      "tokens_seen": 1066532864
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006836910732196589,
      "loss": 1.538,
      "theoretical_loss": 3.655467704712644,
      "tokens_seen": 1066598400
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006836710130391174,
      "loss": 1.455,
      "theoretical_loss": 3.6554467489051925,
      "tokens_seen": 1066663936
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006836509528585757,
      "loss": 1.5363,
      "theoretical_loss": 3.655425794745711,
      "tokens_seen": 1066729472
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006836308926780341,
      "loss": 1.4789,
      "theoretical_loss": 3.6554048422339713,
      "tokens_seen": 1066795008
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006836108324974925,
      "loss": 1.52,
      "theoretical_loss": 3.6553838913697403,
      "tokens_seen": 1066860544
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006835907723169509,
      "loss": 1.5588,
      "theoretical_loss": 3.6553629421527885,
      "tokens_seen": 1066926080
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006835707121364092,
      "loss": 1.5103,
      "theoretical_loss": 3.6553419945828844,
      "tokens_seen": 1066991616
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006835506519558677,
      "loss": 1.4987,
      "theoretical_loss": 3.6553210486597982,
      "tokens_seen": 1067057152
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006835305917753259,
      "loss": 1.5336,
      "theoretical_loss": 3.655300104383299,
      "tokens_seen": 1067122688
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006835105315947844,
      "loss": 1.5246,
      "theoretical_loss": 3.655279161753156,
      "tokens_seen": 1067188224
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006834904714142427,
      "loss": 1.4199,
      "theoretical_loss": 3.6552582207691384,
      "tokens_seen": 1067253760
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006834704112337011,
      "loss": 1.471,
      "theoretical_loss": 3.655237281431017,
      "tokens_seen": 1067319296
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006834503510531595,
      "loss": 1.3372,
      "theoretical_loss": 3.6552163437385596,
      "tokens_seen": 1067384832
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006834302908726178,
      "loss": 1.5869,
      "theoretical_loss": 3.6551954076915374,
      "tokens_seen": 1067450368
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006834102306920762,
      "loss": 1.4602,
      "theoretical_loss": 3.655174473289719,
      "tokens_seen": 1067515904
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006833901705115347,
      "loss": 1.4399,
      "theoretical_loss": 3.6551535405328752,
      "tokens_seen": 1067581440
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006833701103309929,
      "loss": 1.5269,
      "theoretical_loss": 3.6551326094207752,
      "tokens_seen": 1067646976
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006833500501504514,
      "loss": 1.5075,
      "theoretical_loss": 3.655111679953188,
      "tokens_seen": 1067712512
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006833299899699096,
      "loss": 1.4549,
      "theoretical_loss": 3.655090752129885,
      "tokens_seen": 1067778048
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006833099297893681,
      "loss": 1.555,
      "theoretical_loss": 3.655069825950635,
      "tokens_seen": 1067843584
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006832898696088265,
      "loss": 1.502,
      "theoretical_loss": 3.655048901415209,
      "tokens_seen": 1067909120
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006832698094282848,
      "loss": 1.5229,
      "theoretical_loss": 3.6550279785233757,
      "tokens_seen": 1067974656
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006832497492477432,
      "loss": 1.4417,
      "theoretical_loss": 3.655007057274906,
      "tokens_seen": 1068040192
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006832296890672017,
      "loss": 1.5655,
      "theoretical_loss": 3.654986137669569,
      "tokens_seen": 1068105728
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006832096288866599,
      "loss": 1.5192,
      "theoretical_loss": 3.6549652197071367,
      "tokens_seen": 1068171264
    },
    {
      "epoch": 0.32,
      "objective/train/advantage_avg": 0.4904010593891144,
      "objective/train/docs_used": 391774,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5029762983322144,
      "objective/train/original_loss": 1.502976417541504,
      "objective/train/theoretical_loss": 3.654944303387378,
      "objective/train/tokens_used": 1088696800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24100598692893982,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502653121948242,
      "objective/train/weighted_lm_loss": 1.5786162614822388,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.654944303387378,
      "tokens_seen": 1068236800
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006831895687061184,
      "loss": 1.4934,
      "theoretical_loss": 3.654944303387378,
      "tokens_seen": 1068236800
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006831695085255768,
      "loss": 1.5042,
      "theoretical_loss": 3.6549233887100634,
      "tokens_seen": 1068302336
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006831494483450351,
      "loss": 1.4076,
      "theoretical_loss": 3.6549024756749624,
      "tokens_seen": 1068367872
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006831293881644936,
      "loss": 1.5504,
      "theoretical_loss": 3.6548815642818466,
      "tokens_seen": 1068433408
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006831093279839518,
      "loss": 1.5171,
      "theoretical_loss": 3.654860654530486,
      "tokens_seen": 1068498944
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006830892678034103,
      "loss": 1.4856,
      "theoretical_loss": 3.654839746420651,
      "tokens_seen": 1068564480
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006830692076228686,
      "loss": 1.4848,
      "theoretical_loss": 3.6548188399521115,
      "tokens_seen": 1068630016
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000683049147442327,
      "loss": 1.5672,
      "theoretical_loss": 3.6547979351246385,
      "tokens_seen": 1068695552
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006830290872617854,
      "loss": 1.528,
      "theoretical_loss": 3.6547770319380026,
      "tokens_seen": 1068761088
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006830090270812438,
      "loss": 1.3953,
      "theoretical_loss": 3.6547561303919744,
      "tokens_seen": 1068826624
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006829889669007021,
      "loss": 1.3906,
      "theoretical_loss": 3.6547352304863248,
      "tokens_seen": 1068892160
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006829689067201606,
      "loss": 1.5352,
      "theoretical_loss": 3.6547143322208235,
      "tokens_seen": 1068957696
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006829488465396188,
      "loss": 1.5157,
      "theoretical_loss": 3.6546934355952425,
      "tokens_seen": 1069023232
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006829287863590773,
      "loss": 1.5573,
      "theoretical_loss": 3.6546725406093516,
      "tokens_seen": 1069088768
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006829087261785356,
      "loss": 1.5419,
      "theoretical_loss": 3.654651647262922,
      "tokens_seen": 1069154304
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000682888665997994,
      "loss": 1.4173,
      "theoretical_loss": 3.6546307555557247,
      "tokens_seen": 1069219840
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006828686058174524,
      "loss": 1.5487,
      "theoretical_loss": 3.6546098654875303,
      "tokens_seen": 1069285376
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006828485456369108,
      "loss": 1.6328,
      "theoretical_loss": 3.6545889770581104,
      "tokens_seen": 1069350912
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006828284854563691,
      "loss": 1.4297,
      "theoretical_loss": 3.6545680902672353,
      "tokens_seen": 1069416448
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006828084252758275,
      "loss": 1.5311,
      "theoretical_loss": 3.6545472051146763,
      "tokens_seen": 1069481984
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006827883650952859,
      "loss": 1.474,
      "theoretical_loss": 3.654526321600205,
      "tokens_seen": 1069547520
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006827683049147443,
      "loss": 1.4706,
      "theoretical_loss": 3.654505439723592,
      "tokens_seen": 1069613056
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006827482447342026,
      "loss": 1.5452,
      "theoretical_loss": 3.654484559484609,
      "tokens_seen": 1069678592
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000682728184553661,
      "loss": 1.557,
      "theoretical_loss": 3.6544636808830266,
      "tokens_seen": 1069744128
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006827081243731193,
      "loss": 1.3685,
      "theoretical_loss": 3.6544428039186165,
      "tokens_seen": 1069809664
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006826880641925778,
      "loss": 1.4334,
      "theoretical_loss": 3.65442192859115,
      "tokens_seen": 1069875200
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006826680040120361,
      "loss": 1.456,
      "theoretical_loss": 3.654401054900399,
      "tokens_seen": 1069940736
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006826479438314945,
      "loss": 1.4927,
      "theoretical_loss": 3.6543801828461335,
      "tokens_seen": 1070006272
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006826278836509529,
      "loss": 1.4641,
      "theoretical_loss": 3.6543593124281264,
      "tokens_seen": 1070071808
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006826078234704113,
      "loss": 1.5,
      "theoretical_loss": 3.6543384436461483,
      "tokens_seen": 1070137344
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006825877632898696,
      "loss": 1.5629,
      "theoretical_loss": 3.6543175764999716,
      "tokens_seen": 1070202880
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006825677031093281,
      "loss": 1.5135,
      "theoretical_loss": 3.6542967109893674,
      "tokens_seen": 1070268416
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006825476429287863,
      "loss": 1.4799,
      "theoretical_loss": 3.654275847114107,
      "tokens_seen": 1070333952
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006825275827482448,
      "loss": 1.5589,
      "theoretical_loss": 3.6542549848739627,
      "tokens_seen": 1070399488
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006825075225677031,
      "loss": 1.4951,
      "theoretical_loss": 3.654234124268706,
      "tokens_seen": 1070465024
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006824874623871615,
      "loss": 1.6036,
      "theoretical_loss": 3.6542132652981096,
      "tokens_seen": 1070530560
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006824674022066199,
      "loss": 1.5822,
      "theoretical_loss": 3.6541924079619443,
      "tokens_seen": 1070596096
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006824473420260782,
      "loss": 1.5937,
      "theoretical_loss": 3.6541715522599816,
      "tokens_seen": 1070661632
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006824272818455366,
      "loss": 1.5688,
      "theoretical_loss": 3.654150698191995,
      "tokens_seen": 1070727168
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006824072216649951,
      "loss": 1.3952,
      "theoretical_loss": 3.6541298457577547,
      "tokens_seen": 1070792704
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006823871614844533,
      "loss": 1.4878,
      "theoretical_loss": 3.654108994957034,
      "tokens_seen": 1070858240
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006823671013039118,
      "loss": 1.5724,
      "theoretical_loss": 3.6540881457896046,
      "tokens_seen": 1070923776
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00068234704112337,
      "loss": 1.513,
      "theoretical_loss": 3.6540672982552387,
      "tokens_seen": 1070989312
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006823269809428285,
      "loss": 1.558,
      "theoretical_loss": 3.6540464523537084,
      "tokens_seen": 1071054848
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006823069207622869,
      "loss": 1.4895,
      "theoretical_loss": 3.654025608084786,
      "tokens_seen": 1071120384
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006822868605817452,
      "loss": 1.4827,
      "theoretical_loss": 3.654004765448243,
      "tokens_seen": 1071185920
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006822668004012036,
      "loss": 1.6307,
      "theoretical_loss": 3.6539839244438532,
      "tokens_seen": 1071251456
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006822467402206621,
      "loss": 1.4363,
      "theoretical_loss": 3.6539630850713873,
      "tokens_seen": 1071316992
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006822266800401203,
      "loss": 1.5144,
      "theoretical_loss": 3.653942247330619,
      "tokens_seen": 1071382528
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006822066198595788,
      "loss": 1.4328,
      "theoretical_loss": 3.65392141122132,
      "tokens_seen": 1071448064
    },
    {
      "epoch": 0.32,
      "objective/train/advantage_avg": 0.49290400743484497,
      "objective/train/docs_used": 393037,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4883989095687866,
      "objective/train/original_loss": 1.488398790359497,
      "objective/train/theoretical_loss": 3.6539005767432635,
      "objective/train/tokens_used": 1091973600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24328944087028503,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505272150039673,
      "objective/train/weighted_lm_loss": 1.5635225772857666,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6539005767432635,
      "tokens_seen": 1071513600
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000682186559679037,
      "loss": 1.5063,
      "theoretical_loss": 3.6539005767432635,
      "tokens_seen": 1071513600
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006821664994984955,
      "loss": 1.507,
      "theoretical_loss": 3.653879743896221,
      "tokens_seen": 1071579136
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006821464393179539,
      "loss": 1.4985,
      "theoretical_loss": 3.653858912679966,
      "tokens_seen": 1071644672
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006821263791374122,
      "loss": 1.5599,
      "theoretical_loss": 3.6538380830942705,
      "tokens_seen": 1071710208
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006821063189568706,
      "loss": 1.5926,
      "theoretical_loss": 3.6538172551389074,
      "tokens_seen": 1071775744
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000682086258776329,
      "loss": 1.4324,
      "theoretical_loss": 3.65379642881365,
      "tokens_seen": 1071841280
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006820661985957873,
      "loss": 1.4374,
      "theoretical_loss": 3.6537756041182696,
      "tokens_seen": 1071906816
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006820461384152458,
      "loss": 1.5659,
      "theoretical_loss": 3.653754781052541,
      "tokens_seen": 1071972352
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000682026078234704,
      "loss": 1.4138,
      "theoretical_loss": 3.6537339596162353,
      "tokens_seen": 1072037888
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006820060180541625,
      "loss": 1.554,
      "theoretical_loss": 3.653713139809126,
      "tokens_seen": 1072103424
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006819859578736209,
      "loss": 1.4955,
      "theoretical_loss": 3.6536923216309862,
      "tokens_seen": 1072168960
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006819658976930792,
      "loss": 1.5443,
      "theoretical_loss": 3.653671505081589,
      "tokens_seen": 1072234496
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006819458375125376,
      "loss": 1.5494,
      "theoretical_loss": 3.653650690160707,
      "tokens_seen": 1072300032
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000681925777331996,
      "loss": 1.5002,
      "theoretical_loss": 3.6536298768681137,
      "tokens_seen": 1072365568
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006819057171514543,
      "loss": 1.4877,
      "theoretical_loss": 3.653609065203582,
      "tokens_seen": 1072431104
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0006818856569709128,
      "loss": 1.4757,
      "theoretical_loss": 3.653588255166885,
      "tokens_seen": 1072496640
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006818655967903711,
      "loss": 1.5318,
      "theoretical_loss": 3.653567446757796,
      "tokens_seen": 1072562176
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006818455366098295,
      "loss": 1.4693,
      "theoretical_loss": 3.6535466399760885,
      "tokens_seen": 1072627712
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006818254764292878,
      "loss": 1.4397,
      "theoretical_loss": 3.6535258348215356,
      "tokens_seen": 1072693248
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006818054162487463,
      "loss": 1.5623,
      "theoretical_loss": 3.6535050312939106,
      "tokens_seen": 1072758784
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006817853560682046,
      "loss": 1.4721,
      "theoretical_loss": 3.6534842293929866,
      "tokens_seen": 1072824320
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000681765295887663,
      "loss": 1.5773,
      "theoretical_loss": 3.6534634291185375,
      "tokens_seen": 1072889856
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006817452357071213,
      "loss": 1.5121,
      "theoretical_loss": 3.653442630470337,
      "tokens_seen": 1072955392
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006817251755265797,
      "loss": 1.5606,
      "theoretical_loss": 3.6534218334481583,
      "tokens_seen": 1073020928
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006817051153460381,
      "loss": 1.5193,
      "theoretical_loss": 3.6534010380517747,
      "tokens_seen": 1073086464
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006816850551654965,
      "loss": 1.5515,
      "theoretical_loss": 3.65338024428096,
      "tokens_seen": 1073152000
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006816649949849549,
      "loss": 1.4304,
      "theoretical_loss": 3.653359452135488,
      "tokens_seen": 1073217536
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006816449348044133,
      "loss": 1.4724,
      "theoretical_loss": 3.653338661615133,
      "tokens_seen": 1073283072
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006816248746238717,
      "loss": 1.4958,
      "theoretical_loss": 3.6533178727196676,
      "tokens_seen": 1073348608
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00068160481444333,
      "loss": 1.5684,
      "theoretical_loss": 3.6532970854488664,
      "tokens_seen": 1073414144
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006815847542627885,
      "loss": 1.4687,
      "theoretical_loss": 3.653276299802503,
      "tokens_seen": 1073479680
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006815646940822467,
      "loss": 1.502,
      "theoretical_loss": 3.6532555157803515,
      "tokens_seen": 1073545216
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006815446339017052,
      "loss": 1.4845,
      "theoretical_loss": 3.6532347333821855,
      "tokens_seen": 1073610752
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006815245737211635,
      "loss": 1.5676,
      "theoretical_loss": 3.6532139526077785,
      "tokens_seen": 1073676288
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006815045135406219,
      "loss": 1.4557,
      "theoretical_loss": 3.6531931734569056,
      "tokens_seen": 1073741824
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006814844533600803,
      "loss": 1.4194,
      "theoretical_loss": 3.6531723959293414,
      "tokens_seen": 1073807360
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006814643931795386,
      "loss": 1.4528,
      "theoretical_loss": 3.653151620024858,
      "tokens_seen": 1073872896
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000681444332998997,
      "loss": 1.531,
      "theoretical_loss": 3.653130845743231,
      "tokens_seen": 1073938432
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006814242728184555,
      "loss": 1.5334,
      "theoretical_loss": 3.6531100730842336,
      "tokens_seen": 1074003968
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006814042126379137,
      "loss": 1.5127,
      "theoretical_loss": 3.6530893020476416,
      "tokens_seen": 1074069504
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006813841524573722,
      "loss": 1.5554,
      "theoretical_loss": 3.6530685326332284,
      "tokens_seen": 1074135040
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006813640922768304,
      "loss": 1.452,
      "theoretical_loss": 3.6530477648407675,
      "tokens_seen": 1074200576
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006813440320962889,
      "loss": 1.5941,
      "theoretical_loss": 3.653026998670035,
      "tokens_seen": 1074266112
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006813239719157473,
      "loss": 1.6511,
      "theoretical_loss": 3.653006234120804,
      "tokens_seen": 1074331648
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006813039117352056,
      "loss": 1.3835,
      "theoretical_loss": 3.65298547119285,
      "tokens_seen": 1074397184
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000681283851554664,
      "loss": 1.4958,
      "theoretical_loss": 3.6529647098859463,
      "tokens_seen": 1074462720
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006812637913741225,
      "loss": 1.4571,
      "theoretical_loss": 3.652943950199869,
      "tokens_seen": 1074528256
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006812437311935807,
      "loss": 1.481,
      "theoretical_loss": 3.6529231921343914,
      "tokens_seen": 1074593792
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006812236710130392,
      "loss": 1.4999,
      "theoretical_loss": 3.652902435689289,
      "tokens_seen": 1074659328
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006812036108324974,
      "loss": 1.5304,
      "theoretical_loss": 3.652881680864336,
      "tokens_seen": 1074724864
    },
    {
      "epoch": 0.33,
      "objective/train/advantage_avg": 0.49231386184692383,
      "objective/train/docs_used": 394219,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3476676940917969,
      "objective/train/original_loss": 1.3476675748825073,
      "objective/train/theoretical_loss": 3.652860927659307,
      "objective/train/tokens_used": 1095250400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24280966818332672,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050465703010559,
      "objective/train/weighted_lm_loss": 1.4155232906341553,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.652860927659307,
      "tokens_seen": 1074790400
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006811835506519559,
      "loss": 1.4813,
      "theoretical_loss": 3.652860927659307,
      "tokens_seen": 1074790400
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006811634904714143,
      "loss": 1.4962,
      "theoretical_loss": 3.652840176073978,
      "tokens_seen": 1074855936
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006811434302908726,
      "loss": 1.5081,
      "theoretical_loss": 3.652819426108122,
      "tokens_seen": 1074921472
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000681123370110331,
      "loss": 1.4619,
      "theoretical_loss": 3.652798677761515,
      "tokens_seen": 1074987008
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006811033099297894,
      "loss": 1.493,
      "theoretical_loss": 3.6527779310339326,
      "tokens_seen": 1075052544
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006810832497492477,
      "loss": 1.4827,
      "theoretical_loss": 3.6527571859251484,
      "tokens_seen": 1075118080
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006810631895687062,
      "loss": 1.6041,
      "theoretical_loss": 3.6527364424349384,
      "tokens_seen": 1075183616
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006810431293881645,
      "loss": 1.4869,
      "theoretical_loss": 3.6527157005630766,
      "tokens_seen": 1075249152
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006810230692076229,
      "loss": 1.4874,
      "theoretical_loss": 3.652694960309339,
      "tokens_seen": 1075314688
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006810030090270813,
      "loss": 1.5529,
      "theoretical_loss": 3.652674221673501,
      "tokens_seen": 1075380224
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006809829488465396,
      "loss": 1.5706,
      "theoretical_loss": 3.652653484655337,
      "tokens_seen": 1075445760
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000680962888665998,
      "loss": 1.4149,
      "theoretical_loss": 3.652632749254623,
      "tokens_seen": 1075511296
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006809428284854564,
      "loss": 1.5943,
      "theoretical_loss": 3.6526120154711332,
      "tokens_seen": 1075576832
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006809227683049147,
      "loss": 1.4253,
      "theoretical_loss": 3.6525912833046443,
      "tokens_seen": 1075642368
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006809027081243732,
      "loss": 1.4827,
      "theoretical_loss": 3.6525705527549306,
      "tokens_seen": 1075707904
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006808826479438315,
      "loss": 1.5431,
      "theoretical_loss": 3.6525498238217686,
      "tokens_seen": 1075773440
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006808625877632899,
      "loss": 1.4488,
      "theoretical_loss": 3.6525290965049324,
      "tokens_seen": 1075838976
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006808425275827482,
      "loss": 1.5046,
      "theoretical_loss": 3.6525083708041985,
      "tokens_seen": 1075904512
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006808224674022067,
      "loss": 1.4878,
      "theoretical_loss": 3.6524876467193423,
      "tokens_seen": 1075970048
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000680802407221665,
      "loss": 1.6153,
      "theoretical_loss": 3.652466924250139,
      "tokens_seen": 1076035584
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006807823470411234,
      "loss": 1.4387,
      "theoretical_loss": 3.652446203396365,
      "tokens_seen": 1076101120
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006807622868605817,
      "loss": 1.5197,
      "theoretical_loss": 3.6524254841577957,
      "tokens_seen": 1076166656
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006807422266800401,
      "loss": 1.4777,
      "theoretical_loss": 3.6524047665342065,
      "tokens_seen": 1076232192
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006807221664994985,
      "loss": 1.4096,
      "theoretical_loss": 3.652384050525373,
      "tokens_seen": 1076297728
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006807021063189569,
      "loss": 1.5551,
      "theoretical_loss": 3.6523633361310717,
      "tokens_seen": 1076363264
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006806820461384152,
      "loss": 1.5074,
      "theoretical_loss": 3.6523426233510783,
      "tokens_seen": 1076428800
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006806619859578737,
      "loss": 1.4083,
      "theoretical_loss": 3.652321912185169,
      "tokens_seen": 1076494336
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000680641925777332,
      "loss": 1.5249,
      "theoretical_loss": 3.6523012026331187,
      "tokens_seen": 1076559872
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006806218655967904,
      "loss": 1.5726,
      "theoretical_loss": 3.6522804946947045,
      "tokens_seen": 1076625408
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006806018054162487,
      "loss": 1.4643,
      "theoretical_loss": 3.652259788369702,
      "tokens_seen": 1076690944
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006805817452357071,
      "loss": 1.4776,
      "theoretical_loss": 3.6522390836578875,
      "tokens_seen": 1076756480
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006805616850551655,
      "loss": 1.4995,
      "theoretical_loss": 3.6522183805590367,
      "tokens_seen": 1076822016
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006805416248746239,
      "loss": 1.4963,
      "theoretical_loss": 3.6521976790729265,
      "tokens_seen": 1076887552
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006805215646940822,
      "loss": 1.4966,
      "theoretical_loss": 3.6521769791993326,
      "tokens_seen": 1076953088
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006805015045135407,
      "loss": 1.4262,
      "theoretical_loss": 3.6521562809380317,
      "tokens_seen": 1077018624
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006804814443329989,
      "loss": 1.5344,
      "theoretical_loss": 3.652135584288799,
      "tokens_seen": 1077084160
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006804613841524574,
      "loss": 1.479,
      "theoretical_loss": 3.652114889251412,
      "tokens_seen": 1077149696
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006804413239719158,
      "loss": 1.6047,
      "theoretical_loss": 3.6520941958256476,
      "tokens_seen": 1077215232
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006804212637913741,
      "loss": 1.5899,
      "theoretical_loss": 3.652073504011281,
      "tokens_seen": 1077280768
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006804012036108325,
      "loss": 1.5657,
      "theoretical_loss": 3.6520528138080888,
      "tokens_seen": 1077346304
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006803811434302908,
      "loss": 1.4697,
      "theoretical_loss": 3.6520321252158485,
      "tokens_seen": 1077411840
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006803610832497492,
      "loss": 1.4477,
      "theoretical_loss": 3.652011438234336,
      "tokens_seen": 1077477376
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006803410230692077,
      "loss": 1.5189,
      "theoretical_loss": 3.6519907528633277,
      "tokens_seen": 1077542912
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006803209628886659,
      "loss": 1.5113,
      "theoretical_loss": 3.6519700691026014,
      "tokens_seen": 1077608448
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006803009027081244,
      "loss": 1.4098,
      "theoretical_loss": 3.651949386951933,
      "tokens_seen": 1077673984
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006802808425275828,
      "loss": 1.4879,
      "theoretical_loss": 3.6519287064110983,
      "tokens_seen": 1077739520
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006802607823470411,
      "loss": 1.4144,
      "theoretical_loss": 3.651908027479876,
      "tokens_seen": 1077805056
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006802407221664995,
      "loss": 1.506,
      "theoretical_loss": 3.651887350158042,
      "tokens_seen": 1077870592
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006802206619859578,
      "loss": 1.5631,
      "theoretical_loss": 3.6518666744453734,
      "tokens_seen": 1077936128
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006802006018054162,
      "loss": 1.5784,
      "theoretical_loss": 3.6518460003416466,
      "tokens_seen": 1078001664
    },
    {
      "epoch": 0.33,
      "objective/train/advantage_avg": 0.49414318799972534,
      "objective/train/docs_used": 395298,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3375064134597778,
      "objective/train/original_loss": 1.3375062942504883,
      "objective/train/theoretical_loss": 3.6518253278466397,
      "objective/train/tokens_used": 1098527200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24452674388885498,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506573915481567,
      "objective/train/weighted_lm_loss": 1.4052526950836182,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6518253278466397,
      "tokens_seen": 1078067200
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006801805416248747,
      "loss": 1.4462,
      "theoretical_loss": 3.6518253278466397,
      "tokens_seen": 1078067200
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000680160481444333,
      "loss": 1.4994,
      "theoretical_loss": 3.6518046569601292,
      "tokens_seen": 1078132736
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006801404212637914,
      "loss": 1.4552,
      "theoretical_loss": 3.651783987681892,
      "tokens_seen": 1078198272
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006801203610832498,
      "loss": 1.479,
      "theoretical_loss": 3.6517633200117046,
      "tokens_seen": 1078263808
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006801003009027081,
      "loss": 1.5724,
      "theoretical_loss": 3.6517426539493454,
      "tokens_seen": 1078329344
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006800802407221666,
      "loss": 1.5594,
      "theoretical_loss": 3.651721989494591,
      "tokens_seen": 1078394880
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006800601805416249,
      "loss": 1.4045,
      "theoretical_loss": 3.6517013266472187,
      "tokens_seen": 1078460416
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006800401203610833,
      "loss": 1.5851,
      "theoretical_loss": 3.6516806654070058,
      "tokens_seen": 1078525952
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006800200601805417,
      "loss": 1.5795,
      "theoretical_loss": 3.65166000577373,
      "tokens_seen": 1078591488
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00068,
      "loss": 1.4837,
      "theoretical_loss": 3.6516393477471683,
      "tokens_seen": 1078657024
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006799799398194584,
      "loss": 1.3736,
      "theoretical_loss": 3.651618691327098,
      "tokens_seen": 1078722560
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006799598796389168,
      "loss": 1.5296,
      "theoretical_loss": 3.651598036513297,
      "tokens_seen": 1078788096
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006799398194583751,
      "loss": 1.5338,
      "theoretical_loss": 3.6515773833055425,
      "tokens_seen": 1078853632
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006799197592778336,
      "loss": 1.4349,
      "theoretical_loss": 3.6515567317036126,
      "tokens_seen": 1078919168
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006798996990972919,
      "loss": 1.6117,
      "theoretical_loss": 3.651536081707284,
      "tokens_seen": 1078984704
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006798796389167503,
      "loss": 1.5688,
      "theoretical_loss": 3.6515154333163355,
      "tokens_seen": 1079050240
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006798595787362086,
      "loss": 1.5297,
      "theoretical_loss": 3.651494786530544,
      "tokens_seen": 1079115776
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006798395185556671,
      "loss": 1.4753,
      "theoretical_loss": 3.651474141349687,
      "tokens_seen": 1079181312
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006798194583751254,
      "loss": 1.6145,
      "theoretical_loss": 3.651453497773543,
      "tokens_seen": 1079246848
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006797993981945838,
      "loss": 1.5459,
      "theoretical_loss": 3.6514328558018896,
      "tokens_seen": 1079312384
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006797793380140421,
      "loss": 1.5068,
      "theoretical_loss": 3.6514122154345046,
      "tokens_seen": 1079377920
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006797592778335005,
      "loss": 1.4866,
      "theoretical_loss": 3.651391576671166,
      "tokens_seen": 1079443456
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006797392176529589,
      "loss": 1.5681,
      "theoretical_loss": 3.6513709395116516,
      "tokens_seen": 1079508992
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006797191574724173,
      "loss": 1.5728,
      "theoretical_loss": 3.6513503039557396,
      "tokens_seen": 1079574528
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006796990972918756,
      "loss": 1.3209,
      "theoretical_loss": 3.651329670003208,
      "tokens_seen": 1079640064
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006796790371113341,
      "loss": 1.4964,
      "theoretical_loss": 3.651309037653835,
      "tokens_seen": 1079705600
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006796589769307924,
      "loss": 1.4463,
      "theoretical_loss": 3.651288406907399,
      "tokens_seen": 1079771136
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006796389167502508,
      "loss": 1.4502,
      "theoretical_loss": 3.6512677777636773,
      "tokens_seen": 1079836672
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006796188565697091,
      "loss": 1.4089,
      "theoretical_loss": 3.6512471502224493,
      "tokens_seen": 1079902208
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006795987963891675,
      "loss": 1.6451,
      "theoretical_loss": 3.6512265242834916,
      "tokens_seen": 1079967744
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006795787362086259,
      "loss": 1.5442,
      "theoretical_loss": 3.6512058999465844,
      "tokens_seen": 1080033280
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006795586760280843,
      "loss": 1.4668,
      "theoretical_loss": 3.651185277211505,
      "tokens_seen": 1080098816
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006795386158475426,
      "loss": 1.5886,
      "theoretical_loss": 3.6511646560780324,
      "tokens_seen": 1080164352
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006795185556670011,
      "loss": 1.4882,
      "theoretical_loss": 3.6511440365459444,
      "tokens_seen": 1080229888
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006794984954864593,
      "loss": 1.4378,
      "theoretical_loss": 3.6511234186150197,
      "tokens_seen": 1080295424
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006794784353059178,
      "loss": 1.5565,
      "theoretical_loss": 3.651102802285037,
      "tokens_seen": 1080360960
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006794583751253762,
      "loss": 1.5156,
      "theoretical_loss": 3.6510821875557746,
      "tokens_seen": 1080426496
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006794383149448345,
      "loss": 1.5082,
      "theoretical_loss": 3.6510615744270116,
      "tokens_seen": 1080492032
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006794182547642929,
      "loss": 1.4872,
      "theoretical_loss": 3.6510409628985263,
      "tokens_seen": 1080557568
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006793981945837512,
      "loss": 1.4876,
      "theoretical_loss": 3.651020352970098,
      "tokens_seen": 1080623104
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006793781344032096,
      "loss": 1.442,
      "theoretical_loss": 3.6509997446415046,
      "tokens_seen": 1080688640
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006793580742226681,
      "loss": 1.5895,
      "theoretical_loss": 3.6509791379125254,
      "tokens_seen": 1080754176
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006793380140421263,
      "loss": 1.4728,
      "theoretical_loss": 3.6509585327829392,
      "tokens_seen": 1080819712
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006793179538615848,
      "loss": 1.5273,
      "theoretical_loss": 3.650937929252525,
      "tokens_seen": 1080885248
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006792978936810432,
      "loss": 1.3999,
      "theoretical_loss": 3.6509173273210616,
      "tokens_seen": 1080950784
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006792778335005015,
      "loss": 1.4541,
      "theoretical_loss": 3.6508967269883277,
      "tokens_seen": 1081016320
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006792577733199599,
      "loss": 1.6011,
      "theoretical_loss": 3.6508761282541027,
      "tokens_seen": 1081081856
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006792377131394182,
      "loss": 1.5751,
      "theoretical_loss": 3.650855531118166,
      "tokens_seen": 1081147392
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006792176529588766,
      "loss": 1.4911,
      "theoretical_loss": 3.6508349355802956,
      "tokens_seen": 1081212928
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006791975927783351,
      "loss": 1.4212,
      "theoretical_loss": 3.650814341640272,
      "tokens_seen": 1081278464
    },
    {
      "epoch": 0.33,
      "objective/train/advantage_avg": 0.4947785437107086,
      "objective/train/docs_used": 396489,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5168761014938354,
      "objective/train/original_loss": 1.516876220703125,
      "objective/train/theoretical_loss": 3.6507937492978733,
      "objective/train/tokens_used": 1101804000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24508950114250183,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050723910331726,
      "objective/train/weighted_lm_loss": 1.5936757326126099,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6507937492978733,
      "tokens_seen": 1081344000
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006791775325977933,
      "loss": 1.5517,
      "theoretical_loss": 3.6507937492978733,
      "tokens_seen": 1081344000
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006791574724172518,
      "loss": 1.5191,
      "theoretical_loss": 3.6507731585528793,
      "tokens_seen": 1081409536
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006791374122367101,
      "loss": 1.436,
      "theoretical_loss": 3.6507525694050695,
      "tokens_seen": 1081475072
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006791173520561685,
      "loss": 1.4393,
      "theoretical_loss": 3.6507319818542223,
      "tokens_seen": 1081540608
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006790972918756269,
      "loss": 1.4428,
      "theoretical_loss": 3.6507113959001183,
      "tokens_seen": 1081606144
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006790772316950853,
      "loss": 1.5724,
      "theoretical_loss": 3.6506908115425363,
      "tokens_seen": 1081671680
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006790571715145436,
      "loss": 1.4808,
      "theoretical_loss": 3.650670228781255,
      "tokens_seen": 1081737216
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006790371113340021,
      "loss": 1.5624,
      "theoretical_loss": 3.6506496476160555,
      "tokens_seen": 1081802752
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006790170511534603,
      "loss": 1.7071,
      "theoretical_loss": 3.6506290680467166,
      "tokens_seen": 1081868288
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006789969909729188,
      "loss": 1.5227,
      "theoretical_loss": 3.6506084900730174,
      "tokens_seen": 1081933824
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006789769307923771,
      "loss": 1.4366,
      "theoretical_loss": 3.6505879136947383,
      "tokens_seen": 1081999360
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006789568706118355,
      "loss": 1.4778,
      "theoretical_loss": 3.6505673389116584,
      "tokens_seen": 1082064896
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006789368104312939,
      "loss": 1.6074,
      "theoretical_loss": 3.650546765723558,
      "tokens_seen": 1082130432
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006789167502507523,
      "loss": 1.517,
      "theoretical_loss": 3.6505261941302165,
      "tokens_seen": 1082195968
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006788966900702106,
      "loss": 1.5001,
      "theoretical_loss": 3.650505624131414,
      "tokens_seen": 1082261504
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000678876629889669,
      "loss": 1.5629,
      "theoretical_loss": 3.6504850557269295,
      "tokens_seen": 1082327040
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006788565697091273,
      "loss": 1.4005,
      "theoretical_loss": 3.650464488916544,
      "tokens_seen": 1082392576
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006788365095285858,
      "loss": 1.5473,
      "theoretical_loss": 3.6504439237000366,
      "tokens_seen": 1082458112
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006788164493480441,
      "loss": 1.4674,
      "theoretical_loss": 3.6504233600771885,
      "tokens_seen": 1082523648
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006787963891675025,
      "loss": 1.5669,
      "theoretical_loss": 3.6504027980477782,
      "tokens_seen": 1082589184
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006787763289869608,
      "loss": 1.5384,
      "theoretical_loss": 3.650382237611587,
      "tokens_seen": 1082654720
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006787562688064193,
      "loss": 1.6077,
      "theoretical_loss": 3.6503616787683937,
      "tokens_seen": 1082720256
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006787362086258776,
      "loss": 1.547,
      "theoretical_loss": 3.6503411215179797,
      "tokens_seen": 1082785792
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000678716148445336,
      "loss": 1.4616,
      "theoretical_loss": 3.650320565860125,
      "tokens_seen": 1082851328
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006786960882647944,
      "loss": 1.6279,
      "theoretical_loss": 3.65030001179461,
      "tokens_seen": 1082916864
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006786760280842528,
      "loss": 1.5139,
      "theoretical_loss": 3.6502794593212142,
      "tokens_seen": 1082982400
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006786559679037112,
      "loss": 1.5323,
      "theoretical_loss": 3.650258908439718,
      "tokens_seen": 1083047936
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006786359077231695,
      "loss": 1.4924,
      "theoretical_loss": 3.6502383591499026,
      "tokens_seen": 1083113472
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006786158475426279,
      "loss": 1.4726,
      "theoretical_loss": 3.650217811451548,
      "tokens_seen": 1083179008
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006785957873620863,
      "loss": 1.4712,
      "theoretical_loss": 3.650197265344435,
      "tokens_seen": 1083244544
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006785757271815447,
      "loss": 1.4198,
      "theoretical_loss": 3.650176720828343,
      "tokens_seen": 1083310080
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000678555667001003,
      "loss": 1.4345,
      "theoretical_loss": 3.6501561779030536,
      "tokens_seen": 1083375616
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006785356068204615,
      "loss": 1.6273,
      "theoretical_loss": 3.650135636568347,
      "tokens_seen": 1083441152
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006785155466399197,
      "loss": 1.4356,
      "theoretical_loss": 3.6501150968240044,
      "tokens_seen": 1083506688
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006784954864593782,
      "loss": 1.527,
      "theoretical_loss": 3.6500945586698057,
      "tokens_seen": 1083572224
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006784754262788366,
      "loss": 1.4396,
      "theoretical_loss": 3.650074022105532,
      "tokens_seen": 1083637760
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006784553660982949,
      "loss": 1.5911,
      "theoretical_loss": 3.6500534871309642,
      "tokens_seen": 1083703296
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006784353059177533,
      "loss": 1.4993,
      "theoretical_loss": 3.650032953745883,
      "tokens_seen": 1083768832
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006784152457372116,
      "loss": 1.4509,
      "theoretical_loss": 3.6500124219500694,
      "tokens_seen": 1083834368
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00067839518555667,
      "loss": 1.5102,
      "theoretical_loss": 3.649991891743304,
      "tokens_seen": 1083899904
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006783751253761285,
      "loss": 1.5634,
      "theoretical_loss": 3.649971363125368,
      "tokens_seen": 1083965440
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006783550651955867,
      "loss": 1.5312,
      "theoretical_loss": 3.6499508360960418,
      "tokens_seen": 1084030976
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006783350050150452,
      "loss": 1.5185,
      "theoretical_loss": 3.6499303106551073,
      "tokens_seen": 1084096512
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006783149448345036,
      "loss": 1.5129,
      "theoretical_loss": 3.649909786802345,
      "tokens_seen": 1084162048
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006782948846539619,
      "loss": 1.4905,
      "theoretical_loss": 3.6498892645375367,
      "tokens_seen": 1084227584
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006782748244734203,
      "loss": 1.5014,
      "theoretical_loss": 3.649868743860463,
      "tokens_seen": 1084293120
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006782547642928786,
      "loss": 1.5526,
      "theoretical_loss": 3.6498482247709054,
      "tokens_seen": 1084358656
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000678234704112337,
      "loss": 1.4652,
      "theoretical_loss": 3.6498277072686443,
      "tokens_seen": 1084424192
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006782146439317955,
      "loss": 1.4323,
      "theoretical_loss": 3.649807191353462,
      "tokens_seen": 1084489728
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006781945837512537,
      "loss": 1.4663,
      "theoretical_loss": 3.6497866770251397,
      "tokens_seen": 1084555264
    },
    {
      "epoch": 0.33,
      "objective/train/advantage_avg": 0.4904569089412689,
      "objective/train/docs_used": 397685,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3166391849517822,
      "objective/train/original_loss": 1.3166391849517822,
      "objective/train/theoretical_loss": 3.649766164283458,
      "objective/train/tokens_used": 1105080800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24101608991622925,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502707958221436,
      "objective/train/weighted_lm_loss": 1.3828569650650024,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.649766164283458,
      "tokens_seen": 1084620800
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006781745235707122,
      "loss": 1.5095,
      "theoretical_loss": 3.649766164283458,
      "tokens_seen": 1084620800
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006781544633901705,
      "loss": 1.5747,
      "theoretical_loss": 3.649745653128199,
      "tokens_seen": 1084686336
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006781344032096289,
      "loss": 1.5232,
      "theoretical_loss": 3.6497251435591442,
      "tokens_seen": 1084751872
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006781143430290873,
      "loss": 1.5366,
      "theoretical_loss": 3.649704635576075,
      "tokens_seen": 1084817408
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006780942828485457,
      "loss": 1.6367,
      "theoretical_loss": 3.649684129178773,
      "tokens_seen": 1084882944
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000678074222668004,
      "loss": 1.5898,
      "theoretical_loss": 3.6496636243670206,
      "tokens_seen": 1084948480
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006780541624874625,
      "loss": 1.5915,
      "theoretical_loss": 3.6496431211405973,
      "tokens_seen": 1085014016
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006780341023069207,
      "loss": 1.462,
      "theoretical_loss": 3.6496226194992865,
      "tokens_seen": 1085079552
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006780140421263792,
      "loss": 1.4514,
      "theoretical_loss": 3.6496021194428696,
      "tokens_seen": 1085145088
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006779939819458375,
      "loss": 1.5126,
      "theoretical_loss": 3.6495816209711283,
      "tokens_seen": 1085210624
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006779739217652959,
      "loss": 1.424,
      "theoretical_loss": 3.649561124083844,
      "tokens_seen": 1085276160
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006779538615847543,
      "loss": 1.5663,
      "theoretical_loss": 3.649540628780799,
      "tokens_seen": 1085341696
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006779338014042127,
      "loss": 1.5035,
      "theoretical_loss": 3.649520135061776,
      "tokens_seen": 1085407232
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000677913741223671,
      "loss": 1.5945,
      "theoretical_loss": 3.649499642926555,
      "tokens_seen": 1085472768
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006778936810431294,
      "loss": 1.4852,
      "theoretical_loss": 3.6494791523749193,
      "tokens_seen": 1085538304
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006778736208625877,
      "loss": 1.5597,
      "theoretical_loss": 3.6494586634066515,
      "tokens_seen": 1085603840
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006778535606820462,
      "loss": 1.527,
      "theoretical_loss": 3.649438176021532,
      "tokens_seen": 1085669376
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006778335005015045,
      "loss": 1.4906,
      "theoretical_loss": 3.649417690219344,
      "tokens_seen": 1085734912
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006778134403209629,
      "loss": 1.474,
      "theoretical_loss": 3.6493972059998696,
      "tokens_seen": 1085800448
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006777933801404212,
      "loss": 1.5121,
      "theoretical_loss": 3.6493767233628907,
      "tokens_seen": 1085865984
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006777733199598797,
      "loss": 1.5118,
      "theoretical_loss": 3.64935624230819,
      "tokens_seen": 1085931520
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000677753259779338,
      "loss": 1.5066,
      "theoretical_loss": 3.649335762835549,
      "tokens_seen": 1085997056
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006777331995987964,
      "loss": 1.4669,
      "theoretical_loss": 3.649315284944751,
      "tokens_seen": 1086062592
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006777131394182548,
      "loss": 1.459,
      "theoretical_loss": 3.649294808635578,
      "tokens_seen": 1086128128
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006776930792377132,
      "loss": 1.3881,
      "theoretical_loss": 3.6492743339078118,
      "tokens_seen": 1086193664
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006776730190571715,
      "loss": 1.5717,
      "theoretical_loss": 3.6492538607612355,
      "tokens_seen": 1086259200
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00067765295887663,
      "loss": 1.4737,
      "theoretical_loss": 3.649233389195632,
      "tokens_seen": 1086324736
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006776328986960882,
      "loss": 1.5522,
      "theoretical_loss": 3.649212919210783,
      "tokens_seen": 1086390272
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006776128385155467,
      "loss": 1.4988,
      "theoretical_loss": 3.6491924508064715,
      "tokens_seen": 1086455808
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000677592778335005,
      "loss": 1.4813,
      "theoretical_loss": 3.64917198398248,
      "tokens_seen": 1086521344
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006775727181544634,
      "loss": 1.4812,
      "theoretical_loss": 3.6491515187385914,
      "tokens_seen": 1086586880
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006775526579739218,
      "loss": 1.46,
      "theoretical_loss": 3.649131055074588,
      "tokens_seen": 1086652416
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006775325977933801,
      "loss": 1.346,
      "theoretical_loss": 3.6491105929902528,
      "tokens_seen": 1086717952
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006775125376128385,
      "loss": 1.5993,
      "theoretical_loss": 3.649090132485369,
      "tokens_seen": 1086783488
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000677492477432297,
      "loss": 1.4127,
      "theoretical_loss": 3.649069673559719,
      "tokens_seen": 1086849024
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006774724172517552,
      "loss": 1.4291,
      "theoretical_loss": 3.6490492162130854,
      "tokens_seen": 1086914560
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006774523570712137,
      "loss": 1.5217,
      "theoretical_loss": 3.649028760445252,
      "tokens_seen": 1086980096
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006774322968906719,
      "loss": 1.4843,
      "theoretical_loss": 3.6490083062560013,
      "tokens_seen": 1087045632
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006774122367101304,
      "loss": 1.5277,
      "theoretical_loss": 3.648987853645116,
      "tokens_seen": 1087111168
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006773921765295888,
      "loss": 1.4645,
      "theoretical_loss": 3.6489674026123793,
      "tokens_seen": 1087176704
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006773721163490471,
      "loss": 1.3868,
      "theoretical_loss": 3.6489469531575747,
      "tokens_seen": 1087242240
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006773520561685055,
      "loss": 1.4469,
      "theoretical_loss": 3.6489265052804853,
      "tokens_seen": 1087307776
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000677331995987964,
      "loss": 1.5449,
      "theoretical_loss": 3.648906058980894,
      "tokens_seen": 1087373312
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006773119358074222,
      "loss": 1.6172,
      "theoretical_loss": 3.6488856142585844,
      "tokens_seen": 1087438848
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006772918756268807,
      "loss": 1.3578,
      "theoretical_loss": 3.6488651711133393,
      "tokens_seen": 1087504384
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006772718154463389,
      "loss": 1.4546,
      "theoretical_loss": 3.648844729544942,
      "tokens_seen": 1087569920
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006772517552657974,
      "loss": 1.5832,
      "theoretical_loss": 3.6488242895531764,
      "tokens_seen": 1087635456
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006772316950852558,
      "loss": 1.5378,
      "theoretical_loss": 3.648803851137826,
      "tokens_seen": 1087700992
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006772116349047141,
      "loss": 1.4343,
      "theoretical_loss": 3.6487834142986735,
      "tokens_seen": 1087766528
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006771915747241725,
      "loss": 1.49,
      "theoretical_loss": 3.648762979035503,
      "tokens_seen": 1087832064
    },
    {
      "epoch": 0.33,
      "objective/train/advantage_avg": 0.48699241876602173,
      "objective/train/docs_used": 398862,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3465205430984497,
      "objective/train/original_loss": 1.3465203046798706,
      "objective/train/theoretical_loss": 3.6487425453480973,
      "objective/train/tokens_used": 1108357600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23805741965770721,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.049909234046936,
      "objective/train/weighted_lm_loss": 1.4137723445892334,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6487425453480973,
      "tokens_seen": 1087897600
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006771715145436309,
      "loss": 1.4344,
      "theoretical_loss": 3.6487425453480973,
      "tokens_seen": 1087897600
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006771514543630893,
      "loss": 1.4407,
      "theoretical_loss": 3.648722113236241,
      "tokens_seen": 1087963136
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006771313941825477,
      "loss": 1.453,
      "theoretical_loss": 3.6487016826997167,
      "tokens_seen": 1088028672
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006771113340020061,
      "loss": 1.4722,
      "theoretical_loss": 3.6486812537383093,
      "tokens_seen": 1088094208
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006770912738214644,
      "loss": 1.4609,
      "theoretical_loss": 3.648660826351801,
      "tokens_seen": 1088159744
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006770712136409229,
      "loss": 1.4732,
      "theoretical_loss": 3.6486404005399766,
      "tokens_seen": 1088225280
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006770511534603811,
      "loss": 1.394,
      "theoretical_loss": 3.64861997630262,
      "tokens_seen": 1088290816
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006770310932798396,
      "loss": 1.5188,
      "theoretical_loss": 3.648599553639514,
      "tokens_seen": 1088356352
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006770110330992979,
      "loss": 1.6275,
      "theoretical_loss": 3.6485791325504437,
      "tokens_seen": 1088421888
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006769909729187563,
      "loss": 1.4553,
      "theoretical_loss": 3.648558713035192,
      "tokens_seen": 1088487424
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006769709127382147,
      "loss": 1.5107,
      "theoretical_loss": 3.6485382950935437,
      "tokens_seen": 1088552960
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006769508525576731,
      "loss": 1.5323,
      "theoretical_loss": 3.6485178787252823,
      "tokens_seen": 1088618496
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006769307923771314,
      "loss": 1.4612,
      "theoretical_loss": 3.648497463930192,
      "tokens_seen": 1088684032
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006769107321965898,
      "loss": 1.4074,
      "theoretical_loss": 3.6484770507080566,
      "tokens_seen": 1088749568
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006768906720160481,
      "loss": 1.5355,
      "theoretical_loss": 3.6484566390586606,
      "tokens_seen": 1088815104
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006768706118355066,
      "loss": 1.4857,
      "theoretical_loss": 3.6484362289817884,
      "tokens_seen": 1088880640
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006768505516549649,
      "loss": 1.5234,
      "theoretical_loss": 3.6484158204772235,
      "tokens_seen": 1088946176
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006768304914744233,
      "loss": 1.6463,
      "theoretical_loss": 3.6483954135447507,
      "tokens_seen": 1089011712
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006768104312938816,
      "loss": 1.5957,
      "theoretical_loss": 3.6483750081841544,
      "tokens_seen": 1089077248
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006767903711133401,
      "loss": 1.4351,
      "theoretical_loss": 3.648354604395218,
      "tokens_seen": 1089142784
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006767703109327984,
      "loss": 1.4334,
      "theoretical_loss": 3.648334202177727,
      "tokens_seen": 1089208320
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006767502507522568,
      "loss": 1.7271,
      "theoretical_loss": 3.6483138015314656,
      "tokens_seen": 1089273856
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006767301905717152,
      "loss": 1.5031,
      "theoretical_loss": 3.648293402456218,
      "tokens_seen": 1089339392
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006767101303911736,
      "loss": 1.4874,
      "theoretical_loss": 3.6482730049517684,
      "tokens_seen": 1089404928
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006766900702106319,
      "loss": 1.54,
      "theoretical_loss": 3.648252609017902,
      "tokens_seen": 1089470464
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006766700100300903,
      "loss": 1.5703,
      "theoretical_loss": 3.6482322146544033,
      "tokens_seen": 1089536000
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006766499498495486,
      "loss": 1.4827,
      "theoretical_loss": 3.6482118218610564,
      "tokens_seen": 1089601536
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006766298896690071,
      "loss": 1.5849,
      "theoretical_loss": 3.648191430637647,
      "tokens_seen": 1089667072
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006766098294884654,
      "loss": 1.4327,
      "theoretical_loss": 3.648171040983959,
      "tokens_seen": 1089732608
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006765897693079238,
      "loss": 1.5103,
      "theoretical_loss": 3.648150652899777,
      "tokens_seen": 1089798144
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006765697091273822,
      "loss": 1.6159,
      "theoretical_loss": 3.6481302663848867,
      "tokens_seen": 1089863680
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006765496489468405,
      "loss": 1.5631,
      "theoretical_loss": 3.6481098814390727,
      "tokens_seen": 1089929216
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006765295887662989,
      "loss": 1.5739,
      "theoretical_loss": 3.648089498062119,
      "tokens_seen": 1089994752
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006765095285857574,
      "loss": 1.4589,
      "theoretical_loss": 3.6480691162538115,
      "tokens_seen": 1090060288
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006764894684052156,
      "loss": 1.4957,
      "theoretical_loss": 3.6480487360139344,
      "tokens_seen": 1090125824
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006764694082246741,
      "loss": 1.6359,
      "theoretical_loss": 3.648028357342274,
      "tokens_seen": 1090191360
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006764493480441323,
      "loss": 1.5637,
      "theoretical_loss": 3.648007980238614,
      "tokens_seen": 1090256896
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006764292878635908,
      "loss": 1.4761,
      "theoretical_loss": 3.64798760470274,
      "tokens_seen": 1090322432
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006764092276830492,
      "loss": 1.4102,
      "theoretical_loss": 3.6479672307344377,
      "tokens_seen": 1090387968
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006763891675025075,
      "loss": 1.4746,
      "theoretical_loss": 3.6479468583334915,
      "tokens_seen": 1090453504
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006763691073219659,
      "loss": 1.5759,
      "theoretical_loss": 3.6479264874996877,
      "tokens_seen": 1090519040
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006763490471414244,
      "loss": 1.509,
      "theoretical_loss": 3.64790611823281,
      "tokens_seen": 1090584576
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006763289869608826,
      "loss": 1.4725,
      "theoretical_loss": 3.6478857505326445,
      "tokens_seen": 1090650112
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006763089267803411,
      "loss": 1.5734,
      "theoretical_loss": 3.647865384398977,
      "tokens_seen": 1090715648
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006762888665997993,
      "loss": 1.4085,
      "theoretical_loss": 3.6478450198315926,
      "tokens_seen": 1090781184
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006762688064192578,
      "loss": 1.5596,
      "theoretical_loss": 3.647824656830277,
      "tokens_seen": 1090846720
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006762487462387162,
      "loss": 1.5498,
      "theoretical_loss": 3.6478042953948147,
      "tokens_seen": 1090912256
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006762286860581745,
      "loss": 1.644,
      "theoretical_loss": 3.647783935524992,
      "tokens_seen": 1090977792
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006762086258776329,
      "loss": 1.4257,
      "theoretical_loss": 3.6477635772205947,
      "tokens_seen": 1091043328
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006761885656970913,
      "loss": 1.5191,
      "theoretical_loss": 3.647743220481408,
      "tokens_seen": 1091108864
    },
    {
      "epoch": 0.33,
      "objective/train/advantage_avg": 0.49391743540763855,
      "objective/train/docs_used": 400125,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5298376083374023,
      "objective/train/original_loss": 1.5298376083374023,
      "objective/train/theoretical_loss": 3.647722865307218,
      "objective/train/tokens_used": 1111634400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2441958785057068,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050633192062378,
      "objective/train/weighted_lm_loss": 1.6071285009384155,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.647722865307218,
      "tokens_seen": 1091174400
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006761685055165496,
      "loss": 1.495,
      "theoretical_loss": 3.647722865307218,
      "tokens_seen": 1091174400
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006761484453360081,
      "loss": 1.5921,
      "theoretical_loss": 3.6477025116978097,
      "tokens_seen": 1091239936
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006761283851554663,
      "loss": 1.5125,
      "theoretical_loss": 3.647682159652969,
      "tokens_seen": 1091305472
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006761083249749248,
      "loss": 1.4038,
      "theoretical_loss": 3.647661809172482,
      "tokens_seen": 1091371008
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006760882647943831,
      "loss": 1.507,
      "theoretical_loss": 3.6476414602561347,
      "tokens_seen": 1091436544
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006760682046138415,
      "loss": 1.49,
      "theoretical_loss": 3.647621112903713,
      "tokens_seen": 1091502080
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006760481444332999,
      "loss": 1.6492,
      "theoretical_loss": 3.647600767115002,
      "tokens_seen": 1091567616
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006760280842527583,
      "loss": 1.3921,
      "theoretical_loss": 3.6475804228897886,
      "tokens_seen": 1091633152
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006760080240722166,
      "loss": 1.6281,
      "theoretical_loss": 3.6475600802278576,
      "tokens_seen": 1091698688
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006759879638916751,
      "loss": 1.6536,
      "theoretical_loss": 3.6475397391289963,
      "tokens_seen": 1091764224
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006759679037111334,
      "loss": 1.5211,
      "theoretical_loss": 3.6475193995929907,
      "tokens_seen": 1091829760
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006759478435305918,
      "loss": 1.4317,
      "theoretical_loss": 3.6474990616196257,
      "tokens_seen": 1091895296
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006759277833500501,
      "loss": 1.5209,
      "theoretical_loss": 3.6474787252086887,
      "tokens_seen": 1091960832
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006759077231695085,
      "loss": 1.5359,
      "theoretical_loss": 3.6474583903599656,
      "tokens_seen": 1092026368
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006758876629889669,
      "loss": 1.5448,
      "theoretical_loss": 3.6474380570732423,
      "tokens_seen": 1092091904
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006758676028084253,
      "loss": 1.5727,
      "theoretical_loss": 3.6474177253483058,
      "tokens_seen": 1092157440
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006758475426278836,
      "loss": 1.5435,
      "theoretical_loss": 3.6473973951849414,
      "tokens_seen": 1092222976
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000675827482447342,
      "loss": 1.5271,
      "theoretical_loss": 3.647377066582936,
      "tokens_seen": 1092288512
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006758074222668004,
      "loss": 1.4916,
      "theoretical_loss": 3.6473567395420767,
      "tokens_seen": 1092354048
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006757873620862588,
      "loss": 1.5198,
      "theoretical_loss": 3.647336414062149,
      "tokens_seen": 1092419584
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006757673019057171,
      "loss": 1.4289,
      "theoretical_loss": 3.6473160901429393,
      "tokens_seen": 1092485120
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006757472417251756,
      "loss": 1.4338,
      "theoretical_loss": 3.647295767784235,
      "tokens_seen": 1092550656
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006757271815446338,
      "loss": 1.5782,
      "theoretical_loss": 3.647275446985822,
      "tokens_seen": 1092616192
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006757071213640923,
      "loss": 1.4845,
      "theoretical_loss": 3.6472551277474876,
      "tokens_seen": 1092681728
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006756870611835506,
      "loss": 1.5193,
      "theoretical_loss": 3.647234810069018,
      "tokens_seen": 1092747264
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000675667001003009,
      "loss": 1.5629,
      "theoretical_loss": 3.6472144939502,
      "tokens_seen": 1092812800
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006756469408224675,
      "loss": 1.5012,
      "theoretical_loss": 3.64719417939082,
      "tokens_seen": 1092878336
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006756268806419258,
      "loss": 1.5713,
      "theoretical_loss": 3.6471738663906654,
      "tokens_seen": 1092943872
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006756068204613842,
      "loss": 1.5622,
      "theoretical_loss": 3.647153554949522,
      "tokens_seen": 1093009408
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006755867602808426,
      "loss": 1.4912,
      "theoretical_loss": 3.6471332450671783,
      "tokens_seen": 1093074944
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006755667001003009,
      "loss": 1.5816,
      "theoretical_loss": 3.6471129367434205,
      "tokens_seen": 1093140480
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006755466399197593,
      "loss": 1.4023,
      "theoretical_loss": 3.647092629978035,
      "tokens_seen": 1093206016
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006755265797392178,
      "loss": 1.521,
      "theoretical_loss": 3.647072324770809,
      "tokens_seen": 1093271552
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000675506519558676,
      "loss": 1.5213,
      "theoretical_loss": 3.6470520211215303,
      "tokens_seen": 1093337088
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006754864593781345,
      "loss": 1.4253,
      "theoretical_loss": 3.647031719029985,
      "tokens_seen": 1093402624
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006754663991975927,
      "loss": 1.5161,
      "theoretical_loss": 3.6470114184959614,
      "tokens_seen": 1093468160
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006754463390170512,
      "loss": 1.4101,
      "theoretical_loss": 3.646991119519245,
      "tokens_seen": 1093533696
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006754262788365096,
      "loss": 1.5118,
      "theoretical_loss": 3.646970822099625,
      "tokens_seen": 1093599232
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006754062186559679,
      "loss": 1.5352,
      "theoretical_loss": 3.646950526236887,
      "tokens_seen": 1093664768
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006753861584754263,
      "loss": 1.4183,
      "theoretical_loss": 3.6469302319308188,
      "tokens_seen": 1093730304
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006753660982948848,
      "loss": 1.4912,
      "theoretical_loss": 3.6469099391812083,
      "tokens_seen": 1093795840
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000675346038114343,
      "loss": 1.6251,
      "theoretical_loss": 3.6468896479878423,
      "tokens_seen": 1093861376
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006753259779338015,
      "loss": 1.5259,
      "theoretical_loss": 3.6468693583505085,
      "tokens_seen": 1093926912
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006753059177532597,
      "loss": 1.536,
      "theoretical_loss": 3.646849070268994,
      "tokens_seen": 1093992448
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006752858575727182,
      "loss": 1.6278,
      "theoretical_loss": 3.646828783743087,
      "tokens_seen": 1094057984
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006752657973921766,
      "loss": 1.5182,
      "theoretical_loss": 3.6468084987725744,
      "tokens_seen": 1094123520
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006752457372116349,
      "loss": 1.5229,
      "theoretical_loss": 3.646788215357244,
      "tokens_seen": 1094189056
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006752256770310933,
      "loss": 1.4266,
      "theoretical_loss": 3.6467679334968834,
      "tokens_seen": 1094254592
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006752056168505517,
      "loss": 1.5712,
      "theoretical_loss": 3.64674765319128,
      "tokens_seen": 1094320128
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00067518555667001,
      "loss": 1.4797,
      "theoretical_loss": 3.646727374440222,
      "tokens_seen": 1094385664
    },
    {
      "epoch": 0.33,
      "objective/train/advantage_avg": 0.4927752614021301,
      "objective/train/docs_used": 401268,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.367355465888977,
      "objective/train/original_loss": 1.3673555850982666,
      "objective/train/theoretical_loss": 3.646707097243498,
      "objective/train/tokens_used": 1114911200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24321143329143524,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505139827728271,
      "objective/train/weighted_lm_loss": 1.4363179206848145,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.646707097243498,
      "tokens_seen": 1094451200
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006751654964894685,
      "loss": 1.5425,
      "theoretical_loss": 3.646707097243498,
      "tokens_seen": 1094451200
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006751454363089267,
      "loss": 1.5566,
      "theoretical_loss": 3.646686821600894,
      "tokens_seen": 1094516736
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006751253761283852,
      "loss": 1.506,
      "theoretical_loss": 3.646666547512198,
      "tokens_seen": 1094582272
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006751053159478435,
      "loss": 1.3741,
      "theoretical_loss": 3.6466462749771993,
      "tokens_seen": 1094647808
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006750852557673019,
      "loss": 1.5051,
      "theoretical_loss": 3.646626003995685,
      "tokens_seen": 1094713344
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006750651955867603,
      "loss": 1.5773,
      "theoretical_loss": 3.6466057345674425,
      "tokens_seen": 1094778880
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006750451354062187,
      "loss": 1.5567,
      "theoretical_loss": 3.646585466692261,
      "tokens_seen": 1094844416
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000675025075225677,
      "loss": 1.4886,
      "theoretical_loss": 3.646565200369928,
      "tokens_seen": 1094909952
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006750050150451355,
      "loss": 1.5795,
      "theoretical_loss": 3.6465449356002315,
      "tokens_seen": 1094975488
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006749849548645938,
      "loss": 1.4421,
      "theoretical_loss": 3.6465246723829594,
      "tokens_seen": 1095041024
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006749648946840522,
      "loss": 1.5146,
      "theoretical_loss": 3.6465044107179003,
      "tokens_seen": 1095106560
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006749448345035105,
      "loss": 1.4429,
      "theoretical_loss": 3.6464841506048424,
      "tokens_seen": 1095172096
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000674924774322969,
      "loss": 1.4215,
      "theoretical_loss": 3.646463892043574,
      "tokens_seen": 1095237632
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006749047141424273,
      "loss": 1.4552,
      "theoretical_loss": 3.6464436350338825,
      "tokens_seen": 1095303168
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006748846539618857,
      "loss": 1.5209,
      "theoretical_loss": 3.6464233795755576,
      "tokens_seen": 1095368704
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000674864593781344,
      "loss": 1.5718,
      "theoretical_loss": 3.646403125668387,
      "tokens_seen": 1095434240
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006748445336008024,
      "loss": 1.4803,
      "theoretical_loss": 3.6463828733121586,
      "tokens_seen": 1095499776
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006748244734202608,
      "loss": 1.4733,
      "theoretical_loss": 3.646362622506662,
      "tokens_seen": 1095565312
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006748044132397192,
      "loss": 1.4752,
      "theoretical_loss": 3.6463423732516853,
      "tokens_seen": 1095630848
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006747843530591775,
      "loss": 1.5178,
      "theoretical_loss": 3.646322125547017,
      "tokens_seen": 1095696384
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000674764292878636,
      "loss": 1.6401,
      "theoretical_loss": 3.6463018793924453,
      "tokens_seen": 1095761920
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006747442326980942,
      "loss": 1.3761,
      "theoretical_loss": 3.646281634787759,
      "tokens_seen": 1095827456
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006747241725175527,
      "loss": 1.3784,
      "theoretical_loss": 3.6462613917327467,
      "tokens_seen": 1095892992
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000674704112337011,
      "loss": 1.5268,
      "theoretical_loss": 3.6462411502271976,
      "tokens_seen": 1095958528
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006746840521564694,
      "loss": 1.5431,
      "theoretical_loss": 3.6462209102709,
      "tokens_seen": 1096024064
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006746639919759278,
      "loss": 1.4861,
      "theoretical_loss": 3.6462006718636433,
      "tokens_seen": 1096089600
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006746439317953862,
      "loss": 1.4412,
      "theoretical_loss": 3.6461804350052156,
      "tokens_seen": 1096155136
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006746238716148445,
      "loss": 1.47,
      "theoretical_loss": 3.646160199695406,
      "tokens_seen": 1096220672
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000674603811434303,
      "loss": 1.4959,
      "theoretical_loss": 3.6461399659340037,
      "tokens_seen": 1096286208
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006745837512537612,
      "loss": 1.5648,
      "theoretical_loss": 3.6461197337207976,
      "tokens_seen": 1096351744
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006745636910732197,
      "loss": 1.4502,
      "theoretical_loss": 3.6460995030555763,
      "tokens_seen": 1096417280
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000674543630892678,
      "loss": 1.5441,
      "theoretical_loss": 3.646079273938129,
      "tokens_seen": 1096482816
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006745235707121364,
      "loss": 1.4731,
      "theoretical_loss": 3.6460590463682454,
      "tokens_seen": 1096548352
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006745035105315948,
      "loss": 1.5608,
      "theoretical_loss": 3.6460388203457135,
      "tokens_seen": 1096613888
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006744834503510531,
      "loss": 1.3861,
      "theoretical_loss": 3.6460185958703235,
      "tokens_seen": 1096679424
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006744633901705115,
      "loss": 1.4505,
      "theoretical_loss": 3.6459983729418637,
      "tokens_seen": 1096744960
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00067444332998997,
      "loss": 1.5512,
      "theoretical_loss": 3.6459781515601244,
      "tokens_seen": 1096810496
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006744232698094282,
      "loss": 1.5277,
      "theoretical_loss": 3.6459579317248942,
      "tokens_seen": 1096876032
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006744032096288867,
      "loss": 1.5987,
      "theoretical_loss": 3.6459377134359627,
      "tokens_seen": 1096941568
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000674383149448345,
      "loss": 1.569,
      "theoretical_loss": 3.645917496693119,
      "tokens_seen": 1097007104
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006743630892678034,
      "loss": 1.4366,
      "theoretical_loss": 3.6458972814961528,
      "tokens_seen": 1097072640
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006743430290872618,
      "loss": 1.4984,
      "theoretical_loss": 3.6458770678448533,
      "tokens_seen": 1097138176
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006743229689067201,
      "loss": 1.5507,
      "theoretical_loss": 3.6458568557390105,
      "tokens_seen": 1097203712
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006743029087261785,
      "loss": 1.392,
      "theoretical_loss": 3.645836645178413,
      "tokens_seen": 1097269248
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000674282848545637,
      "loss": 1.4723,
      "theoretical_loss": 3.6458164361628516,
      "tokens_seen": 1097334784
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006742627883650952,
      "loss": 1.4773,
      "theoretical_loss": 3.645796228692115,
      "tokens_seen": 1097400320
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006742427281845537,
      "loss": 1.4654,
      "theoretical_loss": 3.645776022765993,
      "tokens_seen": 1097465856
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000674222668004012,
      "loss": 1.4732,
      "theoretical_loss": 3.6457558183842753,
      "tokens_seen": 1097531392
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006742026078234704,
      "loss": 1.4665,
      "theoretical_loss": 3.645735615546752,
      "tokens_seen": 1097596928
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006741825476429288,
      "loss": 1.6162,
      "theoretical_loss": 3.645715414253213,
      "tokens_seen": 1097662464
    },
    {
      "epoch": 0.33,
      "objective/train/advantage_avg": 0.49579912424087524,
      "objective/train/docs_used": 402429,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3258310556411743,
      "objective/train/original_loss": 1.3258309364318848,
      "objective/train/theoretical_loss": 3.645695214503448,
      "objective/train/tokens_used": 1118188000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24606361985206604,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508308410644531,
      "objective/train/weighted_lm_loss": 1.3930492401123047,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.645695214503448,
      "tokens_seen": 1097728000
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006741624874623871,
      "loss": 1.5326,
      "theoretical_loss": 3.645695214503448,
      "tokens_seen": 1097728000
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006741424272818456,
      "loss": 1.5064,
      "theoretical_loss": 3.645675016297246,
      "tokens_seen": 1097793536
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006741223671013039,
      "loss": 1.4899,
      "theoretical_loss": 3.645654819634397,
      "tokens_seen": 1097859072
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006741023069207623,
      "loss": 1.5449,
      "theoretical_loss": 3.6456346245146927,
      "tokens_seen": 1097924608
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006740822467402207,
      "loss": 1.5878,
      "theoretical_loss": 3.645614430937922,
      "tokens_seen": 1097990144
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006740621865596791,
      "loss": 1.4281,
      "theoretical_loss": 3.6455942389038745,
      "tokens_seen": 1098055680
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006740421263791374,
      "loss": 1.4392,
      "theoretical_loss": 3.6455740484123407,
      "tokens_seen": 1098121216
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006740220661985959,
      "loss": 1.4764,
      "theoretical_loss": 3.645553859463111,
      "tokens_seen": 1098186752
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006740020060180542,
      "loss": 1.4742,
      "theoretical_loss": 3.645533672055975,
      "tokens_seen": 1098252288
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006739819458375126,
      "loss": 1.4231,
      "theoretical_loss": 3.6455134861907226,
      "tokens_seen": 1098317824
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006739618856569709,
      "loss": 1.4519,
      "theoretical_loss": 3.645493301867145,
      "tokens_seen": 1098383360
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006739418254764293,
      "loss": 1.4581,
      "theoretical_loss": 3.6454731190850325,
      "tokens_seen": 1098448896
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006739217652958877,
      "loss": 1.6033,
      "theoretical_loss": 3.6454529378441745,
      "tokens_seen": 1098514432
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006739017051153461,
      "loss": 1.5573,
      "theoretical_loss": 3.645432758144362,
      "tokens_seen": 1098579968
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006738816449348044,
      "loss": 1.5138,
      "theoretical_loss": 3.6454125799853854,
      "tokens_seen": 1098645504
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006738615847542628,
      "loss": 1.563,
      "theoretical_loss": 3.645392403367034,
      "tokens_seen": 1098711040
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006738415245737212,
      "loss": 1.5171,
      "theoretical_loss": 3.6453722282891006,
      "tokens_seen": 1098776576
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006738214643931796,
      "loss": 1.4702,
      "theoretical_loss": 3.6453520547513736,
      "tokens_seen": 1098842112
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006738014042126379,
      "loss": 1.4754,
      "theoretical_loss": 3.645331882753645,
      "tokens_seen": 1098907648
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006737813440320964,
      "loss": 1.4873,
      "theoretical_loss": 3.645311712295704,
      "tokens_seen": 1098973184
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006737612838515546,
      "loss": 1.523,
      "theoretical_loss": 3.6452915433773425,
      "tokens_seen": 1099038720
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006737412236710131,
      "loss": 1.4888,
      "theoretical_loss": 3.645271375998351,
      "tokens_seen": 1099104256
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006737211634904714,
      "loss": 1.4898,
      "theoretical_loss": 3.6452512101585195,
      "tokens_seen": 1099169792
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006737011033099298,
      "loss": 1.3337,
      "theoretical_loss": 3.6452310458576394,
      "tokens_seen": 1099235328
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006736810431293882,
      "loss": 1.4228,
      "theoretical_loss": 3.6452108830955012,
      "tokens_seen": 1099300864
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006736609829488466,
      "loss": 1.585,
      "theoretical_loss": 3.645190721871896,
      "tokens_seen": 1099366400
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006736409227683049,
      "loss": 1.3807,
      "theoretical_loss": 3.645170562186615,
      "tokens_seen": 1099431936
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006736208625877634,
      "loss": 1.5915,
      "theoretical_loss": 3.6451504040394482,
      "tokens_seen": 1099497472
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006736008024072216,
      "loss": 1.5875,
      "theoretical_loss": 3.645130247430187,
      "tokens_seen": 1099563008
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006735807422266801,
      "loss": 1.4796,
      "theoretical_loss": 3.645110092358623,
      "tokens_seen": 1099628544
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006735606820461384,
      "loss": 1.4201,
      "theoretical_loss": 3.6450899388245466,
      "tokens_seen": 1099694080
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006735406218655968,
      "loss": 1.5508,
      "theoretical_loss": 3.645069786827749,
      "tokens_seen": 1099759616
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006735205616850552,
      "loss": 1.4636,
      "theoretical_loss": 3.645049636368021,
      "tokens_seen": 1099825152
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006735005015045135,
      "loss": 1.6042,
      "theoretical_loss": 3.6450294874451554,
      "tokens_seen": 1099890688
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006734804413239719,
      "loss": 1.5047,
      "theoretical_loss": 3.645009340058941,
      "tokens_seen": 1099956224
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006734603811434304,
      "loss": 1.4518,
      "theoretical_loss": 3.6449891942091712,
      "tokens_seen": 1100021760
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006734403209628886,
      "loss": 1.5506,
      "theoretical_loss": 3.6449690498956357,
      "tokens_seen": 1100087296
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006734202607823471,
      "loss": 1.4561,
      "theoretical_loss": 3.644948907118127,
      "tokens_seen": 1100152832
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006734002006018055,
      "loss": 1.5508,
      "theoretical_loss": 3.644928765876436,
      "tokens_seen": 1100218368
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006733801404212638,
      "loss": 1.5151,
      "theoretical_loss": 3.644908626170354,
      "tokens_seen": 1100283904
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006733600802407222,
      "loss": 1.4503,
      "theoretical_loss": 3.6448884879996726,
      "tokens_seen": 1100349440
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006733400200601805,
      "loss": 1.3971,
      "theoretical_loss": 3.6448683513641837,
      "tokens_seen": 1100414976
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006733199598796389,
      "loss": 1.5485,
      "theoretical_loss": 3.644848216263678,
      "tokens_seen": 1100480512
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006732998996990974,
      "loss": 1.5164,
      "theoretical_loss": 3.6448280826979484,
      "tokens_seen": 1100546048
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006732798395185556,
      "loss": 1.6342,
      "theoretical_loss": 3.6448079506667845,
      "tokens_seen": 1100611584
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006732597793380141,
      "loss": 1.5644,
      "theoretical_loss": 3.64478782016998,
      "tokens_seen": 1100677120
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006732397191574724,
      "loss": 1.426,
      "theoretical_loss": 3.6447676912073255,
      "tokens_seen": 1100742656
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006732196589769308,
      "loss": 1.5783,
      "theoretical_loss": 3.6447475637786133,
      "tokens_seen": 1100808192
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006731995987963892,
      "loss": 1.5928,
      "theoretical_loss": 3.6447274378836347,
      "tokens_seen": 1100873728
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006731795386158475,
      "loss": 1.4285,
      "theoretical_loss": 3.644707313522182,
      "tokens_seen": 1100939264
    },
    {
      "epoch": 0.33,
      "objective/train/advantage_avg": 0.4861805737018585,
      "objective/train/docs_used": 403530,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6043108701705933,
      "objective/train/original_loss": 1.6043108701705933,
      "objective/train/theoretical_loss": 3.6446871906940466,
      "objective/train/tokens_used": 1121464800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23784573376178741,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0498270988464355,
      "objective/train/weighted_lm_loss": 1.6840763092041016,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6446871906940466,
      "tokens_seen": 1101004800
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006731594784353059,
      "loss": 1.6211,
      "theoretical_loss": 3.6446871906940466,
      "tokens_seen": 1101004800
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006731394182547643,
      "loss": 1.4846,
      "theoretical_loss": 3.6446670693990204,
      "tokens_seen": 1101070336
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006731193580742226,
      "loss": 1.4748,
      "theoretical_loss": 3.644646949636896,
      "tokens_seen": 1101135872
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006730992978936811,
      "loss": 1.5245,
      "theoretical_loss": 3.6446268314074652,
      "tokens_seen": 1101201408
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006730792377131394,
      "loss": 1.4065,
      "theoretical_loss": 3.6446067147105197,
      "tokens_seen": 1101266944
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006730591775325978,
      "loss": 1.4315,
      "theoretical_loss": 3.6445865995458515,
      "tokens_seen": 1101332480
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006730391173520562,
      "loss": 1.3562,
      "theoretical_loss": 3.644566485913253,
      "tokens_seen": 1101398016
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006730190571715146,
      "loss": 1.4938,
      "theoretical_loss": 3.644546373812516,
      "tokens_seen": 1101463552
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006729989969909729,
      "loss": 1.5419,
      "theoretical_loss": 3.644526263243433,
      "tokens_seen": 1101529088
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006729789368104313,
      "loss": 1.5081,
      "theoretical_loss": 3.644506154205797,
      "tokens_seen": 1101594624
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006729588766298896,
      "loss": 1.5642,
      "theoretical_loss": 3.644486046699399,
      "tokens_seen": 1101660160
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006729388164493481,
      "loss": 1.3547,
      "theoretical_loss": 3.644465940724032,
      "tokens_seen": 1101725696
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006729187562688064,
      "loss": 1.6993,
      "theoretical_loss": 3.644445836279488,
      "tokens_seen": 1101791232
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006728986960882648,
      "loss": 1.4868,
      "theoretical_loss": 3.64442573336556,
      "tokens_seen": 1101856768
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006728786359077231,
      "loss": 1.3949,
      "theoretical_loss": 3.6444056319820395,
      "tokens_seen": 1101922304
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006728585757271816,
      "loss": 1.2964,
      "theoretical_loss": 3.6443855321287195,
      "tokens_seen": 1101987840
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006728385155466399,
      "loss": 1.5378,
      "theoretical_loss": 3.644365433805393,
      "tokens_seen": 1102053376
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006728184553660983,
      "loss": 1.3679,
      "theoretical_loss": 3.6443453370118517,
      "tokens_seen": 1102118912
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006727983951855566,
      "loss": 1.4877,
      "theoretical_loss": 3.644325241747889,
      "tokens_seen": 1102184448
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000672778335005015,
      "loss": 1.4747,
      "theoretical_loss": 3.644305148013297,
      "tokens_seen": 1102249984
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006727582748244734,
      "loss": 1.5524,
      "theoretical_loss": 3.6442850558078685,
      "tokens_seen": 1102315520
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006727382146439318,
      "loss": 1.6611,
      "theoretical_loss": 3.6442649651313963,
      "tokens_seen": 1102381056
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006727181544633901,
      "loss": 1.4341,
      "theoretical_loss": 3.644244875983673,
      "tokens_seen": 1102446592
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006726980942828486,
      "loss": 1.5354,
      "theoretical_loss": 3.6442247883644923,
      "tokens_seen": 1102512128
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006726780341023069,
      "loss": 1.6408,
      "theoretical_loss": 3.6442047022736452,
      "tokens_seen": 1102577664
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006726579739217653,
      "loss": 1.5075,
      "theoretical_loss": 3.644184617710926,
      "tokens_seen": 1102643200
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006726379137412238,
      "loss": 1.578,
      "theoretical_loss": 3.6441645346761273,
      "tokens_seen": 1102708736
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000672617853560682,
      "loss": 1.558,
      "theoretical_loss": 3.6441444531690426,
      "tokens_seen": 1102774272
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006725977933801405,
      "loss": 1.5677,
      "theoretical_loss": 3.644124373189464,
      "tokens_seen": 1102839808
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006725777331995988,
      "loss": 1.3814,
      "theoretical_loss": 3.6441042947371844,
      "tokens_seen": 1102905344
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006725576730190572,
      "loss": 1.6479,
      "theoretical_loss": 3.6440842178119985,
      "tokens_seen": 1102970880
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006725376128385156,
      "loss": 1.5759,
      "theoretical_loss": 3.644064142413698,
      "tokens_seen": 1103036416
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006725175526579739,
      "loss": 1.4801,
      "theoretical_loss": 3.644044068542076,
      "tokens_seen": 1103101952
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006724974924774323,
      "loss": 1.4233,
      "theoretical_loss": 3.644023996196926,
      "tokens_seen": 1103167488
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006724774322968908,
      "loss": 1.4964,
      "theoretical_loss": 3.644003925378042,
      "tokens_seen": 1103233024
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000672457372116349,
      "loss": 1.5414,
      "theoretical_loss": 3.643983856085216,
      "tokens_seen": 1103298560
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006724373119358075,
      "loss": 1.4891,
      "theoretical_loss": 3.643963788318242,
      "tokens_seen": 1103364096
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006724172517552659,
      "loss": 1.4655,
      "theoretical_loss": 3.643943722076914,
      "tokens_seen": 1103429632
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006723971915747242,
      "loss": 1.5599,
      "theoretical_loss": 3.6439236573610243,
      "tokens_seen": 1103495168
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006723771313941826,
      "loss": 1.4325,
      "theoretical_loss": 3.643903594170367,
      "tokens_seen": 1103560704
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006723570712136409,
      "loss": 1.4672,
      "theoretical_loss": 3.6438835325047356,
      "tokens_seen": 1103626240
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006723370110330993,
      "loss": 1.4542,
      "theoretical_loss": 3.6438634723639236,
      "tokens_seen": 1103691776
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006723169508525578,
      "loss": 1.4771,
      "theoretical_loss": 3.6438434137477236,
      "tokens_seen": 1103757312
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000672296890672016,
      "loss": 1.5099,
      "theoretical_loss": 3.643823356655931,
      "tokens_seen": 1103822848
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006722768304914745,
      "loss": 1.5177,
      "theoretical_loss": 3.6438033010883375,
      "tokens_seen": 1103888384
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006722567703109328,
      "loss": 1.5008,
      "theoretical_loss": 3.6437832470447384,
      "tokens_seen": 1103953920
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006722367101303912,
      "loss": 1.4975,
      "theoretical_loss": 3.6437631945249267,
      "tokens_seen": 1104019456
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006722166499498496,
      "loss": 1.5229,
      "theoretical_loss": 3.643743143528696,
      "tokens_seen": 1104084992
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000672196589769308,
      "loss": 1.5616,
      "theoretical_loss": 3.643723094055841,
      "tokens_seen": 1104150528
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006721765295887663,
      "loss": 1.5703,
      "theoretical_loss": 3.6437030461061544,
      "tokens_seen": 1104216064
    },
    {
      "epoch": 0.33,
      "objective/train/advantage_avg": 0.4902866780757904,
      "objective/train/docs_used": 404699,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5548778772354126,
      "objective/train/original_loss": 1.554877519607544,
      "objective/train/theoretical_loss": 3.643682999679431,
      "objective/train/tokens_used": 1124741600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24122250080108643,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502548217773438,
      "objective/train/weighted_lm_loss": 1.6329830884933472,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.643682999679431,
      "tokens_seen": 1104281600
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006721564694082247,
      "loss": 1.546,
      "theoretical_loss": 3.643682999679431,
      "tokens_seen": 1104281600
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000672136409227683,
      "loss": 1.5998,
      "theoretical_loss": 3.643662954775464,
      "tokens_seen": 1104347136
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006721163490471415,
      "loss": 1.5458,
      "theoretical_loss": 3.643642911394048,
      "tokens_seen": 1104412672
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006720962888665998,
      "loss": 1.3285,
      "theoretical_loss": 3.6436228695349766,
      "tokens_seen": 1104478208
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006720762286860582,
      "loss": 1.5538,
      "theoretical_loss": 3.643602829198044,
      "tokens_seen": 1104543744
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006720561685055166,
      "loss": 1.5948,
      "theoretical_loss": 3.6435827903830447,
      "tokens_seen": 1104609280
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000672036108324975,
      "loss": 1.5467,
      "theoretical_loss": 3.643562753089772,
      "tokens_seen": 1104674816
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000672036108324975,
      "loss": 1.4746,
      "theoretical_loss": 3.643542717318021,
      "tokens_seen": 1104740352
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006720160481444333,
      "loss": 1.5141,
      "theoretical_loss": 3.6435226830675846,
      "tokens_seen": 1104805888
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006719959879638917,
      "loss": 1.4923,
      "theoretical_loss": 3.6435026503382586,
      "tokens_seen": 1104871424
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00067197592778335,
      "loss": 1.5185,
      "theoretical_loss": 3.6434826191298364,
      "tokens_seen": 1104936960
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006719558676028085,
      "loss": 1.5145,
      "theoretical_loss": 3.6434625894421124,
      "tokens_seen": 1105002496
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006719358074222668,
      "loss": 1.4075,
      "theoretical_loss": 3.643442561274881,
      "tokens_seen": 1105068032
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006719157472417252,
      "loss": 1.557,
      "theoretical_loss": 3.643422534627937,
      "tokens_seen": 1105133568
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006718956870611835,
      "loss": 1.4583,
      "theoretical_loss": 3.6434025095010747,
      "tokens_seen": 1105199104
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000671875626880642,
      "loss": 1.5405,
      "theoretical_loss": 3.6433824858940884,
      "tokens_seen": 1105264640
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006718555667001003,
      "loss": 1.4985,
      "theoretical_loss": 3.6433624638067723,
      "tokens_seen": 1105330176
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0006718355065195587,
      "loss": 1.4929,
      "theoretical_loss": 3.6433424432389216,
      "tokens_seen": 1105395712
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000671815446339017,
      "loss": 1.7293,
      "theoretical_loss": 3.6433224241903304,
      "tokens_seen": 1105461248
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006717953861584754,
      "loss": 1.5674,
      "theoretical_loss": 3.643302406660794,
      "tokens_seen": 1105526784
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006717753259779338,
      "loss": 1.5227,
      "theoretical_loss": 3.643282390650106,
      "tokens_seen": 1105592320
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006717552657973922,
      "loss": 1.6216,
      "theoretical_loss": 3.643262376158063,
      "tokens_seen": 1105657856
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006717352056168505,
      "loss": 1.4265,
      "theoretical_loss": 3.643242363184458,
      "tokens_seen": 1105723392
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000671715145436309,
      "loss": 1.5218,
      "theoretical_loss": 3.6432223517290865,
      "tokens_seen": 1105788928
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006716950852557673,
      "loss": 1.4586,
      "theoretical_loss": 3.643202341791743,
      "tokens_seen": 1105854464
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006716750250752257,
      "loss": 1.4881,
      "theoretical_loss": 3.643182333372223,
      "tokens_seen": 1105920000
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000671654964894684,
      "loss": 1.5907,
      "theoretical_loss": 3.6431623264703212,
      "tokens_seen": 1105985536
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006716349047141424,
      "loss": 1.4305,
      "theoretical_loss": 3.6431423210858322,
      "tokens_seen": 1106051072
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006716148445336008,
      "loss": 1.5379,
      "theoretical_loss": 3.6431223172185514,
      "tokens_seen": 1106116608
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006715947843530592,
      "loss": 1.3536,
      "theoretical_loss": 3.643102314868274,
      "tokens_seen": 1106182144
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006715747241725175,
      "loss": 1.4784,
      "theoretical_loss": 3.6430823140347943,
      "tokens_seen": 1106247680
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000671554663991976,
      "loss": 1.494,
      "theoretical_loss": 3.6430623147179086,
      "tokens_seen": 1106313216
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006715346038114342,
      "loss": 1.4747,
      "theoretical_loss": 3.643042316917411,
      "tokens_seen": 1106378752
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006715145436308927,
      "loss": 1.4485,
      "theoretical_loss": 3.6430223206330976,
      "tokens_seen": 1106444288
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006714944834503511,
      "loss": 1.5419,
      "theoretical_loss": 3.643002325864763,
      "tokens_seen": 1106509824
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006714744232698094,
      "loss": 1.4904,
      "theoretical_loss": 3.6429823326122026,
      "tokens_seen": 1106575360
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006714543630892678,
      "loss": 1.4663,
      "theoretical_loss": 3.6429623408752114,
      "tokens_seen": 1106640896
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006714343029087263,
      "loss": 1.4793,
      "theoretical_loss": 3.642942350653586,
      "tokens_seen": 1106706432
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006714142427281845,
      "loss": 1.5116,
      "theoretical_loss": 3.6429223619471207,
      "tokens_seen": 1106771968
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000671394182547643,
      "loss": 1.6254,
      "theoretical_loss": 3.6429023747556113,
      "tokens_seen": 1106837504
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006713741223671012,
      "loss": 1.4818,
      "theoretical_loss": 3.642882389078853,
      "tokens_seen": 1106903040
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006713540621865597,
      "loss": 1.4762,
      "theoretical_loss": 3.642862404916642,
      "tokens_seen": 1106968576
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006713340020060181,
      "loss": 1.5387,
      "theoretical_loss": 3.6428424222687736,
      "tokens_seen": 1107034112
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006713139418254764,
      "loss": 1.5061,
      "theoretical_loss": 3.642822441135043,
      "tokens_seen": 1107099648
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006712938816449348,
      "loss": 1.4933,
      "theoretical_loss": 3.642802461515246,
      "tokens_seen": 1107165184
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006712738214643932,
      "loss": 1.5127,
      "theoretical_loss": 3.6427824834091784,
      "tokens_seen": 1107230720
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006712537612838515,
      "loss": 1.551,
      "theoretical_loss": 3.642762506816636,
      "tokens_seen": 1107296256
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00067123370110331,
      "loss": 1.3955,
      "theoretical_loss": 3.6427425317374142,
      "tokens_seen": 1107361792
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006712136409227682,
      "loss": 1.4502,
      "theoretical_loss": 3.6427225581713096,
      "tokens_seen": 1107427328
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006711935807422267,
      "loss": 1.4223,
      "theoretical_loss": 3.6427025861181175,
      "tokens_seen": 1107492864
    },
    {
      "epoch": 0.34,
      "objective/train/advantage_avg": 0.4917794167995453,
      "objective/train/docs_used": 405946,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.439220666885376,
      "objective/train/original_loss": 1.4392204284667969,
      "objective/train/theoretical_loss": 3.642682615577634,
      "objective/train/tokens_used": 1128018400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24233192205429077,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504099130630493,
      "objective/train/weighted_lm_loss": 1.511704921722412,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.642682615577634,
      "tokens_seen": 1107558400
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000671173520561685,
      "loss": 1.5764,
      "theoretical_loss": 3.642682615577634,
      "tokens_seen": 1107558400
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006711534603811434,
      "loss": 1.5833,
      "theoretical_loss": 3.6426626465496543,
      "tokens_seen": 1107623936
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006711334002006019,
      "loss": 1.5195,
      "theoretical_loss": 3.642642679033975,
      "tokens_seen": 1107689472
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006711133400200602,
      "loss": 1.5937,
      "theoretical_loss": 3.6426227130303923,
      "tokens_seen": 1107755008
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006710932798395186,
      "loss": 1.5149,
      "theoretical_loss": 3.6426027485387023,
      "tokens_seen": 1107820544
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000671073219658977,
      "loss": 1.5109,
      "theoretical_loss": 3.6425827855587,
      "tokens_seen": 1107886080
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006710531594784354,
      "loss": 1.5412,
      "theoretical_loss": 3.642562824090183,
      "tokens_seen": 1107951616
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006710330992978937,
      "loss": 1.4804,
      "theoretical_loss": 3.6425428641329463,
      "tokens_seen": 1108017152
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006710130391173521,
      "loss": 1.4995,
      "theoretical_loss": 3.6425229056867865,
      "tokens_seen": 1108082688
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006709929789368104,
      "loss": 1.4787,
      "theoretical_loss": 3.6425029487515004,
      "tokens_seen": 1108148224
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006709729187562689,
      "loss": 1.4328,
      "theoretical_loss": 3.6424829933268836,
      "tokens_seen": 1108213760
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006709528585757272,
      "loss": 1.5303,
      "theoretical_loss": 3.6424630394127324,
      "tokens_seen": 1108279296
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006709327983951856,
      "loss": 1.466,
      "theoretical_loss": 3.642443087008844,
      "tokens_seen": 1108344832
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006709127382146439,
      "loss": 1.4864,
      "theoretical_loss": 3.6424231361150134,
      "tokens_seen": 1108410368
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006708926780341024,
      "loss": 1.5152,
      "theoretical_loss": 3.642403186731038,
      "tokens_seen": 1108475904
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006708726178535607,
      "loss": 1.44,
      "theoretical_loss": 3.6423832388567146,
      "tokens_seen": 1108541440
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006708525576730191,
      "loss": 1.4361,
      "theoretical_loss": 3.6423632924918383,
      "tokens_seen": 1108606976
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006708324974924774,
      "loss": 1.4366,
      "theoretical_loss": 3.6423433476362073,
      "tokens_seen": 1108672512
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006708124373119358,
      "loss": 1.4807,
      "theoretical_loss": 3.642323404289617,
      "tokens_seen": 1108738048
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006707923771313942,
      "loss": 1.4816,
      "theoretical_loss": 3.642303462451865,
      "tokens_seen": 1108803584
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006707723169508526,
      "loss": 1.5712,
      "theoretical_loss": 3.6422835221227468,
      "tokens_seen": 1108869120
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006707522567703109,
      "loss": 1.5612,
      "theoretical_loss": 3.6422635833020607,
      "tokens_seen": 1108934656
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006707321965897694,
      "loss": 1.4734,
      "theoretical_loss": 3.642243645989602,
      "tokens_seen": 1109000192
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006707121364092277,
      "loss": 1.538,
      "theoretical_loss": 3.642223710185168,
      "tokens_seen": 1109065728
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006706920762286861,
      "loss": 1.385,
      "theoretical_loss": 3.6422037758885555,
      "tokens_seen": 1109131264
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006706720160481445,
      "loss": 1.5568,
      "theoretical_loss": 3.642183843099562,
      "tokens_seen": 1109196800
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006706519558676028,
      "loss": 1.4302,
      "theoretical_loss": 3.642163911817983,
      "tokens_seen": 1109262336
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006706318956870612,
      "loss": 1.5804,
      "theoretical_loss": 3.642143982043617,
      "tokens_seen": 1109327872
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006706118355065196,
      "loss": 1.5371,
      "theoretical_loss": 3.6421240537762607,
      "tokens_seen": 1109393408
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006705917753259779,
      "loss": 1.6212,
      "theoretical_loss": 3.6421041270157097,
      "tokens_seen": 1109458944
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006705717151454364,
      "loss": 1.5609,
      "theoretical_loss": 3.642084201761763,
      "tokens_seen": 1109524480
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006705516549648946,
      "loss": 1.5456,
      "theoretical_loss": 3.6420642780142165,
      "tokens_seen": 1109590016
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006705315947843531,
      "loss": 1.6036,
      "theoretical_loss": 3.6420443557728674,
      "tokens_seen": 1109655552
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006705115346038115,
      "loss": 1.5047,
      "theoretical_loss": 3.6420244350375137,
      "tokens_seen": 1109721088
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006704914744232698,
      "loss": 1.5695,
      "theoretical_loss": 3.642004515807952,
      "tokens_seen": 1109786624
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006704714142427282,
      "loss": 1.4701,
      "theoretical_loss": 3.641984598083979,
      "tokens_seen": 1109852160
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006704513540621867,
      "loss": 1.5076,
      "theoretical_loss": 3.6419646818653932,
      "tokens_seen": 1109917696
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006704312938816449,
      "loss": 1.4839,
      "theoretical_loss": 3.6419447671519913,
      "tokens_seen": 1109983232
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006704112337011034,
      "loss": 1.6013,
      "theoretical_loss": 3.641924853943571,
      "tokens_seen": 1110048768
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006703911735205616,
      "loss": 1.5839,
      "theoretical_loss": 3.641904942239929,
      "tokens_seen": 1110114304
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006703711133400201,
      "loss": 1.5418,
      "theoretical_loss": 3.641885032040864,
      "tokens_seen": 1110179840
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006703510531594785,
      "loss": 1.4393,
      "theoretical_loss": 3.6418651233461725,
      "tokens_seen": 1110245376
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006703309929789368,
      "loss": 1.5958,
      "theoretical_loss": 3.6418452161556525,
      "tokens_seen": 1110310912
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006703109327983952,
      "loss": 1.5714,
      "theoretical_loss": 3.641825310469101,
      "tokens_seen": 1110376448
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006702908726178536,
      "loss": 1.5345,
      "theoretical_loss": 3.6418054062863163,
      "tokens_seen": 1110441984
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006702708124373119,
      "loss": 1.5439,
      "theoretical_loss": 3.6417855036070956,
      "tokens_seen": 1110507520
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006702507522567704,
      "loss": 1.37,
      "theoretical_loss": 3.641765602431237,
      "tokens_seen": 1110573056
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006702306920762286,
      "loss": 1.579,
      "theoretical_loss": 3.641745702758538,
      "tokens_seen": 1110638592
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006702106318956871,
      "loss": 1.4924,
      "theoretical_loss": 3.6417258045887966,
      "tokens_seen": 1110704128
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006701905717151454,
      "loss": 1.5353,
      "theoretical_loss": 3.6417059079218097,
      "tokens_seen": 1110769664
    },
    {
      "epoch": 0.34,
      "objective/train/advantage_avg": 0.49183207750320435,
      "objective/train/docs_used": 406993,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.438286542892456,
      "objective/train/original_loss": 1.438286304473877,
      "objective/train/theoretical_loss": 3.6416860127573765,
      "objective/train/tokens_used": 1131295200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24259105324745178,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504164695739746,
      "objective/train/weighted_lm_loss": 1.5104471445083618,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.6416860127573765,
      "tokens_seen": 1110835200
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006701705115346038,
      "loss": 1.4714,
      "theoretical_loss": 3.6416860127573765,
      "tokens_seen": 1110835200
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006701504513540622,
      "loss": 1.4823,
      "theoretical_loss": 3.6416661190952944,
      "tokens_seen": 1110900736
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006701303911735206,
      "loss": 1.6122,
      "theoretical_loss": 3.641646226935361,
      "tokens_seen": 1110966272
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006701103309929789,
      "loss": 1.674,
      "theoretical_loss": 3.6416263362773744,
      "tokens_seen": 1111031808
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006700902708124374,
      "loss": 1.4617,
      "theoretical_loss": 3.6416064471211325,
      "tokens_seen": 1111097344
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006700702106318956,
      "loss": 1.492,
      "theoretical_loss": 3.641586559466434,
      "tokens_seen": 1111162880
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006700501504513541,
      "loss": 1.5999,
      "theoretical_loss": 3.641566673313076,
      "tokens_seen": 1111228416
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006700300902708124,
      "loss": 1.5639,
      "theoretical_loss": 3.6415467886608575,
      "tokens_seen": 1111293952
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006700100300902708,
      "loss": 1.6939,
      "theoretical_loss": 3.6415269055095765,
      "tokens_seen": 1111359488
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006699899699097292,
      "loss": 1.5531,
      "theoretical_loss": 3.6415070238590306,
      "tokens_seen": 1111425024
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006699699097291876,
      "loss": 1.5656,
      "theoretical_loss": 3.6414871437090186,
      "tokens_seen": 1111490560
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006699498495486459,
      "loss": 1.5503,
      "theoretical_loss": 3.641467265059339,
      "tokens_seen": 1111556096
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006699297893681043,
      "loss": 1.567,
      "theoretical_loss": 3.6414473879097895,
      "tokens_seen": 1111621632
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006699097291875627,
      "loss": 1.533,
      "theoretical_loss": 3.6414275122601687,
      "tokens_seen": 1111687168
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006698896690070211,
      "loss": 1.5059,
      "theoretical_loss": 3.641407638110275,
      "tokens_seen": 1111752704
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006698696088264794,
      "loss": 1.6057,
      "theoretical_loss": 3.641387765459907,
      "tokens_seen": 1111818240
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006698495486459378,
      "loss": 1.5804,
      "theoretical_loss": 3.641367894308863,
      "tokens_seen": 1111883776
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006698294884653961,
      "loss": 1.4995,
      "theoretical_loss": 3.641348024656942,
      "tokens_seen": 1111949312
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006698094282848546,
      "loss": 1.4966,
      "theoretical_loss": 3.641328156503942,
      "tokens_seen": 1112014848
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006697893681043129,
      "loss": 1.5413,
      "theoretical_loss": 3.6413082898496616,
      "tokens_seen": 1112080384
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006697693079237713,
      "loss": 1.4707,
      "theoretical_loss": 3.6412884246938995,
      "tokens_seen": 1112145920
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006697492477432297,
      "loss": 1.5835,
      "theoretical_loss": 3.641268561036455,
      "tokens_seen": 1112211456
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006697291875626881,
      "loss": 1.603,
      "theoretical_loss": 3.6412486988771255,
      "tokens_seen": 1112276992
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006697091273821464,
      "loss": 1.4749,
      "theoretical_loss": 3.641228838215711,
      "tokens_seen": 1112342528
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006696890672016049,
      "loss": 1.5096,
      "theoretical_loss": 3.6412089790520095,
      "tokens_seen": 1112408064
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006696690070210632,
      "loss": 1.3982,
      "theoretical_loss": 3.6411891213858203,
      "tokens_seen": 1112473600
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006696489468405216,
      "loss": 1.576,
      "theoretical_loss": 3.6411692652169423,
      "tokens_seen": 1112539136
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00066962888665998,
      "loss": 1.5046,
      "theoretical_loss": 3.641149410545174,
      "tokens_seen": 1112604672
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006696088264794383,
      "loss": 1.53,
      "theoretical_loss": 3.6411295573703146,
      "tokens_seen": 1112670208
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006695887662988968,
      "loss": 1.5805,
      "theoretical_loss": 3.6411097056921626,
      "tokens_seen": 1112735744
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000669568706118355,
      "loss": 1.5279,
      "theoretical_loss": 3.641089855510518,
      "tokens_seen": 1112801280
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006695486459378135,
      "loss": 1.554,
      "theoretical_loss": 3.6410700068251796,
      "tokens_seen": 1112866816
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006695285857572719,
      "loss": 1.5354,
      "theoretical_loss": 3.6410501596359457,
      "tokens_seen": 1112932352
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006695085255767302,
      "loss": 1.4089,
      "theoretical_loss": 3.641030313942616,
      "tokens_seen": 1112997888
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006694884653961886,
      "loss": 1.444,
      "theoretical_loss": 3.64101046974499,
      "tokens_seen": 1113063424
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006694684052156471,
      "loss": 1.5257,
      "theoretical_loss": 3.6409906270428665,
      "tokens_seen": 1113128960
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006694483450351053,
      "loss": 1.5058,
      "theoretical_loss": 3.640970785836045,
      "tokens_seen": 1113194496
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006694282848545638,
      "loss": 1.4851,
      "theoretical_loss": 3.6409509461243243,
      "tokens_seen": 1113260032
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000669408224674022,
      "loss": 1.539,
      "theoretical_loss": 3.640931107907504,
      "tokens_seen": 1113325568
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006693881644934805,
      "loss": 1.6791,
      "theoretical_loss": 3.640911271185384,
      "tokens_seen": 1113391104
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006693681043129389,
      "loss": 1.3848,
      "theoretical_loss": 3.640891435957763,
      "tokens_seen": 1113456640
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006693480441323972,
      "loss": 1.5505,
      "theoretical_loss": 3.6408716022244407,
      "tokens_seen": 1113522176
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006693279839518556,
      "loss": 1.6537,
      "theoretical_loss": 3.6408517699852165,
      "tokens_seen": 1113587712
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000669307923771314,
      "loss": 1.5527,
      "theoretical_loss": 3.64083193923989,
      "tokens_seen": 1113653248
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006692878635907723,
      "loss": 1.5077,
      "theoretical_loss": 3.640812109988261,
      "tokens_seen": 1113718784
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006692678034102308,
      "loss": 1.6569,
      "theoretical_loss": 3.6407922822301284,
      "tokens_seen": 1113784320
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000669247743229689,
      "loss": 1.6587,
      "theoretical_loss": 3.640772455965293,
      "tokens_seen": 1113849856
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006692276830491475,
      "loss": 1.5518,
      "theoretical_loss": 3.6407526311935534,
      "tokens_seen": 1113915392
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006692076228686058,
      "loss": 1.4521,
      "theoretical_loss": 3.64073280791471,
      "tokens_seen": 1113980928
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006691875626880642,
      "loss": 1.4879,
      "theoretical_loss": 3.6407129861285625,
      "tokens_seen": 1114046464
    },
    {
      "epoch": 0.34,
      "objective/train/advantage_avg": 0.49177318811416626,
      "objective/train/docs_used": 408181,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3918049335479736,
      "objective/train/original_loss": 1.3918049335479736,
      "objective/train/theoretical_loss": 3.64069316583491,
      "objective/train/tokens_used": 1134572000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24233345687389374,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050409197807312,
      "objective/train/weighted_lm_loss": 1.4619276523590088,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.64069316583491,
      "tokens_seen": 1114112000
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006691675025075226,
      "loss": 1.4236,
      "theoretical_loss": 3.64069316583491,
      "tokens_seen": 1114112000
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000669147442326981,
      "loss": 1.5193,
      "theoretical_loss": 3.6406733470335535,
      "tokens_seen": 1114177536
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006691273821464393,
      "loss": 1.57,
      "theoretical_loss": 3.6406535297242915,
      "tokens_seen": 1114243072
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006691073219658978,
      "loss": 1.501,
      "theoretical_loss": 3.640633713906925,
      "tokens_seen": 1114308608
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000669087261785356,
      "loss": 1.5394,
      "theoretical_loss": 3.640613899581253,
      "tokens_seen": 1114374144
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006690672016048145,
      "loss": 1.4764,
      "theoretical_loss": 3.640594086747077,
      "tokens_seen": 1114439680
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006690471414242728,
      "loss": 1.6153,
      "theoretical_loss": 3.6405742754041954,
      "tokens_seen": 1114505216
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006690270812437312,
      "loss": 1.4947,
      "theoretical_loss": 3.64055446555241,
      "tokens_seen": 1114570752
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006690070210631896,
      "loss": 1.7836,
      "theoretical_loss": 3.6405346571915187,
      "tokens_seen": 1114636288
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000668986960882648,
      "loss": 1.6291,
      "theoretical_loss": 3.640514850321323,
      "tokens_seen": 1114701824
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006689669007021063,
      "loss": 1.5191,
      "theoretical_loss": 3.6404950449416233,
      "tokens_seen": 1114767360
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006689468405215647,
      "loss": 1.4682,
      "theoretical_loss": 3.6404752410522194,
      "tokens_seen": 1114832896
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000668926780341023,
      "loss": 1.5589,
      "theoretical_loss": 3.6404554386529115,
      "tokens_seen": 1114898432
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006689067201604815,
      "loss": 1.5834,
      "theoretical_loss": 3.6404356377435,
      "tokens_seen": 1114963968
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006688866599799398,
      "loss": 1.4854,
      "theoretical_loss": 3.6404158383237855,
      "tokens_seen": 1115029504
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006688665997993982,
      "loss": 1.443,
      "theoretical_loss": 3.640396040393568,
      "tokens_seen": 1115095040
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006688465396188565,
      "loss": 1.3722,
      "theoretical_loss": 3.640376243952648,
      "tokens_seen": 1115160576
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000668826479438315,
      "loss": 1.4882,
      "theoretical_loss": 3.6403564490008256,
      "tokens_seen": 1115226112
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006688064192577733,
      "loss": 1.5665,
      "theoretical_loss": 3.640336655537902,
      "tokens_seen": 1115291648
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006687863590772317,
      "loss": 1.4497,
      "theoretical_loss": 3.640316863563678,
      "tokens_seen": 1115357184
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006687662988966901,
      "loss": 1.3922,
      "theoretical_loss": 3.640297073077953,
      "tokens_seen": 1115422720
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006687462387161485,
      "loss": 1.518,
      "theoretical_loss": 3.6402772840805278,
      "tokens_seen": 1115488256
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006687261785356068,
      "loss": 1.5526,
      "theoretical_loss": 3.640257496571204,
      "tokens_seen": 1115553792
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006687061183550653,
      "loss": 1.6628,
      "theoretical_loss": 3.6402377105497816,
      "tokens_seen": 1115619328
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006686860581745235,
      "loss": 1.6309,
      "theoretical_loss": 3.640217926016061,
      "tokens_seen": 1115684864
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000668665997993982,
      "loss": 1.5087,
      "theoretical_loss": 3.640198142969844,
      "tokens_seen": 1115750400
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006686459378134403,
      "loss": 1.5161,
      "theoretical_loss": 3.6401783614109307,
      "tokens_seen": 1115815936
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006686258776328987,
      "loss": 1.5411,
      "theoretical_loss": 3.6401585813391217,
      "tokens_seen": 1115881472
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006686058174523571,
      "loss": 1.4933,
      "theoretical_loss": 3.6401388027542185,
      "tokens_seen": 1115947008
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006685857572718154,
      "loss": 1.501,
      "theoretical_loss": 3.6401190256560216,
      "tokens_seen": 1116012544
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006685656970912738,
      "loss": 1.4655,
      "theoretical_loss": 3.640099250044332,
      "tokens_seen": 1116078080
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006685456369107323,
      "loss": 1.3823,
      "theoretical_loss": 3.6400794759189514,
      "tokens_seen": 1116143616
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006685255767301905,
      "loss": 1.4341,
      "theoretical_loss": 3.6400597032796798,
      "tokens_seen": 1116209152
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000668505516549649,
      "loss": 1.5672,
      "theoretical_loss": 3.6400399321263186,
      "tokens_seen": 1116274688
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006684854563691072,
      "loss": 1.6057,
      "theoretical_loss": 3.6400201624586694,
      "tokens_seen": 1116340224
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006684653961885657,
      "loss": 1.5134,
      "theoretical_loss": 3.6400003942765324,
      "tokens_seen": 1116405760
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006684453360080241,
      "loss": 1.4611,
      "theoretical_loss": 3.6399806275797095,
      "tokens_seen": 1116471296
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006684252758274824,
      "loss": 1.5756,
      "theoretical_loss": 3.6399608623680018,
      "tokens_seen": 1116536832
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006684052156469408,
      "loss": 1.491,
      "theoretical_loss": 3.63994109864121,
      "tokens_seen": 1116602368
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006683851554663993,
      "loss": 1.5806,
      "theoretical_loss": 3.6399213363991367,
      "tokens_seen": 1116667904
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006683650952858575,
      "loss": 1.5781,
      "theoretical_loss": 3.639901575641582,
      "tokens_seen": 1116733440
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000668345035105316,
      "loss": 1.4311,
      "theoretical_loss": 3.639881816368348,
      "tokens_seen": 1116798976
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006683249749247742,
      "loss": 1.6,
      "theoretical_loss": 3.6398620585792356,
      "tokens_seen": 1116864512
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006683049147442327,
      "loss": 1.5427,
      "theoretical_loss": 3.6398423022740465,
      "tokens_seen": 1116930048
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006682848545636911,
      "loss": 1.5047,
      "theoretical_loss": 3.6398225474525816,
      "tokens_seen": 1116995584
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006682647943831494,
      "loss": 1.4524,
      "theoretical_loss": 3.639802794114644,
      "tokens_seen": 1117061120
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006682447342026078,
      "loss": 1.4831,
      "theoretical_loss": 3.6397830422600332,
      "tokens_seen": 1117126656
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006682246740220662,
      "loss": 1.4988,
      "theoretical_loss": 3.6397632918885527,
      "tokens_seen": 1117192192
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006682046138415245,
      "loss": 1.5099,
      "theoretical_loss": 3.639743543000003,
      "tokens_seen": 1117257728
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000668184553660983,
      "loss": 1.4568,
      "theoretical_loss": 3.639723795594186,
      "tokens_seen": 1117323264
    },
    {
      "epoch": 0.34,
      "objective/train/advantage_avg": 0.4931054711341858,
      "objective/train/docs_used": 409409,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.501597285270691,
      "objective/train/original_loss": 1.5015971660614014,
      "objective/train/theoretical_loss": 3.639704049670904,
      "objective/train/tokens_used": 1137848800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2436681091785431,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505492687225342,
      "objective/train/weighted_lm_loss": 1.5774556398391724,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.639704049670904,
      "tokens_seen": 1117388800
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006681644934804414,
      "loss": 1.4698,
      "theoretical_loss": 3.639704049670904,
      "tokens_seen": 1117388800
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006681444332998997,
      "loss": 1.5725,
      "theoretical_loss": 3.6396843052299577,
      "tokens_seen": 1117454336
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006681243731193582,
      "loss": 1.4831,
      "theoretical_loss": 3.6396645622711494,
      "tokens_seen": 1117519872
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006681043129388164,
      "loss": 1.4928,
      "theoretical_loss": 3.639644820794281,
      "tokens_seen": 1117585408
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006680842527582749,
      "loss": 1.5463,
      "theoretical_loss": 3.6396250807991546,
      "tokens_seen": 1117650944
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006680641925777332,
      "loss": 1.6097,
      "theoretical_loss": 3.639605342285572,
      "tokens_seen": 1117716480
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006680441323971916,
      "loss": 1.6019,
      "theoretical_loss": 3.639585605253335,
      "tokens_seen": 1117782016
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00066802407221665,
      "loss": 1.4866,
      "theoretical_loss": 3.6395658697022455,
      "tokens_seen": 1117847552
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006680040120361084,
      "loss": 1.5032,
      "theoretical_loss": 3.6395461356321053,
      "tokens_seen": 1117913088
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006679839518555667,
      "loss": 1.5896,
      "theoretical_loss": 3.6395264030427166,
      "tokens_seen": 1117978624
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006679638916750251,
      "loss": 1.4753,
      "theoretical_loss": 3.639506671933882,
      "tokens_seen": 1118044160
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006679438314944835,
      "loss": 1.5221,
      "theoretical_loss": 3.6394869423054037,
      "tokens_seen": 1118109696
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006679237713139419,
      "loss": 1.5349,
      "theoretical_loss": 3.639467214157083,
      "tokens_seen": 1118175232
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006679037111334002,
      "loss": 1.4838,
      "theoretical_loss": 3.639447487488723,
      "tokens_seen": 1118240768
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006678836509528586,
      "loss": 1.5044,
      "theoretical_loss": 3.639427762300125,
      "tokens_seen": 1118306304
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006678635907723169,
      "loss": 1.5651,
      "theoretical_loss": 3.6394080385910925,
      "tokens_seen": 1118371840
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006678435305917754,
      "loss": 1.5018,
      "theoretical_loss": 3.6393883163614267,
      "tokens_seen": 1118437376
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006678234704112337,
      "loss": 1.6203,
      "theoretical_loss": 3.6393685956109305,
      "tokens_seen": 1118502912
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006678034102306921,
      "loss": 1.481,
      "theoretical_loss": 3.6393488763394064,
      "tokens_seen": 1118568448
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006677833500501505,
      "loss": 1.5037,
      "theoretical_loss": 3.6393291585466567,
      "tokens_seen": 1118633984
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006677632898696089,
      "loss": 1.4657,
      "theoretical_loss": 3.6393094422324834,
      "tokens_seen": 1118699520
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006677432296890672,
      "loss": 1.5845,
      "theoretical_loss": 3.63928972739669,
      "tokens_seen": 1118765056
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006677231695085257,
      "loss": 1.5,
      "theoretical_loss": 3.639270014039078,
      "tokens_seen": 1118830592
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006677031093279839,
      "loss": 1.6351,
      "theoretical_loss": 3.6392503021594513,
      "tokens_seen": 1118896128
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006676830491474424,
      "loss": 1.4443,
      "theoretical_loss": 3.6392305917576113,
      "tokens_seen": 1118961664
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006676629889669007,
      "loss": 1.5275,
      "theoretical_loss": 3.6392108828333614,
      "tokens_seen": 1119027200
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006676429287863591,
      "loss": 1.4588,
      "theoretical_loss": 3.6391911753865034,
      "tokens_seen": 1119092736
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006676228686058175,
      "loss": 1.549,
      "theoretical_loss": 3.639171469416841,
      "tokens_seen": 1119158272
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006676028084252758,
      "loss": 1.5681,
      "theoretical_loss": 3.639151764924177,
      "tokens_seen": 1119223808
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006675827482447342,
      "loss": 1.4645,
      "theoretical_loss": 3.6391320619083136,
      "tokens_seen": 1119289344
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006675626880641927,
      "loss": 1.6047,
      "theoretical_loss": 3.639112360369054,
      "tokens_seen": 1119354880
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006675426278836509,
      "loss": 1.6076,
      "theoretical_loss": 3.6390926603062006,
      "tokens_seen": 1119420416
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006675225677031094,
      "loss": 1.6338,
      "theoretical_loss": 3.639072961719558,
      "tokens_seen": 1119485952
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006675025075225676,
      "loss": 1.4835,
      "theoretical_loss": 3.6390532646089264,
      "tokens_seen": 1119551488
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006674824473420261,
      "loss": 1.5437,
      "theoretical_loss": 3.6390335689741113,
      "tokens_seen": 1119617024
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006674623871614845,
      "loss": 1.5972,
      "theoretical_loss": 3.6390138748149146,
      "tokens_seen": 1119682560
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006674423269809428,
      "loss": 1.483,
      "theoretical_loss": 3.63899418213114,
      "tokens_seen": 1119748096
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006674222668004012,
      "loss": 1.5184,
      "theoretical_loss": 3.6389744909225894,
      "tokens_seen": 1119813632
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006674022066198597,
      "loss": 1.564,
      "theoretical_loss": 3.638954801189067,
      "tokens_seen": 1119879168
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006673821464393179,
      "loss": 1.5438,
      "theoretical_loss": 3.638935112930376,
      "tokens_seen": 1119944704
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006673620862587764,
      "loss": 1.6092,
      "theoretical_loss": 3.638915426146319,
      "tokens_seen": 1120010240
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006673420260782346,
      "loss": 1.4609,
      "theoretical_loss": 3.6388957408367,
      "tokens_seen": 1120075776
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006673219658976931,
      "loss": 1.6196,
      "theoretical_loss": 3.6388760570013226,
      "tokens_seen": 1120141312
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006673019057171515,
      "loss": 1.5341,
      "theoretical_loss": 3.638856374639989,
      "tokens_seen": 1120206848
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006672818455366098,
      "loss": 1.4712,
      "theoretical_loss": 3.638836693752503,
      "tokens_seen": 1120272384
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006672617853560682,
      "loss": 1.4441,
      "theoretical_loss": 3.6388170143386676,
      "tokens_seen": 1120337920
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006672417251755266,
      "loss": 1.5418,
      "theoretical_loss": 3.6387973363982877,
      "tokens_seen": 1120403456
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006672216649949849,
      "loss": 1.5421,
      "theoretical_loss": 3.638777659931166,
      "tokens_seen": 1120468992
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006672016048144434,
      "loss": 1.3859,
      "theoretical_loss": 3.6387579849371052,
      "tokens_seen": 1120534528
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006671815446339017,
      "loss": 1.5479,
      "theoretical_loss": 3.63873831141591,
      "tokens_seen": 1120600064
    },
    {
      "epoch": 0.34,
      "objective/train/advantage_avg": 0.49377286434173584,
      "objective/train/docs_used": 410712,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4162588119506836,
      "objective/train/original_loss": 1.416258692741394,
      "objective/train/theoretical_loss": 3.638718639367384,
      "objective/train/tokens_used": 1141125600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24424436688423157,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506190061569214,
      "objective/train/weighted_lm_loss": 1.4879770278930664,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.638718639367384,
      "tokens_seen": 1120665600
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006671614844533601,
      "loss": 1.4985,
      "theoretical_loss": 3.638718639367384,
      "tokens_seen": 1120665600
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006671414242728185,
      "loss": 1.6083,
      "theoretical_loss": 3.63869896879133,
      "tokens_seen": 1120731136
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006671213640922768,
      "loss": 1.5754,
      "theoretical_loss": 3.6386792996875523,
      "tokens_seen": 1120796672
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006671013039117352,
      "loss": 1.4914,
      "theoretical_loss": 3.638659632055855,
      "tokens_seen": 1120862208
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006671013039117352,
      "loss": 1.4383,
      "theoretical_loss": 3.638639965896041,
      "tokens_seen": 1120927744
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006670812437311936,
      "loss": 1.5948,
      "theoretical_loss": 3.638620301207914,
      "tokens_seen": 1120993280
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006670611835506519,
      "loss": 1.5867,
      "theoretical_loss": 3.6386006379912788,
      "tokens_seen": 1121058816
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006670411233701104,
      "loss": 1.5531,
      "theoretical_loss": 3.638580976245939,
      "tokens_seen": 1121124352
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006670411233701104,
      "loss": 1.6071,
      "theoretical_loss": 3.638561315971698,
      "tokens_seen": 1121189888
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006670210631895687,
      "loss": 1.4117,
      "theoretical_loss": 3.63854165716836,
      "tokens_seen": 1121255424
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006670010030090271,
      "loss": 1.6251,
      "theoretical_loss": 3.6385219998357297,
      "tokens_seen": 1121320960
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006669809428284854,
      "loss": 1.5006,
      "theoretical_loss": 3.6385023439736095,
      "tokens_seen": 1121386496
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006669608826479439,
      "loss": 1.5169,
      "theoretical_loss": 3.638482689581805,
      "tokens_seen": 1121452032
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006669408224674022,
      "loss": 1.5568,
      "theoretical_loss": 3.63846303666012,
      "tokens_seen": 1121517568
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006669207622868606,
      "loss": 1.4929,
      "theoretical_loss": 3.638443385208358,
      "tokens_seen": 1121583104
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006669007021063189,
      "loss": 1.4969,
      "theoretical_loss": 3.638423735226324,
      "tokens_seen": 1121648640
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006668806419257773,
      "loss": 1.3986,
      "theoretical_loss": 3.6384040867138214,
      "tokens_seen": 1121714176
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006668605817452357,
      "loss": 1.7276,
      "theoretical_loss": 3.638384439670655,
      "tokens_seen": 1121779712
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006668405215646941,
      "loss": 1.47,
      "theoretical_loss": 3.6383647940966286,
      "tokens_seen": 1121845248
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006668204613841524,
      "loss": 1.5032,
      "theoretical_loss": 3.638345149991547,
      "tokens_seen": 1121910784
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006668004012036109,
      "loss": 1.5864,
      "theoretical_loss": 3.6383255073552148,
      "tokens_seen": 1121976320
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006667803410230692,
      "loss": 1.3964,
      "theoretical_loss": 3.6383058661874355,
      "tokens_seen": 1122041856
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006667602808425276,
      "loss": 1.5943,
      "theoretical_loss": 3.638286226488014,
      "tokens_seen": 1122107392
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000666740220661986,
      "loss": 1.6291,
      "theoretical_loss": 3.6382665882567555,
      "tokens_seen": 1122172928
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006667201604814443,
      "loss": 1.5961,
      "theoretical_loss": 3.638246951493463,
      "tokens_seen": 1122238464
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006667001003009027,
      "loss": 1.5283,
      "theoretical_loss": 3.638227316197943,
      "tokens_seen": 1122304000
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006666800401203611,
      "loss": 1.6116,
      "theoretical_loss": 3.6382076823699983,
      "tokens_seen": 1122369536
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006666599799398195,
      "loss": 1.4802,
      "theoretical_loss": 3.6381880500094343,
      "tokens_seen": 1122435072
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006666399197592779,
      "loss": 1.4905,
      "theoretical_loss": 3.6381684191160555,
      "tokens_seen": 1122500608
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006666198595787362,
      "loss": 1.5167,
      "theoretical_loss": 3.6381487896896667,
      "tokens_seen": 1122566144
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006665997993981946,
      "loss": 1.496,
      "theoretical_loss": 3.638129161730073,
      "tokens_seen": 1122631680
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006665797392176531,
      "loss": 1.4268,
      "theoretical_loss": 3.6381095352370787,
      "tokens_seen": 1122697216
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006665596790371113,
      "loss": 1.533,
      "theoretical_loss": 3.638089910210488,
      "tokens_seen": 1122762752
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006665396188565698,
      "loss": 1.5558,
      "theoretical_loss": 3.6380702866501076,
      "tokens_seen": 1122828288
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000666519558676028,
      "loss": 1.543,
      "theoretical_loss": 3.638050664555741,
      "tokens_seen": 1122893824
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006664994984954865,
      "loss": 1.604,
      "theoretical_loss": 3.6380310439271932,
      "tokens_seen": 1122959360
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006664794383149449,
      "loss": 1.5582,
      "theoretical_loss": 3.638011424764269,
      "tokens_seen": 1123024896
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006664593781344032,
      "loss": 1.4186,
      "theoretical_loss": 3.637991807066774,
      "tokens_seen": 1123090432
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006664393179538616,
      "loss": 1.4087,
      "theoretical_loss": 3.637972190834513,
      "tokens_seen": 1123155968
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006664192577733201,
      "loss": 1.5302,
      "theoretical_loss": 3.637952576067291,
      "tokens_seen": 1123221504
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006663991975927783,
      "loss": 1.4838,
      "theoretical_loss": 3.6379329627649137,
      "tokens_seen": 1123287040
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006663791374122368,
      "loss": 1.4943,
      "theoretical_loss": 3.6379133509271853,
      "tokens_seen": 1123352576
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000666359077231695,
      "loss": 1.5055,
      "theoretical_loss": 3.6378937405539116,
      "tokens_seen": 1123418112
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006663390170511535,
      "loss": 1.4844,
      "theoretical_loss": 3.6378741316448977,
      "tokens_seen": 1123483648
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006663189568706119,
      "loss": 1.5155,
      "theoretical_loss": 3.6378545241999487,
      "tokens_seen": 1123549184
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006662988966900702,
      "loss": 1.5629,
      "theoretical_loss": 3.6378349182188696,
      "tokens_seen": 1123614720
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006662788365095286,
      "loss": 1.483,
      "theoretical_loss": 3.637815313701467,
      "tokens_seen": 1123680256
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000666258776328987,
      "loss": 1.6015,
      "theoretical_loss": 3.637795710647545,
      "tokens_seen": 1123745792
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006662387161484453,
      "loss": 1.4915,
      "theoretical_loss": 3.637776109056909,
      "tokens_seen": 1123811328
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006662186559679038,
      "loss": 1.6193,
      "theoretical_loss": 3.6377565089293658,
      "tokens_seen": 1123876864
    },
    {
      "epoch": 0.34,
      "objective/train/advantage_avg": 0.4895123243331909,
      "objective/train/docs_used": 411904,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2212467193603516,
      "objective/train/original_loss": 1.2212467193603516,
      "objective/train/theoretical_loss": 3.637736910264719,
      "objective/train/tokens_used": 1144402400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24087874591350555,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501755475997925,
      "objective/train/weighted_lm_loss": 1.282314419746399,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.7788007855415344,
      "theoretical_loss": 3.637736910264719,
      "tokens_seen": 1123942400
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000666198595787362,
      "loss": 1.3801,
      "theoretical_loss": 3.637736910264719,
      "tokens_seen": 1123942400
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006661785356068205,
      "loss": 1.5557,
      "theoretical_loss": 3.637717313062776,
      "tokens_seen": 1124007936
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006661584754262789,
      "loss": 1.5091,
      "theoretical_loss": 3.6376977173233405,
      "tokens_seen": 1124073472
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006661384152457372,
      "loss": 1.4532,
      "theoretical_loss": 3.6376781230462196,
      "tokens_seen": 1124139008
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006661183550651956,
      "loss": 1.5229,
      "theoretical_loss": 3.6376585302312185,
      "tokens_seen": 1124204544
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000666098294884654,
      "loss": 1.5403,
      "theoretical_loss": 3.6376389388781427,
      "tokens_seen": 1124270080
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006660782347041123,
      "loss": 1.4593,
      "theoretical_loss": 3.6376193489867976,
      "tokens_seen": 1124335616
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006660581745235708,
      "loss": 1.5296,
      "theoretical_loss": 3.6375997605569896,
      "tokens_seen": 1124401152
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006660381143430291,
      "loss": 1.62,
      "theoretical_loss": 3.6375801735885243,
      "tokens_seen": 1124466688
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006660180541624875,
      "loss": 1.5516,
      "theoretical_loss": 3.6375605880812074,
      "tokens_seen": 1124532224
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006659979939819458,
      "loss": 1.4873,
      "theoretical_loss": 3.6375410040348446,
      "tokens_seen": 1124597760
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006659779338014043,
      "loss": 1.4446,
      "theoretical_loss": 3.637521421449242,
      "tokens_seen": 1124663296
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006659578736208626,
      "loss": 1.4611,
      "theoretical_loss": 3.6375018403242056,
      "tokens_seen": 1124728832
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000665937813440321,
      "loss": 1.4969,
      "theoretical_loss": 3.6374822606595414,
      "tokens_seen": 1124794368
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006659177532597793,
      "loss": 1.4332,
      "theoretical_loss": 3.637462682455055,
      "tokens_seen": 1124859904
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006658976930792377,
      "loss": 1.4892,
      "theoretical_loss": 3.6374431057105525,
      "tokens_seen": 1124925440
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006658776328986961,
      "loss": 1.5167,
      "theoretical_loss": 3.6374235304258407,
      "tokens_seen": 1124990976
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006658575727181545,
      "loss": 1.5773,
      "theoretical_loss": 3.637403956600725,
      "tokens_seen": 1125056512
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006658375125376128,
      "loss": 1.4672,
      "theoretical_loss": 3.6373843842350118,
      "tokens_seen": 1125122048
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006658174523570713,
      "loss": 1.6615,
      "theoretical_loss": 3.637364813328507,
      "tokens_seen": 1125187584
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006657973921765296,
      "loss": 1.4925,
      "theoretical_loss": 3.6373452438810174,
      "tokens_seen": 1125253120
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000665777331995988,
      "loss": 1.5937,
      "theoretical_loss": 3.637325675892349,
      "tokens_seen": 1125318656
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006657572718154463,
      "loss": 1.4333,
      "theoretical_loss": 3.637306109362308,
      "tokens_seen": 1125384192
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006657372116349047,
      "loss": 1.5045,
      "theoretical_loss": 3.6372865442907,
      "tokens_seen": 1125449728
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006657171514543631,
      "loss": 1.5736,
      "theoretical_loss": 3.6372669806773334,
      "tokens_seen": 1125515264
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006656970912738215,
      "loss": 1.4393,
      "theoretical_loss": 3.637247418522013,
      "tokens_seen": 1125580800
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006656770310932798,
      "loss": 1.5155,
      "theoretical_loss": 3.6372278578245454,
      "tokens_seen": 1125646336
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006656569709127383,
      "loss": 1.5218,
      "theoretical_loss": 3.637208298584737,
      "tokens_seen": 1125711872
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006656369107321965,
      "loss": 1.5647,
      "theoretical_loss": 3.6371887408023955,
      "tokens_seen": 1125777408
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000665616850551655,
      "loss": 1.4288,
      "theoretical_loss": 3.637169184477326,
      "tokens_seen": 1125842944
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006655967903711134,
      "loss": 1.4995,
      "theoretical_loss": 3.6371496296093357,
      "tokens_seen": 1125908480
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006655767301905717,
      "loss": 1.5453,
      "theoretical_loss": 3.6371300761982317,
      "tokens_seen": 1125974016
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006655566700100301,
      "loss": 1.5403,
      "theoretical_loss": 3.6371105242438198,
      "tokens_seen": 1126039552
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006655366098294884,
      "loss": 1.6351,
      "theoretical_loss": 3.6370909737459067,
      "tokens_seen": 1126105088
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006655165496489468,
      "loss": 1.4208,
      "theoretical_loss": 3.6370714247043003,
      "tokens_seen": 1126170624
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006654964894684053,
      "loss": 1.4991,
      "theoretical_loss": 3.6370518771188065,
      "tokens_seen": 1126236160
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006654764292878635,
      "loss": 1.4554,
      "theoretical_loss": 3.637032330989232,
      "tokens_seen": 1126301696
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000665456369107322,
      "loss": 1.4619,
      "theoretical_loss": 3.6370127863153843,
      "tokens_seen": 1126367232
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006654363089267804,
      "loss": 1.62,
      "theoretical_loss": 3.6369932430970695,
      "tokens_seen": 1126432768
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006654162487462387,
      "loss": 1.5745,
      "theoretical_loss": 3.636973701334095,
      "tokens_seen": 1126498304
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006653961885656971,
      "loss": 1.6293,
      "theoretical_loss": 3.636954161026268,
      "tokens_seen": 1126563840
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006653761283851554,
      "loss": 1.5239,
      "theoretical_loss": 3.636934622173395,
      "tokens_seen": 1126629376
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006653560682046138,
      "loss": 1.5977,
      "theoretical_loss": 3.6369150847752834,
      "tokens_seen": 1126694912
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006653360080240723,
      "loss": 1.5292,
      "theoretical_loss": 3.63689554883174,
      "tokens_seen": 1126760448
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006653159478435305,
      "loss": 1.3729,
      "theoretical_loss": 3.6368760143425716,
      "tokens_seen": 1126825984
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000665295887662989,
      "loss": 1.6491,
      "theoretical_loss": 3.6368564813075865,
      "tokens_seen": 1126891520
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006652758274824473,
      "loss": 1.5743,
      "theoretical_loss": 3.6368369497265913,
      "tokens_seen": 1126957056
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006652557673019057,
      "loss": 1.5287,
      "theoretical_loss": 3.6368174195993923,
      "tokens_seen": 1127022592
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006652357071213641,
      "loss": 1.5089,
      "theoretical_loss": 3.636797890925798,
      "tokens_seen": 1127088128
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006652156469408225,
      "loss": 1.5243,
      "theoretical_loss": 3.6367783637056155,
      "tokens_seen": 1127153664
    },
    {
      "epoch": 0.34,
      "objective/train/advantage_avg": 0.48264655470848083,
      "objective/train/docs_used": 413086,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5066636800765991,
      "objective/train/original_loss": 1.5066636800765991,
      "objective/train/theoretical_loss": 3.6367588379386513,
      "objective/train/tokens_used": 1147679200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23538316786289215,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.049461007118225,
      "objective/train/weighted_lm_loss": 1.581547498703003,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6367588379386513,
      "tokens_seen": 1127219200
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006651955867602808,
      "loss": 1.4409,
      "theoretical_loss": 3.6367588379386513,
      "tokens_seen": 1127219200
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006651755265797393,
      "loss": 1.6045,
      "theoretical_loss": 3.636739313624714,
      "tokens_seen": 1127284736
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006651554663991976,
      "loss": 1.4842,
      "theoretical_loss": 3.63671979076361,
      "tokens_seen": 1127350272
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000665135406218656,
      "loss": 1.5789,
      "theoretical_loss": 3.6367002693551473,
      "tokens_seen": 1127415808
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006651153460381144,
      "loss": 1.5472,
      "theoretical_loss": 3.636680749399133,
      "tokens_seen": 1127481344
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006650952858575727,
      "loss": 1.4914,
      "theoretical_loss": 3.636661230895375,
      "tokens_seen": 1127546880
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006650752256770312,
      "loss": 1.5042,
      "theoretical_loss": 3.6366417138436806,
      "tokens_seen": 1127612416
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006650551654964895,
      "loss": 1.6118,
      "theoretical_loss": 3.636622198243858,
      "tokens_seen": 1127677952
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006650351053159479,
      "loss": 1.4878,
      "theoretical_loss": 3.6366026840957133,
      "tokens_seen": 1127743488
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006650150451354062,
      "loss": 1.5912,
      "theoretical_loss": 3.6365831713990557,
      "tokens_seen": 1127809024
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006649949849548647,
      "loss": 1.515,
      "theoretical_loss": 3.6365636601536924,
      "tokens_seen": 1127874560
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000664974924774323,
      "loss": 1.5722,
      "theoretical_loss": 3.6365441503594305,
      "tokens_seen": 1127940096
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006649548645937814,
      "loss": 1.5701,
      "theoretical_loss": 3.636524642016079,
      "tokens_seen": 1128005632
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006649348044132397,
      "loss": 1.4992,
      "theoretical_loss": 3.636505135123445,
      "tokens_seen": 1128071168
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006649147442326981,
      "loss": 1.5608,
      "theoretical_loss": 3.6364856296813364,
      "tokens_seen": 1128136704
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006648946840521565,
      "loss": 1.5508,
      "theoretical_loss": 3.636466125689561,
      "tokens_seen": 1128202240
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006648746238716149,
      "loss": 1.4929,
      "theoretical_loss": 3.636446623147927,
      "tokens_seen": 1128267776
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006648545636910732,
      "loss": 1.5721,
      "theoretical_loss": 3.636427122056242,
      "tokens_seen": 1128333312
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006648345035105317,
      "loss": 1.5222,
      "theoretical_loss": 3.636407622414314,
      "tokens_seen": 1128398848
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00066481444332999,
      "loss": 1.6117,
      "theoretical_loss": 3.636388124221952,
      "tokens_seen": 1128464384
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006647943831494484,
      "loss": 1.5412,
      "theoretical_loss": 3.6363686274789626,
      "tokens_seen": 1128529920
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006647743229689067,
      "loss": 1.5973,
      "theoretical_loss": 3.6363491321851544,
      "tokens_seen": 1128595456
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006647542627883651,
      "loss": 1.3585,
      "theoretical_loss": 3.6363296383403365,
      "tokens_seen": 1128660992
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006647342026078235,
      "loss": 1.5506,
      "theoretical_loss": 3.636310145944316,
      "tokens_seen": 1128726528
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006647141424272819,
      "loss": 1.4953,
      "theoretical_loss": 3.6362906549969014,
      "tokens_seen": 1128792064
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006646940822467402,
      "loss": 1.545,
      "theoretical_loss": 3.6362711654979005,
      "tokens_seen": 1128857600
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006646740220661987,
      "loss": 1.6261,
      "theoretical_loss": 3.636251677447123,
      "tokens_seen": 1128923136
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006646539618856569,
      "loss": 1.4922,
      "theoretical_loss": 3.6362321908443755,
      "tokens_seen": 1128988672
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006646339017051154,
      "loss": 1.4033,
      "theoretical_loss": 3.6362127056894673,
      "tokens_seen": 1129054208
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006646138415245738,
      "loss": 1.4341,
      "theoretical_loss": 3.6361932219822073,
      "tokens_seen": 1129119744
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006645937813440321,
      "loss": 1.5042,
      "theoretical_loss": 3.6361737397224023,
      "tokens_seen": 1129185280
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006645737211634905,
      "loss": 1.4776,
      "theoretical_loss": 3.6361542589098623,
      "tokens_seen": 1129250816
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006645536609829488,
      "loss": 1.7504,
      "theoretical_loss": 3.6361347795443955,
      "tokens_seen": 1129316352
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006645336008024072,
      "loss": 1.5657,
      "theoretical_loss": 3.6361153016258094,
      "tokens_seen": 1129381888
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006645135406218657,
      "loss": 1.448,
      "theoretical_loss": 3.6360958251539137,
      "tokens_seen": 1129447424
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006644934804413239,
      "loss": 1.5111,
      "theoretical_loss": 3.6360763501285165,
      "tokens_seen": 1129512960
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006644734202607824,
      "loss": 1.5176,
      "theoretical_loss": 3.636056876549427,
      "tokens_seen": 1129578496
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006644533600802408,
      "loss": 1.5054,
      "theoretical_loss": 3.6360374044164536,
      "tokens_seen": 1129644032
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006644332998996991,
      "loss": 1.5127,
      "theoretical_loss": 3.6360179337294047,
      "tokens_seen": 1129709568
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006644132397191575,
      "loss": 1.5905,
      "theoretical_loss": 3.6359984644880896,
      "tokens_seen": 1129775104
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006643931795386158,
      "loss": 1.4408,
      "theoretical_loss": 3.6359789966923164,
      "tokens_seen": 1129840640
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006643731193580742,
      "loss": 1.5381,
      "theoretical_loss": 3.6359595303418946,
      "tokens_seen": 1129906176
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006643530591775327,
      "loss": 1.6096,
      "theoretical_loss": 3.6359400654366327,
      "tokens_seen": 1129971712
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006643329989969909,
      "loss": 1.5907,
      "theoretical_loss": 3.6359206019763395,
      "tokens_seen": 1130037248
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006643129388164494,
      "loss": 1.6017,
      "theoretical_loss": 3.6359011399608243,
      "tokens_seen": 1130102784
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006642928786359077,
      "loss": 1.4441,
      "theoretical_loss": 3.635881679389896,
      "tokens_seen": 1130168320
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006642728184553661,
      "loss": 1.4024,
      "theoretical_loss": 3.6358622202633635,
      "tokens_seen": 1130233856
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006642527582748245,
      "loss": 1.6564,
      "theoretical_loss": 3.635842762581036,
      "tokens_seen": 1130299392
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006642326980942829,
      "loss": 1.6374,
      "theoretical_loss": 3.6358233063427225,
      "tokens_seen": 1130364928
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006642126379137412,
      "loss": 1.4949,
      "theoretical_loss": 3.635803851548232,
      "tokens_seen": 1130430464
    },
    {
      "epoch": 0.34,
      "objective/train/advantage_avg": 0.49409815669059753,
      "objective/train/docs_used": 414307,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5959542989730835,
      "objective/train/original_loss": 1.595954179763794,
      "objective/train/theoretical_loss": 3.635784398197374,
      "objective/train/tokens_used": 1150956000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24457332491874695,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506532192230225,
      "objective/train/weighted_lm_loss": 1.6768319606781006,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.635784398197374,
      "tokens_seen": 1130496000
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006641925777331997,
      "loss": 1.5663,
      "theoretical_loss": 3.635784398197374,
      "tokens_seen": 1130496000
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006641725175526579,
      "loss": 1.4019,
      "theoretical_loss": 3.635764946289957,
      "tokens_seen": 1130561536
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006641524573721164,
      "loss": 1.4147,
      "theoretical_loss": 3.635745495825791,
      "tokens_seen": 1130627072
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006641323971915747,
      "loss": 1.4487,
      "theoretical_loss": 3.635726046804685,
      "tokens_seen": 1130692608
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006641123370110331,
      "loss": 1.4409,
      "theoretical_loss": 3.635706599226448,
      "tokens_seen": 1130758144
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006640922768304915,
      "loss": 1.4457,
      "theoretical_loss": 3.63568715309089,
      "tokens_seen": 1130823680
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006640722166499499,
      "loss": 1.5871,
      "theoretical_loss": 3.63566770839782,
      "tokens_seen": 1130889216
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006640521564694082,
      "loss": 1.5299,
      "theoretical_loss": 3.635648265147047,
      "tokens_seen": 1130954752
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006640320962888666,
      "loss": 1.6043,
      "theoretical_loss": 3.635628823338381,
      "tokens_seen": 1131020288
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000664012036108325,
      "loss": 1.5576,
      "theoretical_loss": 3.635609382971632,
      "tokens_seen": 1131085824
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006639919759277834,
      "loss": 1.3979,
      "theoretical_loss": 3.6355899440466075,
      "tokens_seen": 1131151360
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006639719157472417,
      "loss": 1.4355,
      "theoretical_loss": 3.63557050656312,
      "tokens_seen": 1131216896
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006639518555667001,
      "loss": 1.5444,
      "theoretical_loss": 3.635551070520976,
      "tokens_seen": 1131282432
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006639317953861584,
      "loss": 1.4927,
      "theoretical_loss": 3.635531635919988,
      "tokens_seen": 1131347968
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006639117352056169,
      "loss": 1.5198,
      "theoretical_loss": 3.635512202759964,
      "tokens_seen": 1131413504
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006638916750250752,
      "loss": 1.5611,
      "theoretical_loss": 3.6354927710407132,
      "tokens_seen": 1131479040
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006638716148445336,
      "loss": 1.5329,
      "theoretical_loss": 3.6354733407620468,
      "tokens_seen": 1131544576
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000663851554663992,
      "loss": 1.5282,
      "theoretical_loss": 3.635453911923774,
      "tokens_seen": 1131610112
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006638314944834504,
      "loss": 1.422,
      "theoretical_loss": 3.635434484525704,
      "tokens_seen": 1131675648
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006638114343029087,
      "loss": 1.6011,
      "theoretical_loss": 3.6354150585676477,
      "tokens_seen": 1131741184
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006637913741223671,
      "loss": 1.543,
      "theoretical_loss": 3.6353956340494147,
      "tokens_seen": 1131806720
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006637713139418254,
      "loss": 1.5293,
      "theoretical_loss": 3.635376210970814,
      "tokens_seen": 1131872256
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006637512537612839,
      "loss": 1.5594,
      "theoretical_loss": 3.6353567893316567,
      "tokens_seen": 1131937792
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006637311935807422,
      "loss": 1.49,
      "theoretical_loss": 3.6353373691317525,
      "tokens_seen": 1132003328
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006637111334002006,
      "loss": 1.5612,
      "theoretical_loss": 3.6353179503709114,
      "tokens_seen": 1132068864
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000663691073219659,
      "loss": 1.4841,
      "theoretical_loss": 3.6352985330489425,
      "tokens_seen": 1132134400
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006636710130391173,
      "loss": 1.5291,
      "theoretical_loss": 3.6352791171656573,
      "tokens_seen": 1132199936
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006636509528585758,
      "loss": 1.5272,
      "theoretical_loss": 3.6352597027208655,
      "tokens_seen": 1132265472
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006636308926780342,
      "loss": 1.5654,
      "theoretical_loss": 3.6352402897143774,
      "tokens_seen": 1132331008
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006636108324974925,
      "loss": 1.4974,
      "theoretical_loss": 3.6352208781460025,
      "tokens_seen": 1132396544
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006635907723169509,
      "loss": 1.4923,
      "theoretical_loss": 3.635201468015551,
      "tokens_seen": 1132462080
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006635707121364092,
      "loss": 1.4821,
      "theoretical_loss": 3.6351820593228346,
      "tokens_seen": 1132527616
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006635506519558676,
      "loss": 1.6252,
      "theoretical_loss": 3.6351626520676623,
      "tokens_seen": 1132593152
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006635305917753261,
      "loss": 1.4719,
      "theoretical_loss": 3.635143246249845,
      "tokens_seen": 1132658688
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006635105315947843,
      "loss": 1.4817,
      "theoretical_loss": 3.635123841869193,
      "tokens_seen": 1132724224
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006634904714142428,
      "loss": 1.4959,
      "theoretical_loss": 3.6351044389255165,
      "tokens_seen": 1132789760
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006634704112337012,
      "loss": 1.5964,
      "theoretical_loss": 3.6350850374186265,
      "tokens_seen": 1132855296
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006634503510531595,
      "loss": 1.6481,
      "theoretical_loss": 3.6350656373483328,
      "tokens_seen": 1132920832
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006634302908726179,
      "loss": 1.4177,
      "theoretical_loss": 3.6350462387144464,
      "tokens_seen": 1132986368
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006634102306920762,
      "loss": 1.6254,
      "theoretical_loss": 3.6350268415167775,
      "tokens_seen": 1133051904
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006633901705115346,
      "loss": 1.5289,
      "theoretical_loss": 3.635007445755137,
      "tokens_seen": 1133117440
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006633701103309931,
      "loss": 1.4952,
      "theoretical_loss": 3.6349880514293353,
      "tokens_seen": 1133182976
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006633500501504513,
      "loss": 1.4765,
      "theoretical_loss": 3.634968658539184,
      "tokens_seen": 1133248512
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006633299899699098,
      "loss": 1.6614,
      "theoretical_loss": 3.6349492670844925,
      "tokens_seen": 1133314048
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006633099297893681,
      "loss": 1.5941,
      "theoretical_loss": 3.634929877065072,
      "tokens_seen": 1133379584
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006632898696088265,
      "loss": 1.5764,
      "theoretical_loss": 3.6349104884807337,
      "tokens_seen": 1133445120
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006632698094282849,
      "loss": 1.4683,
      "theoretical_loss": 3.6348911013312883,
      "tokens_seen": 1133510656
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006632497492477433,
      "loss": 1.4854,
      "theoretical_loss": 3.634871715616546,
      "tokens_seen": 1133576192
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006632296890672016,
      "loss": 1.4202,
      "theoretical_loss": 3.634852331336319,
      "tokens_seen": 1133641728
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006632096288866601,
      "loss": 1.5881,
      "theoretical_loss": 3.6348329484904163,
      "tokens_seen": 1133707264
    },
    {
      "epoch": 0.34,
      "objective/train/advantage_avg": 0.4936368763446808,
      "objective/train/docs_used": 415442,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.442875862121582,
      "objective/train/original_loss": 1.442875623703003,
      "objective/train/theoretical_loss": 3.6348135670786506,
      "objective/train/tokens_used": 1154232800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24388210475444794,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506036281585693,
      "objective/train/weighted_lm_loss": 1.5157884359359741,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.6348135670786506,
      "tokens_seen": 1133772800
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006631895687061183,
      "loss": 1.5384,
      "theoretical_loss": 3.6348135670786506,
      "tokens_seen": 1133772800
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006631695085255768,
      "loss": 1.5697,
      "theoretical_loss": 3.634794187100832,
      "tokens_seen": 1133838336
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006631494483450351,
      "loss": 1.4626,
      "theoretical_loss": 3.634774808556772,
      "tokens_seen": 1133903872
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006631293881644935,
      "loss": 1.585,
      "theoretical_loss": 3.6347554314462815,
      "tokens_seen": 1133969408
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006631093279839519,
      "loss": 1.4939,
      "theoretical_loss": 3.6347360557691712,
      "tokens_seen": 1134034944
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006630892678034103,
      "loss": 1.532,
      "theoretical_loss": 3.634716681525253,
      "tokens_seen": 1134100480
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006630692076228686,
      "loss": 1.5383,
      "theoretical_loss": 3.6346973087143377,
      "tokens_seen": 1134166016
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000663049147442327,
      "loss": 1.4681,
      "theoretical_loss": 3.634677937336236,
      "tokens_seen": 1134231552
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006630290872617853,
      "loss": 1.5083,
      "theoretical_loss": 3.63465856739076,
      "tokens_seen": 1134297088
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006630090270812438,
      "loss": 1.5806,
      "theoretical_loss": 3.634639198877721,
      "tokens_seen": 1134362624
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006629889669007021,
      "loss": 1.475,
      "theoretical_loss": 3.63461983179693,
      "tokens_seen": 1134428160
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006629689067201605,
      "loss": 1.5333,
      "theoretical_loss": 3.6346004661481977,
      "tokens_seen": 1134493696
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006629488465396188,
      "loss": 1.5086,
      "theoretical_loss": 3.634581101931336,
      "tokens_seen": 1134559232
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006629287863590773,
      "loss": 1.5998,
      "theoretical_loss": 3.6345617391461573,
      "tokens_seen": 1134624768
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006629087261785356,
      "loss": 1.5175,
      "theoretical_loss": 3.6345423777924717,
      "tokens_seen": 1134690304
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000662888665997994,
      "loss": 1.5022,
      "theoretical_loss": 3.634523017870091,
      "tokens_seen": 1134755840
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006628686058174524,
      "loss": 1.5548,
      "theoretical_loss": 3.6345036593788276,
      "tokens_seen": 1134821376
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006628485456369108,
      "loss": 1.617,
      "theoretical_loss": 3.6344843023184916,
      "tokens_seen": 1134886912
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006628284854563691,
      "loss": 1.5035,
      "theoretical_loss": 3.6344649466888956,
      "tokens_seen": 1134952448
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006628084252758275,
      "loss": 1.6062,
      "theoretical_loss": 3.6344455924898518,
      "tokens_seen": 1135017984
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006627883650952858,
      "loss": 1.4773,
      "theoretical_loss": 3.6344262397211704,
      "tokens_seen": 1135083520
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006627683049147443,
      "loss": 1.6155,
      "theoretical_loss": 3.6344068883826646,
      "tokens_seen": 1135149056
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006627482447342026,
      "loss": 1.5463,
      "theoretical_loss": 3.6343875384741446,
      "tokens_seen": 1135214592
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000662728184553661,
      "loss": 1.5409,
      "theoretical_loss": 3.634368189995423,
      "tokens_seen": 1135280128
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006627081243731194,
      "loss": 1.4537,
      "theoretical_loss": 3.6343488429463124,
      "tokens_seen": 1135345664
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006626880641925777,
      "loss": 1.6322,
      "theoretical_loss": 3.6343294973266236,
      "tokens_seen": 1135411200
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006626680040120361,
      "loss": 1.4818,
      "theoretical_loss": 3.634310153136169,
      "tokens_seen": 1135476736
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006626479438314946,
      "loss": 1.5325,
      "theoretical_loss": 3.6342908103747593,
      "tokens_seen": 1135542272
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006626278836509528,
      "loss": 1.5946,
      "theoretical_loss": 3.634271469042208,
      "tokens_seen": 1135607808
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006626078234704113,
      "loss": 1.5763,
      "theoretical_loss": 3.6342521291383267,
      "tokens_seen": 1135673344
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006625877632898695,
      "loss": 1.4376,
      "theoretical_loss": 3.634232790662927,
      "tokens_seen": 1135738880
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000662567703109328,
      "loss": 1.4785,
      "theoretical_loss": 3.6342134536158217,
      "tokens_seen": 1135804416
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006625476429287864,
      "loss": 1.4799,
      "theoretical_loss": 3.634194117996822,
      "tokens_seen": 1135869952
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006625275827482447,
      "loss": 1.4997,
      "theoretical_loss": 3.634174783805741,
      "tokens_seen": 1135935488
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006625075225677031,
      "loss": 1.4997,
      "theoretical_loss": 3.6341554510423904,
      "tokens_seen": 1136001024
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006624874623871616,
      "loss": 1.4749,
      "theoretical_loss": 3.6341361197065822,
      "tokens_seen": 1136066560
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006624674022066198,
      "loss": 1.566,
      "theoretical_loss": 3.634116789798129,
      "tokens_seen": 1136132096
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006624473420260783,
      "loss": 1.5859,
      "theoretical_loss": 3.634097461316842,
      "tokens_seen": 1136197632
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006624272818455365,
      "loss": 1.5708,
      "theoretical_loss": 3.6340781342625355,
      "tokens_seen": 1136263168
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000662407221664995,
      "loss": 1.6836,
      "theoretical_loss": 3.6340588086350207,
      "tokens_seen": 1136328704
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006623871614844534,
      "loss": 1.4645,
      "theoretical_loss": 3.6340394844341097,
      "tokens_seen": 1136394240
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006623671013039117,
      "loss": 1.4954,
      "theoretical_loss": 3.634020161659616,
      "tokens_seen": 1136459776
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006623470411233701,
      "loss": 1.6437,
      "theoretical_loss": 3.6340008403113506,
      "tokens_seen": 1136525312
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006623269809428285,
      "loss": 1.5553,
      "theoretical_loss": 3.633981520389127,
      "tokens_seen": 1136590848
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006623069207622868,
      "loss": 1.6139,
      "theoretical_loss": 3.6339622018927575,
      "tokens_seen": 1136656384
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006622868605817453,
      "loss": 1.4774,
      "theoretical_loss": 3.633942884822055,
      "tokens_seen": 1136721920
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006622668004012035,
      "loss": 1.5638,
      "theoretical_loss": 3.6339235691768312,
      "tokens_seen": 1136787456
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000662246740220662,
      "loss": 1.4781,
      "theoretical_loss": 3.6339042549569,
      "tokens_seen": 1136852992
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006622266800401203,
      "loss": 1.5447,
      "theoretical_loss": 3.633884942162073,
      "tokens_seen": 1136918528
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006622066198595787,
      "loss": 1.4265,
      "theoretical_loss": 3.6338656307921635,
      "tokens_seen": 1136984064
    },
    {
      "epoch": 0.34,
      "objective/train/advantage_avg": 0.4943179190158844,
      "objective/train/docs_used": 416597,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.522156834602356,
      "objective/train/original_loss": 1.5221569538116455,
      "objective/train/theoretical_loss": 3.633846320846984,
      "objective/train/tokens_used": 1157509600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24462522566318512,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050675392150879,
      "objective/train/weighted_lm_loss": 1.599177360534668,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.633846320846984,
      "tokens_seen": 1137049600
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006621865596790371,
      "loss": 1.5439,
      "theoretical_loss": 3.633846320846984,
      "tokens_seen": 1137049600
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006621664994984955,
      "loss": 1.5055,
      "theoretical_loss": 3.633827012326347,
      "tokens_seen": 1137115136
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006621464393179539,
      "loss": 1.5373,
      "theoretical_loss": 3.6338077052300664,
      "tokens_seen": 1137180672
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006621263791374123,
      "loss": 1.5716,
      "theoretical_loss": 3.633788399557954,
      "tokens_seen": 1137246208
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006621063189568707,
      "loss": 1.4915,
      "theoretical_loss": 3.6337690953098223,
      "tokens_seen": 1137311744
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000662086258776329,
      "loss": 1.507,
      "theoretical_loss": 3.6337497924854856,
      "tokens_seen": 1137377280
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006620661985957874,
      "loss": 1.3709,
      "theoretical_loss": 3.633730491084756,
      "tokens_seen": 1137442816
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006620461384152457,
      "loss": 1.4709,
      "theoretical_loss": 3.6337111911074462,
      "tokens_seen": 1137508352
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006620260782347042,
      "loss": 1.5312,
      "theoretical_loss": 3.6336918925533706,
      "tokens_seen": 1137573888
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006620060180541625,
      "loss": 1.5646,
      "theoretical_loss": 3.633672595422341,
      "tokens_seen": 1137639424
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006619859578736209,
      "loss": 1.5489,
      "theoretical_loss": 3.6336532997141706,
      "tokens_seen": 1137704960
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006619658976930792,
      "loss": 1.5078,
      "theoretical_loss": 3.633634005428673,
      "tokens_seen": 1137770496
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006619458375125377,
      "loss": 1.5488,
      "theoretical_loss": 3.633614712565661,
      "tokens_seen": 1137836032
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000661925777331996,
      "loss": 1.6215,
      "theoretical_loss": 3.633595421124948,
      "tokens_seen": 1137901568
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006619057171514544,
      "loss": 1.7291,
      "theoretical_loss": 3.6335761311063473,
      "tokens_seen": 1137967104
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006618856569709128,
      "loss": 1.5874,
      "theoretical_loss": 3.633556842509672,
      "tokens_seen": 1138032640
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006618655967903712,
      "loss": 1.5728,
      "theoretical_loss": 3.633537555334735,
      "tokens_seen": 1138098176
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006618455366098295,
      "loss": 1.524,
      "theoretical_loss": 3.6335182695813506,
      "tokens_seen": 1138163712
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000661825476429288,
      "loss": 1.5751,
      "theoretical_loss": 3.633498985249332,
      "tokens_seen": 1138229248
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006618054162487462,
      "loss": 1.3827,
      "theoretical_loss": 3.6334797023384917,
      "tokens_seen": 1138294784
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006617853560682047,
      "loss": 1.5908,
      "theoretical_loss": 3.6334604208486443,
      "tokens_seen": 1138360320
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.000661765295887663,
      "loss": 1.558,
      "theoretical_loss": 3.6334411407796026,
      "tokens_seen": 1138425856
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0006617452357071214,
      "loss": 1.5151,
      "theoretical_loss": 3.63342186213118,
      "tokens_seen": 1138491392
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006617251755265798,
      "loss": 1.5057,
      "theoretical_loss": 3.63340258490319,
      "tokens_seen": 1138556928
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006617051153460381,
      "loss": 1.5667,
      "theoretical_loss": 3.6333833090954473,
      "tokens_seen": 1138622464
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006616850551654965,
      "loss": 1.4971,
      "theoretical_loss": 3.633364034707765,
      "tokens_seen": 1138688000
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000661664994984955,
      "loss": 1.4758,
      "theoretical_loss": 3.6333447617399557,
      "tokens_seen": 1138753536
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006616449348044132,
      "loss": 1.5025,
      "theoretical_loss": 3.633325490191834,
      "tokens_seen": 1138819072
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006616248746238717,
      "loss": 1.5655,
      "theoretical_loss": 3.633306220063214,
      "tokens_seen": 1138884608
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006616048144433299,
      "loss": 1.5165,
      "theoretical_loss": 3.6332869513539086,
      "tokens_seen": 1138950144
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006615847542627884,
      "loss": 1.5827,
      "theoretical_loss": 3.6332676840637324,
      "tokens_seen": 1139015680
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006615646940822468,
      "loss": 1.4748,
      "theoretical_loss": 3.6332484181924984,
      "tokens_seen": 1139081216
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006615446339017051,
      "loss": 1.4649,
      "theoretical_loss": 3.633229153740021,
      "tokens_seen": 1139146752
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006615245737211635,
      "loss": 1.6063,
      "theoretical_loss": 3.633209890706114,
      "tokens_seen": 1139212288
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000661504513540622,
      "loss": 1.6715,
      "theoretical_loss": 3.633190629090592,
      "tokens_seen": 1139277824
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006614844533600802,
      "loss": 1.5509,
      "theoretical_loss": 3.633171368893268,
      "tokens_seen": 1139343360
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006614643931795387,
      "loss": 1.5439,
      "theoretical_loss": 3.633152110113956,
      "tokens_seen": 1139408896
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006614443329989969,
      "loss": 1.4997,
      "theoretical_loss": 3.633132852752471,
      "tokens_seen": 1139474432
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006614242728184554,
      "loss": 1.5516,
      "theoretical_loss": 3.6331135968086263,
      "tokens_seen": 1139539968
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006614042126379138,
      "loss": 1.4607,
      "theoretical_loss": 3.633094342282236,
      "tokens_seen": 1139605504
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006613841524573721,
      "loss": 1.6195,
      "theoretical_loss": 3.633075089173115,
      "tokens_seen": 1139671040
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006613640922768305,
      "loss": 1.5376,
      "theoretical_loss": 3.633055837481076,
      "tokens_seen": 1139736576
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006613440320962889,
      "loss": 1.4353,
      "theoretical_loss": 3.633036587205935,
      "tokens_seen": 1139802112
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006613239719157472,
      "loss": 1.5017,
      "theoretical_loss": 3.6330173383475053,
      "tokens_seen": 1139867648
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006613039117352057,
      "loss": 1.5542,
      "theoretical_loss": 3.632998090905601,
      "tokens_seen": 1139933184
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006612838515546639,
      "loss": 1.4944,
      "theoretical_loss": 3.6329788448800375,
      "tokens_seen": 1139998720
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006612637913741224,
      "loss": 1.5667,
      "theoretical_loss": 3.6329596002706275,
      "tokens_seen": 1140064256
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006612437311935807,
      "loss": 1.5029,
      "theoretical_loss": 3.632940357077187,
      "tokens_seen": 1140129792
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006612236710130391,
      "loss": 1.5486,
      "theoretical_loss": 3.6329211152995295,
      "tokens_seen": 1140195328
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006612036108324975,
      "loss": 1.5013,
      "theoretical_loss": 3.6329018749374695,
      "tokens_seen": 1140260864
    },
    {
      "epoch": 0.35,
      "objective/train/advantage_avg": 0.493950754404068,
      "objective/train/docs_used": 417664,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4156428575515747,
      "objective/train/original_loss": 1.4156427383422852,
      "objective/train/theoretical_loss": 3.632882635990822,
      "objective/train/tokens_used": 1160786400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24437083303928375,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506374835968018,
      "objective/train/weighted_lm_loss": 1.487410306930542,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.632882635990822,
      "tokens_seen": 1140326400
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006611835506519559,
      "loss": 1.5338,
      "theoretical_loss": 3.632882635990822,
      "tokens_seen": 1140326400
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006611634904714142,
      "loss": 1.5769,
      "theoretical_loss": 3.632863398459402,
      "tokens_seen": 1140391936
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006611434302908727,
      "loss": 1.5839,
      "theoretical_loss": 3.6328441623430225,
      "tokens_seen": 1140457472
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000661123370110331,
      "loss": 1.4914,
      "theoretical_loss": 3.632824927641499,
      "tokens_seen": 1140523008
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006611033099297894,
      "loss": 1.5883,
      "theoretical_loss": 3.632805694354646,
      "tokens_seen": 1140588544
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006610832497492477,
      "loss": 1.5367,
      "theoretical_loss": 3.632786462482279,
      "tokens_seen": 1140654080
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006610631895687061,
      "loss": 1.5787,
      "theoretical_loss": 3.6327672320242117,
      "tokens_seen": 1140719616
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006610431293881645,
      "loss": 1.5094,
      "theoretical_loss": 3.632748002980259,
      "tokens_seen": 1140785152
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006610230692076229,
      "loss": 1.5405,
      "theoretical_loss": 3.6327287753502358,
      "tokens_seen": 1140850688
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006610030090270812,
      "loss": 1.6194,
      "theoretical_loss": 3.6327095491339567,
      "tokens_seen": 1140916224
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006609829488465396,
      "loss": 1.4677,
      "theoretical_loss": 3.632690324331237,
      "tokens_seen": 1140981760
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000660962888665998,
      "loss": 1.5156,
      "theoretical_loss": 3.6326711009418915,
      "tokens_seen": 1141047296
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006609428284854564,
      "loss": 1.5608,
      "theoretical_loss": 3.632651878965735,
      "tokens_seen": 1141112832
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006609227683049147,
      "loss": 1.5156,
      "theoretical_loss": 3.6326326584025823,
      "tokens_seen": 1141178368
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006609027081243732,
      "loss": 1.482,
      "theoretical_loss": 3.632613439252249,
      "tokens_seen": 1141243904
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006608826479438314,
      "loss": 1.4216,
      "theoretical_loss": 3.632594221514549,
      "tokens_seen": 1141309440
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006608625877632899,
      "loss": 1.6121,
      "theoretical_loss": 3.632575005189299,
      "tokens_seen": 1141374976
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006608425275827482,
      "loss": 1.5825,
      "theoretical_loss": 3.6325557902763124,
      "tokens_seen": 1141440512
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006608224674022066,
      "loss": 1.534,
      "theoretical_loss": 3.632536576775405,
      "tokens_seen": 1141506048
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000660802407221665,
      "loss": 1.5507,
      "theoretical_loss": 3.6325173646863926,
      "tokens_seen": 1141571584
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006607823470411234,
      "loss": 1.599,
      "theoretical_loss": 3.6324981540090895,
      "tokens_seen": 1141637120
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006607622868605817,
      "loss": 1.4923,
      "theoretical_loss": 3.632478944743312,
      "tokens_seen": 1141702656
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006607422266800402,
      "loss": 1.5676,
      "theoretical_loss": 3.6324597368888734,
      "tokens_seen": 1141768192
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006607221664994984,
      "loss": 1.4495,
      "theoretical_loss": 3.6324405304455913,
      "tokens_seen": 1141833728
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006607021063189569,
      "loss": 1.5584,
      "theoretical_loss": 3.6324213254132793,
      "tokens_seen": 1141899264
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006606820461384152,
      "loss": 1.5133,
      "theoretical_loss": 3.632402121791754,
      "tokens_seen": 1141964800
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006606619859578736,
      "loss": 1.571,
      "theoretical_loss": 3.63238291958083,
      "tokens_seen": 1142030336
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006606419257773321,
      "loss": 1.5677,
      "theoretical_loss": 3.632363718780323,
      "tokens_seen": 1142095872
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006606218655967903,
      "loss": 1.55,
      "theoretical_loss": 3.632344519390049,
      "tokens_seen": 1142161408
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006606018054162488,
      "loss": 1.3726,
      "theoretical_loss": 3.6323253214098226,
      "tokens_seen": 1142226944
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006605817452357072,
      "loss": 1.4808,
      "theoretical_loss": 3.63230612483946,
      "tokens_seen": 1142292480
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006605616850551655,
      "loss": 1.5343,
      "theoretical_loss": 3.6322869296787763,
      "tokens_seen": 1142358016
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006605416248746239,
      "loss": 1.5537,
      "theoretical_loss": 3.632267735927588,
      "tokens_seen": 1142423552
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006605215646940824,
      "loss": 1.6257,
      "theoretical_loss": 3.6322485435857095,
      "tokens_seen": 1142489088
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006605015045135406,
      "loss": 1.4507,
      "theoretical_loss": 3.632229352652957,
      "tokens_seen": 1142554624
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006604814443329991,
      "loss": 1.4718,
      "theoretical_loss": 3.632210163129147,
      "tokens_seen": 1142620160
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006604613841524573,
      "loss": 1.5665,
      "theoretical_loss": 3.632190975014094,
      "tokens_seen": 1142685696
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006604413239719158,
      "loss": 1.4778,
      "theoretical_loss": 3.632171788307615,
      "tokens_seen": 1142751232
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006604212637913742,
      "loss": 1.4048,
      "theoretical_loss": 3.6321526030095246,
      "tokens_seen": 1142816768
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006604012036108325,
      "loss": 1.509,
      "theoretical_loss": 3.6321334191196395,
      "tokens_seen": 1142882304
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006603811434302909,
      "loss": 1.5134,
      "theoretical_loss": 3.6321142366377757,
      "tokens_seen": 1142947840
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006603610832497493,
      "loss": 1.5239,
      "theoretical_loss": 3.6320950555637483,
      "tokens_seen": 1143013376
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006603410230692076,
      "loss": 1.423,
      "theoretical_loss": 3.632075875897374,
      "tokens_seen": 1143078912
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006603209628886661,
      "loss": 1.5571,
      "theoretical_loss": 3.6320566976384687,
      "tokens_seen": 1143144448
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006603009027081243,
      "loss": 1.3985,
      "theoretical_loss": 3.6320375207868483,
      "tokens_seen": 1143209984
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006602808425275828,
      "loss": 1.4907,
      "theoretical_loss": 3.6320183453423285,
      "tokens_seen": 1143275520
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006602607823470411,
      "loss": 1.5158,
      "theoretical_loss": 3.6319991713047264,
      "tokens_seen": 1143341056
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006602407221664995,
      "loss": 1.5461,
      "theoretical_loss": 3.6319799986738572,
      "tokens_seen": 1143406592
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006602206619859579,
      "loss": 1.4282,
      "theoretical_loss": 3.6319608274495376,
      "tokens_seen": 1143472128
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006602006018054163,
      "loss": 1.5584,
      "theoretical_loss": 3.6319416576315833,
      "tokens_seen": 1143537664
    },
    {
      "epoch": 0.35,
      "objective/train/advantage_avg": 0.49500709772109985,
      "objective/train/docs_used": 418760,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3706814050674438,
      "objective/train/original_loss": 1.3706812858581543,
      "objective/train/theoretical_loss": 3.6319224892198108,
      "objective/train/tokens_used": 1164063200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2453436702489853,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507479906082153,
      "objective/train/weighted_lm_loss": 1.4402960538864136,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6319224892198108,
      "tokens_seen": 1143603200
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006601805416248746,
      "loss": 1.4774,
      "theoretical_loss": 3.6319224892198108,
      "tokens_seen": 1143603200
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006601604814443331,
      "loss": 1.6082,
      "theoretical_loss": 3.631903322214037,
      "tokens_seen": 1143668736
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006601404212637914,
      "loss": 1.4484,
      "theoretical_loss": 3.6318841566140767,
      "tokens_seen": 1143734272
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006601203610832498,
      "loss": 1.4496,
      "theoretical_loss": 3.6318649924197484,
      "tokens_seen": 1143799808
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006601003009027081,
      "loss": 1.4812,
      "theoretical_loss": 3.6318458296308664,
      "tokens_seen": 1143865344
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006600802407221665,
      "loss": 1.5481,
      "theoretical_loss": 3.631826668247248,
      "tokens_seen": 1143930880
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006600601805416249,
      "loss": 1.4403,
      "theoretical_loss": 3.63180750826871,
      "tokens_seen": 1143996416
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006600401203610833,
      "loss": 1.4521,
      "theoretical_loss": 3.631788349695068,
      "tokens_seen": 1144061952
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006600200601805416,
      "loss": 1.3663,
      "theoretical_loss": 3.6317691925261393,
      "tokens_seen": 1144127488
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00066,
      "loss": 1.3672,
      "theoretical_loss": 3.63175003676174,
      "tokens_seen": 1144193024
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006599799398194584,
      "loss": 1.5593,
      "theoretical_loss": 3.6317308824016874,
      "tokens_seen": 1144258560
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006599598796389168,
      "loss": 1.3891,
      "theoretical_loss": 3.6317117294457972,
      "tokens_seen": 1144324096
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006599398194583751,
      "loss": 1.3479,
      "theoretical_loss": 3.6316925778938867,
      "tokens_seen": 1144389632
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006599197592778336,
      "loss": 1.5413,
      "theoretical_loss": 3.631673427745772,
      "tokens_seen": 1144455168
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006598996990972918,
      "loss": 1.513,
      "theoretical_loss": 3.6316542790012702,
      "tokens_seen": 1144520704
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006598796389167503,
      "loss": 1.557,
      "theoretical_loss": 3.6316351316601976,
      "tokens_seen": 1144586240
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006598595787362086,
      "loss": 1.5342,
      "theoretical_loss": 3.631615985722372,
      "tokens_seen": 1144651776
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000659839518555667,
      "loss": 1.3677,
      "theoretical_loss": 3.6315968411876094,
      "tokens_seen": 1144717312
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006598194583751254,
      "loss": 1.4086,
      "theoretical_loss": 3.631577698055727,
      "tokens_seen": 1144782848
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006597993981945838,
      "loss": 1.3705,
      "theoretical_loss": 3.631558556326542,
      "tokens_seen": 1144848384
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006597793380140421,
      "loss": 1.4904,
      "theoretical_loss": 3.63153941599987,
      "tokens_seen": 1144913920
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006597592778335006,
      "loss": 1.5019,
      "theoretical_loss": 3.6315202770755297,
      "tokens_seen": 1144979456
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006597392176529588,
      "loss": 1.41,
      "theoretical_loss": 3.631501139553337,
      "tokens_seen": 1145044992
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006597191574724173,
      "loss": 1.5035,
      "theoretical_loss": 3.6314820034331095,
      "tokens_seen": 1145110528
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006596990972918756,
      "loss": 1.4706,
      "theoretical_loss": 3.6314628687146637,
      "tokens_seen": 1145176064
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000659679037111334,
      "loss": 1.579,
      "theoretical_loss": 3.6314437353978173,
      "tokens_seen": 1145241600
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006596589769307924,
      "loss": 1.4842,
      "theoretical_loss": 3.6314246034823867,
      "tokens_seen": 1145307136
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006596389167502507,
      "loss": 1.5375,
      "theoretical_loss": 3.6314054729681895,
      "tokens_seen": 1145372672
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006596188565697091,
      "loss": 1.6104,
      "theoretical_loss": 3.631386343855043,
      "tokens_seen": 1145438208
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006595987963891676,
      "loss": 1.4429,
      "theoretical_loss": 3.6313672161427646,
      "tokens_seen": 1145503744
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006595787362086258,
      "loss": 1.5471,
      "theoretical_loss": 3.631348089831171,
      "tokens_seen": 1145569280
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006595586760280843,
      "loss": 1.4931,
      "theoretical_loss": 3.6313289649200797,
      "tokens_seen": 1145634816
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006595386158475427,
      "loss": 1.3731,
      "theoretical_loss": 3.6313098414093083,
      "tokens_seen": 1145700352
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000659518555667001,
      "loss": 1.5673,
      "theoretical_loss": 3.6312907192986748,
      "tokens_seen": 1145765888
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006594984954864594,
      "loss": 1.6092,
      "theoretical_loss": 3.631271598587995,
      "tokens_seen": 1145831424
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006594784353059177,
      "loss": 1.5163,
      "theoretical_loss": 3.6312524792770873,
      "tokens_seen": 1145896960
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006594583751253761,
      "loss": 1.477,
      "theoretical_loss": 3.6312333613657692,
      "tokens_seen": 1145962496
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006594383149448346,
      "loss": 1.5478,
      "theoretical_loss": 3.6312142448538585,
      "tokens_seen": 1146028032
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006594182547642928,
      "loss": 1.5245,
      "theoretical_loss": 3.631195129741172,
      "tokens_seen": 1146093568
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006593981945837513,
      "loss": 1.4932,
      "theoretical_loss": 3.6311760160275277,
      "tokens_seen": 1146159104
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006593781344032096,
      "loss": 1.5628,
      "theoretical_loss": 3.631156903712743,
      "tokens_seen": 1146224640
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000659358074222668,
      "loss": 1.6333,
      "theoretical_loss": 3.631137792796636,
      "tokens_seen": 1146290176
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006593380140421264,
      "loss": 1.4364,
      "theoretical_loss": 3.631118683279024,
      "tokens_seen": 1146355712
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006593179538615847,
      "loss": 1.4994,
      "theoretical_loss": 3.6310995751597246,
      "tokens_seen": 1146421248
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006592978936810431,
      "loss": 1.4735,
      "theoretical_loss": 3.631080468438556,
      "tokens_seen": 1146486784
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006592778335005015,
      "loss": 1.4479,
      "theoretical_loss": 3.6310613631153354,
      "tokens_seen": 1146552320
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006592577733199598,
      "loss": 1.5911,
      "theoretical_loss": 3.6310422591898814,
      "tokens_seen": 1146617856
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006592377131394183,
      "loss": 1.4687,
      "theoretical_loss": 3.6310231566620113,
      "tokens_seen": 1146683392
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006592176529588766,
      "loss": 1.513,
      "theoretical_loss": 3.631004055531543,
      "tokens_seen": 1146748928
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000659197592778335,
      "loss": 1.4784,
      "theoretical_loss": 3.6309849557982945,
      "tokens_seen": 1146814464
    },
    {
      "epoch": 0.35,
      "objective/train/advantage_avg": 0.49184098839759827,
      "objective/train/docs_used": 419954,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3984112739562988,
      "objective/train/original_loss": 1.3984110355377197,
      "objective/train/theoretical_loss": 3.630965857462084,
      "objective/train/tokens_used": 1167340000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24249784648418427,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504168272018433,
      "objective/train/weighted_lm_loss": 1.4686956405639648,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.630965857462084,
      "tokens_seen": 1146880000
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006591775325977934,
      "loss": 1.3878,
      "theoretical_loss": 3.630965857462084,
      "tokens_seen": 1146880000
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006591574724172518,
      "loss": 1.4191,
      "theoretical_loss": 3.630946760522729,
      "tokens_seen": 1146945536
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006591374122367102,
      "loss": 1.5979,
      "theoretical_loss": 3.6309276649800477,
      "tokens_seen": 1147011072
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006591173520561685,
      "loss": 1.605,
      "theoretical_loss": 3.630908570833859,
      "tokens_seen": 1147076608
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000659097291875627,
      "loss": 1.4658,
      "theoretical_loss": 3.6308894780839798,
      "tokens_seen": 1147142144
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006590772316950853,
      "loss": 1.4832,
      "theoretical_loss": 3.630870386730229,
      "tokens_seen": 1147207680
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006590571715145437,
      "loss": 1.4101,
      "theoretical_loss": 3.6308512967724242,
      "tokens_seen": 1147273216
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000659037111334002,
      "loss": 1.4571,
      "theoretical_loss": 3.6308322082103834,
      "tokens_seen": 1147338752
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006590170511534604,
      "loss": 1.5108,
      "theoretical_loss": 3.630813121043926,
      "tokens_seen": 1147404288
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006589969909729188,
      "loss": 1.506,
      "theoretical_loss": 3.6307940352728694,
      "tokens_seen": 1147469824
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006589769307923772,
      "loss": 1.5982,
      "theoretical_loss": 3.630774950897032,
      "tokens_seen": 1147535360
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006589568706118355,
      "loss": 1.5064,
      "theoretical_loss": 3.630755867916232,
      "tokens_seen": 1147600896
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000658936810431294,
      "loss": 1.4379,
      "theoretical_loss": 3.630736786330288,
      "tokens_seen": 1147666432
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006589167502507522,
      "loss": 1.4406,
      "theoretical_loss": 3.6307177061390186,
      "tokens_seen": 1147731968
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006588966900702107,
      "loss": 1.6438,
      "theoretical_loss": 3.630698627342242,
      "tokens_seen": 1147797504
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000658876629889669,
      "loss": 1.5016,
      "theoretical_loss": 3.6306795499397766,
      "tokens_seen": 1147863040
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006588565697091274,
      "loss": 1.5385,
      "theoretical_loss": 3.630660473931441,
      "tokens_seen": 1147928576
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006588365095285858,
      "loss": 1.5811,
      "theoretical_loss": 3.630641399317054,
      "tokens_seen": 1147994112
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006588164493480442,
      "loss": 1.5966,
      "theoretical_loss": 3.630622326096433,
      "tokens_seen": 1148059648
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006587963891675025,
      "loss": 1.5271,
      "theoretical_loss": 3.6306032542693982,
      "tokens_seen": 1148125184
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000658776328986961,
      "loss": 1.4899,
      "theoretical_loss": 3.6305841838357673,
      "tokens_seen": 1148190720
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006587562688064192,
      "loss": 1.4334,
      "theoretical_loss": 3.6305651147953593,
      "tokens_seen": 1148256256
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006587362086258777,
      "loss": 1.5065,
      "theoretical_loss": 3.630546047147993,
      "tokens_seen": 1148321792
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000658716148445336,
      "loss": 1.3569,
      "theoretical_loss": 3.630526980893486,
      "tokens_seen": 1148387328
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006586960882647944,
      "loss": 1.4842,
      "theoretical_loss": 3.630507916031659,
      "tokens_seen": 1148452864
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006586760280842528,
      "loss": 1.4598,
      "theoretical_loss": 3.63048885256233,
      "tokens_seen": 1148518400
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006586559679037111,
      "loss": 1.4761,
      "theoretical_loss": 3.630469790485317,
      "tokens_seen": 1148583936
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006586359077231695,
      "loss": 1.5885,
      "theoretical_loss": 3.6304507298004394,
      "tokens_seen": 1148649472
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000658615847542628,
      "loss": 1.5924,
      "theoretical_loss": 3.630431670507517,
      "tokens_seen": 1148715008
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006585957873620862,
      "loss": 1.4684,
      "theoretical_loss": 3.630412612606367,
      "tokens_seen": 1148780544
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006585757271815447,
      "loss": 1.5399,
      "theoretical_loss": 3.6303935560968106,
      "tokens_seen": 1148846080
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000658555667001003,
      "loss": 1.5319,
      "theoretical_loss": 3.6303745009786645,
      "tokens_seen": 1148911616
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006585356068204614,
      "loss": 1.5311,
      "theoretical_loss": 3.6303554472517496,
      "tokens_seen": 1148977152
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006585155466399198,
      "loss": 1.4134,
      "theoretical_loss": 3.630336394915884,
      "tokens_seen": 1149042688
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006584954864593781,
      "loss": 1.5534,
      "theoretical_loss": 3.630317343970887,
      "tokens_seen": 1149108224
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006584754262788365,
      "loss": 1.4156,
      "theoretical_loss": 3.630298294416578,
      "tokens_seen": 1149173760
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000658455366098295,
      "loss": 1.4704,
      "theoretical_loss": 3.6302792462527758,
      "tokens_seen": 1149239296
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006584353059177532,
      "loss": 1.4268,
      "theoretical_loss": 3.6302601994792996,
      "tokens_seen": 1149304832
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006584152457372117,
      "loss": 1.4593,
      "theoretical_loss": 3.6302411540959687,
      "tokens_seen": 1149370368
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00065839518555667,
      "loss": 1.4028,
      "theoretical_loss": 3.630222110102603,
      "tokens_seen": 1149435904
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006583751253761284,
      "loss": 1.5233,
      "theoretical_loss": 3.6302030674990213,
      "tokens_seen": 1149501440
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006583550651955868,
      "loss": 1.4843,
      "theoretical_loss": 3.6301840262850424,
      "tokens_seen": 1149566976
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006583350050150451,
      "loss": 1.3903,
      "theoretical_loss": 3.6301649864604864,
      "tokens_seen": 1149632512
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006583149448345035,
      "loss": 1.7235,
      "theoretical_loss": 3.6301459480251728,
      "tokens_seen": 1149698048
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006582948846539619,
      "loss": 1.5489,
      "theoretical_loss": 3.6301269109789214,
      "tokens_seen": 1149763584
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006582748244734202,
      "loss": 1.4446,
      "theoretical_loss": 3.63010787532155,
      "tokens_seen": 1149829120
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006582547642928787,
      "loss": 1.4996,
      "theoretical_loss": 3.6300888410528804,
      "tokens_seen": 1149894656
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000658234704112337,
      "loss": 1.4707,
      "theoretical_loss": 3.6300698081727303,
      "tokens_seen": 1149960192
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006582146439317954,
      "loss": 1.458,
      "theoretical_loss": 3.63005077668092,
      "tokens_seen": 1150025728
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006581945837512538,
      "loss": 1.5098,
      "theoretical_loss": 3.630031746577269,
      "tokens_seen": 1150091264
    },
    {
      "epoch": 0.35,
      "objective/train/advantage_avg": 0.4933938682079315,
      "objective/train/docs_used": 421268,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.42300546169281,
      "objective/train/original_loss": 1.4230053424835205,
      "objective/train/theoretical_loss": 3.630012717861597,
      "objective/train/tokens_used": 1170616800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24393661320209503,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05057954788208,
      "objective/train/weighted_lm_loss": 1.4950493574142456,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.630012717861597,
      "tokens_seen": 1150156800
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006581745235707122,
      "loss": 1.4652,
      "theoretical_loss": 3.630012717861597,
      "tokens_seen": 1150156800
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006581544633901705,
      "loss": 1.5233,
      "theoretical_loss": 3.6299936905337242,
      "tokens_seen": 1150222336
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006581344032096289,
      "loss": 1.4103,
      "theoretical_loss": 3.62997466459347,
      "tokens_seen": 1150287872
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006581143430290872,
      "loss": 1.4909,
      "theoretical_loss": 3.629955640040653,
      "tokens_seen": 1150353408
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006580942828485457,
      "loss": 1.4494,
      "theoretical_loss": 3.629936616875095,
      "tokens_seen": 1150418944
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000658074222668004,
      "loss": 1.4434,
      "theoretical_loss": 3.6299175950966145,
      "tokens_seen": 1150484480
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006580541624874624,
      "loss": 1.377,
      "theoretical_loss": 3.629898574705031,
      "tokens_seen": 1150550016
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006580341023069207,
      "loss": 1.5408,
      "theoretical_loss": 3.629879555700166,
      "tokens_seen": 1150615552
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006580140421263792,
      "loss": 1.479,
      "theoretical_loss": 3.6298605380818385,
      "tokens_seen": 1150681088
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006579939819458375,
      "loss": 1.4808,
      "theoretical_loss": 3.629841521849868,
      "tokens_seen": 1150746624
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006579739217652959,
      "loss": 1.5058,
      "theoretical_loss": 3.629822507004075,
      "tokens_seen": 1150812160
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006579538615847542,
      "loss": 1.4666,
      "theoretical_loss": 3.6298034935442796,
      "tokens_seen": 1150877696
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006579338014042126,
      "loss": 1.5837,
      "theoretical_loss": 3.6297844814703017,
      "tokens_seen": 1150943232
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000657913741223671,
      "loss": 1.4509,
      "theoretical_loss": 3.6297654707819618,
      "tokens_seen": 1151008768
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006578936810431294,
      "loss": 1.4068,
      "theoretical_loss": 3.629746461479079,
      "tokens_seen": 1151074304
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006578736208625877,
      "loss": 1.4573,
      "theoretical_loss": 3.6297274535614745,
      "tokens_seen": 1151139840
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006578535606820462,
      "loss": 1.2777,
      "theoretical_loss": 3.6297084470289676,
      "tokens_seen": 1151205376
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006578335005015045,
      "loss": 1.6291,
      "theoretical_loss": 3.6296894418813794,
      "tokens_seen": 1151270912
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006578134403209629,
      "loss": 1.473,
      "theoretical_loss": 3.62967043811853,
      "tokens_seen": 1151336448
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006577933801404213,
      "loss": 1.5142,
      "theoretical_loss": 3.6296514357402394,
      "tokens_seen": 1151401984
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006577733199598796,
      "loss": 1.5524,
      "theoretical_loss": 3.629632434746328,
      "tokens_seen": 1151467520
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000657753259779338,
      "loss": 1.3829,
      "theoretical_loss": 3.6296134351366156,
      "tokens_seen": 1151533056
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006577331995987964,
      "loss": 1.5725,
      "theoretical_loss": 3.629594436910924,
      "tokens_seen": 1151598592
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006577131394182547,
      "loss": 1.543,
      "theoretical_loss": 3.6295754400690723,
      "tokens_seen": 1151664128
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006576930792377132,
      "loss": 1.4144,
      "theoretical_loss": 3.6295564446108814,
      "tokens_seen": 1151729664
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006576730190571714,
      "loss": 1.5205,
      "theoretical_loss": 3.629537450536172,
      "tokens_seen": 1151795200
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006576529588766299,
      "loss": 1.4864,
      "theoretical_loss": 3.6295184578447643,
      "tokens_seen": 1151860736
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006576328986960884,
      "loss": 1.4653,
      "theoretical_loss": 3.6294994665364793,
      "tokens_seen": 1151926272
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006576128385155466,
      "loss": 1.4079,
      "theoretical_loss": 3.6294804766111373,
      "tokens_seen": 1151991808
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006575927783350051,
      "loss": 1.4837,
      "theoretical_loss": 3.629461488068559,
      "tokens_seen": 1152057344
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006575727181544635,
      "loss": 1.3591,
      "theoretical_loss": 3.6294425009085645,
      "tokens_seen": 1152122880
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006575526579739218,
      "loss": 1.4718,
      "theoretical_loss": 3.6294235151309753,
      "tokens_seen": 1152188416
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006575325977933802,
      "loss": 1.5629,
      "theoretical_loss": 3.629404530735612,
      "tokens_seen": 1152253952
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006575125376128385,
      "loss": 1.4265,
      "theoretical_loss": 3.629385547722295,
      "tokens_seen": 1152319488
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006574924774322969,
      "loss": 1.5096,
      "theoretical_loss": 3.6293665660908454,
      "tokens_seen": 1152385024
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006574724172517554,
      "loss": 1.4252,
      "theoretical_loss": 3.629347585841084,
      "tokens_seen": 1152450560
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006574523570712136,
      "loss": 1.4926,
      "theoretical_loss": 3.629328606972831,
      "tokens_seen": 1152516096
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006574322968906721,
      "loss": 1.4961,
      "theoretical_loss": 3.629309629485909,
      "tokens_seen": 1152581632
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006574122367101304,
      "loss": 1.4657,
      "theoretical_loss": 3.6292906533801372,
      "tokens_seen": 1152647168
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006573921765295888,
      "loss": 1.4019,
      "theoretical_loss": 3.629271678655337,
      "tokens_seen": 1152712704
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006573721163490472,
      "loss": 1.5313,
      "theoretical_loss": 3.6292527053113295,
      "tokens_seen": 1152778240
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006573520561685055,
      "loss": 1.4883,
      "theoretical_loss": 3.6292337333479363,
      "tokens_seen": 1152843776
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006573319959879639,
      "loss": 1.6525,
      "theoretical_loss": 3.6292147627649776,
      "tokens_seen": 1152909312
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006573119358074223,
      "loss": 1.5833,
      "theoretical_loss": 3.629195793562275,
      "tokens_seen": 1152974848
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006572918756268806,
      "loss": 1.4269,
      "theoretical_loss": 3.6291768257396493,
      "tokens_seen": 1153040384
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006572718154463391,
      "loss": 1.5407,
      "theoretical_loss": 3.6291578592969223,
      "tokens_seen": 1153105920
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006572517552657974,
      "loss": 1.6303,
      "theoretical_loss": 3.6291388942339147,
      "tokens_seen": 1153171456
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006572316950852558,
      "loss": 1.4647,
      "theoretical_loss": 3.629119930550447,
      "tokens_seen": 1153236992
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006572116349047142,
      "loss": 1.3342,
      "theoretical_loss": 3.629100968246342,
      "tokens_seen": 1153302528
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006571915747241726,
      "loss": 1.5029,
      "theoretical_loss": 3.6290820073214203,
      "tokens_seen": 1153368064
    },
    {
      "epoch": 0.35,
      "objective/train/advantage_avg": 0.4908078908920288,
      "objective/train/docs_used": 422447,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5735915899276733,
      "objective/train/original_loss": 1.5735914707183838,
      "objective/train/theoretical_loss": 3.6290630477755026,
      "objective/train/tokens_used": 1173893600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24156801402568817,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503087043762207,
      "objective/train/weighted_lm_loss": 1.652591347694397,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6290630477755026,
      "tokens_seen": 1153433600
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006571715145436309,
      "loss": 1.5357,
      "theoretical_loss": 3.6290630477755026,
      "tokens_seen": 1153433600
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006571514543630893,
      "loss": 1.3896,
      "theoretical_loss": 3.6290440896084113,
      "tokens_seen": 1153499136
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006571313941825476,
      "loss": 1.5124,
      "theoretical_loss": 3.629025132819967,
      "tokens_seen": 1153564672
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006571113340020061,
      "loss": 1.4602,
      "theoretical_loss": 3.6290061774099915,
      "tokens_seen": 1153630208
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006570912738214644,
      "loss": 1.4918,
      "theoretical_loss": 3.6289872233783065,
      "tokens_seen": 1153695744
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006570712136409228,
      "loss": 1.4224,
      "theoretical_loss": 3.628968270724733,
      "tokens_seen": 1153761280
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006570511534603811,
      "loss": 1.3967,
      "theoretical_loss": 3.6289493194490925,
      "tokens_seen": 1153826816
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006570310932798396,
      "loss": 1.5044,
      "theoretical_loss": 3.628930369551207,
      "tokens_seen": 1153892352
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006570110330992979,
      "loss": 1.4323,
      "theoretical_loss": 3.6289114210308977,
      "tokens_seen": 1153957888
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006569909729187563,
      "loss": 1.4481,
      "theoretical_loss": 3.628892473887987,
      "tokens_seen": 1154023424
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006569709127382146,
      "loss": 1.4825,
      "theoretical_loss": 3.6288735281222957,
      "tokens_seen": 1154088960
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000656950852557673,
      "loss": 1.4966,
      "theoretical_loss": 3.628854583733646,
      "tokens_seen": 1154154496
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006569307923771314,
      "loss": 1.5212,
      "theoretical_loss": 3.628835640721859,
      "tokens_seen": 1154220032
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006569107321965898,
      "loss": 1.3789,
      "theoretical_loss": 3.628816699086757,
      "tokens_seen": 1154285568
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006568906720160481,
      "loss": 1.4672,
      "theoretical_loss": 3.6287977588281617,
      "tokens_seen": 1154351104
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006568706118355066,
      "loss": 1.5228,
      "theoretical_loss": 3.628778819945895,
      "tokens_seen": 1154416640
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006568505516549649,
      "loss": 1.5543,
      "theoretical_loss": 3.6287598824397787,
      "tokens_seen": 1154482176
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006568304914744233,
      "loss": 1.5078,
      "theoretical_loss": 3.6287409463096343,
      "tokens_seen": 1154547712
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006568104312938817,
      "loss": 1.5825,
      "theoretical_loss": 3.6287220115552845,
      "tokens_seen": 1154613248
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00065679037111334,
      "loss": 1.4961,
      "theoretical_loss": 3.6287030781765504,
      "tokens_seen": 1154678784
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006567703109327984,
      "loss": 1.5193,
      "theoretical_loss": 3.6286841461732546,
      "tokens_seen": 1154744320
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006567502507522568,
      "loss": 1.4341,
      "theoretical_loss": 3.628665215545219,
      "tokens_seen": 1154809856
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006567301905717151,
      "loss": 1.4186,
      "theoretical_loss": 3.6286462862922657,
      "tokens_seen": 1154875392
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006567101303911736,
      "loss": 1.4769,
      "theoretical_loss": 3.628627358414217,
      "tokens_seen": 1154940928
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006566900702106318,
      "loss": 1.4898,
      "theoretical_loss": 3.6286084319108944,
      "tokens_seen": 1155006464
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006566700100300903,
      "loss": 1.4173,
      "theoretical_loss": 3.62858950678212,
      "tokens_seen": 1155072000
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006566499498495487,
      "loss": 1.464,
      "theoretical_loss": 3.628570583027717,
      "tokens_seen": 1155137536
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000656629889669007,
      "loss": 1.4712,
      "theoretical_loss": 3.6285516606475072,
      "tokens_seen": 1155203072
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006566098294884654,
      "loss": 1.3858,
      "theoretical_loss": 3.628532739641312,
      "tokens_seen": 1155268608
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006565897693079239,
      "loss": 1.4876,
      "theoretical_loss": 3.628513820008955,
      "tokens_seen": 1155334144
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006565697091273821,
      "loss": 1.3874,
      "theoretical_loss": 3.6284949017502575,
      "tokens_seen": 1155399680
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006565496489468406,
      "loss": 1.5171,
      "theoretical_loss": 3.628475984865042,
      "tokens_seen": 1155465216
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006565295887662988,
      "loss": 1.3477,
      "theoretical_loss": 3.6284570693531317,
      "tokens_seen": 1155530752
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006565095285857573,
      "loss": 1.4582,
      "theoretical_loss": 3.628438155214348,
      "tokens_seen": 1155596288
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006564894684052157,
      "loss": 1.4883,
      "theoretical_loss": 3.6284192424485147,
      "tokens_seen": 1155661824
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000656469408224674,
      "loss": 1.4685,
      "theoretical_loss": 3.6284003310554525,
      "tokens_seen": 1155727360
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006564493480441324,
      "loss": 1.4419,
      "theoretical_loss": 3.6283814210349847,
      "tokens_seen": 1155792896
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006564292878635908,
      "loss": 1.4308,
      "theoretical_loss": 3.6283625123869347,
      "tokens_seen": 1155858432
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006564092276830491,
      "loss": 1.5068,
      "theoretical_loss": 3.628343605111124,
      "tokens_seen": 1155923968
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006563891675025076,
      "loss": 1.4324,
      "theoretical_loss": 3.6283246992073757,
      "tokens_seen": 1155989504
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006563691073219658,
      "loss": 1.4199,
      "theoretical_loss": 3.628305794675512,
      "tokens_seen": 1156055040
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006563490471414243,
      "loss": 1.5363,
      "theoretical_loss": 3.6282868915153568,
      "tokens_seen": 1156120576
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006563289869608826,
      "loss": 1.4379,
      "theoretical_loss": 3.6282679897267314,
      "tokens_seen": 1156186112
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000656308926780341,
      "loss": 1.3656,
      "theoretical_loss": 3.628249089309459,
      "tokens_seen": 1156251648
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006562888665997994,
      "loss": 1.4722,
      "theoretical_loss": 3.6282301902633627,
      "tokens_seen": 1156317184
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006562688064192578,
      "loss": 1.4692,
      "theoretical_loss": 3.628211292588265,
      "tokens_seen": 1156382720
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006562487462387161,
      "loss": 1.364,
      "theoretical_loss": 3.628192396283989,
      "tokens_seen": 1156448256
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006562286860581746,
      "loss": 1.4217,
      "theoretical_loss": 3.628173501350358,
      "tokens_seen": 1156513792
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006562086258776328,
      "loss": 1.4006,
      "theoretical_loss": 3.628154607787194,
      "tokens_seen": 1156579328
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006561885656970913,
      "loss": 1.5849,
      "theoretical_loss": 3.62813571559432,
      "tokens_seen": 1156644864
    },
    {
      "epoch": 0.35,
      "objective/train/advantage_avg": 0.49201664328575134,
      "objective/train/docs_used": 423626,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2265117168426514,
      "objective/train/original_loss": 1.2265117168426514,
      "objective/train/theoretical_loss": 3.62811682477156,
      "objective/train/tokens_used": 1177170400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2423246204853058,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504335165023804,
      "objective/train/weighted_lm_loss": 1.2883967161178589,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 3.62811682477156,
      "tokens_seen": 1156710400
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006561685055165496,
      "loss": 1.3603,
      "theoretical_loss": 3.62811682477156,
      "tokens_seen": 1156710400
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000656148445336008,
      "loss": 1.3862,
      "theoretical_loss": 3.6280979353187357,
      "tokens_seen": 1156775936
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006561283851554665,
      "loss": 1.4208,
      "theoretical_loss": 3.6280790472356705,
      "tokens_seen": 1156841472
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006561083249749248,
      "loss": 1.3841,
      "theoretical_loss": 3.6280601605221885,
      "tokens_seen": 1156907008
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006560882647943832,
      "loss": 1.5108,
      "theoretical_loss": 3.628041275178112,
      "tokens_seen": 1156972544
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006560682046138415,
      "loss": 1.4976,
      "theoretical_loss": 3.6280223912032636,
      "tokens_seen": 1157038080
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006560481444333,
      "loss": 1.3362,
      "theoretical_loss": 3.628003508597468,
      "tokens_seen": 1157103616
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006560280842527583,
      "loss": 1.4971,
      "theoretical_loss": 3.6279846273605467,
      "tokens_seen": 1157169152
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006560080240722167,
      "loss": 1.5312,
      "theoretical_loss": 3.6279657474923246,
      "tokens_seen": 1157234688
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000655987963891675,
      "loss": 1.5567,
      "theoretical_loss": 3.6279468689926233,
      "tokens_seen": 1157300224
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006559679037111334,
      "loss": 1.4131,
      "theoretical_loss": 3.6279279918612675,
      "tokens_seen": 1157365760
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006559478435305918,
      "loss": 1.4279,
      "theoretical_loss": 3.6279091160980803,
      "tokens_seen": 1157431296
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006559277833500502,
      "loss": 1.3345,
      "theoretical_loss": 3.627890241702884,
      "tokens_seen": 1157496832
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006559077231695085,
      "loss": 1.4688,
      "theoretical_loss": 3.6278713686755033,
      "tokens_seen": 1157562368
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000655887662988967,
      "loss": 1.5376,
      "theoretical_loss": 3.6278524970157613,
      "tokens_seen": 1157627904
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006558676028084253,
      "loss": 1.4822,
      "theoretical_loss": 3.627833626723481,
      "tokens_seen": 1157693440
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006558475426278837,
      "loss": 1.4913,
      "theoretical_loss": 3.6278147577984865,
      "tokens_seen": 1157758976
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000655827482447342,
      "loss": 1.4144,
      "theoretical_loss": 3.627795890240601,
      "tokens_seen": 1157824512
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006558074222668004,
      "loss": 1.5887,
      "theoretical_loss": 3.6277770240496476,
      "tokens_seen": 1157890048
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006557873620862588,
      "loss": 1.4489,
      "theoretical_loss": 3.6277581592254515,
      "tokens_seen": 1157955584
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006557673019057172,
      "loss": 1.4616,
      "theoretical_loss": 3.6277392957678343,
      "tokens_seen": 1158021120
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006557472417251755,
      "loss": 1.4108,
      "theoretical_loss": 3.627720433676622,
      "tokens_seen": 1158086656
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000655727181544634,
      "loss": 1.3809,
      "theoretical_loss": 3.6277015729516355,
      "tokens_seen": 1158152192
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006557071213640922,
      "loss": 1.4536,
      "theoretical_loss": 3.627682713592701,
      "tokens_seen": 1158217728
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006556870611835507,
      "loss": 1.497,
      "theoretical_loss": 3.627663855599641,
      "tokens_seen": 1158283264
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006556670010030091,
      "loss": 1.5221,
      "theoretical_loss": 3.627644998972279,
      "tokens_seen": 1158348800
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006556469408224674,
      "loss": 1.3336,
      "theoretical_loss": 3.6276261437104402,
      "tokens_seen": 1158414336
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006556268806419258,
      "loss": 1.4375,
      "theoretical_loss": 3.6276072898139478,
      "tokens_seen": 1158479872
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006556068204613843,
      "loss": 1.6005,
      "theoretical_loss": 3.6275884372826255,
      "tokens_seen": 1158545408
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006555867602808425,
      "loss": 1.5507,
      "theoretical_loss": 3.627569586116297,
      "tokens_seen": 1158610944
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000655566700100301,
      "loss": 1.4881,
      "theoretical_loss": 3.6275507363147868,
      "tokens_seen": 1158676480
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006555466399197592,
      "loss": 1.4737,
      "theoretical_loss": 3.627531887877919,
      "tokens_seen": 1158742016
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006555265797392177,
      "loss": 1.5145,
      "theoretical_loss": 3.6275130408055167,
      "tokens_seen": 1158807552
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006555065195586761,
      "loss": 1.3899,
      "theoretical_loss": 3.6274941950974053,
      "tokens_seen": 1158873088
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006554864593781344,
      "loss": 1.4723,
      "theoretical_loss": 3.6274753507534077,
      "tokens_seen": 1158938624
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006554663991975928,
      "loss": 1.5387,
      "theoretical_loss": 3.6274565077733487,
      "tokens_seen": 1159004160
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006554463390170512,
      "loss": 1.462,
      "theoretical_loss": 3.627437666157052,
      "tokens_seen": 1159069696
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006554262788365095,
      "loss": 1.5726,
      "theoretical_loss": 3.6274188259043427,
      "tokens_seen": 1159135232
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000655406218655968,
      "loss": 1.5469,
      "theoretical_loss": 3.6273999870150444,
      "tokens_seen": 1159200768
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006553861584754262,
      "loss": 1.4165,
      "theoretical_loss": 3.627381149488981,
      "tokens_seen": 1159266304
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006553660982948847,
      "loss": 1.3023,
      "theoretical_loss": 3.6273623133259774,
      "tokens_seen": 1159331840
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000655346038114343,
      "loss": 1.3859,
      "theoretical_loss": 3.6273434785258574,
      "tokens_seen": 1159397376
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006553259779338014,
      "loss": 1.4886,
      "theoretical_loss": 3.627324645088446,
      "tokens_seen": 1159462912
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006553059177532598,
      "loss": 1.5132,
      "theoretical_loss": 3.6273058130135674,
      "tokens_seen": 1159528448
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006552858575727182,
      "loss": 1.4479,
      "theoretical_loss": 3.6272869823010456,
      "tokens_seen": 1159593984
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006552657973921765,
      "loss": 1.5317,
      "theoretical_loss": 3.6272681529507054,
      "tokens_seen": 1159659520
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000655245737211635,
      "loss": 1.42,
      "theoretical_loss": 3.627249324962371,
      "tokens_seen": 1159725056
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006552256770310932,
      "loss": 1.5023,
      "theoretical_loss": 3.6272304983358676,
      "tokens_seen": 1159790592
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006552056168505517,
      "loss": 1.3709,
      "theoretical_loss": 3.6272116730710193,
      "tokens_seen": 1159856128
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00065518555667001,
      "loss": 1.4597,
      "theoretical_loss": 3.62719284916765,
      "tokens_seen": 1159921664
    },
    {
      "epoch": 0.35,
      "objective/train/advantage_avg": 0.4953343868255615,
      "objective/train/docs_used": 424758,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4626121520996094,
      "objective/train/original_loss": 1.4626121520996094,
      "objective/train/theoretical_loss": 3.6271740266255854,
      "objective/train/tokens_used": 1180447200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24556894600391388,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507818460464478,
      "objective/train/weighted_lm_loss": 1.5368435382843018,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6271740266255854,
      "tokens_seen": 1159987200
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006551654964894684,
      "loss": 1.5818,
      "theoretical_loss": 3.6271740266255854,
      "tokens_seen": 1159987200
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006551454363089268,
      "loss": 1.4671,
      "theoretical_loss": 3.6271552054446503,
      "tokens_seen": 1160052736
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006551253761283852,
      "loss": 1.5165,
      "theoretical_loss": 3.6271363856246683,
      "tokens_seen": 1160118272
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006551053159478435,
      "loss": 1.4873,
      "theoretical_loss": 3.627117567165465,
      "tokens_seen": 1160183808
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006550852557673019,
      "loss": 1.4933,
      "theoretical_loss": 3.6270987500668648,
      "tokens_seen": 1160249344
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006550651955867603,
      "loss": 1.4637,
      "theoretical_loss": 3.627079934328692,
      "tokens_seen": 1160314880
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006550451354062187,
      "loss": 1.4812,
      "theoretical_loss": 3.6270611199507723,
      "tokens_seen": 1160380416
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000655025075225677,
      "loss": 1.5158,
      "theoretical_loss": 3.62704230693293,
      "tokens_seen": 1160445952
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006550050150451354,
      "loss": 1.5062,
      "theoretical_loss": 3.6270234952749902,
      "tokens_seen": 1160511488
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006549849548645937,
      "loss": 1.4439,
      "theoretical_loss": 3.627004684976778,
      "tokens_seen": 1160577024
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006549648946840522,
      "loss": 1.5346,
      "theoretical_loss": 3.626985876038118,
      "tokens_seen": 1160642560
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006549448345035105,
      "loss": 1.443,
      "theoretical_loss": 3.626967068458835,
      "tokens_seen": 1160708096
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006549247743229689,
      "loss": 1.4122,
      "theoretical_loss": 3.6269482622387548,
      "tokens_seen": 1160773632
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006549047141424273,
      "loss": 1.4081,
      "theoretical_loss": 3.6269294573777016,
      "tokens_seen": 1160839168
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006548846539618857,
      "loss": 1.4361,
      "theoretical_loss": 3.6269106538755005,
      "tokens_seen": 1160904704
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000654864593781344,
      "loss": 1.5031,
      "theoretical_loss": 3.626891851731977,
      "tokens_seen": 1160970240
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006548445336008025,
      "loss": 1.5959,
      "theoretical_loss": 3.6268730509469567,
      "tokens_seen": 1161035776
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006548244734202607,
      "loss": 1.3638,
      "theoretical_loss": 3.6268542515202635,
      "tokens_seen": 1161101312
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006548044132397192,
      "loss": 1.5414,
      "theoretical_loss": 3.626835453451724,
      "tokens_seen": 1161166848
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006547843530591775,
      "loss": 1.3978,
      "theoretical_loss": 3.626816656741162,
      "tokens_seen": 1161232384
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006547642928786359,
      "loss": 1.4212,
      "theoretical_loss": 3.626797861388404,
      "tokens_seen": 1161297920
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006547442326980943,
      "loss": 1.4016,
      "theoretical_loss": 3.626779067393275,
      "tokens_seen": 1161363456
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006547241725175526,
      "loss": 1.4659,
      "theoretical_loss": 3.6267602747555996,
      "tokens_seen": 1161428992
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000654704112337011,
      "loss": 1.5147,
      "theoretical_loss": 3.626741483475204,
      "tokens_seen": 1161494528
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006546840521564695,
      "loss": 1.5985,
      "theoretical_loss": 3.6267226935519132,
      "tokens_seen": 1161560064
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006546639919759277,
      "loss": 1.4677,
      "theoretical_loss": 3.626703904985553,
      "tokens_seen": 1161625600
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006546439317953862,
      "loss": 1.424,
      "theoretical_loss": 3.6266851177759483,
      "tokens_seen": 1161691136
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006546238716148447,
      "loss": 1.4909,
      "theoretical_loss": 3.626666331922925,
      "tokens_seen": 1161756672
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006546038114343029,
      "loss": 1.551,
      "theoretical_loss": 3.626647547426309,
      "tokens_seen": 1161822208
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006545837512537614,
      "loss": 1.5217,
      "theoretical_loss": 3.6266287642859245,
      "tokens_seen": 1161887744
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006545636910732196,
      "loss": 1.4979,
      "theoretical_loss": 3.6266099825015985,
      "tokens_seen": 1161953280
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006545436308926781,
      "loss": 1.5422,
      "theoretical_loss": 3.626591202073156,
      "tokens_seen": 1162018816
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006545235707121365,
      "loss": 1.421,
      "theoretical_loss": 3.6265724230004226,
      "tokens_seen": 1162084352
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006545035105315948,
      "loss": 1.3755,
      "theoretical_loss": 3.6265536452832245,
      "tokens_seen": 1162149888
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006544834503510532,
      "loss": 1.3901,
      "theoretical_loss": 3.6265348689213868,
      "tokens_seen": 1162215424
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006544633901705116,
      "loss": 1.4938,
      "theoretical_loss": 3.626516093914736,
      "tokens_seen": 1162280960
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006544433299899699,
      "loss": 1.5731,
      "theoretical_loss": 3.6264973202630966,
      "tokens_seen": 1162346496
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006544232698094284,
      "loss": 1.2667,
      "theoretical_loss": 3.626478547966296,
      "tokens_seen": 1162412032
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006544032096288866,
      "loss": 1.6415,
      "theoretical_loss": 3.6264597770241584,
      "tokens_seen": 1162477568
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006543831494483451,
      "loss": 1.3828,
      "theoretical_loss": 3.6264410074365117,
      "tokens_seen": 1162543104
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006543630892678034,
      "loss": 1.4016,
      "theoretical_loss": 3.6264222392031797,
      "tokens_seen": 1162608640
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006543430290872618,
      "loss": 1.4066,
      "theoretical_loss": 3.6264034723239895,
      "tokens_seen": 1162674176
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006543229689067202,
      "loss": 1.3637,
      "theoretical_loss": 3.626384706798767,
      "tokens_seen": 1162739712
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006543029087261786,
      "loss": 1.3711,
      "theoretical_loss": 3.626365942627338,
      "tokens_seen": 1162805248
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006542828485456369,
      "loss": 1.5541,
      "theoretical_loss": 3.6263471798095286,
      "tokens_seen": 1162870784
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006542627883650954,
      "loss": 1.5544,
      "theoretical_loss": 3.6263284183451647,
      "tokens_seen": 1162936320
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006542427281845536,
      "loss": 1.5832,
      "theoretical_loss": 3.6263096582340735,
      "tokens_seen": 1163001856
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006542226680040121,
      "loss": 1.4783,
      "theoretical_loss": 3.6262908994760794,
      "tokens_seen": 1163067392
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006542026078234704,
      "loss": 1.4952,
      "theoretical_loss": 3.6262721420710093,
      "tokens_seen": 1163132928
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006541825476429288,
      "loss": 1.4265,
      "theoretical_loss": 3.62625338601869,
      "tokens_seen": 1163198464
    },
    {
      "epoch": 0.35,
      "objective/train/advantage_avg": 0.49299055337905884,
      "objective/train/docs_used": 425912,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.401368498802185,
      "objective/train/original_loss": 1.4013683795928955,
      "objective/train/theoretical_loss": 3.6262346313189475,
      "objective/train/tokens_used": 1183724000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24344678223133087,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505366325378418,
      "objective/train/weighted_lm_loss": 1.4721384048461914,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6262346313189475,
      "tokens_seen": 1163264000
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006541624874623872,
      "loss": 1.4229,
      "theoretical_loss": 3.6262346313189475,
      "tokens_seen": 1163264000
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006541424272818456,
      "loss": 1.4378,
      "theoretical_loss": 3.626215877971607,
      "tokens_seen": 1163329536
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006541223671013039,
      "loss": 1.5122,
      "theoretical_loss": 3.6261971259764962,
      "tokens_seen": 1163395072
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006541023069207623,
      "loss": 1.4382,
      "theoretical_loss": 3.626178375333441,
      "tokens_seen": 1163460608
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006540822467402207,
      "loss": 1.4533,
      "theoretical_loss": 3.6261596260422673,
      "tokens_seen": 1163526144
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006540621865596791,
      "loss": 1.3055,
      "theoretical_loss": 3.6261408781028015,
      "tokens_seen": 1163591680
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006540421263791374,
      "loss": 1.4861,
      "theoretical_loss": 3.626122131514871,
      "tokens_seen": 1163657216
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006540220661985958,
      "loss": 1.4365,
      "theoretical_loss": 3.6261033862783014,
      "tokens_seen": 1163722752
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006540020060180541,
      "loss": 1.5409,
      "theoretical_loss": 3.626084642392919,
      "tokens_seen": 1163788288
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006539819458375126,
      "loss": 1.4083,
      "theoretical_loss": 3.6260658998585513,
      "tokens_seen": 1163853824
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006539618856569709,
      "loss": 1.544,
      "theoretical_loss": 3.626047158675024,
      "tokens_seen": 1163919360
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006539418254764293,
      "loss": 1.4691,
      "theoretical_loss": 3.6260284188421643,
      "tokens_seen": 1163984896
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006539217652958877,
      "loss": 1.4341,
      "theoretical_loss": 3.6260096803597976,
      "tokens_seen": 1164050432
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006539017051153461,
      "loss": 1.5089,
      "theoretical_loss": 3.625990943227752,
      "tokens_seen": 1164115968
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006538816449348044,
      "loss": 1.3364,
      "theoretical_loss": 3.625972207445854,
      "tokens_seen": 1164181504
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006538615847542629,
      "loss": 1.409,
      "theoretical_loss": 3.6259534730139293,
      "tokens_seen": 1164247040
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006538415245737211,
      "loss": 1.4549,
      "theoretical_loss": 3.6259347399318056,
      "tokens_seen": 1164312576
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006538214643931796,
      "loss": 1.4394,
      "theoretical_loss": 3.625916008199309,
      "tokens_seen": 1164378112
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006538014042126379,
      "loss": 1.3085,
      "theoretical_loss": 3.625897277816267,
      "tokens_seen": 1164443648
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006537813440320963,
      "loss": 1.4518,
      "theoretical_loss": 3.625878548782506,
      "tokens_seen": 1164509184
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006537612838515547,
      "loss": 1.4472,
      "theoretical_loss": 3.625859821097853,
      "tokens_seen": 1164574720
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000653741223671013,
      "loss": 1.5468,
      "theoretical_loss": 3.625841094762135,
      "tokens_seen": 1164640256
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006537211634904714,
      "loss": 1.3583,
      "theoretical_loss": 3.625822369775179,
      "tokens_seen": 1164705792
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006537011033099299,
      "loss": 1.4996,
      "theoretical_loss": 3.625803646136811,
      "tokens_seen": 1164771328
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006536810431293881,
      "loss": 1.5429,
      "theoretical_loss": 3.6257849238468594,
      "tokens_seen": 1164836864
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006536609829488466,
      "loss": 1.4459,
      "theoretical_loss": 3.6257662029051505,
      "tokens_seen": 1164902400
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006536409227683048,
      "loss": 1.5002,
      "theoretical_loss": 3.6257474833115113,
      "tokens_seen": 1164967936
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006536208625877633,
      "loss": 1.507,
      "theoretical_loss": 3.6257287650657695,
      "tokens_seen": 1165033472
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006536008024072217,
      "loss": 1.4534,
      "theoretical_loss": 3.625710048167752,
      "tokens_seen": 1165099008
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00065358074222668,
      "loss": 1.4347,
      "theoretical_loss": 3.6256913326172855,
      "tokens_seen": 1165164544
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006535606820461384,
      "loss": 1.5644,
      "theoretical_loss": 3.625672618414198,
      "tokens_seen": 1165230080
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006535406218655969,
      "loss": 1.5198,
      "theoretical_loss": 3.6256539055583152,
      "tokens_seen": 1165295616
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006535205616850551,
      "loss": 1.4266,
      "theoretical_loss": 3.625635194049466,
      "tokens_seen": 1165361152
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006535005015045136,
      "loss": 1.4222,
      "theoretical_loss": 3.6256164838874767,
      "tokens_seen": 1165426688
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006534804413239718,
      "loss": 1.4758,
      "theoretical_loss": 3.6255977750721753,
      "tokens_seen": 1165492224
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006534603811434303,
      "loss": 1.5499,
      "theoretical_loss": 3.625579067603389,
      "tokens_seen": 1165557760
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006534403209628887,
      "loss": 1.4731,
      "theoretical_loss": 3.625560361480945,
      "tokens_seen": 1165623296
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000653420260782347,
      "loss": 1.5933,
      "theoretical_loss": 3.6255416567046703,
      "tokens_seen": 1165688832
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006534002006018054,
      "loss": 1.4935,
      "theoretical_loss": 3.6255229532743933,
      "tokens_seen": 1165754368
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006533801404212638,
      "loss": 1.3673,
      "theoretical_loss": 3.625504251189941,
      "tokens_seen": 1165819904
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006533600802407221,
      "loss": 1.4222,
      "theoretical_loss": 3.625485550451141,
      "tokens_seen": 1165885440
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006533400200601806,
      "loss": 1.5046,
      "theoretical_loss": 3.62546685105782,
      "tokens_seen": 1165950976
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006533199598796389,
      "loss": 1.5415,
      "theoretical_loss": 3.625448153009807,
      "tokens_seen": 1166016512
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006532998996990973,
      "loss": 1.4154,
      "theoretical_loss": 3.6254294563069287,
      "tokens_seen": 1166082048
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006532798395185557,
      "loss": 1.3942,
      "theoretical_loss": 3.6254107609490127,
      "tokens_seen": 1166147584
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000653259779338014,
      "loss": 1.4405,
      "theoretical_loss": 3.6253920669358872,
      "tokens_seen": 1166213120
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006532397191574724,
      "loss": 1.4019,
      "theoretical_loss": 3.6253733742673795,
      "tokens_seen": 1166278656
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006532196589769308,
      "loss": 1.5028,
      "theoretical_loss": 3.6253546829433176,
      "tokens_seen": 1166344192
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006531995987963891,
      "loss": 1.5961,
      "theoretical_loss": 3.625335992963529,
      "tokens_seen": 1166409728
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006531795386158476,
      "loss": 1.4867,
      "theoretical_loss": 3.625317304327842,
      "tokens_seen": 1166475264
    },
    {
      "epoch": 0.35,
      "objective/train/advantage_avg": 0.49300751090049744,
      "objective/train/docs_used": 427016,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.46001398563385,
      "objective/train/original_loss": 1.4600138664245605,
      "objective/train/theoretical_loss": 3.625298617036084,
      "objective/train/tokens_used": 1187000800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24344559013843536,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505383014678955,
      "objective/train/weighted_lm_loss": 1.5339558124542236,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.625298617036084,
      "tokens_seen": 1166540800
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006531594784353059,
      "loss": 1.5183,
      "theoretical_loss": 3.625298617036084,
      "tokens_seen": 1166540800
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006531394182547643,
      "loss": 1.4337,
      "theoretical_loss": 3.625279931088083,
      "tokens_seen": 1166606336
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006531193580742227,
      "loss": 1.5001,
      "theoretical_loss": 3.6252612464836664,
      "tokens_seen": 1166671872
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000653099297893681,
      "loss": 1.5819,
      "theoretical_loss": 3.625242563222663,
      "tokens_seen": 1166737408
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006530792377131395,
      "loss": 1.4459,
      "theoretical_loss": 3.6252238813049,
      "tokens_seen": 1166802944
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006530591775325978,
      "loss": 1.4767,
      "theoretical_loss": 3.6252052007302056,
      "tokens_seen": 1166868480
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006530391173520562,
      "loss": 1.4498,
      "theoretical_loss": 3.6251865214984083,
      "tokens_seen": 1166934016
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006530190571715145,
      "loss": 1.5213,
      "theoretical_loss": 3.6251678436093355,
      "tokens_seen": 1166999552
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000652998996990973,
      "loss": 1.5505,
      "theoretical_loss": 3.6251491670628155,
      "tokens_seen": 1167065088
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006529789368104313,
      "loss": 1.5106,
      "theoretical_loss": 3.625130491858677,
      "tokens_seen": 1167130624
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006529588766298897,
      "loss": 1.5438,
      "theoretical_loss": 3.625111817996747,
      "tokens_seen": 1167196160
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006529388164493481,
      "loss": 1.4071,
      "theoretical_loss": 3.6250931454768542,
      "tokens_seen": 1167261696
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006529187562688065,
      "loss": 1.3664,
      "theoretical_loss": 3.6250744742988275,
      "tokens_seen": 1167327232
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006528986960882648,
      "loss": 1.4783,
      "theoretical_loss": 3.6250558044624945,
      "tokens_seen": 1167392768
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006528786359077233,
      "loss": 1.4805,
      "theoretical_loss": 3.625037135967683,
      "tokens_seen": 1167458304
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006528585757271815,
      "loss": 1.5128,
      "theoretical_loss": 3.625018468814223,
      "tokens_seen": 1167523840
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00065283851554664,
      "loss": 1.509,
      "theoretical_loss": 3.6249998030019404,
      "tokens_seen": 1167589376
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006528184553660983,
      "loss": 1.4918,
      "theoretical_loss": 3.624981138530665,
      "tokens_seen": 1167654912
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006527983951855567,
      "loss": 1.4182,
      "theoretical_loss": 3.624962475400226,
      "tokens_seen": 1167720448
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006527783350050151,
      "loss": 1.4212,
      "theoretical_loss": 3.62494381361045,
      "tokens_seen": 1167785984
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006527582748244734,
      "loss": 1.4779,
      "theoretical_loss": 3.6249251531611666,
      "tokens_seen": 1167851520
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006527382146439318,
      "loss": 1.3904,
      "theoretical_loss": 3.6249064940522038,
      "tokens_seen": 1167917056
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006527181544633903,
      "loss": 1.4717,
      "theoretical_loss": 3.6248878362833907,
      "tokens_seen": 1167982592
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006526980942828485,
      "loss": 1.4776,
      "theoretical_loss": 3.6248691798545556,
      "tokens_seen": 1168048128
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000652678034102307,
      "loss": 1.5353,
      "theoretical_loss": 3.6248505247655265,
      "tokens_seen": 1168113664
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006526579739217652,
      "loss": 1.5211,
      "theoretical_loss": 3.624831871016133,
      "tokens_seen": 1168179200
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006526379137412237,
      "loss": 1.4767,
      "theoretical_loss": 3.624813218606203,
      "tokens_seen": 1168244736
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006526178535606821,
      "loss": 1.4488,
      "theoretical_loss": 3.6247945675355657,
      "tokens_seen": 1168310272
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006525977933801404,
      "loss": 1.5303,
      "theoretical_loss": 3.6247759178040493,
      "tokens_seen": 1168375808
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006525777331995988,
      "loss": 1.4255,
      "theoretical_loss": 3.624757269411483,
      "tokens_seen": 1168441344
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006525576730190573,
      "loss": 1.3499,
      "theoretical_loss": 3.624738622357696,
      "tokens_seen": 1168506880
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006525376128385155,
      "loss": 1.5218,
      "theoretical_loss": 3.6247199766425156,
      "tokens_seen": 1168572416
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000652517552657974,
      "loss": 1.474,
      "theoretical_loss": 3.624701332265772,
      "tokens_seen": 1168637952
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006524974924774322,
      "loss": 1.4349,
      "theoretical_loss": 3.6246826892272934,
      "tokens_seen": 1168703488
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006524774322968907,
      "loss": 1.4113,
      "theoretical_loss": 3.624664047526909,
      "tokens_seen": 1168769024
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006524573721163491,
      "loss": 1.3558,
      "theoretical_loss": 3.624645407164448,
      "tokens_seen": 1168834560
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006524373119358074,
      "loss": 1.5626,
      "theoretical_loss": 3.6246267681397386,
      "tokens_seen": 1168900096
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006524172517552658,
      "loss": 1.439,
      "theoretical_loss": 3.6246081304526108,
      "tokens_seen": 1168965632
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006523971915747242,
      "loss": 1.4071,
      "theoretical_loss": 3.6245894941028927,
      "tokens_seen": 1169031168
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006523771313941825,
      "loss": 1.6112,
      "theoretical_loss": 3.6245708590904138,
      "tokens_seen": 1169096704
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000652357071213641,
      "loss": 1.3858,
      "theoretical_loss": 3.624552225415003,
      "tokens_seen": 1169162240
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006523370110330993,
      "loss": 1.4301,
      "theoretical_loss": 3.6245335930764897,
      "tokens_seen": 1169227776
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006523169508525577,
      "loss": 1.3584,
      "theoretical_loss": 3.624514962074703,
      "tokens_seen": 1169293312
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006522968906720161,
      "loss": 1.4386,
      "theoretical_loss": 3.624496332409472,
      "tokens_seen": 1169358848
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006522768304914744,
      "loss": 1.459,
      "theoretical_loss": 3.6244777040806255,
      "tokens_seen": 1169424384
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006522567703109328,
      "loss": 1.5557,
      "theoretical_loss": 3.6244590770879936,
      "tokens_seen": 1169489920
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006522367101303912,
      "loss": 1.3897,
      "theoretical_loss": 3.6244404514314055,
      "tokens_seen": 1169555456
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006522166499498495,
      "loss": 1.436,
      "theoretical_loss": 3.6244218271106896,
      "tokens_seen": 1169620992
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000652196589769308,
      "loss": 1.5442,
      "theoretical_loss": 3.624403204125676,
      "tokens_seen": 1169686528
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006521765295887663,
      "loss": 1.5727,
      "theoretical_loss": 3.6243845824761936,
      "tokens_seen": 1169752064
    },
    {
      "epoch": 0.35,
      "objective/train/advantage_avg": 0.4896829128265381,
      "objective/train/docs_used": 428177,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4335746765136719,
      "objective/train/original_loss": 1.4335744380950928,
      "objective/train/theoretical_loss": 3.6243659621620727,
      "objective/train/tokens_used": 1190277600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2403804212808609,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501900911331177,
      "objective/train/weighted_lm_loss": 1.5054765939712524,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6243659621620727,
      "tokens_seen": 1169817600
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006521564694082247,
      "loss": 1.5429,
      "theoretical_loss": 3.6243659621620727,
      "tokens_seen": 1169817600
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000652136409227683,
      "loss": 1.5633,
      "theoretical_loss": 3.6243473431831417,
      "tokens_seen": 1169883136
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006521163490471415,
      "loss": 1.4589,
      "theoretical_loss": 3.6243287255392307,
      "tokens_seen": 1169948672
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006520962888665998,
      "loss": 1.3883,
      "theoretical_loss": 3.624310109230169,
      "tokens_seen": 1170014208
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006520762286860582,
      "loss": 1.5458,
      "theoretical_loss": 3.624291494255786,
      "tokens_seen": 1170079744
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006520561685055165,
      "loss": 1.5789,
      "theoretical_loss": 3.6242728806159112,
      "tokens_seen": 1170145280
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006520361083249749,
      "loss": 1.4217,
      "theoretical_loss": 3.624254268310375,
      "tokens_seen": 1170210816
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006520160481444333,
      "loss": 1.4631,
      "theoretical_loss": 3.624235657339006,
      "tokens_seen": 1170276352
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006519959879638917,
      "loss": 1.3499,
      "theoretical_loss": 3.6242170477016344,
      "tokens_seen": 1170341888
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00065197592778335,
      "loss": 1.4178,
      "theoretical_loss": 3.6241984393980897,
      "tokens_seen": 1170407424
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006519558676028085,
      "loss": 1.4832,
      "theoretical_loss": 3.6241798324282017,
      "tokens_seen": 1170472960
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006519358074222668,
      "loss": 1.5003,
      "theoretical_loss": 3.6241612267918004,
      "tokens_seen": 1170538496
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006519157472417252,
      "loss": 1.5587,
      "theoretical_loss": 3.624142622488715,
      "tokens_seen": 1170604032
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006518956870611835,
      "loss": 1.474,
      "theoretical_loss": 3.624124019518776,
      "tokens_seen": 1170669568
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006518756268806419,
      "loss": 1.4749,
      "theoretical_loss": 3.624105417881813,
      "tokens_seen": 1170735104
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006518555667001003,
      "loss": 1.4336,
      "theoretical_loss": 3.6240868175776555,
      "tokens_seen": 1170800640
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006518355065195587,
      "loss": 1.4671,
      "theoretical_loss": 3.6240682186061335,
      "tokens_seen": 1170866176
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000651815446339017,
      "loss": 1.548,
      "theoretical_loss": 3.624049620967077,
      "tokens_seen": 1170931712
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006517953861584755,
      "loss": 1.5433,
      "theoretical_loss": 3.624031024660317,
      "tokens_seen": 1170997248
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006517753259779337,
      "loss": 1.5002,
      "theoretical_loss": 3.6240124296856813,
      "tokens_seen": 1171062784
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006517552657973922,
      "loss": 1.5381,
      "theoretical_loss": 3.6239938360430024,
      "tokens_seen": 1171128320
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006517352056168506,
      "loss": 1.3738,
      "theoretical_loss": 3.6239752437321084,
      "tokens_seen": 1171193856
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006517151454363089,
      "loss": 1.5952,
      "theoretical_loss": 3.6239566527528306,
      "tokens_seen": 1171259392
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006516950852557673,
      "loss": 1.4659,
      "theoretical_loss": 3.6239380631049984,
      "tokens_seen": 1171324928
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0006516750250752256,
      "loss": 1.5169,
      "theoretical_loss": 3.6239194747884427,
      "tokens_seen": 1171390464
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000651654964894684,
      "loss": 1.4045,
      "theoretical_loss": 3.623900887802993,
      "tokens_seen": 1171456000
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006516349047141425,
      "loss": 1.4622,
      "theoretical_loss": 3.6238823021484796,
      "tokens_seen": 1171521536
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006516148445336008,
      "loss": 1.4945,
      "theoretical_loss": 3.6238637178247335,
      "tokens_seen": 1171587072
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006515947843530592,
      "loss": 1.4766,
      "theoretical_loss": 3.6238451348315843,
      "tokens_seen": 1171652608
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006515747241725177,
      "loss": 1.496,
      "theoretical_loss": 3.623826553168862,
      "tokens_seen": 1171718144
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006515546639919759,
      "loss": 1.4892,
      "theoretical_loss": 3.6238079728363974,
      "tokens_seen": 1171783680
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006515346038114344,
      "loss": 1.4471,
      "theoretical_loss": 3.623789393834021,
      "tokens_seen": 1171849216
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006515145436308926,
      "loss": 1.5545,
      "theoretical_loss": 3.623770816161563,
      "tokens_seen": 1171914752
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006514944834503511,
      "loss": 1.3905,
      "theoretical_loss": 3.623752239818854,
      "tokens_seen": 1171980288
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006514744232698095,
      "loss": 1.4343,
      "theoretical_loss": 3.6237336648057243,
      "tokens_seen": 1172045824
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006514543630892678,
      "loss": 1.5455,
      "theoretical_loss": 3.6237150911220044,
      "tokens_seen": 1172111360
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006514343029087262,
      "loss": 1.3146,
      "theoretical_loss": 3.6236965187675247,
      "tokens_seen": 1172176896
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006514142427281846,
      "loss": 1.4248,
      "theoretical_loss": 3.6236779477421166,
      "tokens_seen": 1172242432
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006513941825476429,
      "loss": 1.3787,
      "theoretical_loss": 3.62365937804561,
      "tokens_seen": 1172307968
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006513741223671014,
      "loss": 1.2507,
      "theoretical_loss": 3.623640809677835,
      "tokens_seen": 1172373504
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006513540621865597,
      "loss": 1.4705,
      "theoretical_loss": 3.623622242638623,
      "tokens_seen": 1172439040
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006513340020060181,
      "loss": 1.4212,
      "theoretical_loss": 3.6236036769278046,
      "tokens_seen": 1172504576
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006513139418254765,
      "loss": 1.4627,
      "theoretical_loss": 3.6235851125452108,
      "tokens_seen": 1172570112
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006512938816449348,
      "loss": 1.477,
      "theoretical_loss": 3.623566549490671,
      "tokens_seen": 1172635648
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006512738214643932,
      "loss": 1.3757,
      "theoretical_loss": 3.623547987764018,
      "tokens_seen": 1172701184
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006512537612838516,
      "loss": 1.5668,
      "theoretical_loss": 3.623529427365081,
      "tokens_seen": 1172766720
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006512337011033099,
      "loss": 1.5023,
      "theoretical_loss": 3.6235108682936916,
      "tokens_seen": 1172832256
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006512136409227684,
      "loss": 1.4518,
      "theoretical_loss": 3.623492310549681,
      "tokens_seen": 1172897792
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006511935807422267,
      "loss": 1.5855,
      "theoretical_loss": 3.6234737541328785,
      "tokens_seen": 1172963328
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006511735205616851,
      "loss": 1.493,
      "theoretical_loss": 3.623455199043117,
      "tokens_seen": 1173028864
    },
    {
      "epoch": 0.36,
      "objective/train/advantage_avg": 0.4950557053089142,
      "objective/train/docs_used": 429267,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4121142625808716,
      "objective/train/original_loss": 1.4121142625808716,
      "objective/train/theoretical_loss": 3.623436645280226,
      "objective/train/tokens_used": 1193554400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2452162802219391,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507522821426392,
      "objective/train/weighted_lm_loss": 1.4837337732315063,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.623436645280226,
      "tokens_seen": 1173094400
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006511534603811434,
      "loss": 1.3935,
      "theoretical_loss": 3.623436645280226,
      "tokens_seen": 1173094400
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006511334002006019,
      "loss": 1.4353,
      "theoretical_loss": 3.6234180928440374,
      "tokens_seen": 1173159936
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006511133400200602,
      "loss": 1.5029,
      "theoretical_loss": 3.623399541734382,
      "tokens_seen": 1173225472
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006510932798395186,
      "loss": 1.488,
      "theoretical_loss": 3.6233809919510906,
      "tokens_seen": 1173291008
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006510732196589769,
      "loss": 1.5184,
      "theoretical_loss": 3.6233624434939946,
      "tokens_seen": 1173356544
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006510531594784353,
      "loss": 1.4739,
      "theoretical_loss": 3.6233438963629245,
      "tokens_seen": 1173422080
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006510330992978937,
      "loss": 1.4608,
      "theoretical_loss": 3.623325350557713,
      "tokens_seen": 1173487616
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006510130391173521,
      "loss": 1.395,
      "theoretical_loss": 3.6233068060781894,
      "tokens_seen": 1173553152
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006509929789368104,
      "loss": 1.4104,
      "theoretical_loss": 3.623288262924186,
      "tokens_seen": 1173618688
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006509729187562689,
      "loss": 1.4656,
      "theoretical_loss": 3.6232697210955336,
      "tokens_seen": 1173684224
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006509528585757272,
      "loss": 1.5279,
      "theoretical_loss": 3.6232511805920637,
      "tokens_seen": 1173749760
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006509327983951856,
      "loss": 1.559,
      "theoretical_loss": 3.6232326414136073,
      "tokens_seen": 1173815296
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000650912738214644,
      "loss": 1.5098,
      "theoretical_loss": 3.623214103559997,
      "tokens_seen": 1173880832
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006508926780341023,
      "loss": 1.4839,
      "theoretical_loss": 3.623195567031062,
      "tokens_seen": 1173946368
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006508726178535607,
      "loss": 1.3842,
      "theoretical_loss": 3.6231770318266356,
      "tokens_seen": 1174011904
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006508525576730191,
      "loss": 1.5446,
      "theoretical_loss": 3.623158497946548,
      "tokens_seen": 1174077440
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006508324974924774,
      "loss": 1.486,
      "theoretical_loss": 3.6231399653906315,
      "tokens_seen": 1174142976
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006508124373119359,
      "loss": 1.4063,
      "theoretical_loss": 3.6231214341587172,
      "tokens_seen": 1174208512
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006507923771313941,
      "loss": 1.3683,
      "theoretical_loss": 3.6231029042506364,
      "tokens_seen": 1174274048
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006507723169508526,
      "loss": 1.4338,
      "theoretical_loss": 3.623084375666221,
      "tokens_seen": 1174339584
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000650752256770311,
      "loss": 1.4194,
      "theoretical_loss": 3.623065848405303,
      "tokens_seen": 1174405120
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006507321965897693,
      "loss": 1.3827,
      "theoretical_loss": 3.6230473224677127,
      "tokens_seen": 1174470656
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006507121364092277,
      "loss": 1.4218,
      "theoretical_loss": 3.6230287978532836,
      "tokens_seen": 1174536192
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000650692076228686,
      "loss": 1.5485,
      "theoretical_loss": 3.623010274561845,
      "tokens_seen": 1174601728
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006506720160481444,
      "loss": 1.5305,
      "theoretical_loss": 3.622991752593231,
      "tokens_seen": 1174667264
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006506519558676029,
      "loss": 1.5622,
      "theoretical_loss": 3.622973231947272,
      "tokens_seen": 1174732800
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006506318956870611,
      "loss": 1.5434,
      "theoretical_loss": 3.6229547126238,
      "tokens_seen": 1174798336
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006506118355065196,
      "loss": 1.4352,
      "theoretical_loss": 3.622936194622647,
      "tokens_seen": 1174863872
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000650591775325978,
      "loss": 1.3948,
      "theoretical_loss": 3.6229176779436445,
      "tokens_seen": 1174929408
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006505717151454363,
      "loss": 1.3093,
      "theoretical_loss": 3.6228991625866245,
      "tokens_seen": 1174994944
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006505516549648947,
      "loss": 1.4633,
      "theoretical_loss": 3.622880648551419,
      "tokens_seen": 1175060480
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000650531594784353,
      "loss": 1.5184,
      "theoretical_loss": 3.6228621358378605,
      "tokens_seen": 1175126016
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006505115346038114,
      "loss": 1.4856,
      "theoretical_loss": 3.6228436244457796,
      "tokens_seen": 1175191552
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006504914744232699,
      "loss": 1.3974,
      "theoretical_loss": 3.6228251143750088,
      "tokens_seen": 1175257088
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006504714142427281,
      "loss": 1.4794,
      "theoretical_loss": 3.6228066056253807,
      "tokens_seen": 1175322624
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006504513540621866,
      "loss": 1.5171,
      "theoretical_loss": 3.622788098196727,
      "tokens_seen": 1175388160
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006504312938816449,
      "loss": 1.5134,
      "theoretical_loss": 3.6227695920888796,
      "tokens_seen": 1175453696
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006504112337011033,
      "loss": 1.4876,
      "theoretical_loss": 3.622751087301671,
      "tokens_seen": 1175519232
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006503911735205617,
      "loss": 1.3791,
      "theoretical_loss": 3.6227325838349325,
      "tokens_seen": 1175584768
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00065037111334002,
      "loss": 1.5402,
      "theoretical_loss": 3.622714081688497,
      "tokens_seen": 1175650304
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006503510531594784,
      "loss": 1.3696,
      "theoretical_loss": 3.622695580862197,
      "tokens_seen": 1175715840
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006503309929789369,
      "loss": 1.6237,
      "theoretical_loss": 3.6226770813558637,
      "tokens_seen": 1175781376
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006503109327983951,
      "loss": 1.3049,
      "theoretical_loss": 3.6226585831693305,
      "tokens_seen": 1175846912
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006502908726178536,
      "loss": 1.5053,
      "theoretical_loss": 3.622640086302429,
      "tokens_seen": 1175912448
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006502708124373119,
      "loss": 1.3178,
      "theoretical_loss": 3.6226215907549912,
      "tokens_seen": 1175977984
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006502507522567703,
      "loss": 1.4515,
      "theoretical_loss": 3.6226030965268503,
      "tokens_seen": 1176043520
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006502306920762287,
      "loss": 1.5335,
      "theoretical_loss": 3.622584603617838,
      "tokens_seen": 1176109056
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006502106318956871,
      "loss": 1.5218,
      "theoretical_loss": 3.6225661120277874,
      "tokens_seen": 1176174592
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006501905717151454,
      "loss": 1.3777,
      "theoretical_loss": 3.6225476217565307,
      "tokens_seen": 1176240128
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006501705115346038,
      "loss": 1.4834,
      "theoretical_loss": 3.6225291328038995,
      "tokens_seen": 1176305664
    },
    {
      "epoch": 0.36,
      "objective/train/advantage_avg": 0.49156454205513,
      "objective/train/docs_used": 430645,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.509208083152771,
      "objective/train/original_loss": 1.509208083152771,
      "objective/train/theoretical_loss": 3.6225106451697275,
      "objective/train/tokens_used": 1196831200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24201633036136627,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503867864608765,
      "objective/train/weighted_lm_loss": 1.585160732269287,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6225106451697275,
      "tokens_seen": 1176371200
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006501504513540621,
      "loss": 1.4768,
      "theoretical_loss": 3.6225106451697275,
      "tokens_seen": 1176371200
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006501303911735206,
      "loss": 1.3556,
      "theoretical_loss": 3.622492158853847,
      "tokens_seen": 1176436736
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000650110330992979,
      "loss": 1.5299,
      "theoretical_loss": 3.6224736738560894,
      "tokens_seen": 1176502272
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006500902708124373,
      "loss": 1.5705,
      "theoretical_loss": 3.622455190176289,
      "tokens_seen": 1176567808
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006500702106318957,
      "loss": 1.5538,
      "theoretical_loss": 3.622436707814277,
      "tokens_seen": 1176633344
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006500501504513541,
      "loss": 1.5042,
      "theoretical_loss": 3.622418226769887,
      "tokens_seen": 1176698880
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006500300902708125,
      "loss": 1.5717,
      "theoretical_loss": 3.6223997470429516,
      "tokens_seen": 1176764416
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006500100300902708,
      "loss": 1.4987,
      "theoretical_loss": 3.6223812686333035,
      "tokens_seen": 1176829952
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006499899699097293,
      "loss": 1.4829,
      "theoretical_loss": 3.622362791540775,
      "tokens_seen": 1176895488
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006499699097291876,
      "loss": 1.4061,
      "theoretical_loss": 3.622344315765199,
      "tokens_seen": 1176961024
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000649949849548646,
      "loss": 1.4562,
      "theoretical_loss": 3.6223258413064086,
      "tokens_seen": 1177026560
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006499297893681043,
      "loss": 1.5578,
      "theoretical_loss": 3.6223073681642366,
      "tokens_seen": 1177092096
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006499097291875627,
      "loss": 1.5844,
      "theoretical_loss": 3.6222888963385156,
      "tokens_seen": 1177157632
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006498896690070211,
      "loss": 1.3392,
      "theoretical_loss": 3.6222704258290785,
      "tokens_seen": 1177223168
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006498696088264795,
      "loss": 1.5346,
      "theoretical_loss": 3.6222519566357585,
      "tokens_seen": 1177288704
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006498495486459378,
      "loss": 1.4917,
      "theoretical_loss": 3.622233488758389,
      "tokens_seen": 1177354240
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006498294884653963,
      "loss": 1.3605,
      "theoretical_loss": 3.622215022196802,
      "tokens_seen": 1177419776
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006498094282848545,
      "loss": 1.4367,
      "theoretical_loss": 3.622196556950831,
      "tokens_seen": 1177485312
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000649789368104313,
      "loss": 1.5452,
      "theoretical_loss": 3.6221780930203096,
      "tokens_seen": 1177550848
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006497693079237714,
      "loss": 1.4972,
      "theoretical_loss": 3.6221596304050694,
      "tokens_seen": 1177616384
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006497492477432297,
      "loss": 1.5357,
      "theoretical_loss": 3.6221411691049448,
      "tokens_seen": 1177681920
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006497291875626881,
      "loss": 1.5154,
      "theoretical_loss": 3.6221227091197687,
      "tokens_seen": 1177747456
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006497091273821464,
      "loss": 1.4437,
      "theoretical_loss": 3.6221042504493743,
      "tokens_seen": 1177812992
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006496890672016048,
      "loss": 1.5876,
      "theoretical_loss": 3.6220857930935946,
      "tokens_seen": 1177878528
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006496690070210633,
      "loss": 1.4544,
      "theoretical_loss": 3.6220673370522625,
      "tokens_seen": 1177944064
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006496489468405215,
      "loss": 1.4789,
      "theoretical_loss": 3.622048882325212,
      "tokens_seen": 1178009600
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00064962888665998,
      "loss": 1.5216,
      "theoretical_loss": 3.622030428912276,
      "tokens_seen": 1178075136
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006496088264794384,
      "loss": 1.3782,
      "theoretical_loss": 3.622011976813288,
      "tokens_seen": 1178140672
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006495887662988967,
      "loss": 1.4234,
      "theoretical_loss": 3.6219935260280813,
      "tokens_seen": 1178206208
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006495687061183551,
      "loss": 1.5531,
      "theoretical_loss": 3.621975076556489,
      "tokens_seen": 1178271744
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006495486459378134,
      "loss": 1.3793,
      "theoretical_loss": 3.6219566283983458,
      "tokens_seen": 1178337280
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006495285857572718,
      "loss": 1.6266,
      "theoretical_loss": 3.6219381815534826,
      "tokens_seen": 1178402816
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006495085255767303,
      "loss": 1.4738,
      "theoretical_loss": 3.6219197360217352,
      "tokens_seen": 1178468352
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006494884653961885,
      "loss": 1.5011,
      "theoretical_loss": 3.6219012918029367,
      "tokens_seen": 1178533888
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000649468405215647,
      "loss": 1.5137,
      "theoretical_loss": 3.6218828488969197,
      "tokens_seen": 1178599424
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006494483450351053,
      "loss": 1.4241,
      "theoretical_loss": 3.6218644073035184,
      "tokens_seen": 1178664960
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006494282848545637,
      "loss": 1.5766,
      "theoretical_loss": 3.6218459670225664,
      "tokens_seen": 1178730496
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006494082246740221,
      "loss": 1.4217,
      "theoretical_loss": 3.621827528053897,
      "tokens_seen": 1178796032
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006493881644934805,
      "loss": 1.3874,
      "theoretical_loss": 3.6218090903973446,
      "tokens_seen": 1178861568
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006493681043129388,
      "loss": 1.5425,
      "theoretical_loss": 3.6217906540527425,
      "tokens_seen": 1178927104
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006493480441323973,
      "loss": 1.6023,
      "theoretical_loss": 3.621772219019924,
      "tokens_seen": 1178992640
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006493279839518555,
      "loss": 1.5271,
      "theoretical_loss": 3.6217537852987234,
      "tokens_seen": 1179058176
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000649307923771314,
      "loss": 1.6079,
      "theoretical_loss": 3.621735352888974,
      "tokens_seen": 1179123712
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006492878635907723,
      "loss": 1.8092,
      "theoretical_loss": 3.6217169217905103,
      "tokens_seen": 1179189248
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006492678034102307,
      "loss": 1.5238,
      "theoretical_loss": 3.6216984920031656,
      "tokens_seen": 1179254784
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006492477432296891,
      "loss": 1.5515,
      "theoretical_loss": 3.6216800635267736,
      "tokens_seen": 1179320320
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006492276830491475,
      "loss": 1.5938,
      "theoretical_loss": 3.621661636361169,
      "tokens_seen": 1179385856
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006492076228686058,
      "loss": 1.5427,
      "theoretical_loss": 3.6216432105061855,
      "tokens_seen": 1179451392
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006491875626880642,
      "loss": 1.492,
      "theoretical_loss": 3.6216247859616564,
      "tokens_seen": 1179516928
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006491675025075225,
      "loss": 1.43,
      "theoretical_loss": 3.621606362727416,
      "tokens_seen": 1179582464
    },
    {
      "epoch": 0.36,
      "objective/train/advantage_avg": 0.49336549639701843,
      "objective/train/docs_used": 431781,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4443635940551758,
      "objective/train/original_loss": 1.4443635940551758,
      "objective/train/theoretical_loss": 3.6215879408032987,
      "objective/train/tokens_used": 1200108000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24383047223091125,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050576090812683,
      "objective/train/weighted_lm_loss": 1.5173157453536987,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6215879408032987,
      "tokens_seen": 1179648000
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000649147442326981,
      "loss": 1.6312,
      "theoretical_loss": 3.6215879408032987,
      "tokens_seen": 1179648000
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006491273821464393,
      "loss": 1.4819,
      "theoretical_loss": 3.621569520189138,
      "tokens_seen": 1179713536
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006491073219658977,
      "loss": 1.4999,
      "theoretical_loss": 3.6215511008847687,
      "tokens_seen": 1179779072
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000649087261785356,
      "loss": 1.5858,
      "theoretical_loss": 3.6215326828900247,
      "tokens_seen": 1179844608
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006490672016048145,
      "loss": 1.5173,
      "theoretical_loss": 3.6215142662047395,
      "tokens_seen": 1179910144
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006490471414242728,
      "loss": 1.4483,
      "theoretical_loss": 3.621495850828748,
      "tokens_seen": 1179975680
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006490270812437312,
      "loss": 1.6071,
      "theoretical_loss": 3.621477436761884,
      "tokens_seen": 1180041216
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006490070210631896,
      "loss": 1.5516,
      "theoretical_loss": 3.6214590240039817,
      "tokens_seen": 1180106752
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000648986960882648,
      "loss": 1.4916,
      "theoretical_loss": 3.621440612554876,
      "tokens_seen": 1180172288
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006489669007021063,
      "loss": 1.4828,
      "theoretical_loss": 3.6214222024144007,
      "tokens_seen": 1180237824
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006489468405215647,
      "loss": 1.5132,
      "theoretical_loss": 3.62140379358239,
      "tokens_seen": 1180303360
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000648926780341023,
      "loss": 1.5801,
      "theoretical_loss": 3.6213853860586784,
      "tokens_seen": 1180368896
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006489067201604815,
      "loss": 1.5096,
      "theoretical_loss": 3.6213669798431005,
      "tokens_seen": 1180434432
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006488866599799398,
      "loss": 1.3671,
      "theoretical_loss": 3.6213485749354906,
      "tokens_seen": 1180499968
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006488665997993982,
      "loss": 1.5179,
      "theoretical_loss": 3.621330171335683,
      "tokens_seen": 1180565504
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006488465396188566,
      "loss": 1.4874,
      "theoretical_loss": 3.621311769043512,
      "tokens_seen": 1180631040
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006488264794383149,
      "loss": 1.5699,
      "theoretical_loss": 3.621293368058813,
      "tokens_seen": 1180696576
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006488064192577733,
      "loss": 1.5197,
      "theoretical_loss": 3.6212749683814196,
      "tokens_seen": 1180762112
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006487863590772318,
      "loss": 1.4683,
      "theoretical_loss": 3.621256570011167,
      "tokens_seen": 1180827648
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00064876629889669,
      "loss": 1.5104,
      "theoretical_loss": 3.621238172947889,
      "tokens_seen": 1180893184
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006487462387161485,
      "loss": 1.5831,
      "theoretical_loss": 3.621219777191421,
      "tokens_seen": 1180958720
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006487261785356067,
      "loss": 1.649,
      "theoretical_loss": 3.6212013827415976,
      "tokens_seen": 1181024256
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006487061183550652,
      "loss": 1.4848,
      "theoretical_loss": 3.621182989598253,
      "tokens_seen": 1181089792
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006486860581745236,
      "loss": 1.5882,
      "theoretical_loss": 3.621164597761222,
      "tokens_seen": 1181155328
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006486659979939819,
      "loss": 1.4067,
      "theoretical_loss": 3.62114620723034,
      "tokens_seen": 1181220864
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006486459378134404,
      "loss": 1.589,
      "theoretical_loss": 3.621127818005441,
      "tokens_seen": 1181286400
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006486258776328988,
      "loss": 1.4644,
      "theoretical_loss": 3.6211094300863604,
      "tokens_seen": 1181351936
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006486058174523571,
      "loss": 1.5332,
      "theoretical_loss": 3.6210910434729326,
      "tokens_seen": 1181417472
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006485857572718155,
      "loss": 1.475,
      "theoretical_loss": 3.621072658164993,
      "tokens_seen": 1181483008
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006485656970912738,
      "loss": 1.6027,
      "theoretical_loss": 3.621054274162376,
      "tokens_seen": 1181548544
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006485456369107322,
      "loss": 1.5402,
      "theoretical_loss": 3.6210358914649166,
      "tokens_seen": 1181614080
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006485255767301907,
      "loss": 1.4259,
      "theoretical_loss": 3.6210175100724498,
      "tokens_seen": 1181679616
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006485055165496489,
      "loss": 1.4532,
      "theoretical_loss": 3.6209991299848108,
      "tokens_seen": 1181745152
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006484854563691074,
      "loss": 1.61,
      "theoretical_loss": 3.6209807512018344,
      "tokens_seen": 1181810688
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006484653961885657,
      "loss": 1.4942,
      "theoretical_loss": 3.620962373723356,
      "tokens_seen": 1181876224
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006484453360080241,
      "loss": 1.5637,
      "theoretical_loss": 3.62094399754921,
      "tokens_seen": 1181941760
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006484252758274825,
      "loss": 1.5275,
      "theoretical_loss": 3.620925622679232,
      "tokens_seen": 1182007296
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006484052156469409,
      "loss": 1.5588,
      "theoretical_loss": 3.620907249113257,
      "tokens_seen": 1182072832
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006483851554663992,
      "loss": 1.6195,
      "theoretical_loss": 3.6208888768511205,
      "tokens_seen": 1182138368
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006483650952858577,
      "loss": 1.4283,
      "theoretical_loss": 3.6208705058926567,
      "tokens_seen": 1182203904
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006483450351053159,
      "loss": 1.4748,
      "theoretical_loss": 3.620852136237702,
      "tokens_seen": 1182269440
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006483249749247744,
      "loss": 1.5668,
      "theoretical_loss": 3.620833767886092,
      "tokens_seen": 1182334976
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006483049147442327,
      "loss": 1.4237,
      "theoretical_loss": 3.6208154008376603,
      "tokens_seen": 1182400512
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006482848545636911,
      "loss": 1.4909,
      "theoretical_loss": 3.6207970350922434,
      "tokens_seen": 1182466048
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006482647943831495,
      "loss": 1.5972,
      "theoretical_loss": 3.620778670649676,
      "tokens_seen": 1182531584
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006482447342026079,
      "loss": 1.5307,
      "theoretical_loss": 3.6207603075097943,
      "tokens_seen": 1182597120
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006482246740220662,
      "loss": 1.4447,
      "theoretical_loss": 3.620741945672433,
      "tokens_seen": 1182662656
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006482046138415246,
      "loss": 1.4192,
      "theoretical_loss": 3.620723585137428,
      "tokens_seen": 1182728192
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006481845536609829,
      "loss": 1.4342,
      "theoretical_loss": 3.620705225904614,
      "tokens_seen": 1182793728
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006481644934804414,
      "loss": 1.5002,
      "theoretical_loss": 3.620686867973828,
      "tokens_seen": 1182859264
    },
    {
      "epoch": 0.36,
      "objective/train/advantage_avg": 0.4936176538467407,
      "objective/train/docs_used": 433029,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4094845056533813,
      "objective/train/original_loss": 1.409484624862671,
      "objective/train/theoretical_loss": 3.6206685113449044,
      "objective/train/tokens_used": 1203384800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24416792392730713,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506032705307007,
      "objective/train/weighted_lm_loss": 1.4806407690048218,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6206685113449044,
      "tokens_seen": 1182924800
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006481444332998997,
      "loss": 1.44,
      "theoretical_loss": 3.6206685113449044,
      "tokens_seen": 1182924800
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006481243731193581,
      "loss": 1.4981,
      "theoretical_loss": 3.6206501560176783,
      "tokens_seen": 1182990336
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006481043129388164,
      "loss": 1.5833,
      "theoretical_loss": 3.620631801991987,
      "tokens_seen": 1183055872
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006480842527582749,
      "loss": 1.512,
      "theoretical_loss": 3.6206134492676645,
      "tokens_seen": 1183121408
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006480641925777332,
      "loss": 1.5799,
      "theoretical_loss": 3.620595097844547,
      "tokens_seen": 1183186944
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006480441323971916,
      "loss": 1.4179,
      "theoretical_loss": 3.62057674772247,
      "tokens_seen": 1183252480
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00064802407221665,
      "loss": 1.6387,
      "theoretical_loss": 3.6205583989012697,
      "tokens_seen": 1183318016
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006480040120361084,
      "loss": 1.5511,
      "theoretical_loss": 3.6205400513807815,
      "tokens_seen": 1183383552
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006479839518555667,
      "loss": 1.4314,
      "theoretical_loss": 3.6205217051608414,
      "tokens_seen": 1183449088
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006479638916750251,
      "loss": 1.5018,
      "theoretical_loss": 3.6205033602412846,
      "tokens_seen": 1183514624
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006479438314944834,
      "loss": 1.6663,
      "theoretical_loss": 3.6204850166219478,
      "tokens_seen": 1183580160
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006479237713139419,
      "loss": 1.4523,
      "theoretical_loss": 3.620466674302666,
      "tokens_seen": 1183645696
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006479037111334002,
      "loss": 1.4661,
      "theoretical_loss": 3.620448333283276,
      "tokens_seen": 1183711232
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006478836509528586,
      "loss": 1.4055,
      "theoretical_loss": 3.620429993563613,
      "tokens_seen": 1183776768
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000647863590772317,
      "loss": 1.423,
      "theoretical_loss": 3.6204116551435126,
      "tokens_seen": 1183842304
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006478435305917753,
      "loss": 1.6615,
      "theoretical_loss": 3.620393318022812,
      "tokens_seen": 1183907840
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006478234704112337,
      "loss": 1.4892,
      "theoretical_loss": 3.6203749822013465,
      "tokens_seen": 1183973376
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006478034102306922,
      "loss": 1.434,
      "theoretical_loss": 3.6203566476789524,
      "tokens_seen": 1184038912
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006477833500501504,
      "loss": 1.5426,
      "theoretical_loss": 3.620338314455465,
      "tokens_seen": 1184104448
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006477632898696089,
      "loss": 1.5556,
      "theoretical_loss": 3.620319982530721,
      "tokens_seen": 1184169984
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006477432296890671,
      "loss": 1.4777,
      "theoretical_loss": 3.6203016519045566,
      "tokens_seen": 1184235520
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006477231695085256,
      "loss": 1.5517,
      "theoretical_loss": 3.6202833225768085,
      "tokens_seen": 1184301056
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000647703109327984,
      "loss": 1.4925,
      "theoretical_loss": 3.6202649945473113,
      "tokens_seen": 1184366592
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006476830491474423,
      "loss": 1.5012,
      "theoretical_loss": 3.6202466678159024,
      "tokens_seen": 1184432128
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006476629889669007,
      "loss": 1.4808,
      "theoretical_loss": 3.6202283423824175,
      "tokens_seen": 1184497664
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006476429287863592,
      "loss": 1.4321,
      "theoretical_loss": 3.620210018246694,
      "tokens_seen": 1184563200
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006476228686058174,
      "loss": 1.5352,
      "theoretical_loss": 3.6201916954085664,
      "tokens_seen": 1184628736
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006476028084252759,
      "loss": 1.4575,
      "theoretical_loss": 3.620173373867873,
      "tokens_seen": 1184694272
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006475827482447341,
      "loss": 1.5131,
      "theoretical_loss": 3.6201550536244484,
      "tokens_seen": 1184759808
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006475626880641926,
      "loss": 1.5683,
      "theoretical_loss": 3.6201367346781304,
      "tokens_seen": 1184825344
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000647542627883651,
      "loss": 1.4233,
      "theoretical_loss": 3.620118417028754,
      "tokens_seen": 1184890880
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006475225677031093,
      "loss": 1.5112,
      "theoretical_loss": 3.620100100676157,
      "tokens_seen": 1184956416
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006475025075225677,
      "loss": 1.5664,
      "theoretical_loss": 3.620081785620175,
      "tokens_seen": 1185021952
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006474824473420261,
      "loss": 1.4779,
      "theoretical_loss": 3.620063471860645,
      "tokens_seen": 1185087488
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006474623871614844,
      "loss": 1.4768,
      "theoretical_loss": 3.620045159397403,
      "tokens_seen": 1185153024
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006474423269809429,
      "loss": 1.4806,
      "theoretical_loss": 3.620026848230286,
      "tokens_seen": 1185218560
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006474222668004011,
      "loss": 1.5108,
      "theoretical_loss": 3.620008538359131,
      "tokens_seen": 1185284096
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006474022066198596,
      "loss": 1.5866,
      "theoretical_loss": 3.619990229783774,
      "tokens_seen": 1185349632
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006473821464393179,
      "loss": 1.5107,
      "theoretical_loss": 3.619971922504052,
      "tokens_seen": 1185415168
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006473620862587763,
      "loss": 1.4007,
      "theoretical_loss": 3.619953616519801,
      "tokens_seen": 1185480704
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006473420260782347,
      "loss": 1.5312,
      "theoretical_loss": 3.619935311830859,
      "tokens_seen": 1185546240
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006473219658976931,
      "loss": 1.5189,
      "theoretical_loss": 3.6199170084370613,
      "tokens_seen": 1185611776
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006473019057171514,
      "loss": 1.4024,
      "theoretical_loss": 3.6198987063382457,
      "tokens_seen": 1185677312
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006472818455366099,
      "loss": 1.5199,
      "theoretical_loss": 3.6198804055342486,
      "tokens_seen": 1185742848
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006472617853560682,
      "loss": 1.6621,
      "theoretical_loss": 3.6198621060249065,
      "tokens_seen": 1185808384
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006472417251755266,
      "loss": 1.5044,
      "theoretical_loss": 3.6198438078100574,
      "tokens_seen": 1185873920
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006472216649949849,
      "loss": 1.534,
      "theoretical_loss": 3.6198255108895374,
      "tokens_seen": 1185939456
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006472016048144433,
      "loss": 1.5118,
      "theoretical_loss": 3.619807215263183,
      "tokens_seen": 1186004992
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006471815446339017,
      "loss": 1.6004,
      "theoretical_loss": 3.6197889209308323,
      "tokens_seen": 1186070528
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006471614844533601,
      "loss": 1.542,
      "theoretical_loss": 3.619770627892321,
      "tokens_seen": 1186136064
    },
    {
      "epoch": 0.36,
      "objective/train/advantage_avg": 0.49382704496383667,
      "objective/train/docs_used": 434243,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5245219469070435,
      "objective/train/original_loss": 1.524521827697754,
      "objective/train/theoretical_loss": 3.6197523361474877,
      "objective/train/tokens_used": 1206661600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24418462812900543,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506240129470825,
      "objective/train/weighted_lm_loss": 1.601709246635437,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6197523361474877,
      "tokens_seen": 1186201600
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006471414242728185,
      "loss": 1.5567,
      "theoretical_loss": 3.6197523361474877,
      "tokens_seen": 1186201600
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006471213640922768,
      "loss": 1.4921,
      "theoretical_loss": 3.619734045696168,
      "tokens_seen": 1186267136
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006471013039117353,
      "loss": 1.4222,
      "theoretical_loss": 3.619715756538199,
      "tokens_seen": 1186332672
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006470812437311936,
      "loss": 1.4482,
      "theoretical_loss": 3.6196974686734187,
      "tokens_seen": 1186398208
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000647061183550652,
      "loss": 1.5808,
      "theoretical_loss": 3.619679182101664,
      "tokens_seen": 1186463744
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006470411233701104,
      "loss": 1.5204,
      "theoretical_loss": 3.6196608968227717,
      "tokens_seen": 1186529280
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006470210631895688,
      "loss": 1.5984,
      "theoretical_loss": 3.6196426128365795,
      "tokens_seen": 1186594816
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006470010030090271,
      "loss": 1.6523,
      "theoretical_loss": 3.6196243301429236,
      "tokens_seen": 1186660352
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006469809428284855,
      "loss": 1.5107,
      "theoretical_loss": 3.619606048741643,
      "tokens_seen": 1186725888
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006469608826479438,
      "loss": 1.4732,
      "theoretical_loss": 3.619587768632573,
      "tokens_seen": 1186791424
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006469408224674023,
      "loss": 1.4639,
      "theoretical_loss": 3.6195694898155524,
      "tokens_seen": 1186856960
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006469207622868606,
      "loss": 1.6962,
      "theoretical_loss": 3.6195512122904177,
      "tokens_seen": 1186922496
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000646900702106319,
      "loss": 1.5452,
      "theoretical_loss": 3.6195329360570065,
      "tokens_seen": 1186988032
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006468806419257774,
      "loss": 1.5766,
      "theoretical_loss": 3.619514661115157,
      "tokens_seen": 1187053568
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006468605817452357,
      "loss": 1.5132,
      "theoretical_loss": 3.619496387464705,
      "tokens_seen": 1187119104
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006468405215646941,
      "loss": 1.4587,
      "theoretical_loss": 3.6194781151054896,
      "tokens_seen": 1187184640
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006468204613841526,
      "loss": 1.5521,
      "theoretical_loss": 3.6194598440373467,
      "tokens_seen": 1187250176
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006468004012036108,
      "loss": 1.4978,
      "theoretical_loss": 3.619441574260115,
      "tokens_seen": 1187315712
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006467803410230693,
      "loss": 1.4008,
      "theoretical_loss": 3.6194233057736316,
      "tokens_seen": 1187381248
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006467602808425275,
      "loss": 1.4592,
      "theoretical_loss": 3.6194050385777343,
      "tokens_seen": 1187446784
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000646740220661986,
      "loss": 1.5678,
      "theoretical_loss": 3.619386772672261,
      "tokens_seen": 1187512320
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006467201604814444,
      "loss": 1.5049,
      "theoretical_loss": 3.6193685080570486,
      "tokens_seen": 1187577856
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006467001003009027,
      "loss": 1.6502,
      "theoretical_loss": 3.619350244731935,
      "tokens_seen": 1187643392
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006466800401203611,
      "loss": 1.5546,
      "theoretical_loss": 3.619331982696758,
      "tokens_seen": 1187708928
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006466599799398196,
      "loss": 1.5904,
      "theoretical_loss": 3.6193137219513556,
      "tokens_seen": 1187774464
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006466399197592778,
      "loss": 1.5576,
      "theoretical_loss": 3.6192954624955647,
      "tokens_seen": 1187840000
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006466198595787363,
      "loss": 1.4618,
      "theoretical_loss": 3.619277204329224,
      "tokens_seen": 1187905536
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006465997993981945,
      "loss": 1.6165,
      "theoretical_loss": 3.619258947452171,
      "tokens_seen": 1187971072
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000646579739217653,
      "loss": 1.5933,
      "theoretical_loss": 3.619240691864243,
      "tokens_seen": 1188036608
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006465596790371114,
      "loss": 1.4737,
      "theoretical_loss": 3.619222437565279,
      "tokens_seen": 1188102144
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006465396188565697,
      "loss": 1.3664,
      "theoretical_loss": 3.619204184555116,
      "tokens_seen": 1188167680
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006465195586760281,
      "loss": 1.4469,
      "theoretical_loss": 3.6191859328335916,
      "tokens_seen": 1188233216
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006464994984954865,
      "loss": 1.5291,
      "theoretical_loss": 3.619167682400545,
      "tokens_seen": 1188298752
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006464794383149448,
      "loss": 1.5127,
      "theoretical_loss": 3.6191494332558136,
      "tokens_seen": 1188364288
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006464593781344033,
      "loss": 1.5788,
      "theoretical_loss": 3.619131185399235,
      "tokens_seen": 1188429824
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006464393179538615,
      "loss": 1.5876,
      "theoretical_loss": 3.6191129388306473,
      "tokens_seen": 1188495360
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00064641925777332,
      "loss": 1.6157,
      "theoretical_loss": 3.619094693549889,
      "tokens_seen": 1188560896
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006463991975927783,
      "loss": 1.6413,
      "theoretical_loss": 3.6190764495567986,
      "tokens_seen": 1188626432
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006463791374122367,
      "loss": 1.4741,
      "theoretical_loss": 3.6190582068512134,
      "tokens_seen": 1188691968
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006463590772316951,
      "loss": 1.4967,
      "theoretical_loss": 3.6190399654329717,
      "tokens_seen": 1188757504
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006463390170511535,
      "loss": 1.3679,
      "theoretical_loss": 3.6190217253019124,
      "tokens_seen": 1188823040
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006463189568706118,
      "loss": 1.5164,
      "theoretical_loss": 3.6190034864578724,
      "tokens_seen": 1188888576
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006462988966900703,
      "loss": 1.5284,
      "theoretical_loss": 3.6189852489006915,
      "tokens_seen": 1188954112
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006462788365095286,
      "loss": 1.7073,
      "theoretical_loss": 3.6189670126302067,
      "tokens_seen": 1189019648
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000646258776328987,
      "loss": 1.3753,
      "theoretical_loss": 3.6189487776462568,
      "tokens_seen": 1189085184
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006462387161484453,
      "loss": 1.4355,
      "theoretical_loss": 3.6189305439486805,
      "tokens_seen": 1189150720
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006462186559679037,
      "loss": 1.4003,
      "theoretical_loss": 3.6189123115373154,
      "tokens_seen": 1189216256
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006461985957873621,
      "loss": 1.5198,
      "theoretical_loss": 3.6188940804120007,
      "tokens_seen": 1189281792
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006461785356068205,
      "loss": 1.427,
      "theoretical_loss": 3.6188758505725738,
      "tokens_seen": 1189347328
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006461584754262788,
      "loss": 1.524,
      "theoretical_loss": 3.6188576220188744,
      "tokens_seen": 1189412864
    },
    {
      "epoch": 0.36,
      "objective/train/advantage_avg": 0.49167755246162415,
      "objective/train/docs_used": 435462,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4333511590957642,
      "objective/train/original_loss": 1.4333510398864746,
      "objective/train/theoretical_loss": 3.61883939475074,
      "objective/train/tokens_used": 1209938400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2430894821882248,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504034757614136,
      "objective/train/weighted_lm_loss": 1.5056500434875488,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.61883939475074,
      "tokens_seen": 1189478400
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006461384152457372,
      "loss": 1.4794,
      "theoretical_loss": 3.61883939475074,
      "tokens_seen": 1189478400
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006461183550651956,
      "loss": 1.6335,
      "theoretical_loss": 3.61882116876801,
      "tokens_seen": 1189543936
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000646098294884654,
      "loss": 1.6601,
      "theoretical_loss": 3.618802944070522,
      "tokens_seen": 1189609472
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006460782347041123,
      "loss": 1.5748,
      "theoretical_loss": 3.6187847206581147,
      "tokens_seen": 1189675008
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006460581745235708,
      "loss": 1.6249,
      "theoretical_loss": 3.6187664985306274,
      "tokens_seen": 1189740544
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000646038114343029,
      "loss": 1.5112,
      "theoretical_loss": 3.6187482776878985,
      "tokens_seen": 1189806080
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006460180541624875,
      "loss": 1.3806,
      "theoretical_loss": 3.618730058129766,
      "tokens_seen": 1189871616
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006459979939819458,
      "loss": 1.4935,
      "theoretical_loss": 3.6187118398560694,
      "tokens_seen": 1189937152
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006459779338014042,
      "loss": 1.5455,
      "theoretical_loss": 3.6186936228666475,
      "tokens_seen": 1190002688
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006459578736208626,
      "loss": 1.5292,
      "theoretical_loss": 3.618675407161338,
      "tokens_seen": 1190068224
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000645937813440321,
      "loss": 1.5325,
      "theoretical_loss": 3.61865719273998,
      "tokens_seen": 1190133760
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006459177532597793,
      "loss": 1.5664,
      "theoretical_loss": 3.618638979602414,
      "tokens_seen": 1190199296
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006458976930792378,
      "loss": 1.5835,
      "theoretical_loss": 3.6186207677484763,
      "tokens_seen": 1190264832
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000645877632898696,
      "loss": 1.3853,
      "theoretical_loss": 3.618602557178007,
      "tokens_seen": 1190330368
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006458575727181545,
      "loss": 1.4969,
      "theoretical_loss": 3.6185843478908453,
      "tokens_seen": 1190395904
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006458375125376128,
      "loss": 1.5456,
      "theoretical_loss": 3.6185661398868296,
      "tokens_seen": 1190461440
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006458174523570712,
      "loss": 1.5068,
      "theoretical_loss": 3.6185479331657993,
      "tokens_seen": 1190526976
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006457973921765296,
      "loss": 1.4651,
      "theoretical_loss": 3.6185297277275925,
      "tokens_seen": 1190592512
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006457773319959879,
      "loss": 1.5334,
      "theoretical_loss": 3.618511523572049,
      "tokens_seen": 1190658048
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006457572718154463,
      "loss": 1.6191,
      "theoretical_loss": 3.6184933206990078,
      "tokens_seen": 1190723584
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006457372116349048,
      "loss": 1.4677,
      "theoretical_loss": 3.6184751191083073,
      "tokens_seen": 1190789120
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000645717151454363,
      "loss": 1.4454,
      "theoretical_loss": 3.6184569187997875,
      "tokens_seen": 1190854656
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006456970912738215,
      "loss": 1.445,
      "theoretical_loss": 3.6184387197732875,
      "tokens_seen": 1190920192
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006456770310932799,
      "loss": 1.6188,
      "theoretical_loss": 3.6184205220286447,
      "tokens_seen": 1190985728
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006456569709127382,
      "loss": 1.6334,
      "theoretical_loss": 3.6184023255657007,
      "tokens_seen": 1191051264
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006456369107321967,
      "loss": 1.4545,
      "theoretical_loss": 3.6183841303842934,
      "tokens_seen": 1191116800
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006456168505516549,
      "loss": 1.6609,
      "theoretical_loss": 3.6183659364842624,
      "tokens_seen": 1191182336
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006455967903711134,
      "loss": 1.4817,
      "theoretical_loss": 3.6183477438654466,
      "tokens_seen": 1191247872
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006455767301905718,
      "loss": 1.4693,
      "theoretical_loss": 3.618329552527686,
      "tokens_seen": 1191313408
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006455566700100301,
      "loss": 1.513,
      "theoretical_loss": 3.618311362470819,
      "tokens_seen": 1191378944
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006455366098294885,
      "loss": 1.6087,
      "theoretical_loss": 3.6182931736946857,
      "tokens_seen": 1191444480
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006455165496489469,
      "loss": 1.4781,
      "theoretical_loss": 3.6182749861991246,
      "tokens_seen": 1191510016
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006454964894684052,
      "loss": 1.4452,
      "theoretical_loss": 3.6182567999839765,
      "tokens_seen": 1191575552
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006454764292878637,
      "loss": 1.5443,
      "theoretical_loss": 3.61823861504908,
      "tokens_seen": 1191641088
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006454563691073219,
      "loss": 1.4579,
      "theoretical_loss": 3.618220431394274,
      "tokens_seen": 1191706624
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006454363089267804,
      "loss": 1.3607,
      "theoretical_loss": 3.6182022490193995,
      "tokens_seen": 1191772160
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006454162487462387,
      "loss": 1.4436,
      "theoretical_loss": 3.618184067924295,
      "tokens_seen": 1191837696
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006453961885656971,
      "loss": 1.3312,
      "theoretical_loss": 3.6181658881087997,
      "tokens_seen": 1191903232
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006453761283851555,
      "loss": 1.64,
      "theoretical_loss": 3.618147709572754,
      "tokens_seen": 1191968768
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006453560682046139,
      "loss": 1.5832,
      "theoretical_loss": 3.6181295323159977,
      "tokens_seen": 1192034304
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006453360080240722,
      "loss": 1.483,
      "theoretical_loss": 3.6181113563383693,
      "tokens_seen": 1192099840
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006453159478435307,
      "loss": 1.4684,
      "theoretical_loss": 3.6180931816397095,
      "tokens_seen": 1192165376
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000645295887662989,
      "loss": 1.519,
      "theoretical_loss": 3.618075008219858,
      "tokens_seen": 1192230912
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006452758274824474,
      "loss": 1.4382,
      "theoretical_loss": 3.6180568360786536,
      "tokens_seen": 1192296448
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006452557673019057,
      "loss": 1.6065,
      "theoretical_loss": 3.618038665215937,
      "tokens_seen": 1192361984
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006452357071213641,
      "loss": 1.4441,
      "theoretical_loss": 3.618020495631548,
      "tokens_seen": 1192427520
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006452156469408225,
      "loss": 1.3862,
      "theoretical_loss": 3.6180023273253252,
      "tokens_seen": 1192493056
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006451955867602809,
      "loss": 1.4119,
      "theoretical_loss": 3.6179841602971097,
      "tokens_seen": 1192558592
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006451755265797392,
      "loss": 1.7448,
      "theoretical_loss": 3.6179659945467417,
      "tokens_seen": 1192624128
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006451554663991976,
      "loss": 1.5167,
      "theoretical_loss": 3.6179478300740593,
      "tokens_seen": 1192689664
    },
    {
      "epoch": 0.36,
      "objective/train/advantage_avg": 0.4904041290283203,
      "objective/train/docs_used": 436657,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5171575546264648,
      "objective/train/original_loss": 1.5171575546264648,
      "objective/train/theoretical_loss": 3.6179296668789043,
      "objective/train/tokens_used": 1213215200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24095970392227173,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502653121948242,
      "objective/train/weighted_lm_loss": 1.5933728218078613,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6179296668789043,
      "tokens_seen": 1192755200
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000645135406218656,
      "loss": 1.506,
      "theoretical_loss": 3.6179296668789043,
      "tokens_seen": 1192755200
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006451153460381144,
      "loss": 1.5113,
      "theoretical_loss": 3.617911504961116,
      "tokens_seen": 1192820736
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006450952858575727,
      "loss": 1.5371,
      "theoretical_loss": 3.6178933443205334,
      "tokens_seen": 1192886272
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006450752256770312,
      "loss": 1.4266,
      "theoretical_loss": 3.6178751849569983,
      "tokens_seen": 1192951808
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006450551654964894,
      "loss": 1.5669,
      "theoretical_loss": 3.6178570268703494,
      "tokens_seen": 1193017344
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006450351053159479,
      "loss": 1.4635,
      "theoretical_loss": 3.6178388700604276,
      "tokens_seen": 1193082880
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006450150451354062,
      "loss": 1.4329,
      "theoretical_loss": 3.617820714527072,
      "tokens_seen": 1193148416
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006449949849548646,
      "loss": 1.5335,
      "theoretical_loss": 3.6178025602701247,
      "tokens_seen": 1193213952
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000644974924774323,
      "loss": 1.5788,
      "theoretical_loss": 3.617784407289424,
      "tokens_seen": 1193279488
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006449548645937814,
      "loss": 1.5858,
      "theoretical_loss": 3.61776625558481,
      "tokens_seen": 1193345024
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006449348044132397,
      "loss": 1.4694,
      "theoretical_loss": 3.6177481051561236,
      "tokens_seen": 1193410560
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006449147442326982,
      "loss": 1.4818,
      "theoretical_loss": 3.6177299560032057,
      "tokens_seen": 1193476096
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006448946840521564,
      "loss": 1.5796,
      "theoretical_loss": 3.617711808125896,
      "tokens_seen": 1193541632
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006448746238716149,
      "loss": 1.5872,
      "theoretical_loss": 3.617693661524034,
      "tokens_seen": 1193607168
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006448545636910732,
      "loss": 1.56,
      "theoretical_loss": 3.617675516197461,
      "tokens_seen": 1193672704
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006448345035105316,
      "loss": 1.5487,
      "theoretical_loss": 3.6176573721460175,
      "tokens_seen": 1193738240
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00064481444332999,
      "loss": 1.4693,
      "theoretical_loss": 3.6176392293695434,
      "tokens_seen": 1193803776
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006447943831494483,
      "loss": 1.5701,
      "theoretical_loss": 3.617621087867879,
      "tokens_seen": 1193869312
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006447743229689067,
      "loss": 1.6187,
      "theoretical_loss": 3.617602947640865,
      "tokens_seen": 1193934848
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006447542627883652,
      "loss": 1.4605,
      "theoretical_loss": 3.617584808688342,
      "tokens_seen": 1194000384
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006447342026078234,
      "loss": 1.501,
      "theoretical_loss": 3.6175666710101506,
      "tokens_seen": 1194065920
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006447141424272819,
      "loss": 1.4605,
      "theoretical_loss": 3.6175485346061307,
      "tokens_seen": 1194131456
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006446940822467403,
      "loss": 1.5909,
      "theoretical_loss": 3.6175303994761236,
      "tokens_seen": 1194196992
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006446740220661986,
      "loss": 1.5377,
      "theoretical_loss": 3.6175122656199696,
      "tokens_seen": 1194262528
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000644653961885657,
      "loss": 1.4199,
      "theoretical_loss": 3.6174941330375097,
      "tokens_seen": 1194328064
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006446339017051153,
      "loss": 1.5727,
      "theoretical_loss": 3.6174760017285834,
      "tokens_seen": 1194393600
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006446138415245737,
      "loss": 1.5728,
      "theoretical_loss": 3.6174578716930323,
      "tokens_seen": 1194459136
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006445937813440322,
      "loss": 1.4584,
      "theoretical_loss": 3.6174397429306975,
      "tokens_seen": 1194524672
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006445737211634904,
      "loss": 1.5543,
      "theoretical_loss": 3.617421615441419,
      "tokens_seen": 1194590208
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006445536609829489,
      "loss": 1.5974,
      "theoretical_loss": 3.6174034892250373,
      "tokens_seen": 1194655744
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006445336008024072,
      "loss": 1.5728,
      "theoretical_loss": 3.617385364281394,
      "tokens_seen": 1194721280
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006445135406218656,
      "loss": 1.5313,
      "theoretical_loss": 3.6173672406103297,
      "tokens_seen": 1194786816
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000644493480441324,
      "loss": 1.6705,
      "theoretical_loss": 3.617349118211685,
      "tokens_seen": 1194852352
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006444734202607823,
      "loss": 1.4363,
      "theoretical_loss": 3.6173309970853014,
      "tokens_seen": 1194917888
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006444533600802407,
      "loss": 1.5225,
      "theoretical_loss": 3.6173128772310186,
      "tokens_seen": 1194983424
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006444332998996991,
      "loss": 1.4952,
      "theoretical_loss": 3.617294758648679,
      "tokens_seen": 1195048960
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006444132397191574,
      "loss": 1.5378,
      "theoretical_loss": 3.6172766413381225,
      "tokens_seen": 1195114496
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006443931795386159,
      "loss": 1.3174,
      "theoretical_loss": 3.6172585252991905,
      "tokens_seen": 1195180032
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006443731193580742,
      "loss": 1.3696,
      "theoretical_loss": 3.617240410531724,
      "tokens_seen": 1195245568
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006443530591775326,
      "loss": 1.6448,
      "theoretical_loss": 3.617222297035564,
      "tokens_seen": 1195311104
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000644332998996991,
      "loss": 1.4328,
      "theoretical_loss": 3.617204184810552,
      "tokens_seen": 1195376640
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006443129388164494,
      "loss": 1.5107,
      "theoretical_loss": 3.6171860738565282,
      "tokens_seen": 1195442176
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006442928786359077,
      "loss": 1.5575,
      "theoretical_loss": 3.6171679641733343,
      "tokens_seen": 1195507712
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006442728184553661,
      "loss": 1.6298,
      "theoretical_loss": 3.617149855760812,
      "tokens_seen": 1195573248
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006442527582748244,
      "loss": 1.3782,
      "theoretical_loss": 3.6171317486188013,
      "tokens_seen": 1195638784
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006442326980942829,
      "loss": 1.5836,
      "theoretical_loss": 3.6171136427471446,
      "tokens_seen": 1195704320
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006442126379137412,
      "loss": 1.5733,
      "theoretical_loss": 3.617095538145682,
      "tokens_seen": 1195769856
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006441925777331996,
      "loss": 1.4481,
      "theoretical_loss": 3.617077434814256,
      "tokens_seen": 1195835392
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006441725175526579,
      "loss": 1.5221,
      "theoretical_loss": 3.6170593327527074,
      "tokens_seen": 1195900928
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006441524573721164,
      "loss": 1.4434,
      "theoretical_loss": 3.617041231960877,
      "tokens_seen": 1195966464
    },
    {
      "epoch": 0.36,
      "objective/train/advantage_avg": 0.4944654405117035,
      "objective/train/docs_used": 437781,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.499495506286621,
      "objective/train/original_loss": 1.4994956254959106,
      "objective/train/theoretical_loss": 3.617023132438607,
      "objective/train/tokens_used": 1216492000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24470704793930054,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506905317306519,
      "objective/train/weighted_lm_loss": 1.575422763824463,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.617023132438607,
      "tokens_seen": 1196032000
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006441323971915748,
      "loss": 1.6074,
      "theoretical_loss": 3.617023132438607,
      "tokens_seen": 1196032000
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006441123370110331,
      "loss": 1.507,
      "theoretical_loss": 3.617005034185738,
      "tokens_seen": 1196097536
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006440922768304916,
      "loss": 1.4131,
      "theoretical_loss": 3.616986937202112,
      "tokens_seen": 1196163072
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006440722166499498,
      "loss": 1.5221,
      "theoretical_loss": 3.6169688414875707,
      "tokens_seen": 1196228608
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006440521564694083,
      "loss": 1.4891,
      "theoretical_loss": 3.6169507470419546,
      "tokens_seen": 1196294144
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006440320962888666,
      "loss": 1.6309,
      "theoretical_loss": 3.6169326538651063,
      "tokens_seen": 1196359680
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000644012036108325,
      "loss": 1.48,
      "theoretical_loss": 3.616914561956867,
      "tokens_seen": 1196425216
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006439919759277834,
      "loss": 1.5175,
      "theoretical_loss": 3.6168964713170775,
      "tokens_seen": 1196490752
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006439719157472418,
      "loss": 1.5304,
      "theoretical_loss": 3.6168783819455803,
      "tokens_seen": 1196556288
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006439518555667001,
      "loss": 1.5987,
      "theoretical_loss": 3.6168602938422167,
      "tokens_seen": 1196621824
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006439317953861586,
      "loss": 1.446,
      "theoretical_loss": 3.6168422070068287,
      "tokens_seen": 1196687360
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006439117352056168,
      "loss": 1.5183,
      "theoretical_loss": 3.616824121439257,
      "tokens_seen": 1196752896
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006438916750250753,
      "loss": 1.4299,
      "theoretical_loss": 3.6168060371393445,
      "tokens_seen": 1196818432
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006438716148445336,
      "loss": 1.5751,
      "theoretical_loss": 3.6167879541069325,
      "tokens_seen": 1196883968
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000643851554663992,
      "loss": 1.4409,
      "theoretical_loss": 3.6167698723418624,
      "tokens_seen": 1196949504
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006438314944834504,
      "loss": 1.4636,
      "theoretical_loss": 3.616751791843976,
      "tokens_seen": 1197015040
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006438114343029087,
      "loss": 1.6584,
      "theoretical_loss": 3.616733712613116,
      "tokens_seen": 1197080576
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006437913741223671,
      "loss": 1.495,
      "theoretical_loss": 3.6167156346491236,
      "tokens_seen": 1197146112
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006437713139418256,
      "loss": 1.6384,
      "theoretical_loss": 3.6166975579518406,
      "tokens_seen": 1197211648
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006437512537612838,
      "loss": 1.4675,
      "theoretical_loss": 3.616679482521109,
      "tokens_seen": 1197277184
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006437311935807423,
      "loss": 1.4893,
      "theoretical_loss": 3.6166614083567707,
      "tokens_seen": 1197342720
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006437111334002007,
      "loss": 1.5542,
      "theoretical_loss": 3.616643335458668,
      "tokens_seen": 1197408256
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000643691073219659,
      "loss": 1.4584,
      "theoretical_loss": 3.6166252638266423,
      "tokens_seen": 1197473792
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006436710130391174,
      "loss": 1.4551,
      "theoretical_loss": 3.6166071934605366,
      "tokens_seen": 1197539328
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006436509528585757,
      "loss": 1.5873,
      "theoretical_loss": 3.6165891243601918,
      "tokens_seen": 1197604864
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006436308926780341,
      "loss": 1.4333,
      "theoretical_loss": 3.6165710565254505,
      "tokens_seen": 1197670400
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006436108324974926,
      "loss": 1.6389,
      "theoretical_loss": 3.616552989956155,
      "tokens_seen": 1197735936
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006435907723169508,
      "loss": 1.5523,
      "theoretical_loss": 3.616534924652147,
      "tokens_seen": 1197801472
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006435707121364093,
      "loss": 1.4616,
      "theoretical_loss": 3.6165168606132694,
      "tokens_seen": 1197867008
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006435506519558676,
      "loss": 1.5595,
      "theoretical_loss": 3.6164987978393635,
      "tokens_seen": 1197932544
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000643530591775326,
      "loss": 1.5409,
      "theoretical_loss": 3.616480736330272,
      "tokens_seen": 1197998080
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006435105315947844,
      "loss": 1.468,
      "theoretical_loss": 3.6164626760858374,
      "tokens_seen": 1198063616
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006434904714142427,
      "loss": 1.4936,
      "theoretical_loss": 3.616444617105901,
      "tokens_seen": 1198129152
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006434704112337011,
      "loss": 1.3861,
      "theoretical_loss": 3.616426559390306,
      "tokens_seen": 1198194688
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006434503510531595,
      "loss": 1.4773,
      "theoretical_loss": 3.616408502938895,
      "tokens_seen": 1198260224
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006434302908726178,
      "loss": 1.4408,
      "theoretical_loss": 3.6163904477515096,
      "tokens_seen": 1198325760
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006434102306920763,
      "loss": 1.4745,
      "theoretical_loss": 3.6163723938279926,
      "tokens_seen": 1198391296
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006433901705115346,
      "loss": 1.4381,
      "theoretical_loss": 3.6163543411681855,
      "tokens_seen": 1198456832
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000643370110330993,
      "loss": 1.397,
      "theoretical_loss": 3.616336289771932,
      "tokens_seen": 1198522368
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006433500501504514,
      "loss": 1.526,
      "theoretical_loss": 3.6163182396390745,
      "tokens_seen": 1198587904
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006433299899699098,
      "loss": 1.432,
      "theoretical_loss": 3.6163001907694547,
      "tokens_seen": 1198653440
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006433099297893681,
      "loss": 1.5204,
      "theoretical_loss": 3.6162821431629153,
      "tokens_seen": 1198718976
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006432898696088265,
      "loss": 1.3641,
      "theoretical_loss": 3.6162640968192994,
      "tokens_seen": 1198784512
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006432698094282848,
      "loss": 1.4491,
      "theoretical_loss": 3.6162460517384494,
      "tokens_seen": 1198850048
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006432497492477433,
      "loss": 1.6159,
      "theoretical_loss": 3.6162280079202076,
      "tokens_seen": 1198915584
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006432296890672016,
      "loss": 1.4349,
      "theoretical_loss": 3.616209965364417,
      "tokens_seen": 1198981120
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00064320962888666,
      "loss": 1.4434,
      "theoretical_loss": 3.61619192407092,
      "tokens_seen": 1199046656
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006431895687061183,
      "loss": 1.4685,
      "theoretical_loss": 3.6161738840395596,
      "tokens_seen": 1199112192
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006431695085255768,
      "loss": 1.5381,
      "theoretical_loss": 3.6161558452701783,
      "tokens_seen": 1199177728
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006431494483450351,
      "loss": 1.3535,
      "theoretical_loss": 3.6161378077626187,
      "tokens_seen": 1199243264
    },
    {
      "epoch": 0.36,
      "objective/train/advantage_avg": 0.49352720379829407,
      "objective/train/docs_used": 438488,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5206199884414673,
      "objective/train/original_loss": 1.5206197500228882,
      "objective/train/theoretical_loss": 3.6161197715167237,
      "objective/train/tokens_used": 1219768800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2444010227918625,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505951642990112,
      "objective/train/weighted_lm_loss": 1.597558856010437,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.6161197715167237,
      "tokens_seen": 1199308800
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006431293881644935,
      "loss": 1.542,
      "theoretical_loss": 3.6161197715167237,
      "tokens_seen": 1199308800
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006431093279839518,
      "loss": 1.4586,
      "theoretical_loss": 3.6161017365323365,
      "tokens_seen": 1199374336
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006430892678034102,
      "loss": 1.471,
      "theoretical_loss": 3.6160837028092994,
      "tokens_seen": 1199439872
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006430692076228686,
      "loss": 1.539,
      "theoretical_loss": 3.6160656703474556,
      "tokens_seen": 1199505408
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000643049147442327,
      "loss": 1.4005,
      "theoretical_loss": 3.616047639146648,
      "tokens_seen": 1199570944
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006430290872617853,
      "loss": 1.5239,
      "theoretical_loss": 3.6160296092067195,
      "tokens_seen": 1199636480
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006430090270812438,
      "loss": 1.5526,
      "theoretical_loss": 3.6160115805275126,
      "tokens_seen": 1199702016
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006429889669007021,
      "loss": 1.4378,
      "theoretical_loss": 3.615993553108871,
      "tokens_seen": 1199767552
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006429689067201605,
      "loss": 1.3499,
      "theoretical_loss": 3.6159755269506375,
      "tokens_seen": 1199833088
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006429488465396189,
      "loss": 1.5604,
      "theoretical_loss": 3.6159575020526553,
      "tokens_seen": 1199898624
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006429287863590772,
      "loss": 1.4094,
      "theoretical_loss": 3.6159394784147665,
      "tokens_seen": 1199964160
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006429087261785356,
      "loss": 1.5057,
      "theoretical_loss": 3.6159214560368156,
      "tokens_seen": 1200029696
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000642888665997994,
      "loss": 1.5594,
      "theoretical_loss": 3.6159034349186445,
      "tokens_seen": 1200095232
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006428686058174523,
      "loss": 1.6178,
      "theoretical_loss": 3.615885415060097,
      "tokens_seen": 1200160768
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006428485456369108,
      "loss": 1.5856,
      "theoretical_loss": 3.6158673964610166,
      "tokens_seen": 1200226304
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000642828485456369,
      "loss": 1.4957,
      "theoretical_loss": 3.6158493791212454,
      "tokens_seen": 1200291840
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006428084252758275,
      "loss": 1.5219,
      "theoretical_loss": 3.615831363040628,
      "tokens_seen": 1200357376
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006427883650952859,
      "loss": 1.4635,
      "theoretical_loss": 3.6158133482190062,
      "tokens_seen": 1200422912
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006427683049147442,
      "loss": 1.5105,
      "theoretical_loss": 3.6157953346562244,
      "tokens_seen": 1200488448
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006427482447342026,
      "loss": 1.3856,
      "theoretical_loss": 3.615777322352126,
      "tokens_seen": 1200553984
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000642728184553661,
      "loss": 1.4599,
      "theoretical_loss": 3.615759311306553,
      "tokens_seen": 1200619520
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006427081243731193,
      "loss": 1.3861,
      "theoretical_loss": 3.6157413015193502,
      "tokens_seen": 1200685056
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006426880641925778,
      "loss": 1.5063,
      "theoretical_loss": 3.615723292990361,
      "tokens_seen": 1200750592
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000642668004012036,
      "loss": 1.487,
      "theoretical_loss": 3.6157052857194274,
      "tokens_seen": 1200816128
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006426479438314945,
      "loss": 1.4787,
      "theoretical_loss": 3.6156872797063944,
      "tokens_seen": 1200881664
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000642627883650953,
      "loss": 1.6271,
      "theoretical_loss": 3.615669274951105,
      "tokens_seen": 1200947200
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006426078234704112,
      "loss": 1.6145,
      "theoretical_loss": 3.615651271453402,
      "tokens_seen": 1201012736
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006425877632898697,
      "loss": 1.4352,
      "theoretical_loss": 3.61563326921313,
      "tokens_seen": 1201078272
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000642567703109328,
      "loss": 1.4193,
      "theoretical_loss": 3.6156152682301324,
      "tokens_seen": 1201143808
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006425476429287864,
      "loss": 1.5573,
      "theoretical_loss": 3.615597268504252,
      "tokens_seen": 1201209344
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006425275827482448,
      "loss": 1.4402,
      "theoretical_loss": 3.615579270035333,
      "tokens_seen": 1201274880
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006425075225677031,
      "loss": 1.5324,
      "theoretical_loss": 3.615561272823219,
      "tokens_seen": 1201340416
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006424874623871615,
      "loss": 1.652,
      "theoretical_loss": 3.6155432768677542,
      "tokens_seen": 1201405952
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006424674022066199,
      "loss": 1.4835,
      "theoretical_loss": 3.6155252821687816,
      "tokens_seen": 1201471488
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006424473420260782,
      "loss": 1.5423,
      "theoretical_loss": 3.615507288726145,
      "tokens_seen": 1201537024
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006424272818455367,
      "loss": 1.4723,
      "theoretical_loss": 3.615489296539688,
      "tokens_seen": 1201602560
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000642407221664995,
      "loss": 1.4552,
      "theoretical_loss": 3.615471305609255,
      "tokens_seen": 1201668096
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006423871614844534,
      "loss": 1.5208,
      "theoretical_loss": 3.6154533159346895,
      "tokens_seen": 1201733632
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006423671013039118,
      "loss": 1.5595,
      "theoretical_loss": 3.6154353275158355,
      "tokens_seen": 1201799168
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006423470411233702,
      "loss": 1.5928,
      "theoretical_loss": 3.6154173403525367,
      "tokens_seen": 1201864704
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006423269809428285,
      "loss": 1.3681,
      "theoretical_loss": 3.6153993544446372,
      "tokens_seen": 1201930240
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006423069207622869,
      "loss": 1.4729,
      "theoretical_loss": 3.6153813697919808,
      "tokens_seen": 1201995776
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006422868605817452,
      "loss": 1.5064,
      "theoretical_loss": 3.6153633863944115,
      "tokens_seen": 1202061312
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006422668004012037,
      "loss": 1.4446,
      "theoretical_loss": 3.615345404251773,
      "tokens_seen": 1202126848
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000642246740220662,
      "loss": 1.5682,
      "theoretical_loss": 3.61532742336391,
      "tokens_seen": 1202192384
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006422266800401204,
      "loss": 1.4201,
      "theoretical_loss": 3.6153094437306654,
      "tokens_seen": 1202257920
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006422066198595787,
      "loss": 1.4042,
      "theoretical_loss": 3.615291465351885,
      "tokens_seen": 1202323456
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006421865596790372,
      "loss": 1.483,
      "theoretical_loss": 3.6152734882274116,
      "tokens_seen": 1202388992
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006421664994984955,
      "loss": 1.4998,
      "theoretical_loss": 3.6152555123570895,
      "tokens_seen": 1202454528
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006421464393179539,
      "loss": 1.451,
      "theoretical_loss": 3.6152375377407626,
      "tokens_seen": 1202520064
    },
    {
      "epoch": 0.36,
      "objective/train/advantage_avg": 0.49168139696121216,
      "objective/train/docs_used": 439824,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4770653247833252,
      "objective/train/original_loss": 1.4770652055740356,
      "objective/train/theoretical_loss": 3.6152195643782763,
      "objective/train/tokens_used": 1223045600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24218137562274933,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503991842269897,
      "objective/train/weighted_lm_loss": 1.5514562129974365,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6152195643782763,
      "tokens_seen": 1202585600
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006421263791374122,
      "loss": 1.3848,
      "theoretical_loss": 3.6152195643782763,
      "tokens_seen": 1202585600
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006421063189568706,
      "loss": 1.3998,
      "theoretical_loss": 3.6152015922694734,
      "tokens_seen": 1202651136
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000642086258776329,
      "loss": 1.3883,
      "theoretical_loss": 3.6151836214141992,
      "tokens_seen": 1202716672
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006420661985957874,
      "loss": 1.4646,
      "theoretical_loss": 3.6151656518122968,
      "tokens_seen": 1202782208
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006420461384152457,
      "loss": 1.5173,
      "theoretical_loss": 3.6151476834636123,
      "tokens_seen": 1202847744
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006420260782347042,
      "loss": 1.5851,
      "theoretical_loss": 3.6151297163679885,
      "tokens_seen": 1202913280
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006420060180541625,
      "loss": 1.4853,
      "theoretical_loss": 3.61511175052527,
      "tokens_seen": 1202978816
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006419859578736209,
      "loss": 1.463,
      "theoretical_loss": 3.615093785935301,
      "tokens_seen": 1203044352
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006419658976930793,
      "loss": 1.4729,
      "theoretical_loss": 3.6150758225979267,
      "tokens_seen": 1203109888
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006419458375125376,
      "loss": 1.4312,
      "theoretical_loss": 3.6150578605129917,
      "tokens_seen": 1203175424
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000641925777331996,
      "loss": 1.4723,
      "theoretical_loss": 3.6150398996803395,
      "tokens_seen": 1203240960
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006419057171514544,
      "loss": 1.4958,
      "theoretical_loss": 3.6150219400998145,
      "tokens_seen": 1203306496
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006418856569709127,
      "loss": 1.4294,
      "theoretical_loss": 3.6150039817712623,
      "tokens_seen": 1203372032
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006418655967903712,
      "loss": 1.5904,
      "theoretical_loss": 3.6149860246945265,
      "tokens_seen": 1203437568
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006418455366098294,
      "loss": 1.5098,
      "theoretical_loss": 3.6149680688694525,
      "tokens_seen": 1203503104
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006418254764292879,
      "loss": 1.5495,
      "theoretical_loss": 3.614950114295884,
      "tokens_seen": 1203568640
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006418054162487463,
      "loss": 1.4231,
      "theoretical_loss": 3.614932160973666,
      "tokens_seen": 1203634176
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006417853560682046,
      "loss": 1.5995,
      "theoretical_loss": 3.614914208902644,
      "tokens_seen": 1203699712
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000641765295887663,
      "loss": 1.4664,
      "theoretical_loss": 3.614896258082661,
      "tokens_seen": 1203765248
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006417452357071215,
      "loss": 1.4346,
      "theoretical_loss": 3.614878308513563,
      "tokens_seen": 1203830784
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006417251755265797,
      "loss": 1.455,
      "theoretical_loss": 3.6148603601951947,
      "tokens_seen": 1203896320
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006417051153460382,
      "loss": 1.402,
      "theoretical_loss": 3.6148424131274,
      "tokens_seen": 1203961856
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006416850551654964,
      "loss": 1.5246,
      "theoretical_loss": 3.614824467310025,
      "tokens_seen": 1204027392
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006416649949849549,
      "loss": 1.4132,
      "theoretical_loss": 3.614806522742913,
      "tokens_seen": 1204092928
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006416449348044133,
      "loss": 1.5133,
      "theoretical_loss": 3.61478857942591,
      "tokens_seen": 1204158464
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006416248746238716,
      "loss": 1.5321,
      "theoretical_loss": 3.6147706373588604,
      "tokens_seen": 1204224000
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00064160481444333,
      "loss": 1.6497,
      "theoretical_loss": 3.614752696541609,
      "tokens_seen": 1204289536
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006415847542627884,
      "loss": 1.499,
      "theoretical_loss": 3.614734756974001,
      "tokens_seen": 1204355072
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006415646940822467,
      "loss": 1.5565,
      "theoretical_loss": 3.6147168186558822,
      "tokens_seen": 1204420608
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0006415446339017052,
      "loss": 1.5602,
      "theoretical_loss": 3.6146988815870955,
      "tokens_seen": 1204486144
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006415245737211634,
      "loss": 1.4066,
      "theoretical_loss": 3.614680945767488,
      "tokens_seen": 1204551680
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006415045135406219,
      "loss": 1.47,
      "theoretical_loss": 3.6146630111969036,
      "tokens_seen": 1204617216
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006414844533600802,
      "loss": 1.6484,
      "theoretical_loss": 3.6146450778751875,
      "tokens_seen": 1204682752
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006414643931795386,
      "loss": 1.4951,
      "theoretical_loss": 3.614627145802185,
      "tokens_seen": 1204748288
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000641444332998997,
      "loss": 1.3429,
      "theoretical_loss": 3.614609214977741,
      "tokens_seen": 1204813824
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006414242728184554,
      "loss": 1.6195,
      "theoretical_loss": 3.6145912854017013,
      "tokens_seen": 1204879360
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006414042126379137,
      "loss": 1.6009,
      "theoretical_loss": 3.6145733570739105,
      "tokens_seen": 1204944896
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006413841524573722,
      "loss": 1.4651,
      "theoretical_loss": 3.6145554299942138,
      "tokens_seen": 1205010432
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006413640922768304,
      "loss": 1.4821,
      "theoretical_loss": 3.614537504162457,
      "tokens_seen": 1205075968
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006413440320962889,
      "loss": 1.5958,
      "theoretical_loss": 3.6145195795784844,
      "tokens_seen": 1205141504
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006413239719157472,
      "loss": 1.4179,
      "theoretical_loss": 3.614501656242142,
      "tokens_seen": 1205207040
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006413039117352056,
      "loss": 1.4984,
      "theoretical_loss": 3.614483734153275,
      "tokens_seen": 1205272576
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000641283851554664,
      "loss": 1.5468,
      "theoretical_loss": 3.614465813311729,
      "tokens_seen": 1205338112
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006412637913741224,
      "loss": 1.528,
      "theoretical_loss": 3.6144478937173488,
      "tokens_seen": 1205403648
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006412437311935807,
      "loss": 1.5136,
      "theoretical_loss": 3.6144299753699802,
      "tokens_seen": 1205469184
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006412236710130391,
      "loss": 1.4376,
      "theoretical_loss": 3.6144120582694685,
      "tokens_seen": 1205534720
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006412036108324975,
      "loss": 1.5471,
      "theoretical_loss": 3.6143941424156596,
      "tokens_seen": 1205600256
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006411835506519559,
      "loss": 1.4584,
      "theoretical_loss": 3.614376227808398,
      "tokens_seen": 1205665792
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006411634904714142,
      "loss": 1.4033,
      "theoretical_loss": 3.6143583144475304,
      "tokens_seen": 1205731328
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006411434302908726,
      "loss": 1.579,
      "theoretical_loss": 3.614340402332901,
      "tokens_seen": 1205796864
    },
    {
      "epoch": 0.37,
      "objective/train/advantage_avg": 0.48837679624557495,
      "objective/train/docs_used": 441116,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4279813766479492,
      "objective/train/original_loss": 1.4279812574386597,
      "objective/train/theoretical_loss": 3.614322491464357,
      "objective/train/tokens_used": 1226322400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23996339738368988,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500574111938477,
      "objective/train/weighted_lm_loss": 1.4998936653137207,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.614322491464357,
      "tokens_seen": 1205862400
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000641123370110331,
      "loss": 1.5175,
      "theoretical_loss": 3.614322491464357,
      "tokens_seen": 1205862400
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006411033099297894,
      "loss": 1.6173,
      "theoretical_loss": 3.6143045818417425,
      "tokens_seen": 1205927936
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006410832497492478,
      "loss": 1.6003,
      "theoretical_loss": 3.6142866734649037,
      "tokens_seen": 1205993472
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006410631895687061,
      "loss": 1.552,
      "theoretical_loss": 3.6142687663336863,
      "tokens_seen": 1206059008
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006410431293881646,
      "loss": 1.4723,
      "theoretical_loss": 3.614250860447936,
      "tokens_seen": 1206124544
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006410230692076229,
      "loss": 1.528,
      "theoretical_loss": 3.6142329558074993,
      "tokens_seen": 1206190080
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006410030090270813,
      "loss": 1.5619,
      "theoretical_loss": 3.61421505241222,
      "tokens_seen": 1206255616
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006409829488465397,
      "loss": 1.5177,
      "theoretical_loss": 3.614197150261946,
      "tokens_seen": 1206321152
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000640962888665998,
      "loss": 1.523,
      "theoretical_loss": 3.6141792493565212,
      "tokens_seen": 1206386688
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006409428284854564,
      "loss": 1.5389,
      "theoretical_loss": 3.6141613496957934,
      "tokens_seen": 1206452224
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006409227683049148,
      "loss": 1.4393,
      "theoretical_loss": 3.6141434512796065,
      "tokens_seen": 1206517760
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006409027081243731,
      "loss": 1.3445,
      "theoretical_loss": 3.6141255541078072,
      "tokens_seen": 1206583296
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006408826479438316,
      "loss": 1.4853,
      "theoretical_loss": 3.6141076581802416,
      "tokens_seen": 1206648832
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006408625877632898,
      "loss": 1.6807,
      "theoretical_loss": 3.614089763496756,
      "tokens_seen": 1206714368
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006408425275827483,
      "loss": 1.3544,
      "theoretical_loss": 3.614071870057195,
      "tokens_seen": 1206779904
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006408224674022067,
      "loss": 1.3661,
      "theoretical_loss": 3.6140539778614063,
      "tokens_seen": 1206845440
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000640802407221665,
      "loss": 1.4781,
      "theoretical_loss": 3.614036086909234,
      "tokens_seen": 1206910976
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006407823470411234,
      "loss": 1.5429,
      "theoretical_loss": 3.614018197200526,
      "tokens_seen": 1206976512
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006407622868605819,
      "loss": 1.4531,
      "theoretical_loss": 3.614000308735127,
      "tokens_seen": 1207042048
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006407422266800401,
      "loss": 1.6118,
      "theoretical_loss": 3.6139824215128837,
      "tokens_seen": 1207107584
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006407221664994986,
      "loss": 1.5393,
      "theoretical_loss": 3.6139645355336425,
      "tokens_seen": 1207173120
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006407021063189568,
      "loss": 1.5636,
      "theoretical_loss": 3.6139466507972484,
      "tokens_seen": 1207238656
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006406820461384153,
      "loss": 1.5329,
      "theoretical_loss": 3.613928767303549,
      "tokens_seen": 1207304192
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006406619859578737,
      "loss": 1.5673,
      "theoretical_loss": 3.61391088505239,
      "tokens_seen": 1207369728
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000640641925777332,
      "loss": 1.5888,
      "theoretical_loss": 3.613893004043617,
      "tokens_seen": 1207435264
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006406218655967904,
      "loss": 1.4545,
      "theoretical_loss": 3.613875124277077,
      "tokens_seen": 1207500800
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006406018054162488,
      "loss": 1.4852,
      "theoretical_loss": 3.613857245752616,
      "tokens_seen": 1207566336
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006405817452357071,
      "loss": 1.4585,
      "theoretical_loss": 3.6138393684700802,
      "tokens_seen": 1207631872
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006405616850551656,
      "loss": 1.5588,
      "theoretical_loss": 3.6138214924293166,
      "tokens_seen": 1207697408
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006405416248746238,
      "loss": 1.6121,
      "theoretical_loss": 3.613803617630171,
      "tokens_seen": 1207762944
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006405215646940823,
      "loss": 1.6272,
      "theoretical_loss": 3.613785744072489,
      "tokens_seen": 1207828480
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006405015045135406,
      "loss": 1.5349,
      "theoretical_loss": 3.613767871756118,
      "tokens_seen": 1207894016
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000640481444332999,
      "loss": 1.5345,
      "theoretical_loss": 3.6137500006809056,
      "tokens_seen": 1207959552
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006404613841524574,
      "loss": 1.4999,
      "theoretical_loss": 3.613732130846696,
      "tokens_seen": 1208025088
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006404413239719158,
      "loss": 1.5456,
      "theoretical_loss": 3.6137142622533367,
      "tokens_seen": 1208090624
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006404212637913741,
      "loss": 1.4484,
      "theoretical_loss": 3.6136963949006744,
      "tokens_seen": 1208156160
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006404012036108326,
      "loss": 1.5749,
      "theoretical_loss": 3.6136785287885553,
      "tokens_seen": 1208221696
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006403811434302908,
      "loss": 1.5378,
      "theoretical_loss": 3.6136606639168267,
      "tokens_seen": 1208287232
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006403610832497493,
      "loss": 1.3642,
      "theoretical_loss": 3.613642800285334,
      "tokens_seen": 1208352768
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006403410230692076,
      "loss": 1.5423,
      "theoretical_loss": 3.613624937893925,
      "tokens_seen": 1208418304
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000640320962888666,
      "loss": 1.6298,
      "theoretical_loss": 3.6136070767424457,
      "tokens_seen": 1208483840
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006403009027081244,
      "loss": 1.5196,
      "theoretical_loss": 3.613589216830743,
      "tokens_seen": 1208549376
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006402808425275828,
      "loss": 1.5142,
      "theoretical_loss": 3.613571358158664,
      "tokens_seen": 1208614912
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006402607823470411,
      "loss": 1.5195,
      "theoretical_loss": 3.613553500726055,
      "tokens_seen": 1208680448
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006402407221664995,
      "loss": 1.5663,
      "theoretical_loss": 3.6135356445327624,
      "tokens_seen": 1208745984
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006402206619859579,
      "loss": 1.4601,
      "theoretical_loss": 3.6135177895786335,
      "tokens_seen": 1208811520
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006402006018054163,
      "loss": 1.3509,
      "theoretical_loss": 3.6134999358635156,
      "tokens_seen": 1208877056
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006401805416248746,
      "loss": 1.3729,
      "theoretical_loss": 3.6134820833872547,
      "tokens_seen": 1208942592
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000640160481444333,
      "loss": 1.4886,
      "theoretical_loss": 3.6134642321496977,
      "tokens_seen": 1209008128
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006401404212637913,
      "loss": 1.4552,
      "theoretical_loss": 3.613446382150692,
      "tokens_seen": 1209073664
    },
    {
      "epoch": 0.37,
      "objective/train/advantage_avg": 0.4915153980255127,
      "objective/train/docs_used": 442331,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2658121585845947,
      "objective/train/original_loss": 1.2658121585845947,
      "objective/train/theoretical_loss": 3.613428533390085,
      "objective/train/tokens_used": 1229599200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24226443469524384,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05038321018219,
      "objective/train/weighted_lm_loss": 1.3296332359313965,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.613428533390085,
      "tokens_seen": 1209139200
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006401203610832498,
      "loss": 1.3874,
      "theoretical_loss": 3.613428533390085,
      "tokens_seen": 1209139200
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006401003009027081,
      "loss": 1.4549,
      "theoretical_loss": 3.613410685867722,
      "tokens_seen": 1209204736
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006400802407221665,
      "loss": 1.508,
      "theoretical_loss": 3.613392839583452,
      "tokens_seen": 1209270272
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006400601805416249,
      "loss": 1.5208,
      "theoretical_loss": 3.6133749945371205,
      "tokens_seen": 1209335808
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006400401203610833,
      "loss": 1.4854,
      "theoretical_loss": 3.613357150728575,
      "tokens_seen": 1209401344
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006400200601805416,
      "loss": 1.5525,
      "theoretical_loss": 3.6133393081576632,
      "tokens_seen": 1209466880
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00064,
      "loss": 1.5417,
      "theoretical_loss": 3.6133214668242317,
      "tokens_seen": 1209532416
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006399799398194583,
      "loss": 1.4774,
      "theoretical_loss": 3.6133036267281273,
      "tokens_seen": 1209597952
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006399598796389168,
      "loss": 1.5002,
      "theoretical_loss": 3.6132857878691977,
      "tokens_seen": 1209663488
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006399398194583751,
      "loss": 1.4159,
      "theoretical_loss": 3.6132679502472898,
      "tokens_seen": 1209729024
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006399197592778335,
      "loss": 1.4145,
      "theoretical_loss": 3.613250113862251,
      "tokens_seen": 1209794560
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006398996990972919,
      "loss": 1.4583,
      "theoretical_loss": 3.613232278713928,
      "tokens_seen": 1209860096
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006398796389167502,
      "loss": 1.5437,
      "theoretical_loss": 3.6132144448021686,
      "tokens_seen": 1209925632
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006398595787362086,
      "loss": 1.4418,
      "theoretical_loss": 3.61319661212682,
      "tokens_seen": 1209991168
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006398395185556671,
      "loss": 1.4946,
      "theoretical_loss": 3.61317878068773,
      "tokens_seen": 1210056704
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006398194583751253,
      "loss": 1.4433,
      "theoretical_loss": 3.613160950484745,
      "tokens_seen": 1210122240
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006397993981945838,
      "loss": 1.4923,
      "theoretical_loss": 3.6131431215177123,
      "tokens_seen": 1210187776
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000639779338014042,
      "loss": 1.5519,
      "theoretical_loss": 3.6131252937864806,
      "tokens_seen": 1210253312
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006397592778335005,
      "loss": 1.3748,
      "theoretical_loss": 3.6131074672908965,
      "tokens_seen": 1210318848
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006397392176529589,
      "loss": 1.4549,
      "theoretical_loss": 3.6130896420308076,
      "tokens_seen": 1210384384
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006397191574724172,
      "loss": 1.4701,
      "theoretical_loss": 3.613071818006061,
      "tokens_seen": 1210449920
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006396990972918756,
      "loss": 1.4565,
      "theoretical_loss": 3.6130539952165046,
      "tokens_seen": 1210515456
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006396790371113341,
      "loss": 1.49,
      "theoretical_loss": 3.6130361736619854,
      "tokens_seen": 1210580992
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006396589769307923,
      "loss": 1.4649,
      "theoretical_loss": 3.613018353342352,
      "tokens_seen": 1210646528
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006396389167502508,
      "loss": 1.5783,
      "theoretical_loss": 3.613000534257451,
      "tokens_seen": 1210712064
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006396188565697092,
      "loss": 1.4745,
      "theoretical_loss": 3.6129827164071306,
      "tokens_seen": 1210777600
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006395987963891675,
      "loss": 1.5494,
      "theoretical_loss": 3.6129648997912382,
      "tokens_seen": 1210843136
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000639578736208626,
      "loss": 1.5336,
      "theoretical_loss": 3.6129470844096216,
      "tokens_seen": 1210908672
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006395586760280842,
      "loss": 1.4692,
      "theoretical_loss": 3.6129292702621285,
      "tokens_seen": 1210974208
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006395386158475427,
      "loss": 1.3986,
      "theoretical_loss": 3.6129114573486065,
      "tokens_seen": 1211039744
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000639518555667001,
      "loss": 1.5278,
      "theoretical_loss": 3.6128936456689034,
      "tokens_seen": 1211105280
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006394984954864594,
      "loss": 1.5297,
      "theoretical_loss": 3.6128758352228667,
      "tokens_seen": 1211170816
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006394784353059178,
      "loss": 1.5211,
      "theoretical_loss": 3.612858026010345,
      "tokens_seen": 1211236352
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006394583751253762,
      "loss": 1.4825,
      "theoretical_loss": 3.6128402180311854,
      "tokens_seen": 1211301888
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006394383149448345,
      "loss": 1.4861,
      "theoretical_loss": 3.612822411285236,
      "tokens_seen": 1211367424
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000639418254764293,
      "loss": 1.5912,
      "theoretical_loss": 3.612804605772345,
      "tokens_seen": 1211432960
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006393981945837512,
      "loss": 1.5664,
      "theoretical_loss": 3.6127868014923594,
      "tokens_seen": 1211498496
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006393781344032097,
      "loss": 1.5527,
      "theoretical_loss": 3.612768998445128,
      "tokens_seen": 1211564032
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000639358074222668,
      "loss": 1.4804,
      "theoretical_loss": 3.612751196630499,
      "tokens_seen": 1211629568
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006393380140421264,
      "loss": 1.4762,
      "theoretical_loss": 3.612733396048319,
      "tokens_seen": 1211695104
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006393179538615848,
      "loss": 1.5078,
      "theoretical_loss": 3.612715596698437,
      "tokens_seen": 1211760640
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006392978936810432,
      "loss": 1.5021,
      "theoretical_loss": 3.612697798580702,
      "tokens_seen": 1211826176
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006392778335005015,
      "loss": 1.4559,
      "theoretical_loss": 3.61268000169496,
      "tokens_seen": 1211891712
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006392577733199599,
      "loss": 1.48,
      "theoretical_loss": 3.612662206041061,
      "tokens_seen": 1211957248
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006392377131394183,
      "loss": 1.474,
      "theoretical_loss": 3.6126444116188514,
      "tokens_seen": 1212022784
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006392176529588767,
      "loss": 1.6025,
      "theoretical_loss": 3.6126266184281812,
      "tokens_seen": 1212088320
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000639197592778335,
      "loss": 1.5397,
      "theoretical_loss": 3.612608826468897,
      "tokens_seen": 1212153856
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006391775325977934,
      "loss": 1.3935,
      "theoretical_loss": 3.6125910357408477,
      "tokens_seen": 1212219392
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006391574724172517,
      "loss": 1.6326,
      "theoretical_loss": 3.612573246243882,
      "tokens_seen": 1212284928
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006391374122367102,
      "loss": 1.4975,
      "theoretical_loss": 3.612555457977847,
      "tokens_seen": 1212350464
    },
    {
      "epoch": 0.37,
      "objective/train/advantage_avg": 0.49347949028015137,
      "objective/train/docs_used": 443456,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4462497234344482,
      "objective/train/original_loss": 1.4462496042251587,
      "objective/train/theoretical_loss": 3.6125376709425923,
      "objective/train/tokens_used": 1232876000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24396289885044098,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505881309509277,
      "objective/train/weighted_lm_loss": 1.5192829370498657,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6125376709425923,
      "tokens_seen": 1212416000
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006391173520561685,
      "loss": 1.5428,
      "theoretical_loss": 3.6125376709425923,
      "tokens_seen": 1212416000
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006390972918756269,
      "loss": 1.4604,
      "theoretical_loss": 3.6125198851379654,
      "tokens_seen": 1212481536
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006390772316950853,
      "loss": 1.3559,
      "theoretical_loss": 3.6125021005638143,
      "tokens_seen": 1212547072
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006390571715145437,
      "loss": 1.4453,
      "theoretical_loss": 3.6124843172199883,
      "tokens_seen": 1212612608
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000639037111334002,
      "loss": 1.3611,
      "theoretical_loss": 3.6124665351063356,
      "tokens_seen": 1212678144
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006390170511534605,
      "loss": 1.4355,
      "theoretical_loss": 3.6124487542227044,
      "tokens_seen": 1212743680
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006389969909729187,
      "loss": 1.5464,
      "theoretical_loss": 3.6124309745689427,
      "tokens_seen": 1212809216
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006389769307923772,
      "loss": 1.5067,
      "theoretical_loss": 3.6124131961448995,
      "tokens_seen": 1212874752
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006389568706118355,
      "loss": 1.5837,
      "theoretical_loss": 3.612395418950424,
      "tokens_seen": 1212940288
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006389368104312939,
      "loss": 1.4858,
      "theoretical_loss": 3.612377642985363,
      "tokens_seen": 1213005824
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006389167502507523,
      "loss": 1.5875,
      "theoretical_loss": 3.612359868249567,
      "tokens_seen": 1213071360
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006388966900702106,
      "loss": 1.4645,
      "theoretical_loss": 3.612342094742883,
      "tokens_seen": 1213136896
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000638876629889669,
      "loss": 1.5663,
      "theoretical_loss": 3.6123243224651604,
      "tokens_seen": 1213202432
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006388565697091275,
      "loss": 1.3972,
      "theoretical_loss": 3.612306551416248,
      "tokens_seen": 1213267968
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006388365095285857,
      "loss": 1.4907,
      "theoretical_loss": 3.6122887815959936,
      "tokens_seen": 1213333504
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006388164493480442,
      "loss": 1.5514,
      "theoretical_loss": 3.6122710130042472,
      "tokens_seen": 1213399040
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006387963891675024,
      "loss": 1.4602,
      "theoretical_loss": 3.612253245640856,
      "tokens_seen": 1213464576
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006387763289869609,
      "loss": 1.5116,
      "theoretical_loss": 3.6122354795056704,
      "tokens_seen": 1213530112
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006387562688064193,
      "loss": 1.4621,
      "theoretical_loss": 3.612217714598538,
      "tokens_seen": 1213595648
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006387362086258776,
      "loss": 1.4126,
      "theoretical_loss": 3.6121999509193077,
      "tokens_seen": 1213661184
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000638716148445336,
      "loss": 1.4927,
      "theoretical_loss": 3.6121821884678287,
      "tokens_seen": 1213726720
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006386960882647945,
      "loss": 1.5269,
      "theoretical_loss": 3.6121644272439495,
      "tokens_seen": 1213792256
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006386760280842527,
      "loss": 1.4852,
      "theoretical_loss": 3.6121466672475195,
      "tokens_seen": 1213857792
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006386559679037112,
      "loss": 1.5479,
      "theoretical_loss": 3.6121289084783874,
      "tokens_seen": 1213923328
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006386359077231694,
      "loss": 1.549,
      "theoretical_loss": 3.6121111509364017,
      "tokens_seen": 1213988864
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006386158475426279,
      "loss": 1.5622,
      "theoretical_loss": 3.6120933946214118,
      "tokens_seen": 1214054400
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006385957873620863,
      "loss": 1.6136,
      "theoretical_loss": 3.612075639533267,
      "tokens_seen": 1214119936
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006385757271815446,
      "loss": 1.474,
      "theoretical_loss": 3.6120578856718155,
      "tokens_seen": 1214185472
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000638555667001003,
      "loss": 1.503,
      "theoretical_loss": 3.6120401330369067,
      "tokens_seen": 1214251008
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006385356068204614,
      "loss": 1.3705,
      "theoretical_loss": 3.61202238162839,
      "tokens_seen": 1214316544
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006385155466399197,
      "loss": 1.529,
      "theoretical_loss": 3.612004631446114,
      "tokens_seen": 1214382080
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006384954864593782,
      "loss": 1.5173,
      "theoretical_loss": 3.611986882489928,
      "tokens_seen": 1214447616
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006384754262788365,
      "loss": 1.4983,
      "theoretical_loss": 3.6119691347596814,
      "tokens_seen": 1214513152
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006384553660982949,
      "loss": 1.5245,
      "theoretical_loss": 3.611951388255223,
      "tokens_seen": 1214578688
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006384353059177532,
      "loss": 1.5477,
      "theoretical_loss": 3.6119336429764024,
      "tokens_seen": 1214644224
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006384152457372116,
      "loss": 1.5833,
      "theoretical_loss": 3.6119158989230686,
      "tokens_seen": 1214709760
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00063839518555667,
      "loss": 1.4098,
      "theoretical_loss": 3.611898156095071,
      "tokens_seen": 1214775296
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006383751253761284,
      "loss": 1.5674,
      "theoretical_loss": 3.6118804144922585,
      "tokens_seen": 1214840832
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006383550651955867,
      "loss": 1.523,
      "theoretical_loss": 3.6118626741144806,
      "tokens_seen": 1214906368
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006383350050150452,
      "loss": 1.5419,
      "theoretical_loss": 3.611844934961587,
      "tokens_seen": 1214971904
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006383149448345035,
      "loss": 1.4073,
      "theoretical_loss": 3.6118271970334264,
      "tokens_seen": 1215037440
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006382948846539619,
      "loss": 1.3651,
      "theoretical_loss": 3.611809460329849,
      "tokens_seen": 1215102976
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006382748244734202,
      "loss": 1.3845,
      "theoretical_loss": 3.611791724850703,
      "tokens_seen": 1215168512
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006382547642928787,
      "loss": 1.6816,
      "theoretical_loss": 3.6117739905958395,
      "tokens_seen": 1215234048
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000638234704112337,
      "loss": 1.5004,
      "theoretical_loss": 3.6117562575651068,
      "tokens_seen": 1215299584
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006382146439317954,
      "loss": 1.5599,
      "theoretical_loss": 3.6117385257583545,
      "tokens_seen": 1215365120
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006381945837512537,
      "loss": 1.4215,
      "theoretical_loss": 3.611720795175432,
      "tokens_seen": 1215430656
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006381745235707121,
      "loss": 1.4682,
      "theoretical_loss": 3.6117030658161897,
      "tokens_seen": 1215496192
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006381544633901705,
      "loss": 1.5266,
      "theoretical_loss": 3.611685337680477,
      "tokens_seen": 1215561728
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006381344032096289,
      "loss": 1.5474,
      "theoretical_loss": 3.6116676107681425,
      "tokens_seen": 1215627264
    },
    {
      "epoch": 0.37,
      "objective/train/advantage_avg": 0.496288925409317,
      "objective/train/docs_used": 444688,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3462483882904053,
      "objective/train/original_loss": 1.3462481498718262,
      "objective/train/theoretical_loss": 3.6116498850790366,
      "objective/train/tokens_used": 1236152800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2464585304260254,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508819818496704,
      "objective/train/weighted_lm_loss": 1.4147021770477295,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6116498850790366,
      "tokens_seen": 1215692800
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006381143430290873,
      "loss": 1.4614,
      "theoretical_loss": 3.6116498850790366,
      "tokens_seen": 1215692800
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006380942828485457,
      "loss": 1.4183,
      "theoretical_loss": 3.6116321606130084,
      "tokens_seen": 1215758336
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006380742226680041,
      "loss": 1.4873,
      "theoretical_loss": 3.6116144373699086,
      "tokens_seen": 1215823872
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006380541624874624,
      "loss": 1.5919,
      "theoretical_loss": 3.6115967153495863,
      "tokens_seen": 1215889408
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006380341023069209,
      "loss": 1.4939,
      "theoretical_loss": 3.6115789945518912,
      "tokens_seen": 1215954944
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006380140421263791,
      "loss": 1.4924,
      "theoretical_loss": 3.611561274976673,
      "tokens_seen": 1216020480
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006379939819458376,
      "loss": 1.4131,
      "theoretical_loss": 3.611543556623782,
      "tokens_seen": 1216086016
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006379739217652959,
      "loss": 1.4785,
      "theoretical_loss": 3.6115258394930674,
      "tokens_seen": 1216151552
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006379538615847543,
      "loss": 1.4541,
      "theoretical_loss": 3.611508123584379,
      "tokens_seen": 1216217088
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006379338014042127,
      "loss": 1.6499,
      "theoretical_loss": 3.6114904088975672,
      "tokens_seen": 1216282624
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000637913741223671,
      "loss": 1.4095,
      "theoretical_loss": 3.6114726954324814,
      "tokens_seen": 1216348160
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006378936810431294,
      "loss": 1.4369,
      "theoretical_loss": 3.6114549831889717,
      "tokens_seen": 1216413696
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006378736208625879,
      "loss": 1.4489,
      "theoretical_loss": 3.6114372721668886,
      "tokens_seen": 1216479232
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006378535606820461,
      "loss": 1.5974,
      "theoretical_loss": 3.6114195623660814,
      "tokens_seen": 1216544768
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006378335005015046,
      "loss": 1.4779,
      "theoretical_loss": 3.6114018537864,
      "tokens_seen": 1216610304
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006378134403209628,
      "loss": 1.4998,
      "theoretical_loss": 3.611384146427695,
      "tokens_seen": 1216675840
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006377933801404213,
      "loss": 1.56,
      "theoretical_loss": 3.611366440289816,
      "tokens_seen": 1216741376
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006377733199598797,
      "loss": 1.4365,
      "theoretical_loss": 3.611348735372613,
      "tokens_seen": 1216806912
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000637753259779338,
      "loss": 1.4566,
      "theoretical_loss": 3.6113310316759373,
      "tokens_seen": 1216872448
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006377331995987964,
      "loss": 1.409,
      "theoretical_loss": 3.6113133291996373,
      "tokens_seen": 1216937984
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006377131394182549,
      "loss": 1.4593,
      "theoretical_loss": 3.611295627943564,
      "tokens_seen": 1217003520
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006376930792377131,
      "loss": 1.3576,
      "theoretical_loss": 3.611277927907568,
      "tokens_seen": 1217069056
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006376730190571716,
      "loss": 1.4456,
      "theoretical_loss": 3.6112602290914984,
      "tokens_seen": 1217134592
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006376529588766298,
      "loss": 1.4722,
      "theoretical_loss": 3.611242531495207,
      "tokens_seen": 1217200128
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006376328986960883,
      "loss": 1.4786,
      "theoretical_loss": 3.6112248351185423,
      "tokens_seen": 1217265664
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006376128385155467,
      "loss": 1.5234,
      "theoretical_loss": 3.6112071399613557,
      "tokens_seen": 1217331200
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000637592778335005,
      "loss": 1.4056,
      "theoretical_loss": 3.611189446023497,
      "tokens_seen": 1217396736
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006375727181544634,
      "loss": 1.3411,
      "theoretical_loss": 3.6111717533048173,
      "tokens_seen": 1217462272
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006375526579739218,
      "loss": 1.4951,
      "theoretical_loss": 3.6111540618051663,
      "tokens_seen": 1217527808
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006375325977933801,
      "loss": 1.5132,
      "theoretical_loss": 3.6111363715243945,
      "tokens_seen": 1217593344
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006375125376128386,
      "loss": 1.6048,
      "theoretical_loss": 3.6111186824623527,
      "tokens_seen": 1217658880
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006374924774322969,
      "loss": 1.4726,
      "theoretical_loss": 3.611100994618891,
      "tokens_seen": 1217724416
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006374724172517553,
      "loss": 1.4593,
      "theoretical_loss": 3.6110833079938596,
      "tokens_seen": 1217789952
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006374523570712136,
      "loss": 1.5428,
      "theoretical_loss": 3.61106562258711,
      "tokens_seen": 1217855488
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000637432296890672,
      "loss": 1.5968,
      "theoretical_loss": 3.6110479383984915,
      "tokens_seen": 1217921024
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006374122367101304,
      "loss": 1.4472,
      "theoretical_loss": 3.6110302554278553,
      "tokens_seen": 1217986560
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006373921765295888,
      "loss": 1.4787,
      "theoretical_loss": 3.611012573675052,
      "tokens_seen": 1218052096
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006373721163490471,
      "loss": 1.6243,
      "theoretical_loss": 3.610994893139932,
      "tokens_seen": 1218117632
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006373520561685056,
      "loss": 1.433,
      "theoretical_loss": 3.6109772138223466,
      "tokens_seen": 1218183168
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006373319959879639,
      "loss": 1.5123,
      "theoretical_loss": 3.610959535722145,
      "tokens_seen": 1218248704
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006373119358074223,
      "loss": 1.4521,
      "theoretical_loss": 3.6109418588391797,
      "tokens_seen": 1218314240
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006372918756268806,
      "loss": 1.538,
      "theoretical_loss": 3.6109241831733003,
      "tokens_seen": 1218379776
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000637271815446339,
      "loss": 1.5223,
      "theoretical_loss": 3.6109065087243577,
      "tokens_seen": 1218445312
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006372517552657974,
      "loss": 1.4826,
      "theoretical_loss": 3.6108888354922026,
      "tokens_seen": 1218510848
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006372316950852558,
      "loss": 1.4566,
      "theoretical_loss": 3.610871163476686,
      "tokens_seen": 1218576384
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006372116349047141,
      "loss": 1.4557,
      "theoretical_loss": 3.610853492677659,
      "tokens_seen": 1218641920
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006371915747241725,
      "loss": 1.4721,
      "theoretical_loss": 3.610835823094972,
      "tokens_seen": 1218707456
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006371715145436309,
      "loss": 1.5733,
      "theoretical_loss": 3.610818154728476,
      "tokens_seen": 1218772992
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006371514543630893,
      "loss": 1.4978,
      "theoretical_loss": 3.6108004875780217,
      "tokens_seen": 1218838528
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006371313941825476,
      "loss": 1.3696,
      "theoretical_loss": 3.6107828216434603,
      "tokens_seen": 1218904064
    },
    {
      "epoch": 0.37,
      "objective/train/advantage_avg": 0.48686742782592773,
      "objective/train/docs_used": 445814,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5334632396697998,
      "objective/train/original_loss": 1.5334631204605103,
      "objective/train/theoretical_loss": 3.6107651569246424,
      "objective/train/tokens_used": 1239429600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2382122129201889,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0498976707458496,
      "objective/train/weighted_lm_loss": 1.6100496053695679,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.6107651569246424,
      "tokens_seen": 1218969600
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006371113340020061,
      "loss": 1.5414,
      "theoretical_loss": 3.6107651569246424,
      "tokens_seen": 1218969600
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006370912738214644,
      "loss": 1.4315,
      "theoretical_loss": 3.610747493421419,
      "tokens_seen": 1219035136
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006370712136409228,
      "loss": 1.3999,
      "theoretical_loss": 3.610729831133642,
      "tokens_seen": 1219100672
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006370511534603811,
      "loss": 1.4091,
      "theoretical_loss": 3.610712170061162,
      "tokens_seen": 1219166208
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006370310932798395,
      "loss": 1.5903,
      "theoretical_loss": 3.610694510203829,
      "tokens_seen": 1219231744
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006370110330992979,
      "loss": 1.4187,
      "theoretical_loss": 3.6106768515614953,
      "tokens_seen": 1219297280
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006369909729187563,
      "loss": 1.4265,
      "theoretical_loss": 3.610659194134012,
      "tokens_seen": 1219362816
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006369709127382146,
      "loss": 1.544,
      "theoretical_loss": 3.610641537921229,
      "tokens_seen": 1219428352
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006369508525576731,
      "loss": 1.5031,
      "theoretical_loss": 3.610623882922999,
      "tokens_seen": 1219493888
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006369307923771313,
      "loss": 1.5261,
      "theoretical_loss": 3.6106062291391723,
      "tokens_seen": 1219559424
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006369107321965898,
      "loss": 1.4952,
      "theoretical_loss": 3.6105885765696004,
      "tokens_seen": 1219624960
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006368906720160482,
      "loss": 1.4527,
      "theoretical_loss": 3.6105709252141343,
      "tokens_seen": 1219690496
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006368706118355065,
      "loss": 1.4149,
      "theoretical_loss": 3.6105532750726255,
      "tokens_seen": 1219756032
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006368505516549649,
      "loss": 1.4902,
      "theoretical_loss": 3.6105356261449253,
      "tokens_seen": 1219821568
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006368304914744232,
      "loss": 1.4691,
      "theoretical_loss": 3.610517978430885,
      "tokens_seen": 1219887104
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006368104312938816,
      "loss": 1.5404,
      "theoretical_loss": 3.6105003319303557,
      "tokens_seen": 1219952640
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006367903711133401,
      "loss": 1.3823,
      "theoretical_loss": 3.6104826866431887,
      "tokens_seen": 1220018176
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006367703109327983,
      "loss": 1.5048,
      "theoretical_loss": 3.6104650425692366,
      "tokens_seen": 1220083712
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006367502507522568,
      "loss": 1.4589,
      "theoretical_loss": 3.610447399708349,
      "tokens_seen": 1220149248
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006367301905717152,
      "loss": 1.4925,
      "theoretical_loss": 3.610429758060379,
      "tokens_seen": 1220214784
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006367101303911735,
      "loss": 1.4216,
      "theoretical_loss": 3.6104121176251764,
      "tokens_seen": 1220280320
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006366900702106319,
      "loss": 1.4654,
      "theoretical_loss": 3.6103944784025943,
      "tokens_seen": 1220345856
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006366700100300902,
      "loss": 1.5588,
      "theoretical_loss": 3.610376840392483,
      "tokens_seen": 1220411392
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006366499498495486,
      "loss": 1.4058,
      "theoretical_loss": 3.610359203594695,
      "tokens_seen": 1220476928
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006366298896690071,
      "loss": 1.5305,
      "theoretical_loss": 3.6103415680090816,
      "tokens_seen": 1220542464
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006366098294884654,
      "loss": 1.405,
      "theoretical_loss": 3.610323933635494,
      "tokens_seen": 1220608000
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006365897693079238,
      "loss": 1.4798,
      "theoretical_loss": 3.610306300473784,
      "tokens_seen": 1220673536
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006365697091273822,
      "loss": 1.6366,
      "theoretical_loss": 3.610288668523804,
      "tokens_seen": 1220739072
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006365496489468405,
      "loss": 1.5493,
      "theoretical_loss": 3.6102710377854046,
      "tokens_seen": 1220804608
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000636529588766299,
      "loss": 1.4351,
      "theoretical_loss": 3.610253408258438,
      "tokens_seen": 1220870144
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006365095285857573,
      "loss": 1.4863,
      "theoretical_loss": 3.610235779942756,
      "tokens_seen": 1220935680
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006364894684052157,
      "loss": 1.4587,
      "theoretical_loss": 3.6102181528382102,
      "tokens_seen": 1221001216
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000636469408224674,
      "loss": 1.4148,
      "theoretical_loss": 3.610200526944652,
      "tokens_seen": 1221066752
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006364493480441324,
      "loss": 1.3556,
      "theoretical_loss": 3.6101829022619345,
      "tokens_seen": 1221132288
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006364292878635908,
      "loss": 1.507,
      "theoretical_loss": 3.610165278789909,
      "tokens_seen": 1221197824
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006364092276830492,
      "loss": 1.5081,
      "theoretical_loss": 3.6101476565284263,
      "tokens_seen": 1221263360
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006363891675025075,
      "loss": 1.5577,
      "theoretical_loss": 3.610130035477339,
      "tokens_seen": 1221328896
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000636369107321966,
      "loss": 1.5777,
      "theoretical_loss": 3.6101124156365,
      "tokens_seen": 1221394432
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006363490471414243,
      "loss": 1.6385,
      "theoretical_loss": 3.6100947970057593,
      "tokens_seen": 1221459968
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006363289869608827,
      "loss": 1.4161,
      "theoretical_loss": 3.610077179584971,
      "tokens_seen": 1221525504
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000636308926780341,
      "loss": 1.5369,
      "theoretical_loss": 3.6100595633739854,
      "tokens_seen": 1221591040
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006362888665997995,
      "loss": 1.5454,
      "theoretical_loss": 3.610041948372655,
      "tokens_seen": 1221656576
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006362688064192578,
      "loss": 1.4155,
      "theoretical_loss": 3.6100243345808325,
      "tokens_seen": 1221722112
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006362487462387162,
      "loss": 1.4635,
      "theoretical_loss": 3.610006721998369,
      "tokens_seen": 1221787648
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006362286860581745,
      "loss": 1.5527,
      "theoretical_loss": 3.6099891106251176,
      "tokens_seen": 1221853184
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006362086258776329,
      "loss": 1.5255,
      "theoretical_loss": 3.6099715004609294,
      "tokens_seen": 1221918720
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006361885656970913,
      "loss": 1.4761,
      "theoretical_loss": 3.6099538915056577,
      "tokens_seen": 1221984256
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006361685055165497,
      "loss": 1.4583,
      "theoretical_loss": 3.6099362837591533,
      "tokens_seen": 1222049792
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000636148445336008,
      "loss": 1.6776,
      "theoretical_loss": 3.6099186772212697,
      "tokens_seen": 1222115328
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006361283851554665,
      "loss": 1.4902,
      "theoretical_loss": 3.6099010718918585,
      "tokens_seen": 1222180864
    },
    {
      "epoch": 0.37,
      "objective/train/advantage_avg": 0.4875772297382355,
      "objective/train/docs_used": 446976,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4010878801345825,
      "objective/train/original_loss": 1.4010876417160034,
      "objective/train/theoretical_loss": 3.6098834677707723,
      "objective/train/tokens_used": 1242706400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2399894744157791,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0499775409698486,
      "objective/train/weighted_lm_loss": 1.4715443849563599,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6098834677707723,
      "tokens_seen": 1222246400
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006361083249749248,
      "loss": 1.4545,
      "theoretical_loss": 3.6098834677707723,
      "tokens_seen": 1222246400
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006360882647943832,
      "loss": 1.4968,
      "theoretical_loss": 3.6098658648578628,
      "tokens_seen": 1222311936
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006360682046138415,
      "loss": 1.3861,
      "theoretical_loss": 3.6098482631529825,
      "tokens_seen": 1222377472
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006360481444332999,
      "loss": 1.436,
      "theoretical_loss": 3.609830662655984,
      "tokens_seen": 1222443008
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006360280842527583,
      "loss": 1.4834,
      "theoretical_loss": 3.60981306336672,
      "tokens_seen": 1222508544
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006360080240722167,
      "loss": 1.5654,
      "theoretical_loss": 3.6097954652850417,
      "tokens_seen": 1222574080
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000635987963891675,
      "loss": 1.4835,
      "theoretical_loss": 3.6097778684108026,
      "tokens_seen": 1222639616
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006359679037111335,
      "loss": 1.4731,
      "theoretical_loss": 3.6097602727438547,
      "tokens_seen": 1222705152
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006359478435305917,
      "loss": 1.3735,
      "theoretical_loss": 3.609742678284051,
      "tokens_seen": 1222770688
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006359277833500502,
      "loss": 1.5609,
      "theoretical_loss": 3.609725085031243,
      "tokens_seen": 1222836224
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006359077231695086,
      "loss": 1.4821,
      "theoretical_loss": 3.609707492985284,
      "tokens_seen": 1222901760
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006358876629889669,
      "loss": 1.4125,
      "theoretical_loss": 3.6096899021460263,
      "tokens_seen": 1222967296
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006358676028084253,
      "loss": 1.4551,
      "theoretical_loss": 3.6096723125133225,
      "tokens_seen": 1223032832
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006358475426278836,
      "loss": 1.4492,
      "theoretical_loss": 3.609654724087025,
      "tokens_seen": 1223098368
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000635827482447342,
      "loss": 1.5368,
      "theoretical_loss": 3.609637136866987,
      "tokens_seen": 1223163904
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006358074222668005,
      "loss": 1.5107,
      "theoretical_loss": 3.6096195508530604,
      "tokens_seen": 1223229440
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006357873620862587,
      "loss": 1.4318,
      "theoretical_loss": 3.6096019660450986,
      "tokens_seen": 1223294976
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006357673019057172,
      "loss": 1.3526,
      "theoretical_loss": 3.6095843824429537,
      "tokens_seen": 1223360512
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006357472417251756,
      "loss": 1.4268,
      "theoretical_loss": 3.609566800046478,
      "tokens_seen": 1223426048
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006357271815446339,
      "loss": 1.5602,
      "theoretical_loss": 3.609549218855526,
      "tokens_seen": 1223491584
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006357071213640923,
      "loss": 1.4812,
      "theoretical_loss": 3.6095316388699485,
      "tokens_seen": 1223557120
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006356870611835506,
      "loss": 1.357,
      "theoretical_loss": 3.6095140600896,
      "tokens_seen": 1223622656
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000635667001003009,
      "loss": 1.4327,
      "theoretical_loss": 3.609496482514332,
      "tokens_seen": 1223688192
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006356469408224675,
      "loss": 1.5766,
      "theoretical_loss": 3.6094789061439982,
      "tokens_seen": 1223753728
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006356268806419257,
      "loss": 1.5137,
      "theoretical_loss": 3.609461330978451,
      "tokens_seen": 1223819264
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006356068204613842,
      "loss": 1.4779,
      "theoretical_loss": 3.609443757017543,
      "tokens_seen": 1223884800
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006355867602808425,
      "loss": 1.4392,
      "theoretical_loss": 3.609426184261128,
      "tokens_seen": 1223950336
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006355667001003009,
      "loss": 1.5558,
      "theoretical_loss": 3.6094086127090588,
      "tokens_seen": 1224015872
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006355466399197593,
      "loss": 1.4861,
      "theoretical_loss": 3.609391042361188,
      "tokens_seen": 1224081408
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006355265797392177,
      "loss": 1.4862,
      "theoretical_loss": 3.609373473217368,
      "tokens_seen": 1224146944
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000635506519558676,
      "loss": 1.5309,
      "theoretical_loss": 3.6093559052774538,
      "tokens_seen": 1224212480
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006354864593781344,
      "loss": 1.5066,
      "theoretical_loss": 3.6093383385412965,
      "tokens_seen": 1224278016
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006354663991975927,
      "loss": 1.4171,
      "theoretical_loss": 3.6093207730087498,
      "tokens_seen": 1224343552
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006354463390170512,
      "loss": 1.4584,
      "theoretical_loss": 3.609303208679667,
      "tokens_seen": 1224409088
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006354262788365095,
      "loss": 1.5221,
      "theoretical_loss": 3.6092856455539017,
      "tokens_seen": 1224474624
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006354062186559679,
      "loss": 1.454,
      "theoretical_loss": 3.6092680836313065,
      "tokens_seen": 1224540160
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006353861584754263,
      "loss": 1.5206,
      "theoretical_loss": 3.6092505229117338,
      "tokens_seen": 1224605696
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006353660982948847,
      "loss": 1.5739,
      "theoretical_loss": 3.6092329633950384,
      "tokens_seen": 1224671232
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000635346038114343,
      "loss": 1.5792,
      "theoretical_loss": 3.6092154050810725,
      "tokens_seen": 1224736768
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006353259779338014,
      "loss": 1.5401,
      "theoretical_loss": 3.60919784796969,
      "tokens_seen": 1224802304
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006353059177532597,
      "loss": 1.4517,
      "theoretical_loss": 3.6091802920607434,
      "tokens_seen": 1224867840
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006352858575727182,
      "loss": 1.399,
      "theoretical_loss": 3.6091627373540867,
      "tokens_seen": 1224933376
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006352657973921765,
      "loss": 1.6073,
      "theoretical_loss": 3.6091451838495727,
      "tokens_seen": 1224998912
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006352457372116349,
      "loss": 1.4098,
      "theoretical_loss": 3.6091276315470555,
      "tokens_seen": 1225064448
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006352256770310932,
      "loss": 1.4736,
      "theoretical_loss": 3.609110080446388,
      "tokens_seen": 1225129984
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006352056168505517,
      "loss": 1.5497,
      "theoretical_loss": 3.6090925305474237,
      "tokens_seen": 1225195520
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00063518555667001,
      "loss": 1.4537,
      "theoretical_loss": 3.609074981850016,
      "tokens_seen": 1225261056
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006351654964894684,
      "loss": 1.5602,
      "theoretical_loss": 3.6090574343540185,
      "tokens_seen": 1225326592
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006351454363089268,
      "loss": 1.5507,
      "theoretical_loss": 3.609039888059285,
      "tokens_seen": 1225392128
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006351253761283852,
      "loss": 1.5809,
      "theoretical_loss": 3.609022342965668,
      "tokens_seen": 1225457664
    },
    {
      "epoch": 0.37,
      "objective/train/advantage_avg": 0.4946627914905548,
      "objective/train/docs_used": 448161,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4292793273925781,
      "objective/train/original_loss": 1.4292793273925781,
      "objective/train/theoretical_loss": 3.6090047990730216,
      "objective/train/tokens_used": 1245983200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24554826319217682,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507146120071411,
      "objective/train/weighted_lm_loss": 1.5018622875213623,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6090047990730216,
      "tokens_seen": 1225523200
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006351053159478436,
      "loss": 1.4582,
      "theoretical_loss": 3.6090047990730216,
      "tokens_seen": 1225523200
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006350852557673019,
      "loss": 1.4318,
      "theoretical_loss": 3.6089872563811998,
      "tokens_seen": 1225588736
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006350651955867603,
      "loss": 1.4644,
      "theoretical_loss": 3.608969714890056,
      "tokens_seen": 1225654272
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006350451354062187,
      "loss": 1.49,
      "theoretical_loss": 3.6089521745994437,
      "tokens_seen": 1225719808
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006350250752256771,
      "loss": 1.4645,
      "theoretical_loss": 3.6089346355092164,
      "tokens_seen": 1225785344
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006350050150451354,
      "loss": 1.4749,
      "theoretical_loss": 3.608917097619228,
      "tokens_seen": 1225850880
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006349849548645939,
      "loss": 1.56,
      "theoretical_loss": 3.6088995609293324,
      "tokens_seen": 1225916416
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006349648946840521,
      "loss": 1.4637,
      "theoretical_loss": 3.608882025439383,
      "tokens_seen": 1225981952
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006349448345035106,
      "loss": 1.5557,
      "theoretical_loss": 3.6088644911492334,
      "tokens_seen": 1226047488
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000634924774322969,
      "loss": 1.5015,
      "theoretical_loss": 3.608846958058738,
      "tokens_seen": 1226113024
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006349047141424273,
      "loss": 1.5425,
      "theoretical_loss": 3.6088294261677496,
      "tokens_seen": 1226178560
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006348846539618857,
      "loss": 1.4498,
      "theoretical_loss": 3.608811895476123,
      "tokens_seen": 1226244096
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000634864593781344,
      "loss": 1.5379,
      "theoretical_loss": 3.6087943659837114,
      "tokens_seen": 1226309632
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006348445336008024,
      "loss": 1.4772,
      "theoretical_loss": 3.6087768376903693,
      "tokens_seen": 1226375168
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006348244734202609,
      "loss": 1.5127,
      "theoretical_loss": 3.608759310595951,
      "tokens_seen": 1226440704
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006348044132397191,
      "loss": 1.4604,
      "theoretical_loss": 3.6087417847003085,
      "tokens_seen": 1226506240
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006347843530591776,
      "loss": 1.3823,
      "theoretical_loss": 3.6087242600032976,
      "tokens_seen": 1226571776
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000634764292878636,
      "loss": 1.4417,
      "theoretical_loss": 3.6087067365047716,
      "tokens_seen": 1226637312
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006347442326980943,
      "loss": 1.4207,
      "theoretical_loss": 3.6086892142045848,
      "tokens_seen": 1226702848
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006347241725175527,
      "loss": 1.5733,
      "theoretical_loss": 3.608671693102591,
      "tokens_seen": 1226768384
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000634704112337011,
      "loss": 1.4952,
      "theoretical_loss": 3.6086541731986443,
      "tokens_seen": 1226833920
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006346840521564694,
      "loss": 1.3822,
      "theoretical_loss": 3.6086366544925985,
      "tokens_seen": 1226899456
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006346639919759279,
      "loss": 1.4459,
      "theoretical_loss": 3.6086191369843084,
      "tokens_seen": 1226964992
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006346439317953861,
      "loss": 1.3985,
      "theoretical_loss": 3.6086016206736278,
      "tokens_seen": 1227030528
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006346238716148446,
      "loss": 1.4113,
      "theoretical_loss": 3.6085841055604106,
      "tokens_seen": 1227096064
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006346038114343029,
      "loss": 1.6051,
      "theoretical_loss": 3.608566591644511,
      "tokens_seen": 1227161600
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006345837512537613,
      "loss": 1.5366,
      "theoretical_loss": 3.6085490789257837,
      "tokens_seen": 1227227136
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006345636910732197,
      "loss": 1.4519,
      "theoretical_loss": 3.608531567404083,
      "tokens_seen": 1227292672
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000634543630892678,
      "loss": 1.4051,
      "theoretical_loss": 3.608514057079262,
      "tokens_seen": 1227358208
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006345235707121364,
      "loss": 1.4239,
      "theoretical_loss": 3.6084965479511766,
      "tokens_seen": 1227423744
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006345035105315948,
      "loss": 1.5469,
      "theoretical_loss": 3.60847904001968,
      "tokens_seen": 1227489280
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006344834503510531,
      "loss": 1.4317,
      "theoretical_loss": 3.6084615332846264,
      "tokens_seen": 1227554816
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006344633901705116,
      "loss": 1.464,
      "theoretical_loss": 3.6084440277458714,
      "tokens_seen": 1227620352
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006344433299899699,
      "loss": 1.4173,
      "theoretical_loss": 3.608426523403269,
      "tokens_seen": 1227685888
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006344232698094283,
      "loss": 1.491,
      "theoretical_loss": 3.608409020256672,
      "tokens_seen": 1227751424
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006344032096288867,
      "loss": 1.3989,
      "theoretical_loss": 3.608391518305937,
      "tokens_seen": 1227816960
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006343831494483451,
      "loss": 1.4666,
      "theoretical_loss": 3.6083740175509176,
      "tokens_seen": 1227882496
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006343630892678034,
      "loss": 1.5054,
      "theoretical_loss": 3.608356517991468,
      "tokens_seen": 1227948032
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006343430290872618,
      "loss": 1.6256,
      "theoretical_loss": 3.608339019627443,
      "tokens_seen": 1228013568
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006343229689067201,
      "loss": 1.4961,
      "theoretical_loss": 3.6083215224586973,
      "tokens_seen": 1228079104
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006343029087261786,
      "loss": 1.5305,
      "theoretical_loss": 3.6083040264850856,
      "tokens_seen": 1228144640
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006342828485456369,
      "loss": 1.46,
      "theoretical_loss": 3.6082865317064616,
      "tokens_seen": 1228210176
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006342627883650953,
      "loss": 1.4123,
      "theoretical_loss": 3.6082690381226814,
      "tokens_seen": 1228275712
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006342427281845536,
      "loss": 1.4081,
      "theoretical_loss": 3.608251545733598,
      "tokens_seen": 1228341248
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006342226680040121,
      "loss": 1.5691,
      "theoretical_loss": 3.6082340545390674,
      "tokens_seen": 1228406784
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006342026078234704,
      "loss": 1.6289,
      "theoretical_loss": 3.6082165645389432,
      "tokens_seen": 1228472320
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006341825476429288,
      "loss": 1.5281,
      "theoretical_loss": 3.608199075733082,
      "tokens_seen": 1228537856
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006341624874623872,
      "loss": 1.4136,
      "theoretical_loss": 3.608181588121336,
      "tokens_seen": 1228603392
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006341424272818456,
      "loss": 1.4747,
      "theoretical_loss": 3.6081641017035615,
      "tokens_seen": 1228668928
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006341223671013039,
      "loss": 1.5676,
      "theoretical_loss": 3.6081466164796137,
      "tokens_seen": 1228734464
    },
    {
      "epoch": 0.37,
      "objective/train/advantage_avg": 0.4903579354286194,
      "objective/train/docs_used": 449484,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.518437385559082,
      "objective/train/original_loss": 1.518437385559082,
      "objective/train/theoretical_loss": 3.6081291324493456,
      "objective/train/tokens_used": 1249260000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24099566042423248,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502609014511108,
      "objective/train/weighted_lm_loss": 1.5948468446731567,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.6081291324493456,
      "tokens_seen": 1228800000
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006341023069207623,
      "loss": 1.5905,
      "theoretical_loss": 3.6081291324493456,
      "tokens_seen": 1228800000
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006340822467402206,
      "loss": 1.5016,
      "theoretical_loss": 3.6081116496126144,
      "tokens_seen": 1228865536
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006340621865596791,
      "loss": 1.4949,
      "theoretical_loss": 3.608094167969273,
      "tokens_seen": 1228931072
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006340421263791374,
      "loss": 1.4643,
      "theoretical_loss": 3.6080766875191776,
      "tokens_seen": 1228996608
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006340220661985958,
      "loss": 1.492,
      "theoretical_loss": 3.6080592082621825,
      "tokens_seen": 1229062144
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006340020060180542,
      "loss": 1.4782,
      "theoretical_loss": 3.608041730198143,
      "tokens_seen": 1229127680
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006339819458375125,
      "loss": 1.4503,
      "theoretical_loss": 3.6080242533269136,
      "tokens_seen": 1229193216
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006339618856569709,
      "loss": 1.4499,
      "theoretical_loss": 3.6080067776483493,
      "tokens_seen": 1229258752
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006339418254764294,
      "loss": 1.4778,
      "theoretical_loss": 3.607989303162306,
      "tokens_seen": 1229324288
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006339217652958876,
      "loss": 1.4375,
      "theoretical_loss": 3.607971829868638,
      "tokens_seen": 1229389824
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006339017051153461,
      "loss": 1.3998,
      "theoretical_loss": 3.6079543577672,
      "tokens_seen": 1229455360
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006338816449348043,
      "loss": 1.4751,
      "theoretical_loss": 3.607936886857848,
      "tokens_seen": 1229520896
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006338615847542628,
      "loss": 1.5057,
      "theoretical_loss": 3.6079194171404367,
      "tokens_seen": 1229586432
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006338415245737212,
      "loss": 1.5469,
      "theoretical_loss": 3.607901948614822,
      "tokens_seen": 1229651968
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006338214643931795,
      "loss": 1.5463,
      "theoretical_loss": 3.6078844812808577,
      "tokens_seen": 1229717504
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006338014042126379,
      "loss": 1.5246,
      "theoretical_loss": 3.6078670151384,
      "tokens_seen": 1229783040
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006337813440320964,
      "loss": 1.5761,
      "theoretical_loss": 3.607849550187304,
      "tokens_seen": 1229848576
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006337612838515546,
      "loss": 1.5624,
      "theoretical_loss": 3.6078320864274245,
      "tokens_seen": 1229914112
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006337412236710131,
      "loss": 1.5555,
      "theoretical_loss": 3.6078146238586175,
      "tokens_seen": 1229979648
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006337211634904713,
      "loss": 1.4829,
      "theoretical_loss": 3.607797162480738,
      "tokens_seen": 1230045184
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006337011033099298,
      "loss": 1.5408,
      "theoretical_loss": 3.607779702293641,
      "tokens_seen": 1230110720
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006336810431293882,
      "loss": 1.4496,
      "theoretical_loss": 3.6077622432971816,
      "tokens_seen": 1230176256
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006336609829488465,
      "loss": 1.577,
      "theoretical_loss": 3.6077447854912164,
      "tokens_seen": 1230241792
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006336409227683049,
      "loss": 1.3644,
      "theoretical_loss": 3.6077273288755998,
      "tokens_seen": 1230307328
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006336208625877633,
      "loss": 1.5274,
      "theoretical_loss": 3.6077098734501876,
      "tokens_seen": 1230372864
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006336008024072217,
      "loss": 1.502,
      "theoretical_loss": 3.6076924192148354,
      "tokens_seen": 1230438400
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006335807422266801,
      "loss": 1.5419,
      "theoretical_loss": 3.6076749661693985,
      "tokens_seen": 1230503936
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006335606820461385,
      "loss": 1.4903,
      "theoretical_loss": 3.6076575143137326,
      "tokens_seen": 1230569472
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006335406218655968,
      "loss": 1.5736,
      "theoretical_loss": 3.6076400636476924,
      "tokens_seen": 1230635008
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006335205616850552,
      "loss": 1.4192,
      "theoretical_loss": 3.6076226141711345,
      "tokens_seen": 1230700544
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006335005015045135,
      "loss": 1.6146,
      "theoretical_loss": 3.607605165883914,
      "tokens_seen": 1230766080
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000633480441323972,
      "loss": 1.4717,
      "theoretical_loss": 3.6075877187858865,
      "tokens_seen": 1230831616
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006334603811434303,
      "loss": 1.5028,
      "theoretical_loss": 3.607570272876908,
      "tokens_seen": 1230897152
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006334403209628887,
      "loss": 1.5014,
      "theoretical_loss": 3.607552828156834,
      "tokens_seen": 1230962688
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006334202607823471,
      "loss": 1.5242,
      "theoretical_loss": 3.60753538462552,
      "tokens_seen": 1231028224
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006334002006018055,
      "loss": 1.5473,
      "theoretical_loss": 3.6075179422828216,
      "tokens_seen": 1231093760
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006333801404212638,
      "loss": 1.4911,
      "theoretical_loss": 3.6075005011285945,
      "tokens_seen": 1231159296
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006333600802407222,
      "loss": 1.5101,
      "theoretical_loss": 3.6074830611626956,
      "tokens_seen": 1231224832
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006333400200601805,
      "loss": 1.3882,
      "theoretical_loss": 3.607465622384979,
      "tokens_seen": 1231290368
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000633319959879639,
      "loss": 1.4083,
      "theoretical_loss": 3.6074481847953015,
      "tokens_seen": 1231355904
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006332998996990973,
      "loss": 1.5396,
      "theoretical_loss": 3.6074307483935186,
      "tokens_seen": 1231421440
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006332798395185557,
      "loss": 1.4971,
      "theoretical_loss": 3.607413313179487,
      "tokens_seen": 1231486976
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000633259779338014,
      "loss": 1.5787,
      "theoretical_loss": 3.6073958791530614,
      "tokens_seen": 1231552512
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006332397191574725,
      "loss": 1.3911,
      "theoretical_loss": 3.6073784463140983,
      "tokens_seen": 1231618048
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006332196589769308,
      "loss": 1.5488,
      "theoretical_loss": 3.6073610146624535,
      "tokens_seen": 1231683584
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006331995987963892,
      "loss": 1.4918,
      "theoretical_loss": 3.6073435841979835,
      "tokens_seen": 1231749120
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006331795386158476,
      "loss": 1.5384,
      "theoretical_loss": 3.607326154920543,
      "tokens_seen": 1231814656
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000633159478435306,
      "loss": 1.4415,
      "theoretical_loss": 3.6073087268299893,
      "tokens_seen": 1231880192
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006331394182547643,
      "loss": 1.476,
      "theoretical_loss": 3.607291299926178,
      "tokens_seen": 1231945728
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006331193580742227,
      "loss": 1.4356,
      "theoretical_loss": 3.607273874208965,
      "tokens_seen": 1232011264
    },
    {
      "epoch": 0.37,
      "objective/train/advantage_avg": 0.49268966913223267,
      "objective/train/docs_used": 450737,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4987081289291382,
      "objective/train/original_loss": 1.4987080097198486,
      "objective/train/theoretical_loss": 3.6072564496782062,
      "objective/train/tokens_used": 1252536800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24331818521022797,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505058765411377,
      "objective/train/weighted_lm_loss": 1.5744669437408447,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6072564496782062,
      "tokens_seen": 1232076800
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000633099297893681,
      "loss": 1.5357,
      "theoretical_loss": 3.6072564496782062,
      "tokens_seen": 1232076800
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006330792377131395,
      "loss": 1.4186,
      "theoretical_loss": 3.607239026333759,
      "tokens_seen": 1232142336
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006330591775325978,
      "loss": 1.4141,
      "theoretical_loss": 3.6072216041754777,
      "tokens_seen": 1232207872
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006330391173520562,
      "loss": 1.5352,
      "theoretical_loss": 3.60720418320322,
      "tokens_seen": 1232273408
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006330190571715146,
      "loss": 1.4017,
      "theoretical_loss": 3.607186763416841,
      "tokens_seen": 1232338944
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006329989969909729,
      "loss": 1.4815,
      "theoretical_loss": 3.6071693448161977,
      "tokens_seen": 1232404480
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006329789368104313,
      "loss": 1.5283,
      "theoretical_loss": 3.607151927401145,
      "tokens_seen": 1232470016
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006329588766298898,
      "loss": 1.4727,
      "theoretical_loss": 3.6071345111715414,
      "tokens_seen": 1232535552
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000632938816449348,
      "loss": 1.5012,
      "theoretical_loss": 3.607117096127242,
      "tokens_seen": 1232601088
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006329187562688065,
      "loss": 1.4751,
      "theoretical_loss": 3.6070996822681023,
      "tokens_seen": 1232666624
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006328986960882647,
      "loss": 1.5272,
      "theoretical_loss": 3.6070822695939793,
      "tokens_seen": 1232732160
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006328786359077232,
      "loss": 1.5261,
      "theoretical_loss": 3.6070648581047298,
      "tokens_seen": 1232797696
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006328585757271816,
      "loss": 1.5715,
      "theoretical_loss": 3.60704744780021,
      "tokens_seen": 1232863232
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006328385155466399,
      "loss": 1.6252,
      "theoretical_loss": 3.6070300386802767,
      "tokens_seen": 1232928768
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006328184553660983,
      "loss": 1.4211,
      "theoretical_loss": 3.6070126307447845,
      "tokens_seen": 1232994304
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006327983951855568,
      "loss": 1.423,
      "theoretical_loss": 3.6069952239935925,
      "tokens_seen": 1233059840
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000632778335005015,
      "loss": 1.4581,
      "theoretical_loss": 3.606977818426555,
      "tokens_seen": 1233125376
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006327582748244735,
      "loss": 1.5906,
      "theoretical_loss": 3.60696041404353,
      "tokens_seen": 1233190912
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006327382146439317,
      "loss": 1.5293,
      "theoretical_loss": 3.606943010844373,
      "tokens_seen": 1233256448
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006327181544633902,
      "loss": 1.4924,
      "theoretical_loss": 3.6069256088289414,
      "tokens_seen": 1233321984
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006326980942828486,
      "loss": 1.5223,
      "theoretical_loss": 3.6069082079970913,
      "tokens_seen": 1233387520
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006326780341023069,
      "loss": 1.5191,
      "theoretical_loss": 3.6068908083486795,
      "tokens_seen": 1233453056
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006326579739217653,
      "loss": 1.4456,
      "theoretical_loss": 3.606873409883562,
      "tokens_seen": 1233518592
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006326379137412237,
      "loss": 1.4624,
      "theoretical_loss": 3.6068560126015967,
      "tokens_seen": 1233584128
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000632617853560682,
      "loss": 1.6534,
      "theoretical_loss": 3.60683861650264,
      "tokens_seen": 1233649664
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006325977933801405,
      "loss": 1.4447,
      "theoretical_loss": 3.606821221586547,
      "tokens_seen": 1233715200
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006325777331995987,
      "loss": 1.3757,
      "theoretical_loss": 3.6068038278531764,
      "tokens_seen": 1233780736
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006325576730190572,
      "loss": 1.5312,
      "theoretical_loss": 3.606786435302384,
      "tokens_seen": 1233846272
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006325376128385155,
      "loss": 1.4667,
      "theoretical_loss": 3.606769043934027,
      "tokens_seen": 1233911808
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006325175526579739,
      "loss": 1.5038,
      "theoretical_loss": 3.6067516537479625,
      "tokens_seen": 1233977344
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006324974924774323,
      "loss": 1.4942,
      "theoretical_loss": 3.6067342647440457,
      "tokens_seen": 1234042880
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006324774322968907,
      "loss": 1.5394,
      "theoretical_loss": 3.606716876922136,
      "tokens_seen": 1234108416
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000632457372116349,
      "loss": 1.5193,
      "theoretical_loss": 3.606699490282088,
      "tokens_seen": 1234173952
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006324373119358075,
      "loss": 1.4187,
      "theoretical_loss": 3.6066821048237605,
      "tokens_seen": 1234239488
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006324172517552657,
      "loss": 1.448,
      "theoretical_loss": 3.6066647205470086,
      "tokens_seen": 1234305024
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006323971915747242,
      "loss": 1.473,
      "theoretical_loss": 3.6066473374516903,
      "tokens_seen": 1234370560
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006323771313941825,
      "loss": 1.4112,
      "theoretical_loss": 3.606629955537663,
      "tokens_seen": 1234436096
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006323570712136409,
      "loss": 1.345,
      "theoretical_loss": 3.6066125748047826,
      "tokens_seen": 1234501632
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006323370110330993,
      "loss": 1.4973,
      "theoretical_loss": 3.606595195252907,
      "tokens_seen": 1234567168
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006323169508525577,
      "loss": 1.4764,
      "theoretical_loss": 3.6065778168818934,
      "tokens_seen": 1234632704
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000632296890672016,
      "loss": 1.5128,
      "theoretical_loss": 3.606560439691598,
      "tokens_seen": 1234698240
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006322768304914744,
      "loss": 1.4465,
      "theoretical_loss": 3.6065430636818783,
      "tokens_seen": 1234763776
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006322567703109328,
      "loss": 1.5305,
      "theoretical_loss": 3.606525688852592,
      "tokens_seen": 1234829312
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006322367101303912,
      "loss": 1.4989,
      "theoretical_loss": 3.6065083152035955,
      "tokens_seen": 1234894848
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006322166499498495,
      "loss": 1.4761,
      "theoretical_loss": 3.6064909427347462,
      "tokens_seen": 1234960384
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000632196589769308,
      "loss": 1.5659,
      "theoretical_loss": 3.6064735714459015,
      "tokens_seen": 1235025920
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006321765295887662,
      "loss": 1.5011,
      "theoretical_loss": 3.6064562013369184,
      "tokens_seen": 1235091456
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006321564694082247,
      "loss": 1.4454,
      "theoretical_loss": 3.6064388324076546,
      "tokens_seen": 1235156992
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000632136409227683,
      "loss": 1.5326,
      "theoretical_loss": 3.6064214646579673,
      "tokens_seen": 1235222528
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006321163490471414,
      "loss": 1.5369,
      "theoretical_loss": 3.606404098087713,
      "tokens_seen": 1235288064
    },
    {
      "epoch": 0.37,
      "objective/train/advantage_avg": 0.4912415146827698,
      "objective/train/docs_used": 451987,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4503008127212524,
      "objective/train/original_loss": 1.450300931930542,
      "objective/train/theoretical_loss": 3.60638673269675,
      "objective/train/tokens_used": 1255813600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24178089201450348,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503530502319336,
      "objective/train/weighted_lm_loss": 1.5232105255126953,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.60638673269675,
      "tokens_seen": 1235353600
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006320962888665999,
      "loss": 1.5725,
      "theoretical_loss": 3.60638673269675,
      "tokens_seen": 1235353600
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006320762286860582,
      "loss": 1.4775,
      "theoretical_loss": 3.606369368484935,
      "tokens_seen": 1235419136
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006320561685055166,
      "loss": 1.538,
      "theoretical_loss": 3.606352005452126,
      "tokens_seen": 1235484672
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000632036108324975,
      "loss": 1.4018,
      "theoretical_loss": 3.6063346435981805,
      "tokens_seen": 1235550208
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006320160481444333,
      "loss": 1.6188,
      "theoretical_loss": 3.606317282922955,
      "tokens_seen": 1235615744
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006319959879638917,
      "loss": 1.4869,
      "theoretical_loss": 3.606299923426308,
      "tokens_seen": 1235681280
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006319759277833502,
      "loss": 1.4906,
      "theoretical_loss": 3.6062825651080956,
      "tokens_seen": 1235746816
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006319558676028084,
      "loss": 1.5212,
      "theoretical_loss": 3.606265207968177,
      "tokens_seen": 1235812352
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006319358074222669,
      "loss": 1.5981,
      "theoretical_loss": 3.6062478520064087,
      "tokens_seen": 1235877888
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006319157472417251,
      "loss": 1.4429,
      "theoretical_loss": 3.606230497222649,
      "tokens_seen": 1235943424
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006318956870611836,
      "loss": 1.5024,
      "theoretical_loss": 3.6062131436167544,
      "tokens_seen": 1236008960
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000631875626880642,
      "loss": 1.4902,
      "theoretical_loss": 3.606195791188583,
      "tokens_seen": 1236074496
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006318555667001003,
      "loss": 1.3771,
      "theoretical_loss": 3.606178439937993,
      "tokens_seen": 1236140032
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006318355065195587,
      "loss": 1.49,
      "theoretical_loss": 3.6061610898648415,
      "tokens_seen": 1236205568
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006318154463390172,
      "loss": 1.5054,
      "theoretical_loss": 3.606143740968986,
      "tokens_seen": 1236271104
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006317953861584754,
      "loss": 1.6036,
      "theoretical_loss": 3.6061263932502854,
      "tokens_seen": 1236336640
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006317753259779339,
      "loss": 1.5738,
      "theoretical_loss": 3.6061090467085957,
      "tokens_seen": 1236402176
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006317552657973921,
      "loss": 1.4869,
      "theoretical_loss": 3.6060917013437757,
      "tokens_seen": 1236467712
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006317352056168506,
      "loss": 1.4822,
      "theoretical_loss": 3.6060743571556833,
      "tokens_seen": 1236533248
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000631715145436309,
      "loss": 1.4391,
      "theoretical_loss": 3.6060570141441755,
      "tokens_seen": 1236598784
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006316950852557673,
      "loss": 1.4819,
      "theoretical_loss": 3.6060396723091106,
      "tokens_seen": 1236664320
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006316750250752257,
      "loss": 1.5844,
      "theoretical_loss": 3.6060223316503475,
      "tokens_seen": 1236729856
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006316549648946841,
      "loss": 1.5955,
      "theoretical_loss": 3.606004992167742,
      "tokens_seen": 1236795392
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006316349047141424,
      "loss": 1.4726,
      "theoretical_loss": 3.6059876538611535,
      "tokens_seen": 1236860928
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006316148445336009,
      "loss": 1.458,
      "theoretical_loss": 3.6059703167304393,
      "tokens_seen": 1236926464
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006315947843530591,
      "loss": 1.4162,
      "theoretical_loss": 3.605952980775458,
      "tokens_seen": 1236992000
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006315747241725176,
      "loss": 1.5392,
      "theoretical_loss": 3.605935645996067,
      "tokens_seen": 1237057536
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006315546639919759,
      "loss": 1.3804,
      "theoretical_loss": 3.6059183123921246,
      "tokens_seen": 1237123072
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006315346038114343,
      "loss": 1.4855,
      "theoretical_loss": 3.6059009799634882,
      "tokens_seen": 1237188608
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006315145436308927,
      "loss": 1.3834,
      "theoretical_loss": 3.6058836487100168,
      "tokens_seen": 1237254144
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006314944834503511,
      "loss": 1.4558,
      "theoretical_loss": 3.605866318631568,
      "tokens_seen": 1237319680
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006314744232698094,
      "loss": 1.5114,
      "theoretical_loss": 3.605848989728,
      "tokens_seen": 1237385216
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0006314543630892679,
      "loss": 1.4402,
      "theoretical_loss": 3.6058316619991704,
      "tokens_seen": 1237450752
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006314343029087262,
      "loss": 1.4907,
      "theoretical_loss": 3.605814335444938,
      "tokens_seen": 1237516288
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006314142427281846,
      "loss": 1.3999,
      "theoretical_loss": 3.605797010065161,
      "tokens_seen": 1237581824
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006313941825476429,
      "loss": 1.4321,
      "theoretical_loss": 3.6057796858596975,
      "tokens_seen": 1237647360
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006313741223671013,
      "loss": 1.4346,
      "theoretical_loss": 3.6057623628284055,
      "tokens_seen": 1237712896
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006313540621865597,
      "loss": 1.5332,
      "theoretical_loss": 3.6057450409711436,
      "tokens_seen": 1237778432
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006313340020060181,
      "loss": 1.4745,
      "theoretical_loss": 3.6057277202877698,
      "tokens_seen": 1237843968
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006313139418254764,
      "loss": 1.5031,
      "theoretical_loss": 3.6057104007781424,
      "tokens_seen": 1237909504
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006312938816449348,
      "loss": 1.4763,
      "theoretical_loss": 3.6056930824421194,
      "tokens_seen": 1237975040
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006312738214643932,
      "loss": 1.4494,
      "theoretical_loss": 3.6056757652795604,
      "tokens_seen": 1238040576
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006312537612838516,
      "loss": 1.4698,
      "theoretical_loss": 3.6056584492903223,
      "tokens_seen": 1238106112
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006312337011033099,
      "loss": 1.5569,
      "theoretical_loss": 3.6056411344742645,
      "tokens_seen": 1238171648
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006312136409227684,
      "loss": 1.3596,
      "theoretical_loss": 3.6056238208312448,
      "tokens_seen": 1238237184
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006311935807422266,
      "loss": 1.4885,
      "theoretical_loss": 3.605606508361122,
      "tokens_seen": 1238302720
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006311735205616851,
      "loss": 1.4622,
      "theoretical_loss": 3.6055891970637544,
      "tokens_seen": 1238368256
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006311534603811434,
      "loss": 1.4606,
      "theoretical_loss": 3.6055718869390008,
      "tokens_seen": 1238433792
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006311334002006018,
      "loss": 1.6067,
      "theoretical_loss": 3.6055545779867195,
      "tokens_seen": 1238499328
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006311133400200602,
      "loss": 1.5083,
      "theoretical_loss": 3.605537270206769,
      "tokens_seen": 1238564864
    },
    {
      "epoch": 0.38,
      "objective/train/advantage_avg": 0.49321281909942627,
      "objective/train/docs_used": 453261,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5835257768630981,
      "objective/train/original_loss": 1.5835256576538086,
      "objective/train/theoretical_loss": 3.605519963599008,
      "objective/train/tokens_used": 1259090400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2438763529062271,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505610704421997,
      "objective/train/weighted_lm_loss": 1.663584589958191,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.605519963599008,
      "tokens_seen": 1238630400
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006310932798395186,
      "loss": 1.5933,
      "theoretical_loss": 3.605519963599008,
      "tokens_seen": 1238630400
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006310732196589769,
      "loss": 1.4785,
      "theoretical_loss": 3.605502658163295,
      "tokens_seen": 1238695936
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006310531594784354,
      "loss": 1.5471,
      "theoretical_loss": 3.605485353899488,
      "tokens_seen": 1238761472
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006310330992978936,
      "loss": 1.4699,
      "theoretical_loss": 3.6054680508074473,
      "tokens_seen": 1238827008
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006310130391173521,
      "loss": 1.4608,
      "theoretical_loss": 3.60545074888703,
      "tokens_seen": 1238892544
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006309929789368104,
      "loss": 1.3587,
      "theoretical_loss": 3.6054334481380956,
      "tokens_seen": 1238958080
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006309729187562688,
      "loss": 1.5618,
      "theoretical_loss": 3.6054161485605025,
      "tokens_seen": 1239023616
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006309528585757272,
      "loss": 1.4331,
      "theoretical_loss": 3.60539885015411,
      "tokens_seen": 1239089152
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006309327983951855,
      "loss": 1.5303,
      "theoretical_loss": 3.6053815529187756,
      "tokens_seen": 1239154688
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006309127382146439,
      "loss": 1.4268,
      "theoretical_loss": 3.60536425685436,
      "tokens_seen": 1239220224
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006308926780341024,
      "loss": 1.6579,
      "theoretical_loss": 3.60534696196072,
      "tokens_seen": 1239285760
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006308726178535606,
      "loss": 1.3859,
      "theoretical_loss": 3.605329668237715,
      "tokens_seen": 1239351296
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006308525576730191,
      "loss": 1.6145,
      "theoretical_loss": 3.6053123756852052,
      "tokens_seen": 1239416832
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006308324974924775,
      "loss": 1.5419,
      "theoretical_loss": 3.605295084303048,
      "tokens_seen": 1239482368
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006308124373119358,
      "loss": 1.4061,
      "theoretical_loss": 3.605277794091103,
      "tokens_seen": 1239547904
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006307923771313942,
      "loss": 1.447,
      "theoretical_loss": 3.6052605050492286,
      "tokens_seen": 1239613440
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006307723169508525,
      "loss": 1.4394,
      "theoretical_loss": 3.605243217177285,
      "tokens_seen": 1239678976
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006307522567703109,
      "loss": 1.4678,
      "theoretical_loss": 3.6052259304751297,
      "tokens_seen": 1239744512
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006307321965897694,
      "loss": 1.4114,
      "theoretical_loss": 3.605208644942622,
      "tokens_seen": 1239810048
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006307121364092276,
      "loss": 1.3983,
      "theoretical_loss": 3.6051913605796218,
      "tokens_seen": 1239875584
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006306920762286861,
      "loss": 1.5512,
      "theoretical_loss": 3.6051740773859877,
      "tokens_seen": 1239941120
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006306720160481443,
      "loss": 1.4946,
      "theoretical_loss": 3.6051567953615784,
      "tokens_seen": 1240006656
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006306519558676028,
      "loss": 1.5333,
      "theoretical_loss": 3.605139514506254,
      "tokens_seen": 1240072192
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006306318956870612,
      "loss": 1.4741,
      "theoretical_loss": 3.6051222348198726,
      "tokens_seen": 1240137728
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006306118355065195,
      "loss": 1.4888,
      "theoretical_loss": 3.605104956302293,
      "tokens_seen": 1240203264
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000630591775325978,
      "loss": 1.3985,
      "theoretical_loss": 3.6050876789533763,
      "tokens_seen": 1240268800
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006305717151454363,
      "loss": 1.5083,
      "theoretical_loss": 3.60507040277298,
      "tokens_seen": 1240334336
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006305516549648947,
      "loss": 1.4896,
      "theoretical_loss": 3.605053127760964,
      "tokens_seen": 1240399872
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006305315947843531,
      "loss": 1.594,
      "theoretical_loss": 3.605035853917187,
      "tokens_seen": 1240465408
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006305115346038115,
      "loss": 1.5371,
      "theoretical_loss": 3.6050185812415094,
      "tokens_seen": 1240530944
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006304914744232698,
      "loss": 1.5929,
      "theoretical_loss": 3.6050013097337894,
      "tokens_seen": 1240596480
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006304714142427283,
      "loss": 1.4829,
      "theoretical_loss": 3.604984039393887,
      "tokens_seen": 1240662016
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006304513540621866,
      "loss": 1.5154,
      "theoretical_loss": 3.604966770221661,
      "tokens_seen": 1240727552
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000630431293881645,
      "loss": 1.4422,
      "theoretical_loss": 3.6049495022169715,
      "tokens_seen": 1240793088
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006304112337011033,
      "loss": 1.3866,
      "theoretical_loss": 3.604932235379677,
      "tokens_seen": 1240858624
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006303911735205617,
      "loss": 1.4863,
      "theoretical_loss": 3.6049149697096374,
      "tokens_seen": 1240924160
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006303711133400201,
      "loss": 1.3951,
      "theoretical_loss": 3.604897705206713,
      "tokens_seen": 1240989696
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006303510531594785,
      "loss": 1.5153,
      "theoretical_loss": 3.604880441870762,
      "tokens_seen": 1241055232
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006303309929789368,
      "loss": 1.4702,
      "theoretical_loss": 3.6048631797016437,
      "tokens_seen": 1241120768
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006303109327983952,
      "loss": 1.4564,
      "theoretical_loss": 3.604845918699219,
      "tokens_seen": 1241186304
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006302908726178536,
      "loss": 1.5176,
      "theoretical_loss": 3.6048286588633465,
      "tokens_seen": 1241251840
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000630270812437312,
      "loss": 1.5514,
      "theoretical_loss": 3.604811400193886,
      "tokens_seen": 1241317376
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006302507522567703,
      "loss": 1.4779,
      "theoretical_loss": 3.6047941426906975,
      "tokens_seen": 1241382912
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006302306920762288,
      "loss": 1.517,
      "theoretical_loss": 3.6047768863536396,
      "tokens_seen": 1241448448
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000630210631895687,
      "loss": 1.612,
      "theoretical_loss": 3.6047596311825725,
      "tokens_seen": 1241513984
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006301905717151455,
      "loss": 1.5555,
      "theoretical_loss": 3.604742377177357,
      "tokens_seen": 1241579520
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006301705115346038,
      "loss": 1.5098,
      "theoretical_loss": 3.6047251243378504,
      "tokens_seen": 1241645056
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006301504513540622,
      "loss": 1.5631,
      "theoretical_loss": 3.604707872663915,
      "tokens_seen": 1241710592
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006301303911735206,
      "loss": 1.5056,
      "theoretical_loss": 3.6046906221554087,
      "tokens_seen": 1241776128
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000630110330992979,
      "loss": 1.5396,
      "theoretical_loss": 3.604673372812192,
      "tokens_seen": 1241841664
    },
    {
      "epoch": 0.38,
      "objective/train/advantage_avg": 0.49360406398773193,
      "objective/train/docs_used": 454384,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.440698504447937,
      "objective/train/original_loss": 1.4406983852386475,
      "objective/train/theoretical_loss": 3.604656124634124,
      "objective/train/tokens_used": 1262367200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24405449628829956,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506012439727783,
      "objective/train/weighted_lm_loss": 1.5135451555252075,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.604656124634124,
      "tokens_seen": 1241907200
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006300902708124373,
      "loss": 1.4601,
      "theoretical_loss": 3.604656124634124,
      "tokens_seen": 1241907200
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006300702106318958,
      "loss": 1.5467,
      "theoretical_loss": 3.604638877621066,
      "tokens_seen": 1241972736
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000630050150451354,
      "loss": 1.514,
      "theoretical_loss": 3.6046216317728765,
      "tokens_seen": 1242038272
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006300300902708125,
      "loss": 1.5008,
      "theoretical_loss": 3.6046043870894158,
      "tokens_seen": 1242103808
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006300100300902708,
      "loss": 1.5091,
      "theoretical_loss": 3.6045871435705434,
      "tokens_seen": 1242169344
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006299899699097292,
      "loss": 1.4921,
      "theoretical_loss": 3.6045699012161205,
      "tokens_seen": 1242234880
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006299699097291876,
      "loss": 1.4446,
      "theoretical_loss": 3.6045526600260054,
      "tokens_seen": 1242300416
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006299498495486459,
      "loss": 1.502,
      "theoretical_loss": 3.6045354200000594,
      "tokens_seen": 1242365952
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006299297893681043,
      "loss": 1.4059,
      "theoretical_loss": 3.6045181811381415,
      "tokens_seen": 1242431488
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006299097291875628,
      "loss": 1.5413,
      "theoretical_loss": 3.6045009434401125,
      "tokens_seen": 1242497024
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000629889669007021,
      "loss": 1.4132,
      "theoretical_loss": 3.6044837069058318,
      "tokens_seen": 1242562560
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006298696088264795,
      "loss": 1.6524,
      "theoretical_loss": 3.6044664715351593,
      "tokens_seen": 1242628096
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006298495486459379,
      "loss": 1.5069,
      "theoretical_loss": 3.604449237327956,
      "tokens_seen": 1242693632
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006298294884653962,
      "loss": 1.5428,
      "theoretical_loss": 3.604432004284081,
      "tokens_seen": 1242759168
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006298094282848546,
      "loss": 1.467,
      "theoretical_loss": 3.6044147724033957,
      "tokens_seen": 1242824704
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006297893681043129,
      "loss": 1.5746,
      "theoretical_loss": 3.604397541685759,
      "tokens_seen": 1242890240
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006297693079237713,
      "loss": 1.5853,
      "theoretical_loss": 3.6043803121310316,
      "tokens_seen": 1242955776
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006297492477432298,
      "loss": 1.4616,
      "theoretical_loss": 3.6043630837390737,
      "tokens_seen": 1243021312
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000629729187562688,
      "loss": 1.5739,
      "theoretical_loss": 3.6043458565097453,
      "tokens_seen": 1243086848
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006297091273821465,
      "loss": 1.5425,
      "theoretical_loss": 3.6043286304429074,
      "tokens_seen": 1243152384
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006296890672016047,
      "loss": 1.4742,
      "theoretical_loss": 3.6043114055384198,
      "tokens_seen": 1243217920
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006296690070210632,
      "loss": 1.4422,
      "theoretical_loss": 3.604294181796142,
      "tokens_seen": 1243283456
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006296489468405216,
      "loss": 1.2968,
      "theoretical_loss": 3.6042769592159356,
      "tokens_seen": 1243348992
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006296288866599799,
      "loss": 1.4835,
      "theoretical_loss": 3.60425973779766,
      "tokens_seen": 1243414528
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006296088264794383,
      "loss": 1.4431,
      "theoretical_loss": 3.6042425175411763,
      "tokens_seen": 1243480064
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006295887662988967,
      "loss": 1.5383,
      "theoretical_loss": 3.6042252984463445,
      "tokens_seen": 1243545600
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000629568706118355,
      "loss": 1.4984,
      "theoretical_loss": 3.6042080805130254,
      "tokens_seen": 1243611136
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006295486459378135,
      "loss": 1.5257,
      "theoretical_loss": 3.6041908637410787,
      "tokens_seen": 1243676672
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006295285857572718,
      "loss": 1.5617,
      "theoretical_loss": 3.6041736481303657,
      "tokens_seen": 1243742208
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006295085255767302,
      "loss": 1.5728,
      "theoretical_loss": 3.604156433680746,
      "tokens_seen": 1243807744
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006294884653961886,
      "loss": 1.4923,
      "theoretical_loss": 3.6041392203920815,
      "tokens_seen": 1243873280
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000629468405215647,
      "loss": 1.5104,
      "theoretical_loss": 3.6041220082642313,
      "tokens_seen": 1243938816
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006294483450351053,
      "loss": 1.5299,
      "theoretical_loss": 3.6041047972970564,
      "tokens_seen": 1244004352
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006294282848545637,
      "loss": 1.4862,
      "theoretical_loss": 3.604087587490418,
      "tokens_seen": 1244069888
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000629408224674022,
      "loss": 1.617,
      "theoretical_loss": 3.6040703788441757,
      "tokens_seen": 1244135424
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006293881644934805,
      "loss": 1.3895,
      "theoretical_loss": 3.604053171358191,
      "tokens_seen": 1244200960
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006293681043129388,
      "loss": 1.6952,
      "theoretical_loss": 3.604035965032325,
      "tokens_seen": 1244266496
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006293480441323972,
      "loss": 1.5762,
      "theoretical_loss": 3.6040187598664364,
      "tokens_seen": 1244332032
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006293279839518555,
      "loss": 1.4543,
      "theoretical_loss": 3.6040015558603877,
      "tokens_seen": 1244397568
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000629307923771314,
      "loss": 1.5639,
      "theoretical_loss": 3.6039843530140394,
      "tokens_seen": 1244463104
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006292878635907723,
      "loss": 1.3668,
      "theoretical_loss": 3.6039671513272515,
      "tokens_seen": 1244528640
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006292678034102307,
      "loss": 1.5125,
      "theoretical_loss": 3.6039499507998856,
      "tokens_seen": 1244594176
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000629247743229689,
      "loss": 1.5589,
      "theoretical_loss": 3.603932751431802,
      "tokens_seen": 1244659712
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006292276830491474,
      "loss": 1.4556,
      "theoretical_loss": 3.603915553222862,
      "tokens_seen": 1244725248
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006292076228686058,
      "loss": 1.4525,
      "theoretical_loss": 3.603898356172926,
      "tokens_seen": 1244790784
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006291875626880642,
      "loss": 1.4719,
      "theoretical_loss": 3.6038811602818543,
      "tokens_seen": 1244856320
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006291675025075225,
      "loss": 1.5188,
      "theoretical_loss": 3.6038639655495093,
      "tokens_seen": 1244921856
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000629147442326981,
      "loss": 1.5017,
      "theoretical_loss": 3.6038467719757508,
      "tokens_seen": 1244987392
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006291273821464394,
      "loss": 1.4474,
      "theoretical_loss": 3.6038295795604403,
      "tokens_seen": 1245052928
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006291073219658977,
      "loss": 1.4079,
      "theoretical_loss": 3.6038123883034388,
      "tokens_seen": 1245118464
    },
    {
      "epoch": 0.38,
      "objective/train/advantage_avg": 0.4870053827762604,
      "objective/train/docs_used": 455531,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4811385869979858,
      "objective/train/original_loss": 1.481138825416565,
      "objective/train/theoretical_loss": 3.6037951982046064,
      "objective/train/tokens_used": 1265644000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23820097744464874,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0499112606048584,
      "objective/train/weighted_lm_loss": 1.554864525794983,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.6037951982046064,
      "tokens_seen": 1245184000
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006290872617853562,
      "loss": 1.6124,
      "theoretical_loss": 3.6037951982046064,
      "tokens_seen": 1245184000
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006290672016048144,
      "loss": 1.4617,
      "theoretical_loss": 3.6037780092638054,
      "tokens_seen": 1245249536
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006290471414242729,
      "loss": 1.6059,
      "theoretical_loss": 3.603760821480896,
      "tokens_seen": 1245315072
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006290270812437312,
      "loss": 1.4993,
      "theoretical_loss": 3.6037436348557397,
      "tokens_seen": 1245380608
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006290070210631896,
      "loss": 1.4229,
      "theoretical_loss": 3.6037264493881973,
      "tokens_seen": 1245446144
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000628986960882648,
      "loss": 1.6258,
      "theoretical_loss": 3.6037092650781304,
      "tokens_seen": 1245511680
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006289669007021063,
      "loss": 1.4177,
      "theoretical_loss": 3.6036920819253995,
      "tokens_seen": 1245577216
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006289468405215647,
      "loss": 1.6491,
      "theoretical_loss": 3.6036748999298656,
      "tokens_seen": 1245642752
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006289267803410232,
      "loss": 1.4724,
      "theoretical_loss": 3.603657719091391,
      "tokens_seen": 1245708288
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006289067201604814,
      "loss": 1.4671,
      "theoretical_loss": 3.6036405394098363,
      "tokens_seen": 1245773824
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006288866599799399,
      "loss": 1.5748,
      "theoretical_loss": 3.603623360885062,
      "tokens_seen": 1245839360
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006288665997993983,
      "loss": 1.4276,
      "theoretical_loss": 3.6036061835169306,
      "tokens_seen": 1245904896
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006288465396188566,
      "loss": 1.5539,
      "theoretical_loss": 3.603589007305303,
      "tokens_seen": 1245970432
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000628826479438315,
      "loss": 1.4093,
      "theoretical_loss": 3.60357183225004,
      "tokens_seen": 1246035968
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006288064192577733,
      "loss": 1.5166,
      "theoretical_loss": 3.603554658351004,
      "tokens_seen": 1246101504
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006287863590772317,
      "loss": 1.6103,
      "theoretical_loss": 3.6035374856080544,
      "tokens_seen": 1246167040
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006287662988966902,
      "loss": 1.6155,
      "theoretical_loss": 3.603520314021055,
      "tokens_seen": 1246232576
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006287462387161484,
      "loss": 1.5538,
      "theoretical_loss": 3.6035031435898657,
      "tokens_seen": 1246298112
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006287261785356069,
      "loss": 1.4363,
      "theoretical_loss": 3.6034859743143484,
      "tokens_seen": 1246363648
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006287061183550651,
      "loss": 1.3961,
      "theoretical_loss": 3.603468806194364,
      "tokens_seen": 1246429184
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006286860581745236,
      "loss": 1.5196,
      "theoretical_loss": 3.6034516392297746,
      "tokens_seen": 1246494720
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000628665997993982,
      "loss": 1.544,
      "theoretical_loss": 3.6034344734204415,
      "tokens_seen": 1246560256
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006286459378134403,
      "loss": 1.5275,
      "theoretical_loss": 3.6034173087662267,
      "tokens_seen": 1246625792
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006286258776328987,
      "loss": 1.6393,
      "theoretical_loss": 3.603400145266991,
      "tokens_seen": 1246691328
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006286058174523571,
      "loss": 1.5265,
      "theoretical_loss": 3.6033829829225965,
      "tokens_seen": 1246756864
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006285857572718154,
      "loss": 1.5235,
      "theoretical_loss": 3.603365821732904,
      "tokens_seen": 1246822400
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006285656970912739,
      "loss": 1.4385,
      "theoretical_loss": 3.603348661697776,
      "tokens_seen": 1246887936
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006285456369107322,
      "loss": 1.5423,
      "theoretical_loss": 3.603331502817074,
      "tokens_seen": 1246953472
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006285255767301906,
      "loss": 1.6609,
      "theoretical_loss": 3.6033143450906593,
      "tokens_seen": 1247019008
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000628505516549649,
      "loss": 1.5236,
      "theoretical_loss": 3.603297188518394,
      "tokens_seen": 1247084544
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006284854563691074,
      "loss": 1.5369,
      "theoretical_loss": 3.6032800331001393,
      "tokens_seen": 1247150080
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006284653961885657,
      "loss": 1.4591,
      "theoretical_loss": 3.603262878835758,
      "tokens_seen": 1247215616
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006284453360080241,
      "loss": 1.4744,
      "theoretical_loss": 3.6032457257251105,
      "tokens_seen": 1247281152
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006284252758274824,
      "loss": 1.4589,
      "theoretical_loss": 3.6032285737680594,
      "tokens_seen": 1247346688
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006284052156469409,
      "loss": 1.389,
      "theoretical_loss": 3.603211422964466,
      "tokens_seen": 1247412224
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006283851554663992,
      "loss": 1.5177,
      "theoretical_loss": 3.603194273314193,
      "tokens_seen": 1247477760
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006283650952858576,
      "loss": 1.3845,
      "theoretical_loss": 3.603177124817101,
      "tokens_seen": 1247543296
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006283450351053159,
      "loss": 1.5301,
      "theoretical_loss": 3.6031599774730534,
      "tokens_seen": 1247608832
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006283249749247744,
      "loss": 1.4894,
      "theoretical_loss": 3.6031428312819114,
      "tokens_seen": 1247674368
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006283049147442327,
      "loss": 1.4049,
      "theoretical_loss": 3.603125686243536,
      "tokens_seen": 1247739904
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006282848545636911,
      "loss": 1.5361,
      "theoretical_loss": 3.603108542357791,
      "tokens_seen": 1247805440
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006282647943831494,
      "loss": 1.4376,
      "theoretical_loss": 3.603091399624537,
      "tokens_seen": 1247870976
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006282447342026078,
      "loss": 1.6681,
      "theoretical_loss": 3.603074258043636,
      "tokens_seen": 1247936512
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006282246740220662,
      "loss": 1.4459,
      "theoretical_loss": 3.6030571176149513,
      "tokens_seen": 1248002048
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006282046138415246,
      "loss": 1.5477,
      "theoretical_loss": 3.603039978338343,
      "tokens_seen": 1248067584
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006281845536609829,
      "loss": 1.4392,
      "theoretical_loss": 3.603022840213675,
      "tokens_seen": 1248133120
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006281644934804414,
      "loss": 1.3718,
      "theoretical_loss": 3.6030057032408083,
      "tokens_seen": 1248198656
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006281444332998997,
      "loss": 1.5599,
      "theoretical_loss": 3.6029885674196054,
      "tokens_seen": 1248264192
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006281243731193581,
      "loss": 1.494,
      "theoretical_loss": 3.6029714327499285,
      "tokens_seen": 1248329728
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006281043129388165,
      "loss": 1.5881,
      "theoretical_loss": 3.60295429923164,
      "tokens_seen": 1248395264
    },
    {
      "epoch": 0.38,
      "objective/train/advantage_avg": 0.49346864223480225,
      "objective/train/docs_used": 456712,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3770898580551147,
      "objective/train/original_loss": 1.3770899772644043,
      "objective/train/theoretical_loss": 3.602937166864601,
      "objective/train/tokens_used": 1268920800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2438165247440338,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050586223602295,
      "objective/train/weighted_lm_loss": 1.4467809200286865,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.602937166864601,
      "tokens_seen": 1248460800
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006280842527582748,
      "loss": 1.4668,
      "theoretical_loss": 3.602937166864601,
      "tokens_seen": 1248460800
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006280641925777332,
      "loss": 1.5587,
      "theoretical_loss": 3.6029200356486752,
      "tokens_seen": 1248526336
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006280441323971916,
      "loss": 1.5804,
      "theoretical_loss": 3.602902905583724,
      "tokens_seen": 1248591872
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006280240722166499,
      "loss": 1.4664,
      "theoretical_loss": 3.60288577666961,
      "tokens_seen": 1248657408
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006280040120361084,
      "loss": 1.5357,
      "theoretical_loss": 3.6028686489061945,
      "tokens_seen": 1248722944
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006279839518555666,
      "loss": 1.4111,
      "theoretical_loss": 3.6028515222933413,
      "tokens_seen": 1248788480
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006279638916750251,
      "loss": 1.4836,
      "theoretical_loss": 3.602834396830912,
      "tokens_seen": 1248854016
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006279438314944835,
      "loss": 1.5339,
      "theoretical_loss": 3.6028172725187693,
      "tokens_seen": 1248919552
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006279237713139418,
      "loss": 1.4483,
      "theoretical_loss": 3.6028001493567747,
      "tokens_seen": 1248985088
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006279037111334002,
      "loss": 1.5788,
      "theoretical_loss": 3.602783027344791,
      "tokens_seen": 1249050624
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006278836509528587,
      "loss": 1.3948,
      "theoretical_loss": 3.6027659064826816,
      "tokens_seen": 1249116160
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006278635907723169,
      "loss": 1.5557,
      "theoretical_loss": 3.6027487867703076,
      "tokens_seen": 1249181696
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006278435305917754,
      "loss": 1.6063,
      "theoretical_loss": 3.6027316682075323,
      "tokens_seen": 1249247232
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006278234704112336,
      "loss": 1.5193,
      "theoretical_loss": 3.6027145507942175,
      "tokens_seen": 1249312768
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006278034102306921,
      "loss": 1.4314,
      "theoretical_loss": 3.602697434530227,
      "tokens_seen": 1249378304
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006277833500501505,
      "loss": 1.566,
      "theoretical_loss": 3.6026803194154216,
      "tokens_seen": 1249443840
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006277632898696088,
      "loss": 1.5281,
      "theoretical_loss": 3.6026632054496655,
      "tokens_seen": 1249509376
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006277432296890672,
      "loss": 1.4892,
      "theoretical_loss": 3.6026460926328205,
      "tokens_seen": 1249574912
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006277231695085255,
      "loss": 1.5391,
      "theoretical_loss": 3.6026289809647487,
      "tokens_seen": 1249640448
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006277031093279839,
      "loss": 1.5931,
      "theoretical_loss": 3.602611870445314,
      "tokens_seen": 1249705984
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006276830491474424,
      "loss": 1.4958,
      "theoretical_loss": 3.6025947610743785,
      "tokens_seen": 1249771520
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006276629889669006,
      "loss": 1.509,
      "theoretical_loss": 3.6025776528518048,
      "tokens_seen": 1249837056
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006276429287863591,
      "loss": 1.4848,
      "theoretical_loss": 3.6025605457774548,
      "tokens_seen": 1249902592
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006276228686058175,
      "loss": 1.5011,
      "theoretical_loss": 3.6025434398511926,
      "tokens_seen": 1249968128
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006276028084252758,
      "loss": 1.5259,
      "theoretical_loss": 3.6025263350728807,
      "tokens_seen": 1250033664
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006275827482447343,
      "loss": 1.5116,
      "theoretical_loss": 3.6025092314423817,
      "tokens_seen": 1250099200
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006275626880641926,
      "loss": 1.3877,
      "theoretical_loss": 3.602492128959558,
      "tokens_seen": 1250164736
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000627542627883651,
      "loss": 1.5652,
      "theoretical_loss": 3.6024750276242727,
      "tokens_seen": 1250230272
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006275225677031094,
      "loss": 1.5697,
      "theoretical_loss": 3.602457927436389,
      "tokens_seen": 1250295808
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006275025075225678,
      "loss": 1.4534,
      "theoretical_loss": 3.6024408283957694,
      "tokens_seen": 1250361344
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006274824473420261,
      "loss": 1.3931,
      "theoretical_loss": 3.6024237305022764,
      "tokens_seen": 1250426880
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006274623871614845,
      "loss": 1.4818,
      "theoretical_loss": 3.6024066337557743,
      "tokens_seen": 1250492416
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006274423269809428,
      "loss": 1.4936,
      "theoretical_loss": 3.602389538156124,
      "tokens_seen": 1250557952
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006274222668004013,
      "loss": 1.5158,
      "theoretical_loss": 3.6023724437031905,
      "tokens_seen": 1250623488
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006274022066198596,
      "loss": 1.576,
      "theoretical_loss": 3.602355350396836,
      "tokens_seen": 1250689024
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000627382146439318,
      "loss": 1.4273,
      "theoretical_loss": 3.602338258236923,
      "tokens_seen": 1250754560
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006273620862587763,
      "loss": 1.4833,
      "theoretical_loss": 3.6023211672233155,
      "tokens_seen": 1250820096
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006273420260782348,
      "loss": 1.4844,
      "theoretical_loss": 3.6023040773558757,
      "tokens_seen": 1250885632
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006273219658976931,
      "loss": 1.4992,
      "theoretical_loss": 3.602286988634467,
      "tokens_seen": 1250951168
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006273019057171515,
      "loss": 1.5982,
      "theoretical_loss": 3.602269901058953,
      "tokens_seen": 1251016704
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006272818455366098,
      "loss": 1.38,
      "theoretical_loss": 3.602252814629196,
      "tokens_seen": 1251082240
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006272617853560682,
      "loss": 1.4312,
      "theoretical_loss": 3.60223572934506,
      "tokens_seen": 1251147776
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006272417251755266,
      "loss": 1.5147,
      "theoretical_loss": 3.6022186452064076,
      "tokens_seen": 1251213312
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000627221664994985,
      "loss": 1.5636,
      "theoretical_loss": 3.602201562213102,
      "tokens_seen": 1251278848
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006272016048144433,
      "loss": 1.437,
      "theoretical_loss": 3.6021844803650067,
      "tokens_seen": 1251344384
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006271815446339018,
      "loss": 1.397,
      "theoretical_loss": 3.602167399661985,
      "tokens_seen": 1251409920
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006271614844533601,
      "loss": 1.4996,
      "theoretical_loss": 3.6021503201039,
      "tokens_seen": 1251475456
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006271414242728185,
      "loss": 1.3987,
      "theoretical_loss": 3.6021332416906153,
      "tokens_seen": 1251540992
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006271213640922769,
      "loss": 1.4975,
      "theoretical_loss": 3.6021161644219934,
      "tokens_seen": 1251606528
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006271013039117352,
      "loss": 1.4132,
      "theoretical_loss": 3.6020990882978983,
      "tokens_seen": 1251672064
    },
    {
      "epoch": 0.38,
      "objective/train/advantage_avg": 0.4943554103374481,
      "objective/train/docs_used": 458046,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4159032106399536,
      "objective/train/original_loss": 1.4159033298492432,
      "objective/train/theoretical_loss": 3.6020820133181934,
      "objective/train/tokens_used": 1272197600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2446722388267517,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506794452667236,
      "objective/train/weighted_lm_loss": 1.487627625465393,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.6020820133181934,
      "tokens_seen": 1251737600
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006270812437311936,
      "loss": 1.5075,
      "theoretical_loss": 3.6020820133181934,
      "tokens_seen": 1251737600
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000627061183550652,
      "loss": 1.5078,
      "theoretical_loss": 3.602064939482742,
      "tokens_seen": 1251803136
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006270411233701103,
      "loss": 1.5767,
      "theoretical_loss": 3.6020478667914078,
      "tokens_seen": 1251868672
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006270210631895688,
      "loss": 1.4495,
      "theoretical_loss": 3.602030795244054,
      "tokens_seen": 1251934208
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000627001003009027,
      "loss": 1.5677,
      "theoretical_loss": 3.6020137248405435,
      "tokens_seen": 1251999744
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006269809428284855,
      "loss": 1.4171,
      "theoretical_loss": 3.6019966555807406,
      "tokens_seen": 1252065280
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006269608826479439,
      "loss": 1.4895,
      "theoretical_loss": 3.601979587464508,
      "tokens_seen": 1252130816
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006269408224674022,
      "loss": 1.5497,
      "theoretical_loss": 3.6019625204917105,
      "tokens_seen": 1252196352
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006269207622868606,
      "loss": 1.568,
      "theoretical_loss": 3.6019454546622107,
      "tokens_seen": 1252261888
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000626900702106319,
      "loss": 1.3889,
      "theoretical_loss": 3.601928389975873,
      "tokens_seen": 1252327424
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006268806419257773,
      "loss": 1.4986,
      "theoretical_loss": 3.6019113264325595,
      "tokens_seen": 1252392960
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006268605817452358,
      "loss": 1.4568,
      "theoretical_loss": 3.601894264032135,
      "tokens_seen": 1252458496
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000626840521564694,
      "loss": 1.5098,
      "theoretical_loss": 3.601877202774463,
      "tokens_seen": 1252524032
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006268204613841525,
      "loss": 1.4634,
      "theoretical_loss": 3.6018601426594072,
      "tokens_seen": 1252589568
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006268004012036109,
      "loss": 1.5565,
      "theoretical_loss": 3.601843083686831,
      "tokens_seen": 1252655104
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006267803410230692,
      "loss": 1.4147,
      "theoretical_loss": 3.6018260258565986,
      "tokens_seen": 1252720640
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006267602808425276,
      "loss": 1.5432,
      "theoretical_loss": 3.601808969168573,
      "tokens_seen": 1252786176
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000626740220661986,
      "loss": 1.4009,
      "theoretical_loss": 3.6017919136226184,
      "tokens_seen": 1252851712
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006267201604814443,
      "loss": 1.5429,
      "theoretical_loss": 3.601774859218599,
      "tokens_seen": 1252917248
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006267001003009028,
      "loss": 1.5376,
      "theoretical_loss": 3.601757805956378,
      "tokens_seen": 1252982784
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000626680040120361,
      "loss": 1.4511,
      "theoretical_loss": 3.60174075383582,
      "tokens_seen": 1253048320
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006266599799398195,
      "loss": 1.5443,
      "theoretical_loss": 3.6017237028567877,
      "tokens_seen": 1253113856
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006266399197592778,
      "loss": 1.5842,
      "theoretical_loss": 3.6017066530191455,
      "tokens_seen": 1253179392
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006266198595787362,
      "loss": 1.6054,
      "theoretical_loss": 3.601689604322758,
      "tokens_seen": 1253244928
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006265997993981946,
      "loss": 1.463,
      "theoretical_loss": 3.6016725567674883,
      "tokens_seen": 1253310464
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000626579739217653,
      "loss": 1.4374,
      "theoretical_loss": 3.6016555103532006,
      "tokens_seen": 1253376000
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006265596790371113,
      "loss": 1.5401,
      "theoretical_loss": 3.601638465079759,
      "tokens_seen": 1253441536
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006265396188565698,
      "loss": 1.4951,
      "theoretical_loss": 3.6016214209470276,
      "tokens_seen": 1253507072
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000626519558676028,
      "loss": 1.4986,
      "theoretical_loss": 3.60160437795487,
      "tokens_seen": 1253572608
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006264994984954865,
      "loss": 1.506,
      "theoretical_loss": 3.601587336103151,
      "tokens_seen": 1253638144
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006264794383149448,
      "loss": 1.5167,
      "theoretical_loss": 3.6015702953917335,
      "tokens_seen": 1253703680
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006264593781344032,
      "loss": 1.472,
      "theoretical_loss": 3.601553255820483,
      "tokens_seen": 1253769216
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006264393179538616,
      "loss": 1.5053,
      "theoretical_loss": 3.6015362173892624,
      "tokens_seen": 1253834752
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00062641925777332,
      "loss": 1.5038,
      "theoretical_loss": 3.6015191800979363,
      "tokens_seen": 1253900288
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006263991975927783,
      "loss": 1.5252,
      "theoretical_loss": 3.601502143946369,
      "tokens_seen": 1253965824
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006263791374122367,
      "loss": 1.5798,
      "theoretical_loss": 3.6014851089344244,
      "tokens_seen": 1254031360
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000626359077231695,
      "loss": 1.4862,
      "theoretical_loss": 3.6014680750619674,
      "tokens_seen": 1254096896
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006263390170511535,
      "loss": 1.5001,
      "theoretical_loss": 3.6014510423288613,
      "tokens_seen": 1254162432
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006263189568706118,
      "loss": 1.4472,
      "theoretical_loss": 3.6014340107349714,
      "tokens_seen": 1254227968
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006262988966900702,
      "loss": 1.4658,
      "theoretical_loss": 3.6014169802801606,
      "tokens_seen": 1254293504
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006262788365095285,
      "loss": 1.5561,
      "theoretical_loss": 3.6013999509642947,
      "tokens_seen": 1254359040
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000626258776328987,
      "loss": 1.5036,
      "theoretical_loss": 3.601382922787237,
      "tokens_seen": 1254424576
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006262387161484453,
      "loss": 1.4723,
      "theoretical_loss": 3.6013658957488524,
      "tokens_seen": 1254490112
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006262186559679037,
      "loss": 1.4964,
      "theoretical_loss": 3.601348869849005,
      "tokens_seen": 1254555648
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006261985957873621,
      "loss": 1.4732,
      "theoretical_loss": 3.601331845087559,
      "tokens_seen": 1254621184
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006261785356068205,
      "loss": 1.5542,
      "theoretical_loss": 3.6013148214643795,
      "tokens_seen": 1254686720
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006261584754262788,
      "loss": 1.403,
      "theoretical_loss": 3.6012977989793304,
      "tokens_seen": 1254752256
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006261384152457373,
      "loss": 1.5857,
      "theoretical_loss": 3.601280777632277,
      "tokens_seen": 1254817792
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006261183550651956,
      "loss": 1.4521,
      "theoretical_loss": 3.601263757423082,
      "tokens_seen": 1254883328
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000626098294884654,
      "loss": 1.4456,
      "theoretical_loss": 3.6012467383516116,
      "tokens_seen": 1254948864
    },
    {
      "epoch": 0.38,
      "objective/train/advantage_avg": 0.49449729919433594,
      "objective/train/docs_used": 459153,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4806921482086182,
      "objective/train/original_loss": 1.4806920289993286,
      "objective/train/theoretical_loss": 3.601229720417729,
      "objective/train/tokens_used": 1275474400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2448570877313614,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506945848464966,
      "objective/train/weighted_lm_loss": 1.5554972887039185,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.601229720417729,
      "tokens_seen": 1255014400
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006260782347041124,
      "loss": 1.5629,
      "theoretical_loss": 3.601229720417729,
      "tokens_seen": 1255014400
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006260581745235707,
      "loss": 1.4356,
      "theoretical_loss": 3.6012127036213,
      "tokens_seen": 1255079936
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006260381143430292,
      "loss": 1.5326,
      "theoretical_loss": 3.601195687962189,
      "tokens_seen": 1255145472
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006260180541624874,
      "loss": 1.4382,
      "theoretical_loss": 3.60117867344026,
      "tokens_seen": 1255211008
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006259979939819459,
      "loss": 1.4912,
      "theoretical_loss": 3.6011616600553786,
      "tokens_seen": 1255276544
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006259779338014043,
      "loss": 1.4646,
      "theoretical_loss": 3.6011446478074083,
      "tokens_seen": 1255342080
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006259578736208626,
      "loss": 1.4967,
      "theoretical_loss": 3.601127636696214,
      "tokens_seen": 1255407616
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000625937813440321,
      "loss": 1.5021,
      "theoretical_loss": 3.6011106267216615,
      "tokens_seen": 1255473152
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006259177532597795,
      "loss": 1.5093,
      "theoretical_loss": 3.601093617883614,
      "tokens_seen": 1255538688
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006258976930792377,
      "loss": 1.4872,
      "theoretical_loss": 3.601076610181937,
      "tokens_seen": 1255604224
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006258776328986962,
      "loss": 1.4765,
      "theoretical_loss": 3.6010596036164957,
      "tokens_seen": 1255669760
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006258575727181544,
      "loss": 1.4967,
      "theoretical_loss": 3.6010425981871546,
      "tokens_seen": 1255735296
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006258375125376129,
      "loss": 1.4394,
      "theoretical_loss": 3.6010255938937776,
      "tokens_seen": 1255800832
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006258174523570713,
      "loss": 1.544,
      "theoretical_loss": 3.601008590736231,
      "tokens_seen": 1255866368
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006257973921765296,
      "loss": 1.4545,
      "theoretical_loss": 3.600991588714378,
      "tokens_seen": 1255931904
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000625777331995988,
      "loss": 1.4194,
      "theoretical_loss": 3.600974587828085,
      "tokens_seen": 1255997440
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006257572718154464,
      "loss": 1.4873,
      "theoretical_loss": 3.6009575880772164,
      "tokens_seen": 1256062976
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006257372116349047,
      "loss": 1.4207,
      "theoretical_loss": 3.6009405894616373,
      "tokens_seen": 1256128512
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006257171514543632,
      "loss": 1.4893,
      "theoretical_loss": 3.6009235919812124,
      "tokens_seen": 1256194048
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006256970912738214,
      "loss": 1.4878,
      "theoretical_loss": 3.6009065956358066,
      "tokens_seen": 1256259584
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006256770310932799,
      "loss": 1.4318,
      "theoretical_loss": 3.6008896004252846,
      "tokens_seen": 1256325120
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006256569709127382,
      "loss": 1.4691,
      "theoretical_loss": 3.6008726063495127,
      "tokens_seen": 1256390656
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006256369107321966,
      "loss": 1.4882,
      "theoretical_loss": 3.6008556134083545,
      "tokens_seen": 1256456192
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000625616850551655,
      "loss": 1.4813,
      "theoretical_loss": 3.6008386216016754,
      "tokens_seen": 1256521728
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006255967903711134,
      "loss": 1.6147,
      "theoretical_loss": 3.6008216309293415,
      "tokens_seen": 1256587264
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006255767301905717,
      "loss": 1.5022,
      "theoretical_loss": 3.6008046413912167,
      "tokens_seen": 1256652800
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006255566700100302,
      "loss": 1.4189,
      "theoretical_loss": 3.600787652987167,
      "tokens_seen": 1256718336
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006255366098294884,
      "loss": 1.4726,
      "theoretical_loss": 3.600770665717057,
      "tokens_seen": 1256783872
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006255165496489469,
      "loss": 1.4664,
      "theoretical_loss": 3.600753679580752,
      "tokens_seen": 1256849408
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006254964894684052,
      "loss": 1.5266,
      "theoretical_loss": 3.600736694578117,
      "tokens_seen": 1256914944
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006254764292878636,
      "loss": 1.4042,
      "theoretical_loss": 3.6007197107090185,
      "tokens_seen": 1256980480
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000625456369107322,
      "loss": 1.4893,
      "theoretical_loss": 3.60070272797332,
      "tokens_seen": 1257046016
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006254363089267804,
      "loss": 1.5358,
      "theoretical_loss": 3.600685746370888,
      "tokens_seen": 1257111552
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006254162487462387,
      "loss": 1.5758,
      "theoretical_loss": 3.6006687659015872,
      "tokens_seen": 1257177088
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006253961885656971,
      "loss": 1.451,
      "theoretical_loss": 3.600651786565283,
      "tokens_seen": 1257242624
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006253761283851554,
      "loss": 1.488,
      "theoretical_loss": 3.600634808361841,
      "tokens_seen": 1257308160
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006253560682046139,
      "loss": 1.4425,
      "theoretical_loss": 3.6006178312911263,
      "tokens_seen": 1257373696
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006253360080240722,
      "loss": 1.5382,
      "theoretical_loss": 3.6006008553530044,
      "tokens_seen": 1257439232
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006253159478435306,
      "loss": 1.4387,
      "theoretical_loss": 3.6005838805473402,
      "tokens_seen": 1257504768
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006252958876629889,
      "loss": 1.5394,
      "theoretical_loss": 3.6005669068740005,
      "tokens_seen": 1257570304
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006252758274824474,
      "loss": 1.4722,
      "theoretical_loss": 3.6005499343328498,
      "tokens_seen": 1257635840
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006252557673019057,
      "loss": 1.3972,
      "theoretical_loss": 3.6005329629237535,
      "tokens_seen": 1257701376
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006252357071213641,
      "loss": 1.4945,
      "theoretical_loss": 3.600515992646577,
      "tokens_seen": 1257766912
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006252156469408225,
      "loss": 1.4343,
      "theoretical_loss": 3.600499023501187,
      "tokens_seen": 1257832448
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006251955867602809,
      "loss": 1.4564,
      "theoretical_loss": 3.600482055487448,
      "tokens_seen": 1257897984
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006251755265797392,
      "loss": 1.5397,
      "theoretical_loss": 3.600465088605225,
      "tokens_seen": 1257963520
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006251554663991977,
      "loss": 1.4813,
      "theoretical_loss": 3.6004481228543854,
      "tokens_seen": 1258029056
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006251354062186559,
      "loss": 1.3923,
      "theoretical_loss": 3.600431158234793,
      "tokens_seen": 1258094592
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006251153460381144,
      "loss": 1.4404,
      "theoretical_loss": 3.600414194746315,
      "tokens_seen": 1258160128
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006250952858575727,
      "loss": 1.4567,
      "theoretical_loss": 3.600397232388816,
      "tokens_seen": 1258225664
    },
    {
      "epoch": 0.38,
      "objective/train/advantage_avg": 0.4913419187068939,
      "objective/train/docs_used": 460354,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5437325239181519,
      "objective/train/original_loss": 1.5437326431274414,
      "objective/train/theoretical_loss": 3.600380271162162,
      "objective/train/tokens_used": 1278751200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24238081276416779,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503662824630737,
      "objective/train/weighted_lm_loss": 1.6213421821594238,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.600380271162162,
      "tokens_seen": 1258291200
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006250752256770311,
      "loss": 1.5316,
      "theoretical_loss": 3.600380271162162,
      "tokens_seen": 1258291200
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006250551654964895,
      "loss": 1.4208,
      "theoretical_loss": 3.6003633110662188,
      "tokens_seen": 1258356736
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006250351053159478,
      "loss": 1.3874,
      "theoretical_loss": 3.600346352100852,
      "tokens_seen": 1258422272
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006250150451354062,
      "loss": 1.4404,
      "theoretical_loss": 3.600329394265928,
      "tokens_seen": 1258487808
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006249949849548647,
      "loss": 1.4259,
      "theoretical_loss": 3.6003124375613114,
      "tokens_seen": 1258553344
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006249749247743229,
      "loss": 1.5382,
      "theoretical_loss": 3.6002954819868687,
      "tokens_seen": 1258618880
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006249548645937814,
      "loss": 1.535,
      "theoretical_loss": 3.600278527542466,
      "tokens_seen": 1258684416
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006249348044132396,
      "loss": 1.5445,
      "theoretical_loss": 3.6002615742279684,
      "tokens_seen": 1258749952
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006249147442326981,
      "loss": 1.524,
      "theoretical_loss": 3.6002446220432427,
      "tokens_seen": 1258815488
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006248946840521565,
      "loss": 1.5243,
      "theoretical_loss": 3.600227670988154,
      "tokens_seen": 1258881024
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006248746238716148,
      "loss": 1.3861,
      "theoretical_loss": 3.6002107210625685,
      "tokens_seen": 1258946560
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006248545636910732,
      "loss": 1.5428,
      "theoretical_loss": 3.600193772266352,
      "tokens_seen": 1259012096
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006248345035105317,
      "loss": 1.6294,
      "theoretical_loss": 3.6001768245993713,
      "tokens_seen": 1259077632
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006248144433299899,
      "loss": 1.5099,
      "theoretical_loss": 3.6001598780614916,
      "tokens_seen": 1259143168
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006247943831494484,
      "loss": 1.4396,
      "theoretical_loss": 3.6001429326525787,
      "tokens_seen": 1259208704
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006247743229689066,
      "loss": 1.5423,
      "theoretical_loss": 3.600125988372499,
      "tokens_seen": 1259274240
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006247542627883651,
      "loss": 1.5279,
      "theoretical_loss": 3.600109045221119,
      "tokens_seen": 1259339776
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006247342026078235,
      "loss": 1.5332,
      "theoretical_loss": 3.600092103198304,
      "tokens_seen": 1259405312
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006247141424272818,
      "loss": 1.5593,
      "theoretical_loss": 3.60007516230392,
      "tokens_seen": 1259470848
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006246940822467402,
      "loss": 1.5275,
      "theoretical_loss": 3.600058222537834,
      "tokens_seen": 1259536384
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006246740220661986,
      "loss": 1.5266,
      "theoretical_loss": 3.600041283899912,
      "tokens_seen": 1259601920
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006246539618856569,
      "loss": 1.4666,
      "theoretical_loss": 3.60002434639002,
      "tokens_seen": 1259667456
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006246339017051154,
      "loss": 1.463,
      "theoretical_loss": 3.6000074100080237,
      "tokens_seen": 1259732992
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006246138415245738,
      "loss": 1.518,
      "theoretical_loss": 3.5999904747537896,
      "tokens_seen": 1259798528
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006245937813440321,
      "loss": 1.4125,
      "theoretical_loss": 3.599973540627184,
      "tokens_seen": 1259864064
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006245737211634906,
      "loss": 1.5982,
      "theoretical_loss": 3.599956607628074,
      "tokens_seen": 1259929600
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006245536609829488,
      "loss": 1.537,
      "theoretical_loss": 3.5999396757563242,
      "tokens_seen": 1259995136
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006245336008024073,
      "loss": 1.5398,
      "theoretical_loss": 3.5999227450118023,
      "tokens_seen": 1260060672
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006245135406218656,
      "loss": 1.4107,
      "theoretical_loss": 3.599905815394374,
      "tokens_seen": 1260126208
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000624493480441324,
      "loss": 1.5225,
      "theoretical_loss": 3.599888886903906,
      "tokens_seen": 1260191744
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006244734202607824,
      "loss": 1.5496,
      "theoretical_loss": 3.5998719595402644,
      "tokens_seen": 1260257280
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006244533600802408,
      "loss": 1.4212,
      "theoretical_loss": 3.5998550333033155,
      "tokens_seen": 1260322816
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006244332998996991,
      "loss": 1.5532,
      "theoretical_loss": 3.599838108192926,
      "tokens_seen": 1260388352
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006244132397191575,
      "loss": 1.4704,
      "theoretical_loss": 3.5998211842089622,
      "tokens_seen": 1260453888
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006243931795386159,
      "loss": 1.5934,
      "theoretical_loss": 3.5998042613512906,
      "tokens_seen": 1260519424
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006243731193580743,
      "loss": 1.5056,
      "theoretical_loss": 3.599787339619778,
      "tokens_seen": 1260584960
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006243530591775326,
      "loss": 1.3758,
      "theoretical_loss": 3.5997704190142907,
      "tokens_seen": 1260650496
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000624332998996991,
      "loss": 1.4619,
      "theoretical_loss": 3.599753499534695,
      "tokens_seen": 1260716032
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006243129388164493,
      "loss": 1.4812,
      "theoretical_loss": 3.5997365811808577,
      "tokens_seen": 1260781568
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006242928786359078,
      "loss": 1.5946,
      "theoretical_loss": 3.599719663952645,
      "tokens_seen": 1260847104
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006242728184553661,
      "loss": 1.5647,
      "theoretical_loss": 3.599702747849925,
      "tokens_seen": 1260912640
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006242527582748245,
      "loss": 1.5839,
      "theoretical_loss": 3.599685832872562,
      "tokens_seen": 1260978176
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006242326980942829,
      "loss": 1.3983,
      "theoretical_loss": 3.599668919020424,
      "tokens_seen": 1261043712
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006242126379137413,
      "loss": 1.5439,
      "theoretical_loss": 3.5996520062933777,
      "tokens_seen": 1261109248
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006241925777331996,
      "loss": 1.4759,
      "theoretical_loss": 3.5996350946912896,
      "tokens_seen": 1261174784
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000624172517552658,
      "loss": 1.5444,
      "theoretical_loss": 3.5996181842140267,
      "tokens_seen": 1261240320
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006241524573721163,
      "loss": 1.4815,
      "theoretical_loss": 3.599601274861455,
      "tokens_seen": 1261305856
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006241323971915748,
      "loss": 1.3926,
      "theoretical_loss": 3.599584366633442,
      "tokens_seen": 1261371392
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006241123370110331,
      "loss": 1.4409,
      "theoretical_loss": 3.5995674595298537,
      "tokens_seen": 1261436928
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006240922768304915,
      "loss": 1.4248,
      "theoretical_loss": 3.599550553550558,
      "tokens_seen": 1261502464
    },
    {
      "epoch": 0.38,
      "objective/train/advantage_avg": 0.49468809366226196,
      "objective/train/docs_used": 461530,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3616101741790771,
      "objective/train/original_loss": 1.3616101741790771,
      "objective/train/theoretical_loss": 3.599533648695421,
      "objective/train/tokens_used": 1282028000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24501526355743408,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507144927978516,
      "objective/train/weighted_lm_loss": 1.4308217763900757,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.599533648695421,
      "tokens_seen": 1261568000
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006240722166499499,
      "loss": 1.4029,
      "theoretical_loss": 3.599533648695421,
      "tokens_seen": 1261568000
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006240521564694082,
      "loss": 1.5058,
      "theoretical_loss": 3.5995167449643097,
      "tokens_seen": 1261633536
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006240320962888666,
      "loss": 1.5555,
      "theoretical_loss": 3.5994998423570914,
      "tokens_seen": 1261699072
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006240120361083251,
      "loss": 1.4751,
      "theoretical_loss": 3.5994829408736324,
      "tokens_seen": 1261764608
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006239919759277833,
      "loss": 1.456,
      "theoretical_loss": 3.599466040513799,
      "tokens_seen": 1261830144
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006239719157472418,
      "loss": 1.571,
      "theoretical_loss": 3.59944914127746,
      "tokens_seen": 1261895680
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006239518555667,
      "loss": 1.3518,
      "theoretical_loss": 3.599432243164481,
      "tokens_seen": 1261961216
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006239317953861585,
      "loss": 1.4308,
      "theoretical_loss": 3.5994153461747294,
      "tokens_seen": 1262026752
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006239117352056169,
      "loss": 1.556,
      "theoretical_loss": 3.5993984503080725,
      "tokens_seen": 1262092288
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006238916750250752,
      "loss": 1.4508,
      "theoretical_loss": 3.599381555564376,
      "tokens_seen": 1262157824
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006238716148445336,
      "loss": 1.4345,
      "theoretical_loss": 3.599364661943509,
      "tokens_seen": 1262223360
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006238515546639921,
      "loss": 1.5632,
      "theoretical_loss": 3.599347769445337,
      "tokens_seen": 1262288896
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006238314944834503,
      "loss": 1.3921,
      "theoretical_loss": 3.5993308780697277,
      "tokens_seen": 1262354432
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006238114343029088,
      "loss": 1.4766,
      "theoretical_loss": 3.599313987816548,
      "tokens_seen": 1262419968
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000623791374122367,
      "loss": 1.534,
      "theoretical_loss": 3.599297098685666,
      "tokens_seen": 1262485504
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006237713139418255,
      "loss": 1.3713,
      "theoretical_loss": 3.5992802106769473,
      "tokens_seen": 1262551040
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006237512537612839,
      "loss": 1.5112,
      "theoretical_loss": 3.5992633237902605,
      "tokens_seen": 1262616576
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006237311935807422,
      "loss": 1.4552,
      "theoretical_loss": 3.599246438025472,
      "tokens_seen": 1262682112
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006237111334002006,
      "loss": 1.4899,
      "theoretical_loss": 3.599229553382449,
      "tokens_seen": 1262747648
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000623691073219659,
      "loss": 1.4916,
      "theoretical_loss": 3.5992126698610596,
      "tokens_seen": 1262813184
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006236710130391173,
      "loss": 1.4938,
      "theoretical_loss": 3.59919578746117,
      "tokens_seen": 1262878720
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006236509528585758,
      "loss": 1.5082,
      "theoretical_loss": 3.5991789061826482,
      "tokens_seen": 1262944256
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000623630892678034,
      "loss": 1.4782,
      "theoretical_loss": 3.599162026025361,
      "tokens_seen": 1263009792
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006236108324974925,
      "loss": 1.4337,
      "theoretical_loss": 3.5991451469891764,
      "tokens_seen": 1263075328
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006235907723169508,
      "loss": 1.4429,
      "theoretical_loss": 3.5991282690739617,
      "tokens_seen": 1263140864
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006235707121364092,
      "loss": 1.4925,
      "theoretical_loss": 3.5991113922795837,
      "tokens_seen": 1263206400
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006235506519558676,
      "loss": 1.438,
      "theoretical_loss": 3.599094516605911,
      "tokens_seen": 1263271936
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000623530591775326,
      "loss": 1.4518,
      "theoretical_loss": 3.5990776420528094,
      "tokens_seen": 1263337472
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006235105315947843,
      "loss": 1.5623,
      "theoretical_loss": 3.5990607686201477,
      "tokens_seen": 1263403008
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006234904714142428,
      "loss": 1.4977,
      "theoretical_loss": 3.5990438963077924,
      "tokens_seen": 1263468544
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006234704112337011,
      "loss": 1.5592,
      "theoretical_loss": 3.5990270251156122,
      "tokens_seen": 1263534080
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006234503510531595,
      "loss": 1.5605,
      "theoretical_loss": 3.5990101550434734,
      "tokens_seen": 1263599616
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006234302908726178,
      "loss": 1.5524,
      "theoretical_loss": 3.5989932860912446,
      "tokens_seen": 1263665152
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006234102306920763,
      "loss": 1.5215,
      "theoretical_loss": 3.5989764182587924,
      "tokens_seen": 1263730688
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006233901705115346,
      "loss": 1.499,
      "theoretical_loss": 3.5989595515459856,
      "tokens_seen": 1263796224
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000623370110330993,
      "loss": 1.5136,
      "theoretical_loss": 3.5989426859526907,
      "tokens_seen": 1263861760
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006233500501504513,
      "loss": 1.5047,
      "theoretical_loss": 3.598925821478776,
      "tokens_seen": 1263927296
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006233299899699097,
      "loss": 1.5665,
      "theoretical_loss": 3.5989089581241087,
      "tokens_seen": 1263992832
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006233099297893681,
      "loss": 1.5173,
      "theoretical_loss": 3.598892095888557,
      "tokens_seen": 1264058368
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006232898696088265,
      "loss": 1.332,
      "theoretical_loss": 3.598875234771988,
      "tokens_seen": 1264123904
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006232698094282848,
      "loss": 1.5262,
      "theoretical_loss": 3.59885837477427,
      "tokens_seen": 1264189440
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006232497492477433,
      "loss": 1.5035,
      "theoretical_loss": 3.5988415158952707,
      "tokens_seen": 1264254976
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006232296890672016,
      "loss": 1.5194,
      "theoretical_loss": 3.5988246581348573,
      "tokens_seen": 1264320512
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00062320962888666,
      "loss": 1.5032,
      "theoretical_loss": 3.5988078014928986,
      "tokens_seen": 1264386048
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006231895687061183,
      "loss": 1.4819,
      "theoretical_loss": 3.5987909459692617,
      "tokens_seen": 1264451584
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006231695085255767,
      "loss": 1.5943,
      "theoretical_loss": 3.5987740915638144,
      "tokens_seen": 1264517120
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006231494483450351,
      "loss": 1.42,
      "theoretical_loss": 3.5987572382764252,
      "tokens_seen": 1264582656
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006231293881644935,
      "loss": 1.4792,
      "theoretical_loss": 3.5987403861069613,
      "tokens_seen": 1264648192
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006231093279839519,
      "loss": 1.4204,
      "theoretical_loss": 3.5987235350552913,
      "tokens_seen": 1264713728
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006230892678034103,
      "loss": 1.5521,
      "theoretical_loss": 3.5987066851212823,
      "tokens_seen": 1264779264
    },
    {
      "epoch": 0.38,
      "objective/train/advantage_avg": 0.49422839283943176,
      "objective/train/docs_used": 462829,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4624844789505005,
      "objective/train/original_loss": 1.462484359741211,
      "objective/train/theoretical_loss": 3.5986898363048025,
      "objective/train/tokens_used": 1285304800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24454471468925476,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506659746170044,
      "objective/train/weighted_lm_loss": 1.5365638732910156,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5986898363048025,
      "tokens_seen": 1264844800
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006230692076228686,
      "loss": 1.498,
      "theoretical_loss": 3.5986898363048025,
      "tokens_seen": 1264844800
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000623049147442327,
      "loss": 1.4612,
      "theoretical_loss": 3.598672988605721,
      "tokens_seen": 1264910336
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006230290872617855,
      "loss": 1.3746,
      "theoretical_loss": 3.598656142023904,
      "tokens_seen": 1264975872
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006230090270812437,
      "loss": 1.4859,
      "theoretical_loss": 3.598639296559221,
      "tokens_seen": 1265041408
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006229889669007022,
      "loss": 1.5303,
      "theoretical_loss": 3.5986224522115395,
      "tokens_seen": 1265106944
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006229689067201604,
      "loss": 1.4343,
      "theoretical_loss": 3.5986056089807272,
      "tokens_seen": 1265172480
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006229488465396189,
      "loss": 1.5158,
      "theoretical_loss": 3.598588766866653,
      "tokens_seen": 1265238016
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006229287863590773,
      "loss": 1.4352,
      "theoretical_loss": 3.5985719258691846,
      "tokens_seen": 1265303552
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006229087261785356,
      "loss": 1.4034,
      "theoretical_loss": 3.59855508598819,
      "tokens_seen": 1265369088
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000622888665997994,
      "loss": 1.4493,
      "theoretical_loss": 3.598538247223537,
      "tokens_seen": 1265434624
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006228686058174525,
      "loss": 1.5025,
      "theoretical_loss": 3.598521409575095,
      "tokens_seen": 1265500160
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006228485456369107,
      "loss": 1.5577,
      "theoretical_loss": 3.5985045730427307,
      "tokens_seen": 1265565696
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006228284854563692,
      "loss": 1.5542,
      "theoretical_loss": 3.5984877376263142,
      "tokens_seen": 1265631232
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006228084252758274,
      "loss": 1.4023,
      "theoretical_loss": 3.598470903325712,
      "tokens_seen": 1265696768
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006227883650952859,
      "loss": 1.4096,
      "theoretical_loss": 3.598454070140793,
      "tokens_seen": 1265762304
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006227683049147443,
      "loss": 1.5033,
      "theoretical_loss": 3.598437238071426,
      "tokens_seen": 1265827840
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006227482447342026,
      "loss": 1.4315,
      "theoretical_loss": 3.5984204071174783,
      "tokens_seen": 1265893376
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000622728184553661,
      "loss": 1.51,
      "theoretical_loss": 3.5984035772788197,
      "tokens_seen": 1265958912
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006227081243731194,
      "loss": 1.4947,
      "theoretical_loss": 3.598386748555317,
      "tokens_seen": 1266024448
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006226880641925777,
      "loss": 1.591,
      "theoretical_loss": 3.598369920946839,
      "tokens_seen": 1266089984
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006226680040120362,
      "loss": 1.6828,
      "theoretical_loss": 3.598353094453255,
      "tokens_seen": 1266155520
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006226479438314944,
      "loss": 1.595,
      "theoretical_loss": 3.5983362690744327,
      "tokens_seen": 1266221056
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006226278836509529,
      "loss": 1.391,
      "theoretical_loss": 3.5983194448102402,
      "tokens_seen": 1266286592
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006226078234704112,
      "loss": 1.545,
      "theoretical_loss": 3.5983026216605474,
      "tokens_seen": 1266352128
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006225877632898696,
      "loss": 1.4267,
      "theoretical_loss": 3.598285799625221,
      "tokens_seen": 1266417664
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000622567703109328,
      "loss": 1.5612,
      "theoretical_loss": 3.5982689787041306,
      "tokens_seen": 1266483200
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006225476429287864,
      "loss": 1.4698,
      "theoretical_loss": 3.598252158897145,
      "tokens_seen": 1266548736
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006225275827482447,
      "loss": 1.39,
      "theoretical_loss": 3.5982353402041314,
      "tokens_seen": 1266614272
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006225075225677032,
      "loss": 1.5058,
      "theoretical_loss": 3.5982185226249594,
      "tokens_seen": 1266679808
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006224874623871615,
      "loss": 1.4889,
      "theoretical_loss": 3.5982017061594984,
      "tokens_seen": 1266745344
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006224674022066199,
      "loss": 1.4286,
      "theoretical_loss": 3.598184890807615,
      "tokens_seen": 1266810880
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006224473420260782,
      "loss": 1.4521,
      "theoretical_loss": 3.5981680765691797,
      "tokens_seen": 1266876416
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006224272818455367,
      "loss": 1.518,
      "theoretical_loss": 3.5981512634440604,
      "tokens_seen": 1266941952
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000622407221664995,
      "loss": 1.4362,
      "theoretical_loss": 3.5981344514321254,
      "tokens_seen": 1267007488
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006223871614844534,
      "loss": 1.5318,
      "theoretical_loss": 3.5981176405332445,
      "tokens_seen": 1267073024
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006223671013039117,
      "loss": 1.5041,
      "theoretical_loss": 3.5981008307472857,
      "tokens_seen": 1267138560
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006223470411233701,
      "loss": 1.4844,
      "theoretical_loss": 3.5980840220741177,
      "tokens_seen": 1267204096
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006223269809428285,
      "loss": 1.5852,
      "theoretical_loss": 3.5980672145136094,
      "tokens_seen": 1267269632
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006223069207622869,
      "loss": 1.6188,
      "theoretical_loss": 3.59805040806563,
      "tokens_seen": 1267335168
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006222868605817452,
      "loss": 1.4331,
      "theoretical_loss": 3.5980336027300477,
      "tokens_seen": 1267400704
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006222668004012037,
      "loss": 1.5215,
      "theoretical_loss": 3.5980167985067317,
      "tokens_seen": 1267466240
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000622246740220662,
      "loss": 1.5234,
      "theoretical_loss": 3.597999995395551,
      "tokens_seen": 1267531776
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006222266800401204,
      "loss": 1.5869,
      "theoretical_loss": 3.5979831933963746,
      "tokens_seen": 1267597312
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006222066198595787,
      "loss": 1.3786,
      "theoretical_loss": 3.597966392509071,
      "tokens_seen": 1267662848
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006221865596790371,
      "loss": 1.5779,
      "theoretical_loss": 3.5979495927335092,
      "tokens_seen": 1267728384
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006221664994984955,
      "loss": 1.4727,
      "theoretical_loss": 3.5979327940695582,
      "tokens_seen": 1267793920
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006221464393179539,
      "loss": 1.5028,
      "theoretical_loss": 3.597915996517087,
      "tokens_seen": 1267859456
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006221263791374122,
      "loss": 1.5057,
      "theoretical_loss": 3.5978992000759655,
      "tokens_seen": 1267924992
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006221063189568707,
      "loss": 1.4717,
      "theoretical_loss": 3.5978824047460614,
      "tokens_seen": 1267990528
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006220862587763289,
      "loss": 1.5626,
      "theoretical_loss": 3.5978656105272444,
      "tokens_seen": 1268056064
    },
    {
      "epoch": 0.38,
      "objective/train/advantage_avg": 0.4927498400211334,
      "objective/train/docs_used": 464032,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4968583583831787,
      "objective/train/original_loss": 1.4968582391738892,
      "objective/train/theoretical_loss": 3.5978488174193832,
      "objective/train/tokens_used": 1288581600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24323022365570068,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505115985870361,
      "objective/train/weighted_lm_loss": 1.5722885131835938,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5978488174193832,
      "tokens_seen": 1268121600
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006220661985957874,
      "loss": 1.5361,
      "theoretical_loss": 3.5978488174193832,
      "tokens_seen": 1268121600
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006220461384152458,
      "loss": 1.4734,
      "theoretical_loss": 3.5978320254223473,
      "tokens_seen": 1268187136
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006220260782347041,
      "loss": 1.4517,
      "theoretical_loss": 3.5978152345360055,
      "tokens_seen": 1268252672
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006220060180541625,
      "loss": 1.4688,
      "theoretical_loss": 3.597798444760228,
      "tokens_seen": 1268318208
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006219859578736208,
      "loss": 1.5721,
      "theoretical_loss": 3.5977816560948823,
      "tokens_seen": 1268383744
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006219658976930792,
      "loss": 1.3808,
      "theoretical_loss": 3.597764868539839,
      "tokens_seen": 1268449280
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006219458375125377,
      "loss": 1.5267,
      "theoretical_loss": 3.5977480820949657,
      "tokens_seen": 1268514816
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006219257773319959,
      "loss": 1.5927,
      "theoretical_loss": 3.5977312967601334,
      "tokens_seen": 1268580352
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006219057171514544,
      "loss": 1.5074,
      "theoretical_loss": 3.5977145125352106,
      "tokens_seen": 1268645888
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006218856569709128,
      "loss": 1.4885,
      "theoretical_loss": 3.5976977294200667,
      "tokens_seen": 1268711424
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006218655967903711,
      "loss": 1.4941,
      "theoretical_loss": 3.5976809474145703,
      "tokens_seen": 1268776960
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006218455366098295,
      "loss": 1.4662,
      "theoretical_loss": 3.5976641665185918,
      "tokens_seen": 1268842496
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006218254764292878,
      "loss": 1.4308,
      "theoretical_loss": 3.5976473867319996,
      "tokens_seen": 1268908032
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006218054162487462,
      "loss": 1.6264,
      "theoretical_loss": 3.597630608054664,
      "tokens_seen": 1268973568
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006217853560682047,
      "loss": 1.4916,
      "theoretical_loss": 3.597613830486454,
      "tokens_seen": 1269039104
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006217652958876629,
      "loss": 1.4801,
      "theoretical_loss": 3.597597054027238,
      "tokens_seen": 1269104640
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006217452357071214,
      "loss": 1.4787,
      "theoretical_loss": 3.5975802786768876,
      "tokens_seen": 1269170176
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006217251755265797,
      "loss": 1.4369,
      "theoretical_loss": 3.5975635044352705,
      "tokens_seen": 1269235712
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006217051153460381,
      "loss": 1.5098,
      "theoretical_loss": 3.5975467313022564,
      "tokens_seen": 1269301248
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006216850551654965,
      "loss": 1.4841,
      "theoretical_loss": 3.597529959277715,
      "tokens_seen": 1269366784
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006216649949849548,
      "loss": 1.5725,
      "theoretical_loss": 3.5975131883615163,
      "tokens_seen": 1269432320
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006216449348044132,
      "loss": 1.4725,
      "theoretical_loss": 3.597496418553529,
      "tokens_seen": 1269497856
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006216248746238716,
      "loss": 1.5276,
      "theoretical_loss": 3.5974796498536237,
      "tokens_seen": 1269563392
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00062160481444333,
      "loss": 1.4051,
      "theoretical_loss": 3.597462882261669,
      "tokens_seen": 1269628928
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006215847542627884,
      "loss": 1.5178,
      "theoretical_loss": 3.597446115777535,
      "tokens_seen": 1269694464
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006215646940822468,
      "loss": 1.4325,
      "theoretical_loss": 3.5974293504010912,
      "tokens_seen": 1269760000
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006215446339017051,
      "loss": 1.4584,
      "theoretical_loss": 3.597412586132208,
      "tokens_seen": 1269825536
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006215245737211636,
      "loss": 1.536,
      "theoretical_loss": 3.5973958229707534,
      "tokens_seen": 1269891072
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006215045135406219,
      "loss": 1.4214,
      "theoretical_loss": 3.597379060916598,
      "tokens_seen": 1269956608
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006214844533600803,
      "loss": 1.4724,
      "theoretical_loss": 3.597362299969612,
      "tokens_seen": 1270022144
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006214643931795386,
      "loss": 1.5445,
      "theoretical_loss": 3.5973455401296652,
      "tokens_seen": 1270087680
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000621444332998997,
      "loss": 1.4956,
      "theoretical_loss": 3.5973287813966266,
      "tokens_seen": 1270153216
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006214242728184554,
      "loss": 1.4749,
      "theoretical_loss": 3.5973120237703657,
      "tokens_seen": 1270218752
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006214042126379138,
      "loss": 1.3847,
      "theoretical_loss": 3.5972952672507534,
      "tokens_seen": 1270284288
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006213841524573721,
      "loss": 1.4075,
      "theoretical_loss": 3.597278511837659,
      "tokens_seen": 1270349824
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006213640922768305,
      "loss": 1.4794,
      "theoretical_loss": 3.5972617575309522,
      "tokens_seen": 1270415360
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0006213440320962889,
      "loss": 1.4643,
      "theoretical_loss": 3.597245004330503,
      "tokens_seen": 1270480896
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006213239719157473,
      "loss": 1.4354,
      "theoretical_loss": 3.5972282522361816,
      "tokens_seen": 1270546432
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006213039117352056,
      "loss": 1.4943,
      "theoretical_loss": 3.597211501247858,
      "tokens_seen": 1270611968
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006212838515546641,
      "loss": 1.44,
      "theoretical_loss": 3.5971947513654015,
      "tokens_seen": 1270677504
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006212637913741224,
      "loss": 1.4725,
      "theoretical_loss": 3.597178002588682,
      "tokens_seen": 1270743040
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006212437311935808,
      "loss": 1.4393,
      "theoretical_loss": 3.5971612549175704,
      "tokens_seen": 1270808576
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006212236710130391,
      "loss": 1.4339,
      "theoretical_loss": 3.5971445083519358,
      "tokens_seen": 1270874112
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006212036108324975,
      "loss": 1.406,
      "theoretical_loss": 3.597127762891649,
      "tokens_seen": 1270939648
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006211835506519559,
      "loss": 1.5077,
      "theoretical_loss": 3.5971110185365793,
      "tokens_seen": 1271005184
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006211634904714143,
      "loss": 1.5337,
      "theoretical_loss": 3.597094275286597,
      "tokens_seen": 1271070720
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006211434302908726,
      "loss": 1.5918,
      "theoretical_loss": 3.5970775331415723,
      "tokens_seen": 1271136256
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006211233701103311,
      "loss": 1.4479,
      "theoretical_loss": 3.5970607921013755,
      "tokens_seen": 1271201792
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006211033099297893,
      "loss": 1.3888,
      "theoretical_loss": 3.5970440521658764,
      "tokens_seen": 1271267328
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006210832497492478,
      "loss": 1.4589,
      "theoretical_loss": 3.5970273133349453,
      "tokens_seen": 1271332864
    },
    {
      "epoch": 0.39,
      "objective/train/advantage_avg": 0.4927798807621002,
      "objective/train/docs_used": 465264,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4413305521011353,
      "objective/train/original_loss": 1.4413305521011353,
      "objective/train/theoretical_loss": 3.5970105756084525,
      "objective/train/tokens_used": 1291858400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2432139664888382,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505144596099854,
      "objective/train/weighted_lm_loss": 1.513904094696045,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.5970105756084525,
      "tokens_seen": 1271398400
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006210631895687062,
      "loss": 1.489,
      "theoretical_loss": 3.5970105756084525,
      "tokens_seen": 1271398400
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006210431293881645,
      "loss": 1.5309,
      "theoretical_loss": 3.5969938389862683,
      "tokens_seen": 1271463936
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006210230692076229,
      "loss": 1.5387,
      "theoretical_loss": 3.5969771034682627,
      "tokens_seen": 1271529472
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006210030090270812,
      "loss": 1.4311,
      "theoretical_loss": 3.596960369054306,
      "tokens_seen": 1271595008
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006209829488465396,
      "loss": 1.486,
      "theoretical_loss": 3.5969436357442683,
      "tokens_seen": 1271660544
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006209628886659981,
      "loss": 1.484,
      "theoretical_loss": 3.5969269035380202,
      "tokens_seen": 1271726080
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006209428284854563,
      "loss": 1.4912,
      "theoretical_loss": 3.5969101724354315,
      "tokens_seen": 1271791616
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006209227683049148,
      "loss": 1.4685,
      "theoretical_loss": 3.5968934424363734,
      "tokens_seen": 1271857152
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006209027081243732,
      "loss": 1.47,
      "theoretical_loss": 3.5968767135407154,
      "tokens_seen": 1271922688
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006208826479438315,
      "loss": 1.4895,
      "theoretical_loss": 3.5968599857483285,
      "tokens_seen": 1271988224
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006208625877632899,
      "loss": 1.5138,
      "theoretical_loss": 3.5968432590590824,
      "tokens_seen": 1272053760
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006208425275827482,
      "loss": 1.4152,
      "theoretical_loss": 3.596826533472848,
      "tokens_seen": 1272119296
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006208224674022066,
      "loss": 1.4274,
      "theoretical_loss": 3.596809808989496,
      "tokens_seen": 1272184832
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006208024072216651,
      "loss": 1.4911,
      "theoretical_loss": 3.5967930856088968,
      "tokens_seen": 1272250368
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006207823470411233,
      "loss": 1.5279,
      "theoretical_loss": 3.5967763633309207,
      "tokens_seen": 1272315904
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006207622868605818,
      "loss": 1.5663,
      "theoretical_loss": 3.5967596421554378,
      "tokens_seen": 1272381440
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006207422266800401,
      "loss": 1.4038,
      "theoretical_loss": 3.596742922082319,
      "tokens_seen": 1272446976
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006207221664994985,
      "loss": 1.587,
      "theoretical_loss": 3.5967262031114346,
      "tokens_seen": 1272512512
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006207021063189569,
      "loss": 1.3841,
      "theoretical_loss": 3.596709485242656,
      "tokens_seen": 1272578048
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006206820461384153,
      "loss": 1.4373,
      "theoretical_loss": 3.5966927684758527,
      "tokens_seen": 1272643584
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006206619859578736,
      "loss": 1.3819,
      "theoretical_loss": 3.5966760528108965,
      "tokens_seen": 1272709120
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000620641925777332,
      "loss": 1.4885,
      "theoretical_loss": 3.596659338247657,
      "tokens_seen": 1272774656
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006206218655967903,
      "loss": 1.5436,
      "theoretical_loss": 3.5966426247860053,
      "tokens_seen": 1272840192
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006206018054162488,
      "loss": 1.5274,
      "theoretical_loss": 3.596625912425812,
      "tokens_seen": 1272905728
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006205817452357071,
      "loss": 1.415,
      "theoretical_loss": 3.596609201166948,
      "tokens_seen": 1272971264
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006205616850551655,
      "loss": 1.5545,
      "theoretical_loss": 3.5965924910092837,
      "tokens_seen": 1273036800
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006205416248746239,
      "loss": 1.5563,
      "theoretical_loss": 3.5965757819526902,
      "tokens_seen": 1273102336
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006205215646940823,
      "loss": 1.3664,
      "theoretical_loss": 3.596559073997038,
      "tokens_seen": 1273167872
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006205015045135406,
      "loss": 1.4985,
      "theoretical_loss": 3.596542367142198,
      "tokens_seen": 1273233408
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000620481444332999,
      "loss": 1.5485,
      "theoretical_loss": 3.596525661388041,
      "tokens_seen": 1273298944
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006204613841524573,
      "loss": 1.4961,
      "theoretical_loss": 3.5965089567344384,
      "tokens_seen": 1273364480
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006204413239719158,
      "loss": 1.4567,
      "theoretical_loss": 3.5964922531812595,
      "tokens_seen": 1273430016
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006204212637913741,
      "loss": 1.4849,
      "theoretical_loss": 3.5964755507283774,
      "tokens_seen": 1273495552
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006204012036108325,
      "loss": 1.387,
      "theoretical_loss": 3.5964588493756606,
      "tokens_seen": 1273561088
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006203811434302908,
      "loss": 1.5088,
      "theoretical_loss": 3.596442149122982,
      "tokens_seen": 1273626624
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006203610832497493,
      "loss": 1.6389,
      "theoretical_loss": 3.5964254499702117,
      "tokens_seen": 1273692160
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006203410230692076,
      "loss": 1.5408,
      "theoretical_loss": 3.5964087519172203,
      "tokens_seen": 1273757696
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000620320962888666,
      "loss": 1.4432,
      "theoretical_loss": 3.59639205496388,
      "tokens_seen": 1273823232
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006203009027081243,
      "loss": 1.5461,
      "theoretical_loss": 3.5963753591100605,
      "tokens_seen": 1273888768
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006202808425275828,
      "loss": 1.6316,
      "theoretical_loss": 3.596358664355634,
      "tokens_seen": 1273954304
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006202607823470411,
      "loss": 1.535,
      "theoretical_loss": 3.59634197070047,
      "tokens_seen": 1274019840
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006202407221664995,
      "loss": 1.5376,
      "theoretical_loss": 3.596325278144441,
      "tokens_seen": 1274085376
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006202206619859578,
      "loss": 1.5688,
      "theoretical_loss": 3.5963085866874174,
      "tokens_seen": 1274150912
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006202006018054163,
      "loss": 1.4174,
      "theoretical_loss": 3.596291896329271,
      "tokens_seen": 1274216448
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006201805416248746,
      "loss": 1.6514,
      "theoretical_loss": 3.5962752070698727,
      "tokens_seen": 1274281984
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000620160481444333,
      "loss": 1.425,
      "theoretical_loss": 3.5962585189090923,
      "tokens_seen": 1274347520
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006201404212637914,
      "loss": 1.422,
      "theoretical_loss": 3.596241831846803,
      "tokens_seen": 1274413056
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006201203610832497,
      "loss": 1.4004,
      "theoretical_loss": 3.596225145882875,
      "tokens_seen": 1274478592
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006201003009027082,
      "loss": 1.4735,
      "theoretical_loss": 3.59620846101718,
      "tokens_seen": 1274544128
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006200802407221666,
      "loss": 1.5558,
      "theoretical_loss": 3.5961917772495884,
      "tokens_seen": 1274609664
    },
    {
      "epoch": 0.39,
      "objective/train/advantage_avg": 0.48994871973991394,
      "objective/train/docs_used": 466434,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3847399950027466,
      "objective/train/original_loss": 1.3847399950027466,
      "objective/train/theoretical_loss": 3.5961750945799724,
      "objective/train/tokens_used": 1295135200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24063628911972046,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502179861068726,
      "objective/train/weighted_lm_loss": 1.4542152881622314,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5961750945799724,
      "tokens_seen": 1274675200
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006200601805416249,
      "loss": 1.3945,
      "theoretical_loss": 3.5961750945799724,
      "tokens_seen": 1274675200
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006200401203610833,
      "loss": 1.4847,
      "theoretical_loss": 3.5961584130082027,
      "tokens_seen": 1274740736
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006200200601805416,
      "loss": 1.5441,
      "theoretical_loss": 3.596141732534151,
      "tokens_seen": 1274806272
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00062,
      "loss": 1.5861,
      "theoretical_loss": 3.5961250531576883,
      "tokens_seen": 1274871808
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006199799398194585,
      "loss": 1.5325,
      "theoretical_loss": 3.5961083748786864,
      "tokens_seen": 1274937344
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006199598796389167,
      "loss": 1.5728,
      "theoretical_loss": 3.596091697697016,
      "tokens_seen": 1275002880
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006199398194583752,
      "loss": 1.4752,
      "theoretical_loss": 3.596075021612549,
      "tokens_seen": 1275068416
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006199197592778336,
      "loss": 1.4622,
      "theoretical_loss": 3.5960583466251568,
      "tokens_seen": 1275133952
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006198996990972919,
      "loss": 1.538,
      "theoretical_loss": 3.5960416727347106,
      "tokens_seen": 1275199488
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006198796389167503,
      "loss": 1.4627,
      "theoretical_loss": 3.5960249999410827,
      "tokens_seen": 1275265024
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006198595787362086,
      "loss": 1.6031,
      "theoretical_loss": 3.5960083282441433,
      "tokens_seen": 1275330560
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000619839518555667,
      "loss": 1.5361,
      "theoretical_loss": 3.5959916576437645,
      "tokens_seen": 1275396096
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006198194583751255,
      "loss": 1.6073,
      "theoretical_loss": 3.595974988139818,
      "tokens_seen": 1275461632
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006197993981945837,
      "loss": 1.4794,
      "theoretical_loss": 3.5959583197321754,
      "tokens_seen": 1275527168
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006197793380140422,
      "loss": 1.498,
      "theoretical_loss": 3.5959416524207084,
      "tokens_seen": 1275592704
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006197592778335005,
      "loss": 1.5499,
      "theoretical_loss": 3.5959249862052873,
      "tokens_seen": 1275658240
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006197392176529589,
      "loss": 1.5567,
      "theoretical_loss": 3.5959083210857856,
      "tokens_seen": 1275723776
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006197191574724173,
      "loss": 1.4893,
      "theoretical_loss": 3.595891657062074,
      "tokens_seen": 1275789312
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006196990972918757,
      "loss": 1.444,
      "theoretical_loss": 3.5958749941340242,
      "tokens_seen": 1275854848
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000619679037111334,
      "loss": 1.5203,
      "theoretical_loss": 3.595858332301508,
      "tokens_seen": 1275920384
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006196589769307924,
      "loss": 1.5416,
      "theoretical_loss": 3.595841671564397,
      "tokens_seen": 1275985920
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006196389167502507,
      "loss": 1.4716,
      "theoretical_loss": 3.595825011922563,
      "tokens_seen": 1276051456
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006196188565697092,
      "loss": 1.5179,
      "theoretical_loss": 3.595808353375877,
      "tokens_seen": 1276116992
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006195987963891675,
      "loss": 1.5164,
      "theoretical_loss": 3.595791695924212,
      "tokens_seen": 1276182528
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006195787362086259,
      "loss": 1.5957,
      "theoretical_loss": 3.5957750395674397,
      "tokens_seen": 1276248064
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006195586760280843,
      "loss": 1.4705,
      "theoretical_loss": 3.595758384305431,
      "tokens_seen": 1276313600
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006195386158475427,
      "loss": 1.3663,
      "theoretical_loss": 3.5957417301380588,
      "tokens_seen": 1276379136
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000619518555667001,
      "loss": 1.5334,
      "theoretical_loss": 3.595725077065194,
      "tokens_seen": 1276444672
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006194984954864594,
      "loss": 1.5003,
      "theoretical_loss": 3.5957084250867086,
      "tokens_seen": 1276510208
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006194784353059177,
      "loss": 1.5056,
      "theoretical_loss": 3.595691774202475,
      "tokens_seen": 1276575744
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006194583751253762,
      "loss": 1.5142,
      "theoretical_loss": 3.595675124412365,
      "tokens_seen": 1276641280
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006194383149448345,
      "loss": 1.6231,
      "theoretical_loss": 3.5956584757162506,
      "tokens_seen": 1276706816
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006194182547642929,
      "loss": 1.6415,
      "theoretical_loss": 3.5956418281140032,
      "tokens_seen": 1276772352
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006193981945837512,
      "loss": 1.473,
      "theoretical_loss": 3.595625181605495,
      "tokens_seen": 1276837888
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006193781344032097,
      "loss": 1.5461,
      "theoretical_loss": 3.595608536190599,
      "tokens_seen": 1276903424
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000619358074222668,
      "loss": 1.4043,
      "theoretical_loss": 3.595591891869186,
      "tokens_seen": 1276968960
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006193380140421264,
      "loss": 1.4932,
      "theoretical_loss": 3.5955752486411283,
      "tokens_seen": 1277034496
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006193179538615847,
      "loss": 1.4585,
      "theoretical_loss": 3.595558606506298,
      "tokens_seen": 1277100032
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006192978936810432,
      "loss": 1.6136,
      "theoretical_loss": 3.595541965464568,
      "tokens_seen": 1277165568
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006192778335005015,
      "loss": 1.4549,
      "theoretical_loss": 3.5955253255158093,
      "tokens_seen": 1277231104
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006192577733199599,
      "loss": 1.376,
      "theoretical_loss": 3.595508686659895,
      "tokens_seen": 1277296640
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006192377131394182,
      "loss": 1.4503,
      "theoretical_loss": 3.5954920488966966,
      "tokens_seen": 1277362176
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006192176529588767,
      "loss": 1.499,
      "theoretical_loss": 3.5954754122260866,
      "tokens_seen": 1277427712
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000619197592778335,
      "loss": 1.5083,
      "theoretical_loss": 3.5954587766479364,
      "tokens_seen": 1277493248
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006191775325977934,
      "loss": 1.5156,
      "theoretical_loss": 3.595442142162119,
      "tokens_seen": 1277558784
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006191574724172518,
      "loss": 1.5138,
      "theoretical_loss": 3.595425508768507,
      "tokens_seen": 1277624320
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006191374122367101,
      "loss": 1.5475,
      "theoretical_loss": 3.595408876466972,
      "tokens_seen": 1277689856
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006191173520561685,
      "loss": 1.4824,
      "theoretical_loss": 3.5953922452573863,
      "tokens_seen": 1277755392
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000619097291875627,
      "loss": 1.4487,
      "theoretical_loss": 3.5953756151396226,
      "tokens_seen": 1277820928
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006190772316950852,
      "loss": 1.5828,
      "theoretical_loss": 3.595358986113553,
      "tokens_seen": 1277886464
    },
    {
      "epoch": 0.39,
      "objective/train/advantage_avg": 0.4862214922904968,
      "objective/train/docs_used": 467639,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3731242418289185,
      "objective/train/original_loss": 1.373124361038208,
      "objective/train/theoretical_loss": 3.59534235817905,
      "objective/train/tokens_used": 1298412000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23886923491954803,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.049836277961731,
      "objective/train/weighted_lm_loss": 1.4405465126037598,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.59534235817905,
      "tokens_seen": 1277952000
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006190571715145437,
      "loss": 1.4757,
      "theoretical_loss": 3.59534235817905,
      "tokens_seen": 1277952000
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006190371113340019,
      "loss": 1.5093,
      "theoretical_loss": 3.5953257313359854,
      "tokens_seen": 1278017536
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006190170511534604,
      "loss": 1.4668,
      "theoretical_loss": 3.595309105584233,
      "tokens_seen": 1278083072
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006189969909729188,
      "loss": 1.5242,
      "theoretical_loss": 3.5952924809236633,
      "tokens_seen": 1278148608
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006189769307923771,
      "loss": 1.5039,
      "theoretical_loss": 3.5952758573541503,
      "tokens_seen": 1278214144
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006189568706118355,
      "loss": 1.527,
      "theoretical_loss": 3.5952592348755656,
      "tokens_seen": 1278279680
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000618936810431294,
      "loss": 1.369,
      "theoretical_loss": 3.5952426134877817,
      "tokens_seen": 1278345216
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006189167502507522,
      "loss": 1.4163,
      "theoretical_loss": 3.5952259931906716,
      "tokens_seen": 1278410752
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006188966900702107,
      "loss": 1.617,
      "theoretical_loss": 3.5952093739841073,
      "tokens_seen": 1278476288
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006188766298896689,
      "loss": 1.4779,
      "theoretical_loss": 3.5951927558679624,
      "tokens_seen": 1278541824
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006188565697091274,
      "loss": 1.5676,
      "theoretical_loss": 3.595176138842108,
      "tokens_seen": 1278607360
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006188365095285858,
      "loss": 1.5355,
      "theoretical_loss": 3.5951595229064175,
      "tokens_seen": 1278672896
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006188164493480441,
      "loss": 1.5376,
      "theoretical_loss": 3.5951429080607635,
      "tokens_seen": 1278738432
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006187963891675025,
      "loss": 1.4693,
      "theoretical_loss": 3.5951262943050186,
      "tokens_seen": 1278803968
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006187763289869609,
      "loss": 1.4233,
      "theoretical_loss": 3.5951096816390553,
      "tokens_seen": 1278869504
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006187562688064192,
      "loss": 1.4336,
      "theoretical_loss": 3.595093070062746,
      "tokens_seen": 1278935040
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006187362086258777,
      "loss": 1.5822,
      "theoretical_loss": 3.5950764595759646,
      "tokens_seen": 1279000576
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006187161484453359,
      "loss": 1.4308,
      "theoretical_loss": 3.5950598501785826,
      "tokens_seen": 1279066112
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006186960882647944,
      "loss": 1.5318,
      "theoretical_loss": 3.5950432418704725,
      "tokens_seen": 1279131648
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006186760280842527,
      "loss": 1.5269,
      "theoretical_loss": 3.5950266346515085,
      "tokens_seen": 1279197184
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006186559679037111,
      "loss": 1.4806,
      "theoretical_loss": 3.5950100285215623,
      "tokens_seen": 1279262720
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006186359077231695,
      "loss": 1.4243,
      "theoretical_loss": 3.5949934234805068,
      "tokens_seen": 1279328256
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006186158475426279,
      "loss": 1.4088,
      "theoretical_loss": 3.594976819528215,
      "tokens_seen": 1279393792
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006185957873620863,
      "loss": 1.499,
      "theoretical_loss": 3.5949602166645596,
      "tokens_seen": 1279459328
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006185757271815447,
      "loss": 1.467,
      "theoretical_loss": 3.594943614889414,
      "tokens_seen": 1279524864
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006185556670010031,
      "loss": 1.6048,
      "theoretical_loss": 3.59492701420265,
      "tokens_seen": 1279590400
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006185356068204614,
      "loss": 1.5494,
      "theoretical_loss": 3.5949104146041417,
      "tokens_seen": 1279655936
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006185155466399198,
      "loss": 1.4369,
      "theoretical_loss": 3.5948938160937614,
      "tokens_seen": 1279721472
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006184954864593781,
      "loss": 1.5569,
      "theoretical_loss": 3.5948772186713827,
      "tokens_seen": 1279787008
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006184754262788366,
      "loss": 1.474,
      "theoretical_loss": 3.594860622336877,
      "tokens_seen": 1279852544
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006184553660982949,
      "loss": 1.461,
      "theoretical_loss": 3.5948440270901187,
      "tokens_seen": 1279918080
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006184353059177533,
      "loss": 1.4984,
      "theoretical_loss": 3.5948274329309813,
      "tokens_seen": 1279983616
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006184152457372116,
      "loss": 1.5546,
      "theoretical_loss": 3.5948108398593357,
      "tokens_seen": 1280049152
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006183951855566701,
      "loss": 1.5552,
      "theoretical_loss": 3.594794247875057,
      "tokens_seen": 1280114688
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006183751253761284,
      "loss": 1.4736,
      "theoretical_loss": 3.5947776569780174,
      "tokens_seen": 1280180224
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006183550651955868,
      "loss": 1.5274,
      "theoretical_loss": 3.5947610671680907,
      "tokens_seen": 1280245760
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006183350050150452,
      "loss": 1.498,
      "theoretical_loss": 3.5947444784451488,
      "tokens_seen": 1280311296
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006183149448345036,
      "loss": 1.5625,
      "theoretical_loss": 3.5947278908090654,
      "tokens_seen": 1280376832
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006182948846539619,
      "loss": 1.458,
      "theoretical_loss": 3.594711304259714,
      "tokens_seen": 1280442368
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006182748244734203,
      "loss": 1.4259,
      "theoretical_loss": 3.594694718796968,
      "tokens_seen": 1280507904
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006182547642928786,
      "loss": 1.4393,
      "theoretical_loss": 3.5946781344206995,
      "tokens_seen": 1280573440
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006182347041123371,
      "loss": 1.5479,
      "theoretical_loss": 3.5946615511307827,
      "tokens_seen": 1280638976
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006182146439317954,
      "loss": 1.5885,
      "theoretical_loss": 3.5946449689270903,
      "tokens_seen": 1280704512
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006181945837512538,
      "loss": 1.393,
      "theoretical_loss": 3.5946283878094967,
      "tokens_seen": 1280770048
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006181745235707122,
      "loss": 1.6013,
      "theoretical_loss": 3.5946118077778735,
      "tokens_seen": 1280835584
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006181544633901705,
      "loss": 1.6187,
      "theoretical_loss": 3.5945952288320946,
      "tokens_seen": 1280901120
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006181344032096289,
      "loss": 1.5098,
      "theoretical_loss": 3.594578650972034,
      "tokens_seen": 1280966656
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006181143430290874,
      "loss": 1.5839,
      "theoretical_loss": 3.5945620741975643,
      "tokens_seen": 1281032192
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006180942828485456,
      "loss": 1.5197,
      "theoretical_loss": 3.59454549850856,
      "tokens_seen": 1281097728
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006180742226680041,
      "loss": 1.4657,
      "theoretical_loss": 3.594528923904893,
      "tokens_seen": 1281163264
    },
    {
      "epoch": 0.39,
      "objective/train/advantage_avg": 0.4932679533958435,
      "objective/train/docs_used": 468763,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.6008708477020264,
      "objective/train/original_loss": 1.6008708477020264,
      "objective/train/theoretical_loss": 3.594512350386438,
      "objective/train/tokens_used": 1301688800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2438812255859375,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505666732788086,
      "objective/train/weighted_lm_loss": 1.6816649436950684,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.594512350386438,
      "tokens_seen": 1281228800
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006180541624874623,
      "loss": 1.5942,
      "theoretical_loss": 3.594512350386438,
      "tokens_seen": 1281228800
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006180341023069208,
      "loss": 1.5437,
      "theoretical_loss": 3.5944957779530675,
      "tokens_seen": 1281294336
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006180140421263792,
      "loss": 1.4598,
      "theoretical_loss": 3.594479206604655,
      "tokens_seen": 1281359872
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006179939819458375,
      "loss": 1.6272,
      "theoretical_loss": 3.594462636341075,
      "tokens_seen": 1281425408
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006179739217652959,
      "loss": 1.4597,
      "theoretical_loss": 3.5944460671622,
      "tokens_seen": 1281490944
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006179538615847544,
      "loss": 1.5661,
      "theoretical_loss": 3.5944294990679033,
      "tokens_seen": 1281556480
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006179338014042126,
      "loss": 1.4894,
      "theoretical_loss": 3.59441293205806,
      "tokens_seen": 1281622016
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006179137412236711,
      "loss": 1.4961,
      "theoretical_loss": 3.5943963661325418,
      "tokens_seen": 1281687552
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006178936810431293,
      "loss": 1.4485,
      "theoretical_loss": 3.5943798012912236,
      "tokens_seen": 1281753088
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006178736208625878,
      "loss": 1.5641,
      "theoretical_loss": 3.5943632375339787,
      "tokens_seen": 1281818624
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006178535606820462,
      "loss": 1.4634,
      "theoretical_loss": 3.59434667486068,
      "tokens_seen": 1281884160
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006178335005015045,
      "loss": 1.4654,
      "theoretical_loss": 3.5943301132712024,
      "tokens_seen": 1281949696
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006178134403209629,
      "loss": 1.4846,
      "theoretical_loss": 3.594313552765419,
      "tokens_seen": 1282015232
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006177933801404213,
      "loss": 1.5476,
      "theoretical_loss": 3.594296993343203,
      "tokens_seen": 1282080768
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006177733199598796,
      "loss": 1.4802,
      "theoretical_loss": 3.5942804350044284,
      "tokens_seen": 1282146304
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006177532597793381,
      "loss": 1.5236,
      "theoretical_loss": 3.5942638777489697,
      "tokens_seen": 1282211840
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006177331995987963,
      "loss": 1.5038,
      "theoretical_loss": 3.5942473215766997,
      "tokens_seen": 1282277376
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006177131394182548,
      "loss": 1.4459,
      "theoretical_loss": 3.5942307664874926,
      "tokens_seen": 1282342912
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006176930792377131,
      "loss": 1.4571,
      "theoretical_loss": 3.594214212481222,
      "tokens_seen": 1282408448
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006176730190571715,
      "loss": 1.4817,
      "theoretical_loss": 3.594197659557762,
      "tokens_seen": 1282473984
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006176529588766299,
      "loss": 1.3514,
      "theoretical_loss": 3.5941811077169863,
      "tokens_seen": 1282539520
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006176328986960883,
      "loss": 1.6033,
      "theoretical_loss": 3.5941645569587686,
      "tokens_seen": 1282605056
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006176128385155466,
      "loss": 1.3385,
      "theoretical_loss": 3.594148007282983,
      "tokens_seen": 1282670592
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006175927783350051,
      "loss": 1.3861,
      "theoretical_loss": 3.5941314586895037,
      "tokens_seen": 1282736128
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006175727181544633,
      "loss": 1.3734,
      "theoretical_loss": 3.594114911178204,
      "tokens_seen": 1282801664
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006175526579739218,
      "loss": 1.3016,
      "theoretical_loss": 3.594098364748958,
      "tokens_seen": 1282867200
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006175325977933801,
      "loss": 1.4481,
      "theoretical_loss": 3.59408181940164,
      "tokens_seen": 1282932736
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006175125376128385,
      "loss": 1.4714,
      "theoretical_loss": 3.594065275136124,
      "tokens_seen": 1282998272
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006174924774322969,
      "loss": 1.4075,
      "theoretical_loss": 3.5940487319522836,
      "tokens_seen": 1283063808
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006174724172517553,
      "loss": 1.5239,
      "theoretical_loss": 3.594032189849993,
      "tokens_seen": 1283129344
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006174523570712136,
      "loss": 1.4471,
      "theoretical_loss": 3.5940156488291266,
      "tokens_seen": 1283194880
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000617432296890672,
      "loss": 1.5064,
      "theoretical_loss": 3.593999108889558,
      "tokens_seen": 1283260416
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006174122367101304,
      "loss": 1.4211,
      "theoretical_loss": 3.5939825700311614,
      "tokens_seen": 1283325952
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006173921765295888,
      "loss": 1.4755,
      "theoretical_loss": 3.593966032253811,
      "tokens_seen": 1283391488
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006173721163490471,
      "loss": 1.3954,
      "theoretical_loss": 3.593949495557381,
      "tokens_seen": 1283457024
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006173520561685056,
      "loss": 1.4398,
      "theoretical_loss": 3.5939329599417458,
      "tokens_seen": 1283522560
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006173319959879638,
      "loss": 1.431,
      "theoretical_loss": 3.5939164254067792,
      "tokens_seen": 1283588096
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006173119358074223,
      "loss": 1.5255,
      "theoretical_loss": 3.593899891952355,
      "tokens_seen": 1283653632
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006172918756268806,
      "loss": 1.5273,
      "theoretical_loss": 3.5938833595783484,
      "tokens_seen": 1283719168
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000617271815446339,
      "loss": 1.4313,
      "theoretical_loss": 3.593866828284633,
      "tokens_seen": 1283784704
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006172517552657974,
      "loss": 1.6491,
      "theoretical_loss": 3.593850298071083,
      "tokens_seen": 1283850240
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006172316950852558,
      "loss": 1.4934,
      "theoretical_loss": 3.5938337689375737,
      "tokens_seen": 1283915776
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006172116349047141,
      "loss": 1.351,
      "theoretical_loss": 3.593817240883978,
      "tokens_seen": 1283981312
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006171915747241726,
      "loss": 1.4609,
      "theoretical_loss": 3.593800713910171,
      "tokens_seen": 1284046848
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006171715145436308,
      "loss": 1.5312,
      "theoretical_loss": 3.593784188016027,
      "tokens_seen": 1284112384
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006171514543630893,
      "loss": 1.4617,
      "theoretical_loss": 3.59376766320142,
      "tokens_seen": 1284177920
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006171313941825476,
      "loss": 1.4745,
      "theoretical_loss": 3.5937511394662254,
      "tokens_seen": 1284243456
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000617111334002006,
      "loss": 1.5321,
      "theoretical_loss": 3.593734616810316,
      "tokens_seen": 1284308992
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006170912738214645,
      "loss": 1.377,
      "theoretical_loss": 3.5937180952335677,
      "tokens_seen": 1284374528
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006170712136409227,
      "loss": 1.5813,
      "theoretical_loss": 3.593701574735854,
      "tokens_seen": 1284440064
    },
    {
      "epoch": 0.39,
      "objective/train/advantage_avg": 0.4954558312892914,
      "objective/train/docs_used": 470107,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4378275871276855,
      "objective/train/original_loss": 1.437827706336975,
      "objective/train/theoretical_loss": 3.5936850553170503,
      "objective/train/tokens_used": 1304965600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24565638601779938,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507944822311401,
      "objective/train/weighted_lm_loss": 1.5107450485229492,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.5936850553170503,
      "tokens_seen": 1284505600
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006170511534603812,
      "loss": 1.3972,
      "theoretical_loss": 3.5936850553170503,
      "tokens_seen": 1284505600
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006170310932798396,
      "loss": 1.5083,
      "theoretical_loss": 3.59366853697703,
      "tokens_seen": 1284571136
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006170110330992979,
      "loss": 1.5545,
      "theoretical_loss": 3.5936520197156683,
      "tokens_seen": 1284636672
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006169909729187563,
      "loss": 1.4173,
      "theoretical_loss": 3.59363550353284,
      "tokens_seen": 1284702208
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006169709127382148,
      "loss": 1.5604,
      "theoretical_loss": 3.593618988428419,
      "tokens_seen": 1284767744
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000616950852557673,
      "loss": 1.4656,
      "theoretical_loss": 3.59360247440228,
      "tokens_seen": 1284833280
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006169307923771315,
      "loss": 1.4442,
      "theoretical_loss": 3.593585961454298,
      "tokens_seen": 1284898816
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006169107321965897,
      "loss": 1.4172,
      "theoretical_loss": 3.593569449584347,
      "tokens_seen": 1284964352
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006168906720160482,
      "loss": 1.4396,
      "theoretical_loss": 3.593552938792303,
      "tokens_seen": 1285029888
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006168706118355066,
      "loss": 1.4291,
      "theoretical_loss": 3.5935364290780387,
      "tokens_seen": 1285095424
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006168505516549649,
      "loss": 1.5044,
      "theoretical_loss": 3.5935199204414303,
      "tokens_seen": 1285160960
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006168304914744233,
      "loss": 1.4552,
      "theoretical_loss": 3.5935034128823515,
      "tokens_seen": 1285226496
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006168104312938817,
      "loss": 1.494,
      "theoretical_loss": 3.5934869064006785,
      "tokens_seen": 1285292032
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00061679037111334,
      "loss": 1.489,
      "theoretical_loss": 3.593470400996284,
      "tokens_seen": 1285357568
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006167703109327985,
      "loss": 1.5658,
      "theoretical_loss": 3.5934538966690446,
      "tokens_seen": 1285423104
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006167502507522567,
      "loss": 1.5179,
      "theoretical_loss": 3.593437393418834,
      "tokens_seen": 1285488640
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006167301905717152,
      "loss": 1.4506,
      "theoretical_loss": 3.593420891245527,
      "tokens_seen": 1285554176
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006167101303911735,
      "loss": 1.5246,
      "theoretical_loss": 3.593404390149,
      "tokens_seen": 1285619712
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006166900702106319,
      "loss": 1.5078,
      "theoretical_loss": 3.593387890129126,
      "tokens_seen": 1285685248
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006166700100300903,
      "loss": 1.6277,
      "theoretical_loss": 3.59337139118578,
      "tokens_seen": 1285750784
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006166499498495487,
      "loss": 1.4918,
      "theoretical_loss": 3.5933548933188377,
      "tokens_seen": 1285816320
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000616629889669007,
      "loss": 1.3921,
      "theoretical_loss": 3.593338396528174,
      "tokens_seen": 1285881856
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006166098294884655,
      "loss": 1.4667,
      "theoretical_loss": 3.593321900813663,
      "tokens_seen": 1285947392
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006165897693079237,
      "loss": 1.5043,
      "theoretical_loss": 3.5933054061751806,
      "tokens_seen": 1286012928
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006165697091273822,
      "loss": 1.5022,
      "theoretical_loss": 3.5932889126126017,
      "tokens_seen": 1286078464
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006165496489468405,
      "loss": 1.6201,
      "theoretical_loss": 3.5932724201258006,
      "tokens_seen": 1286144000
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006165295887662989,
      "loss": 1.4292,
      "theoretical_loss": 3.593255928714653,
      "tokens_seen": 1286209536
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006165095285857573,
      "loss": 1.4231,
      "theoretical_loss": 3.5932394383790336,
      "tokens_seen": 1286275072
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006164894684052157,
      "loss": 1.597,
      "theoretical_loss": 3.5932229491188172,
      "tokens_seen": 1286340608
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000616469408224674,
      "loss": 1.4441,
      "theoretical_loss": 3.59320646093388,
      "tokens_seen": 1286406144
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006164493480441324,
      "loss": 1.3605,
      "theoretical_loss": 3.5931899738240958,
      "tokens_seen": 1286471680
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006164292878635908,
      "loss": 1.5197,
      "theoretical_loss": 3.5931734877893398,
      "tokens_seen": 1286537216
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006164092276830492,
      "loss": 1.5072,
      "theoretical_loss": 3.5931570028294884,
      "tokens_seen": 1286602752
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006163891675025075,
      "loss": 1.4561,
      "theoretical_loss": 3.5931405189444154,
      "tokens_seen": 1286668288
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000616369107321966,
      "loss": 1.4637,
      "theoretical_loss": 3.593124036133997,
      "tokens_seen": 1286733824
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006163490471414242,
      "loss": 1.5062,
      "theoretical_loss": 3.5931075543981077,
      "tokens_seen": 1286799360
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006163289869608827,
      "loss": 1.5083,
      "theoretical_loss": 3.5930910737366233,
      "tokens_seen": 1286864896
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000616308926780341,
      "loss": 1.6134,
      "theoretical_loss": 3.5930745941494187,
      "tokens_seen": 1286930432
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006162888665997994,
      "loss": 1.4863,
      "theoretical_loss": 3.593058115636369,
      "tokens_seen": 1286995968
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006162688064192578,
      "loss": 1.3839,
      "theoretical_loss": 3.5930416381973496,
      "tokens_seen": 1287061504
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006162487462387162,
      "loss": 1.5464,
      "theoretical_loss": 3.5930251618322364,
      "tokens_seen": 1287127040
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006162286860581745,
      "loss": 1.3719,
      "theoretical_loss": 3.593008686540904,
      "tokens_seen": 1287192576
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000616208625877633,
      "loss": 1.4806,
      "theoretical_loss": 3.592992212323228,
      "tokens_seen": 1287258112
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006161885656970912,
      "loss": 1.5262,
      "theoretical_loss": 3.5929757391790837,
      "tokens_seen": 1287323648
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006161685055165497,
      "loss": 1.5605,
      "theoretical_loss": 3.5929592671083466,
      "tokens_seen": 1287389184
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000616148445336008,
      "loss": 1.5553,
      "theoretical_loss": 3.5929427961108926,
      "tokens_seen": 1287454720
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006161283851554664,
      "loss": 1.5273,
      "theoretical_loss": 3.5929263261865962,
      "tokens_seen": 1287520256
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006161083249749248,
      "loss": 1.3654,
      "theoretical_loss": 3.592909857335333,
      "tokens_seen": 1287585792
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006160882647943831,
      "loss": 1.4562,
      "theoretical_loss": 3.5928933895569792,
      "tokens_seen": 1287651328
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006160682046138415,
      "loss": 1.3844,
      "theoretical_loss": 3.59287692285141,
      "tokens_seen": 1287716864
    },
    {
      "epoch": 0.39,
      "objective/train/advantage_avg": 0.4860357940196991,
      "objective/train/docs_used": 471101,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3308324813842773,
      "objective/train/original_loss": 1.3308324813842773,
      "objective/train/theoretical_loss": 3.592860457218501,
      "objective/train/tokens_used": 1308242400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23733286559581757,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0498099327087402,
      "objective/train/weighted_lm_loss": 1.396789789199829,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.592860457218501,
      "tokens_seen": 1287782400
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006160481444333,
      "loss": 1.4951,
      "theoretical_loss": 3.592860457218501,
      "tokens_seen": 1287782400
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006160280842527582,
      "loss": 1.4165,
      "theoretical_loss": 3.592843992658127,
      "tokens_seen": 1287847936
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006160080240722167,
      "loss": 1.4679,
      "theoretical_loss": 3.5928275291701643,
      "tokens_seen": 1287913472
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000615987963891675,
      "loss": 1.4554,
      "theoretical_loss": 3.5928110667544884,
      "tokens_seen": 1287979008
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006159679037111334,
      "loss": 1.4668,
      "theoretical_loss": 3.5927946054109747,
      "tokens_seen": 1288044544
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006159478435305918,
      "loss": 1.5731,
      "theoretical_loss": 3.592778145139499,
      "tokens_seen": 1288110080
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006159277833500501,
      "loss": 1.3664,
      "theoretical_loss": 3.592761685939937,
      "tokens_seen": 1288175616
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006159077231695085,
      "loss": 1.3979,
      "theoretical_loss": 3.592745227812164,
      "tokens_seen": 1288241152
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000615887662988967,
      "loss": 1.5741,
      "theoretical_loss": 3.5927287707560565,
      "tokens_seen": 1288306688
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006158676028084252,
      "loss": 1.4821,
      "theoretical_loss": 3.592712314771489,
      "tokens_seen": 1288372224
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006158475426278837,
      "loss": 1.4514,
      "theoretical_loss": 3.5926958598583383,
      "tokens_seen": 1288437760
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000615827482447342,
      "loss": 1.3717,
      "theoretical_loss": 3.5926794060164795,
      "tokens_seen": 1288503296
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006158074222668004,
      "loss": 1.4448,
      "theoretical_loss": 3.5926629532457883,
      "tokens_seen": 1288568832
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006157873620862588,
      "loss": 1.4229,
      "theoretical_loss": 3.5926465015461413,
      "tokens_seen": 1288634368
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006157673019057171,
      "loss": 1.4591,
      "theoretical_loss": 3.5926300509174136,
      "tokens_seen": 1288699904
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006157472417251755,
      "loss": 1.4327,
      "theoretical_loss": 3.5926136013594814,
      "tokens_seen": 1288765440
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006157271815446339,
      "loss": 1.4683,
      "theoretical_loss": 3.5925971528722203,
      "tokens_seen": 1288830976
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006157071213640922,
      "loss": 1.4303,
      "theoretical_loss": 3.5925807054555063,
      "tokens_seen": 1288896512
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006156870611835507,
      "loss": 1.5045,
      "theoretical_loss": 3.5925642591092153,
      "tokens_seen": 1288962048
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000615667001003009,
      "loss": 1.3996,
      "theoretical_loss": 3.5925478138332227,
      "tokens_seen": 1289027584
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006156469408224674,
      "loss": 1.3208,
      "theoretical_loss": 3.5925313696274053,
      "tokens_seen": 1289093120
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006156268806419258,
      "loss": 1.4176,
      "theoretical_loss": 3.5925149264916385,
      "tokens_seen": 1289158656
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006156068204613841,
      "loss": 1.5721,
      "theoretical_loss": 3.592498484425799,
      "tokens_seen": 1289224192
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006155867602808426,
      "loss": 1.4936,
      "theoretical_loss": 3.5924820434297615,
      "tokens_seen": 1289289728
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006155667001003009,
      "loss": 1.437,
      "theoretical_loss": 3.592465603503403,
      "tokens_seen": 1289355264
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006155466399197593,
      "loss": 1.6111,
      "theoretical_loss": 3.5924491646465997,
      "tokens_seen": 1289420800
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006155265797392177,
      "loss": 1.534,
      "theoretical_loss": 3.5924327268592267,
      "tokens_seen": 1289486336
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006155065195586761,
      "loss": 1.4,
      "theoretical_loss": 3.592416290141161,
      "tokens_seen": 1289551872
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006154864593781344,
      "loss": 1.4678,
      "theoretical_loss": 3.592399854492278,
      "tokens_seen": 1289617408
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006154663991975928,
      "loss": 1.4528,
      "theoretical_loss": 3.5923834199124545,
      "tokens_seen": 1289682944
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006154463390170512,
      "loss": 1.4229,
      "theoretical_loss": 3.5923669864015664,
      "tokens_seen": 1289748480
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006154262788365096,
      "loss": 1.4458,
      "theoretical_loss": 3.5923505539594895,
      "tokens_seen": 1289814016
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006154062186559679,
      "loss": 1.3894,
      "theoretical_loss": 3.5923341225861005,
      "tokens_seen": 1289879552
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006153861584754264,
      "loss": 1.4473,
      "theoretical_loss": 3.5923176922812754,
      "tokens_seen": 1289945088
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006153660982948846,
      "loss": 1.4692,
      "theoretical_loss": 3.59230126304489,
      "tokens_seen": 1290010624
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006153460381143431,
      "loss": 1.5096,
      "theoretical_loss": 3.5922848348768213,
      "tokens_seen": 1290076160
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006153259779338014,
      "loss": 1.5731,
      "theoretical_loss": 3.592268407776945,
      "tokens_seen": 1290141696
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006153059177532598,
      "loss": 1.4124,
      "theoretical_loss": 3.5922519817451377,
      "tokens_seen": 1290207232
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006152858575727182,
      "loss": 1.5286,
      "theoretical_loss": 3.592235556781276,
      "tokens_seen": 1290272768
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006152657973921766,
      "loss": 1.337,
      "theoretical_loss": 3.592219132885235,
      "tokens_seen": 1290338304
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006152457372116349,
      "loss": 1.5065,
      "theoretical_loss": 3.5922027100568923,
      "tokens_seen": 1290403840
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006152256770310934,
      "loss": 1.4479,
      "theoretical_loss": 3.5921862882961237,
      "tokens_seen": 1290469376
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006152056168505516,
      "loss": 1.5041,
      "theoretical_loss": 3.5921698676028058,
      "tokens_seen": 1290534912
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006151855566700101,
      "loss": 1.4433,
      "theoretical_loss": 3.5921534479768154,
      "tokens_seen": 1290600448
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006151654964894684,
      "loss": 1.333,
      "theoretical_loss": 3.592137029418028,
      "tokens_seen": 1290665984
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006151454363089268,
      "loss": 1.4371,
      "theoretical_loss": 3.5921206119263207,
      "tokens_seen": 1290731520
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006151253761283852,
      "loss": 1.4068,
      "theoretical_loss": 3.59210419550157,
      "tokens_seen": 1290797056
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006151053159478435,
      "loss": 1.3383,
      "theoretical_loss": 3.592087780143652,
      "tokens_seen": 1290862592
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006150852557673019,
      "loss": 1.6343,
      "theoretical_loss": 3.5920713658524432,
      "tokens_seen": 1290928128
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006150651955867604,
      "loss": 1.4595,
      "theoretical_loss": 3.5920549526278207,
      "tokens_seen": 1290993664
    },
    {
      "epoch": 0.39,
      "objective/train/advantage_avg": 0.49279874563217163,
      "objective/train/docs_used": 472355,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4964741468429565,
      "objective/train/original_loss": 1.496474266052246,
      "objective/train/theoretical_loss": 3.5920385404696606,
      "objective/train/tokens_used": 1311519200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24349567294120789,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505177974700928,
      "objective/train/weighted_lm_loss": 1.572201132774353,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5920385404696606,
      "tokens_seen": 1291059200
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006150451354062186,
      "loss": 1.4985,
      "theoretical_loss": 3.5920385404696606,
      "tokens_seen": 1291059200
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006150250752256771,
      "loss": 1.5096,
      "theoretical_loss": 3.5920221293778396,
      "tokens_seen": 1291124736
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006150050150451355,
      "loss": 1.3804,
      "theoretical_loss": 3.592005719352234,
      "tokens_seen": 1291190272
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006149849548645938,
      "loss": 1.4752,
      "theoretical_loss": 3.591989310392721,
      "tokens_seen": 1291255808
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006149648946840522,
      "loss": 1.5077,
      "theoretical_loss": 3.591972902499177,
      "tokens_seen": 1291321344
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006149448345035105,
      "loss": 1.4894,
      "theoretical_loss": 3.5919564956714787,
      "tokens_seen": 1291386880
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006149247743229689,
      "loss": 1.4296,
      "theoretical_loss": 3.591940089909502,
      "tokens_seen": 1291452416
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006149047141424274,
      "loss": 1.433,
      "theoretical_loss": 3.591923685213125,
      "tokens_seen": 1291517952
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006148846539618856,
      "loss": 1.5303,
      "theoretical_loss": 3.5919072815822233,
      "tokens_seen": 1291583488
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006148645937813441,
      "loss": 1.5167,
      "theoretical_loss": 3.5918908790166744,
      "tokens_seen": 1291649024
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006148445336008023,
      "loss": 1.496,
      "theoretical_loss": 3.5918744775163542,
      "tokens_seen": 1291714560
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006148244734202608,
      "loss": 1.4964,
      "theoretical_loss": 3.5918580770811404,
      "tokens_seen": 1291780096
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006148044132397192,
      "loss": 1.3589,
      "theoretical_loss": 3.5918416777109092,
      "tokens_seen": 1291845632
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006147843530591775,
      "loss": 1.466,
      "theoretical_loss": 3.591825279405538,
      "tokens_seen": 1291911168
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006147642928786359,
      "loss": 1.4693,
      "theoretical_loss": 3.5918088821649024,
      "tokens_seen": 1291976704
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006147442326980943,
      "loss": 1.341,
      "theoretical_loss": 3.5917924859888806,
      "tokens_seen": 1292042240
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006147241725175526,
      "loss": 1.4421,
      "theoretical_loss": 3.591776090877349,
      "tokens_seen": 1292107776
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006147041123370111,
      "loss": 1.5815,
      "theoretical_loss": 3.5917596968301844,
      "tokens_seen": 1292173312
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006146840521564694,
      "loss": 1.4846,
      "theoretical_loss": 3.591743303847264,
      "tokens_seen": 1292238848
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006146639919759278,
      "loss": 1.5661,
      "theoretical_loss": 3.5917269119284643,
      "tokens_seen": 1292304384
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006146439317953862,
      "loss": 1.5385,
      "theoretical_loss": 3.5917105210736624,
      "tokens_seen": 1292369920
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006146238716148445,
      "loss": 1.4882,
      "theoretical_loss": 3.5916941312827357,
      "tokens_seen": 1292435456
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006146038114343029,
      "loss": 1.5288,
      "theoretical_loss": 3.591677742555561,
      "tokens_seen": 1292500992
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006145837512537613,
      "loss": 1.4909,
      "theoretical_loss": 3.5916613548920147,
      "tokens_seen": 1292566528
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006145636910732196,
      "loss": 1.4093,
      "theoretical_loss": 3.591644968291975,
      "tokens_seen": 1292632064
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006145436308926781,
      "loss": 1.4811,
      "theoretical_loss": 3.5916285827553187,
      "tokens_seen": 1292697600
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006145235707121364,
      "loss": 1.4813,
      "theoretical_loss": 3.5916121982819216,
      "tokens_seen": 1292763136
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006145035105315948,
      "loss": 1.3349,
      "theoretical_loss": 3.5915958148716625,
      "tokens_seen": 1292828672
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006144834503510531,
      "loss": 1.516,
      "theoretical_loss": 3.5915794325244175,
      "tokens_seen": 1292894208
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006144633901705116,
      "loss": 1.5646,
      "theoretical_loss": 3.5915630512400636,
      "tokens_seen": 1292959744
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006144433299899699,
      "loss": 1.4927,
      "theoretical_loss": 3.591546671018479,
      "tokens_seen": 1293025280
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006144232698094283,
      "loss": 1.4271,
      "theoretical_loss": 3.5915302918595398,
      "tokens_seen": 1293090816
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006144032096288866,
      "loss": 1.4414,
      "theoretical_loss": 3.5915139137631242,
      "tokens_seen": 1293156352
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000614383149448345,
      "loss": 1.4091,
      "theoretical_loss": 3.5914975367291087,
      "tokens_seen": 1293221888
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006143630892678034,
      "loss": 1.406,
      "theoretical_loss": 3.591481160757371,
      "tokens_seen": 1293287424
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006143430290872618,
      "loss": 1.4339,
      "theoretical_loss": 3.591464785847788,
      "tokens_seen": 1293352960
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006143229689067201,
      "loss": 1.5095,
      "theoretical_loss": 3.5914484120002372,
      "tokens_seen": 1293418496
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006143029087261786,
      "loss": 1.3986,
      "theoretical_loss": 3.5914320392145953,
      "tokens_seen": 1293484032
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006142828485456369,
      "loss": 1.4986,
      "theoretical_loss": 3.591415667490741,
      "tokens_seen": 1293549568
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006142627883650953,
      "loss": 1.5345,
      "theoretical_loss": 3.59139929682855,
      "tokens_seen": 1293615104
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006142427281845536,
      "loss": 1.465,
      "theoretical_loss": 3.591382927227902,
      "tokens_seen": 1293680640
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000614222668004012,
      "loss": 1.4994,
      "theoretical_loss": 3.5913665586886716,
      "tokens_seen": 1293746176
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006142026078234704,
      "loss": 1.5255,
      "theoretical_loss": 3.591350191210738,
      "tokens_seen": 1293811712
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006141825476429288,
      "loss": 1.4495,
      "theoretical_loss": 3.591333824793978,
      "tokens_seen": 1293877248
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006141624874623871,
      "loss": 1.6615,
      "theoretical_loss": 3.5913174594382693,
      "tokens_seen": 1293942784
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006141424272818456,
      "loss": 1.4252,
      "theoretical_loss": 3.5913010951434896,
      "tokens_seen": 1294008320
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006141223671013038,
      "loss": 1.5514,
      "theoretical_loss": 3.591284731909516,
      "tokens_seen": 1294073856
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006141023069207623,
      "loss": 1.5313,
      "theoretical_loss": 3.591268369736226,
      "tokens_seen": 1294139392
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006140822467402208,
      "loss": 1.4596,
      "theoretical_loss": 3.5912520086234974,
      "tokens_seen": 1294204928
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000614062186559679,
      "loss": 1.5663,
      "theoretical_loss": 3.591235648571207,
      "tokens_seen": 1294270464
    },
    {
      "epoch": 0.39,
      "objective/train/advantage_avg": 0.49386945366859436,
      "objective/train/docs_used": 473604,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4118523597717285,
      "objective/train/original_loss": 1.411852240562439,
      "objective/train/theoretical_loss": 3.591219289579233,
      "objective/train/tokens_used": 1314796000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24412734806537628,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506280660629272,
      "objective/train/weighted_lm_loss": 1.4832409620285034,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 3.591219289579233,
      "tokens_seen": 1294336000
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006140421263791375,
      "loss": 1.4342,
      "theoretical_loss": 3.591219289579233,
      "tokens_seen": 1294336000
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006140220661985959,
      "loss": 1.4502,
      "theoretical_loss": 3.591202931647454,
      "tokens_seen": 1294401536
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006140020060180542,
      "loss": 1.579,
      "theoretical_loss": 3.5911865747757457,
      "tokens_seen": 1294467072
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006139819458375126,
      "loss": 1.4593,
      "theoretical_loss": 3.5911702189639874,
      "tokens_seen": 1294532608
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006139618856569709,
      "loss": 1.4527,
      "theoretical_loss": 3.591153864212055,
      "tokens_seen": 1294598144
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006139418254764293,
      "loss": 1.4339,
      "theoretical_loss": 3.591137510519828,
      "tokens_seen": 1294663680
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006139217652958878,
      "loss": 1.5333,
      "theoretical_loss": 3.5911211578871827,
      "tokens_seen": 1294729216
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000613901705115346,
      "loss": 1.4081,
      "theoretical_loss": 3.5911048063139983,
      "tokens_seen": 1294794752
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006138816449348045,
      "loss": 1.4726,
      "theoretical_loss": 3.5910884558001506,
      "tokens_seen": 1294860288
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006138615847542627,
      "loss": 1.4387,
      "theoretical_loss": 3.591072106345519,
      "tokens_seen": 1294925824
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006138415245737212,
      "loss": 1.4435,
      "theoretical_loss": 3.591055757949981,
      "tokens_seen": 1294991360
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006138214643931796,
      "loss": 1.6442,
      "theoretical_loss": 3.591039410613413,
      "tokens_seen": 1295056896
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006138014042126379,
      "loss": 1.4791,
      "theoretical_loss": 3.591023064335695,
      "tokens_seen": 1295122432
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006137813440320963,
      "loss": 1.5173,
      "theoretical_loss": 3.591006719116703,
      "tokens_seen": 1295187968
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006137612838515547,
      "loss": 1.4427,
      "theoretical_loss": 3.590990374956316,
      "tokens_seen": 1295253504
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000613741223671013,
      "loss": 1.5982,
      "theoretical_loss": 3.590974031854411,
      "tokens_seen": 1295319040
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006137211634904715,
      "loss": 1.462,
      "theoretical_loss": 3.590957689810867,
      "tokens_seen": 1295384576
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006137011033099298,
      "loss": 1.5352,
      "theoretical_loss": 3.5909413488255613,
      "tokens_seen": 1295450112
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006136810431293882,
      "loss": 1.4919,
      "theoretical_loss": 3.5909250088983713,
      "tokens_seen": 1295515648
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006136609829488466,
      "loss": 1.5255,
      "theoretical_loss": 3.590908670029176,
      "tokens_seen": 1295581184
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000613640922768305,
      "loss": 1.5196,
      "theoretical_loss": 3.5908923322178525,
      "tokens_seen": 1295646720
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006136208625877633,
      "loss": 1.5791,
      "theoretical_loss": 3.5908759954642795,
      "tokens_seen": 1295712256
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006136008024072217,
      "loss": 1.459,
      "theoretical_loss": 3.5908596597683347,
      "tokens_seen": 1295777792
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00061358074222668,
      "loss": 1.346,
      "theoretical_loss": 3.590843325129896,
      "tokens_seen": 1295843328
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006135606820461385,
      "loss": 1.5194,
      "theoretical_loss": 3.590826991548842,
      "tokens_seen": 1295908864
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006135406218655968,
      "loss": 1.458,
      "theoretical_loss": 3.59081065902505,
      "tokens_seen": 1295974400
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006135205616850552,
      "loss": 1.5398,
      "theoretical_loss": 3.590794327558399,
      "tokens_seen": 1296039936
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006135005015045135,
      "loss": 1.3746,
      "theoretical_loss": 3.5907779971487663,
      "tokens_seen": 1296105472
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000613480441323972,
      "loss": 1.437,
      "theoretical_loss": 3.5907616677960306,
      "tokens_seen": 1296171008
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006134603811434303,
      "loss": 1.4561,
      "theoretical_loss": 3.5907453395000695,
      "tokens_seen": 1296236544
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006134403209628887,
      "loss": 1.4657,
      "theoretical_loss": 3.590729012260762,
      "tokens_seen": 1296302080
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000613420260782347,
      "loss": 1.4892,
      "theoretical_loss": 3.5907126860779854,
      "tokens_seen": 1296367616
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006134002006018054,
      "loss": 1.5106,
      "theoretical_loss": 3.5906963609516183,
      "tokens_seen": 1296433152
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006133801404212638,
      "loss": 1.5126,
      "theoretical_loss": 3.5906800368815395,
      "tokens_seen": 1296498688
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006133600802407222,
      "loss": 1.5802,
      "theoretical_loss": 3.5906637138676265,
      "tokens_seen": 1296564224
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006133400200601805,
      "loss": 1.5107,
      "theoretical_loss": 3.590647391909758,
      "tokens_seen": 1296629760
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000613319959879639,
      "loss": 1.4534,
      "theoretical_loss": 3.590631071007812,
      "tokens_seen": 1296695296
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006132998996990973,
      "loss": 1.4605,
      "theoretical_loss": 3.5906147511616675,
      "tokens_seen": 1296760832
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006132798395185557,
      "loss": 1.4124,
      "theoretical_loss": 3.590598432371202,
      "tokens_seen": 1296826368
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000613259779338014,
      "loss": 1.4656,
      "theoretical_loss": 3.5905821146362946,
      "tokens_seen": 1296891904
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006132397191574724,
      "loss": 1.4636,
      "theoretical_loss": 3.590565797956822,
      "tokens_seen": 1296957440
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006132196589769308,
      "loss": 1.4017,
      "theoretical_loss": 3.5905494823326656,
      "tokens_seen": 1297022976
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006131995987963892,
      "loss": 1.464,
      "theoretical_loss": 3.5905331677637013,
      "tokens_seen": 1297088512
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006131795386158475,
      "loss": 1.4814,
      "theoretical_loss": 3.5905168542498087,
      "tokens_seen": 1297154048
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000613159478435306,
      "loss": 1.4851,
      "theoretical_loss": 3.5905005417908655,
      "tokens_seen": 1297219584
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006131394182547642,
      "loss": 1.4448,
      "theoretical_loss": 3.5904842303867515,
      "tokens_seen": 1297285120
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006131193580742227,
      "loss": 1.4461,
      "theoretical_loss": 3.5904679200373435,
      "tokens_seen": 1297350656
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006130992978936811,
      "loss": 1.4371,
      "theoretical_loss": 3.5904516107425213,
      "tokens_seen": 1297416192
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006130792377131394,
      "loss": 1.4534,
      "theoretical_loss": 3.5904353025021627,
      "tokens_seen": 1297481728
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006130591775325978,
      "loss": 1.4262,
      "theoretical_loss": 3.590418995316147,
      "tokens_seen": 1297547264
    },
    {
      "epoch": 0.39,
      "objective/train/advantage_avg": 0.49484673142433167,
      "objective/train/docs_used": 474727,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3615150451660156,
      "objective/train/original_loss": 1.3615148067474365,
      "objective/train/theoretical_loss": 3.5904026891843523,
      "objective/train/tokens_used": 1318072800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24520966410636902,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507313013076782,
      "objective/train/weighted_lm_loss": 1.4305227994918823,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.5904026891843523,
      "tokens_seen": 1297612800
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006130391173520563,
      "loss": 1.476,
      "theoretical_loss": 3.5904026891843523,
      "tokens_seen": 1297612800
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006130190571715145,
      "loss": 1.4232,
      "theoretical_loss": 3.5903863841066572,
      "tokens_seen": 1297678336
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000612998996990973,
      "loss": 1.3696,
      "theoretical_loss": 3.5903700800829403,
      "tokens_seen": 1297743872
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006129789368104312,
      "loss": 1.4494,
      "theoretical_loss": 3.5903537771130805,
      "tokens_seen": 1297809408
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006129588766298897,
      "loss": 1.3644,
      "theoretical_loss": 3.5903374751969563,
      "tokens_seen": 1297874944
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006129388164493481,
      "loss": 1.3584,
      "theoretical_loss": 3.5903211743344468,
      "tokens_seen": 1297940480
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006129187562688064,
      "loss": 1.3884,
      "theoretical_loss": 3.5903048745254296,
      "tokens_seen": 1298006016
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006128986960882648,
      "loss": 1.4327,
      "theoretical_loss": 3.590288575769785,
      "tokens_seen": 1298071552
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006128786359077231,
      "loss": 1.487,
      "theoretical_loss": 3.59027227806739,
      "tokens_seen": 1298137088
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006128585757271815,
      "loss": 1.5533,
      "theoretical_loss": 3.5902559814181254,
      "tokens_seen": 1298202624
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00061283851554664,
      "loss": 1.4426,
      "theoretical_loss": 3.590239685821868,
      "tokens_seen": 1298268160
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006128184553660982,
      "loss": 1.4798,
      "theoretical_loss": 3.590223391278498,
      "tokens_seen": 1298333696
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006127983951855567,
      "loss": 1.3497,
      "theoretical_loss": 3.5902070977878937,
      "tokens_seen": 1298399232
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000612778335005015,
      "loss": 1.5744,
      "theoretical_loss": 3.5901908053499336,
      "tokens_seen": 1298464768
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006127582748244734,
      "loss": 1.3599,
      "theoretical_loss": 3.590174513964497,
      "tokens_seen": 1298530304
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006127382146439318,
      "loss": 1.4539,
      "theoretical_loss": 3.5901582236314633,
      "tokens_seen": 1298595840
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006127181544633902,
      "loss": 1.3697,
      "theoretical_loss": 3.5901419343507106,
      "tokens_seen": 1298661376
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006126980942828485,
      "loss": 1.3218,
      "theoretical_loss": 3.5901256461221176,
      "tokens_seen": 1298726912
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000612678034102307,
      "loss": 1.6308,
      "theoretical_loss": 3.590109358945564,
      "tokens_seen": 1298792448
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006126579739217652,
      "loss": 1.4383,
      "theoretical_loss": 3.5900930728209284,
      "tokens_seen": 1298857984
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006126379137412237,
      "loss": 1.393,
      "theoretical_loss": 3.5900767877480906,
      "tokens_seen": 1298923520
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000612617853560682,
      "loss": 1.4797,
      "theoretical_loss": 3.590060503726928,
      "tokens_seen": 1298989056
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006125977933801404,
      "loss": 1.3843,
      "theoretical_loss": 3.5900442207573207,
      "tokens_seen": 1299054592
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006125777331995989,
      "loss": 1.4598,
      "theoretical_loss": 3.590027938839148,
      "tokens_seen": 1299120128
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006125576730190572,
      "loss": 1.3987,
      "theoretical_loss": 3.5900116579722883,
      "tokens_seen": 1299185664
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006125376128385156,
      "loss": 1.4098,
      "theoretical_loss": 3.5899953781566207,
      "tokens_seen": 1299251200
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006125175526579739,
      "loss": 1.558,
      "theoretical_loss": 3.589979099392025,
      "tokens_seen": 1299316736
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006124974924774324,
      "loss": 1.4071,
      "theoretical_loss": 3.5899628216783794,
      "tokens_seen": 1299382272
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006124774322968907,
      "loss": 1.5051,
      "theoretical_loss": 3.5899465450155637,
      "tokens_seen": 1299447808
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006124573721163491,
      "loss": 1.4781,
      "theoretical_loss": 3.5899302694034567,
      "tokens_seen": 1299513344
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006124373119358074,
      "loss": 1.5439,
      "theoretical_loss": 3.5899139948419383,
      "tokens_seen": 1299578880
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006124172517552658,
      "loss": 1.4847,
      "theoretical_loss": 3.5898977213308867,
      "tokens_seen": 1299644416
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006123971915747242,
      "loss": 1.4704,
      "theoretical_loss": 3.589881448870182,
      "tokens_seen": 1299709952
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006123771313941826,
      "loss": 1.3473,
      "theoretical_loss": 3.5898651774597026,
      "tokens_seen": 1299775488
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006123570712136409,
      "loss": 1.4964,
      "theoretical_loss": 3.589848907099329,
      "tokens_seen": 1299841024
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006123370110330994,
      "loss": 1.5368,
      "theoretical_loss": 3.5898326377889394,
      "tokens_seen": 1299906560
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006123169508525577,
      "loss": 1.5539,
      "theoretical_loss": 3.589816369528413,
      "tokens_seen": 1299972096
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006122968906720161,
      "loss": 1.4628,
      "theoretical_loss": 3.5898001023176302,
      "tokens_seen": 1300037632
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006122768304914744,
      "loss": 1.42,
      "theoretical_loss": 3.589783836156469,
      "tokens_seen": 1300103168
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006122567703109328,
      "loss": 1.4212,
      "theoretical_loss": 3.58976757104481,
      "tokens_seen": 1300168704
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006122367101303912,
      "loss": 1.4716,
      "theoretical_loss": 3.5897513069825324,
      "tokens_seen": 1300234240
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006122166499498496,
      "loss": 1.4437,
      "theoretical_loss": 3.5897350439695144,
      "tokens_seen": 1300299776
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006121965897693079,
      "loss": 1.5316,
      "theoretical_loss": 3.589718782005637,
      "tokens_seen": 1300365312
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006121765295887664,
      "loss": 1.4168,
      "theoretical_loss": 3.589702521090779,
      "tokens_seen": 1300430848
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006121564694082246,
      "loss": 1.5325,
      "theoretical_loss": 3.589686261224819,
      "tokens_seen": 1300496384
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006121364092276831,
      "loss": 1.4342,
      "theoretical_loss": 3.5896700024076384,
      "tokens_seen": 1300561920
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006121163490471415,
      "loss": 1.4699,
      "theoretical_loss": 3.589653744639115,
      "tokens_seen": 1300627456
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006120962888665998,
      "loss": 1.5681,
      "theoretical_loss": 3.5896374879191293,
      "tokens_seen": 1300692992
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006120762286860582,
      "loss": 1.4326,
      "theoretical_loss": 3.5896212322475605,
      "tokens_seen": 1300758528
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006120561685055167,
      "loss": 1.4532,
      "theoretical_loss": 3.589604977624288,
      "tokens_seen": 1300824064
    },
    {
      "epoch": 0.39,
      "objective/train/advantage_avg": 0.4960155189037323,
      "objective/train/docs_used": 475730,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3575814962387085,
      "objective/train/original_loss": 1.3575811386108398,
      "objective/train/theoretical_loss": 3.5895887240491917,
      "objective/train/tokens_used": 1321349600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24630206823349,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508538484573364,
      "objective/train/weighted_lm_loss": 1.4266254901885986,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5895887240491917,
      "tokens_seen": 1300889600
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006120361083249749,
      "loss": 1.5063,
      "theoretical_loss": 3.5895887240491917,
      "tokens_seen": 1300889600
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006120160481444334,
      "loss": 1.3876,
      "theoretical_loss": 3.589572471522151,
      "tokens_seen": 1300955136
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006119959879638916,
      "loss": 1.5121,
      "theoretical_loss": 3.589556220043046,
      "tokens_seen": 1301020672
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006119759277833501,
      "loss": 1.4597,
      "theoretical_loss": 3.5895399696117556,
      "tokens_seen": 1301086208
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006119558676028085,
      "loss": 1.5546,
      "theoretical_loss": 3.5895237202281596,
      "tokens_seen": 1301151744
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006119358074222668,
      "loss": 1.4002,
      "theoretical_loss": 3.589507471892138,
      "tokens_seen": 1301217280
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006119157472417252,
      "loss": 1.4638,
      "theoretical_loss": 3.589491224603571,
      "tokens_seen": 1301282816
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006118956870611835,
      "loss": 1.458,
      "theoretical_loss": 3.5894749783623374,
      "tokens_seen": 1301348352
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006118756268806419,
      "loss": 1.5404,
      "theoretical_loss": 3.589458733168317,
      "tokens_seen": 1301413888
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006118555667001004,
      "loss": 1.4635,
      "theoretical_loss": 3.5894424890213905,
      "tokens_seen": 1301479424
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006118355065195586,
      "loss": 1.4206,
      "theoretical_loss": 3.5894262459214366,
      "tokens_seen": 1301544960
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006118154463390171,
      "loss": 1.3871,
      "theoretical_loss": 3.589410003868336,
      "tokens_seen": 1301610496
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006117953861584754,
      "loss": 1.5326,
      "theoretical_loss": 3.5893937628619677,
      "tokens_seen": 1301676032
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006117753259779338,
      "loss": 1.4681,
      "theoretical_loss": 3.589377522902212,
      "tokens_seen": 1301741568
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006117552657973922,
      "loss": 1.3505,
      "theoretical_loss": 3.589361283988948,
      "tokens_seen": 1301807104
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006117352056168506,
      "loss": 1.5717,
      "theoretical_loss": 3.5893450461220575,
      "tokens_seen": 1301872640
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006117151454363089,
      "loss": 1.4255,
      "theoretical_loss": 3.589328809301419,
      "tokens_seen": 1301938176
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006116950852557674,
      "loss": 1.4345,
      "theoretical_loss": 3.589312573526912,
      "tokens_seen": 1302003712
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006116750250752256,
      "loss": 1.4196,
      "theoretical_loss": 3.589296338798418,
      "tokens_seen": 1302069248
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006116549648946841,
      "loss": 1.4472,
      "theoretical_loss": 3.589280105115815,
      "tokens_seen": 1302134784
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006116349047141424,
      "loss": 1.521,
      "theoretical_loss": 3.5892638724789845,
      "tokens_seen": 1302200320
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006116148445336008,
      "loss": 1.5678,
      "theoretical_loss": 3.5892476408878062,
      "tokens_seen": 1302265856
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006115947843530592,
      "loss": 1.4158,
      "theoretical_loss": 3.5892314103421596,
      "tokens_seen": 1302331392
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006115747241725176,
      "loss": 1.4063,
      "theoretical_loss": 3.5892151808419257,
      "tokens_seen": 1302396928
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006115546639919759,
      "loss": 1.3836,
      "theoretical_loss": 3.589198952386983,
      "tokens_seen": 1302462464
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006115346038114343,
      "loss": 1.485,
      "theoretical_loss": 3.5891827249772135,
      "tokens_seen": 1302528000
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006115145436308926,
      "loss": 1.5998,
      "theoretical_loss": 3.589166498612496,
      "tokens_seen": 1302593536
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006114944834503511,
      "loss": 1.4472,
      "theoretical_loss": 3.5891502732927107,
      "tokens_seen": 1302659072
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006114744232698094,
      "loss": 1.4552,
      "theoretical_loss": 3.589134049017739,
      "tokens_seen": 1302724608
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006114543630892678,
      "loss": 1.4909,
      "theoretical_loss": 3.5891178257874587,
      "tokens_seen": 1302790144
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006114343029087261,
      "loss": 1.5308,
      "theoretical_loss": 3.589101603601752,
      "tokens_seen": 1302855680
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006114142427281846,
      "loss": 1.432,
      "theoretical_loss": 3.5890853824604987,
      "tokens_seen": 1302921216
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006113941825476429,
      "loss": 1.5417,
      "theoretical_loss": 3.589069162363579,
      "tokens_seen": 1302986752
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006113741223671013,
      "loss": 1.5069,
      "theoretical_loss": 3.5890529433108727,
      "tokens_seen": 1303052288
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006113540621865597,
      "loss": 1.5418,
      "theoretical_loss": 3.58903672530226,
      "tokens_seen": 1303117824
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006113340020060181,
      "loss": 1.4096,
      "theoretical_loss": 3.5890205083376214,
      "tokens_seen": 1303183360
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006113139418254764,
      "loss": 1.5043,
      "theoretical_loss": 3.589004292416838,
      "tokens_seen": 1303248896
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006112938816449349,
      "loss": 1.4556,
      "theoretical_loss": 3.588988077539789,
      "tokens_seen": 1303314432
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006112738214643931,
      "loss": 1.5531,
      "theoretical_loss": 3.5889718637063552,
      "tokens_seen": 1303379968
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0006112537612838516,
      "loss": 1.4696,
      "theoretical_loss": 3.588955650916417,
      "tokens_seen": 1303445504
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006112337011033099,
      "loss": 1.4658,
      "theoretical_loss": 3.5889394391698546,
      "tokens_seen": 1303511040
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006112136409227683,
      "loss": 1.4948,
      "theoretical_loss": 3.5889232284665487,
      "tokens_seen": 1303576576
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006111935807422267,
      "loss": 1.3591,
      "theoretical_loss": 3.5889070188063794,
      "tokens_seen": 1303642112
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000611173520561685,
      "loss": 1.4951,
      "theoretical_loss": 3.588890810189227,
      "tokens_seen": 1303707648
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006111534603811434,
      "loss": 1.533,
      "theoretical_loss": 3.5888746026149727,
      "tokens_seen": 1303773184
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006111334002006019,
      "loss": 1.437,
      "theoretical_loss": 3.588858396083496,
      "tokens_seen": 1303838720
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006111133400200601,
      "loss": 1.5875,
      "theoretical_loss": 3.5888421905946783,
      "tokens_seen": 1303904256
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006110932798395186,
      "loss": 1.5784,
      "theoretical_loss": 3.5888259861483993,
      "tokens_seen": 1303969792
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000611073219658977,
      "loss": 1.4353,
      "theoretical_loss": 3.5888097827445407,
      "tokens_seen": 1304035328
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006110531594784353,
      "loss": 1.4195,
      "theoretical_loss": 3.5887935803829816,
      "tokens_seen": 1304100864
    },
    {
      "epoch": 0.4,
      "objective/train/advantage_avg": 0.4923355281352997,
      "objective/train/docs_used": 476908,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3218961954116821,
      "objective/train/original_loss": 1.3218963146209717,
      "objective/train/theoretical_loss": 3.588777379063604,
      "objective/train/tokens_used": 1324626400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24285028874874115,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05046808719635,
      "objective/train/weighted_lm_loss": 1.3884955644607544,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.588777379063604,
      "tokens_seen": 1304166400
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006110330992978938,
      "loss": 1.3665,
      "theoretical_loss": 3.588777379063604,
      "tokens_seen": 1304166400
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000611013039117352,
      "loss": 1.6092,
      "theoretical_loss": 3.588761178786287,
      "tokens_seen": 1304231936
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006109929789368105,
      "loss": 1.4508,
      "theoretical_loss": 3.5887449795509125,
      "tokens_seen": 1304297472
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006109729187562689,
      "loss": 1.3941,
      "theoretical_loss": 3.5887287813573607,
      "tokens_seen": 1304363008
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006109528585757272,
      "loss": 1.6414,
      "theoretical_loss": 3.5887125842055116,
      "tokens_seen": 1304428544
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006109327983951856,
      "loss": 1.477,
      "theoretical_loss": 3.588696388095247,
      "tokens_seen": 1304494080
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000610912738214644,
      "loss": 1.3862,
      "theoretical_loss": 3.5886801930264474,
      "tokens_seen": 1304559616
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006108926780341023,
      "loss": 1.4944,
      "theoretical_loss": 3.5886639989989932,
      "tokens_seen": 1304625152
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006108726178535608,
      "loss": 1.4228,
      "theoretical_loss": 3.588647806012765,
      "tokens_seen": 1304690688
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000610852557673019,
      "loss": 1.4914,
      "theoretical_loss": 3.5886316140676433,
      "tokens_seen": 1304756224
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006108324974924775,
      "loss": 1.4314,
      "theoretical_loss": 3.5886154231635095,
      "tokens_seen": 1304821760
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006108124373119358,
      "loss": 1.3798,
      "theoretical_loss": 3.588599233300244,
      "tokens_seen": 1304887296
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006107923771313942,
      "loss": 1.5918,
      "theoretical_loss": 3.588583044477728,
      "tokens_seen": 1304952832
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006107723169508526,
      "loss": 1.5529,
      "theoretical_loss": 3.5885668566958415,
      "tokens_seen": 1305018368
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000610752256770311,
      "loss": 1.5405,
      "theoretical_loss": 3.5885506699544667,
      "tokens_seen": 1305083904
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006107321965897693,
      "loss": 1.5434,
      "theoretical_loss": 3.5885344842534836,
      "tokens_seen": 1305149440
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006107121364092278,
      "loss": 1.4648,
      "theoretical_loss": 3.5885182995927734,
      "tokens_seen": 1305214976
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000610692076228686,
      "loss": 1.4889,
      "theoretical_loss": 3.588502115972217,
      "tokens_seen": 1305280512
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006106720160481445,
      "loss": 1.4819,
      "theoretical_loss": 3.588485933391694,
      "tokens_seen": 1305346048
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006106519558676028,
      "loss": 1.5029,
      "theoretical_loss": 3.588469751851088,
      "tokens_seen": 1305411584
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006106318956870612,
      "loss": 1.3937,
      "theoretical_loss": 3.5884535713502776,
      "tokens_seen": 1305477120
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006106118355065196,
      "loss": 1.4655,
      "theoretical_loss": 3.588437391889145,
      "tokens_seen": 1305542656
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000610591775325978,
      "loss": 1.404,
      "theoretical_loss": 3.5884212134675706,
      "tokens_seen": 1305608192
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006105717151454363,
      "loss": 1.3834,
      "theoretical_loss": 3.588405036085436,
      "tokens_seen": 1305673728
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006105516549648947,
      "loss": 1.5406,
      "theoretical_loss": 3.588388859742622,
      "tokens_seen": 1305739264
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000610531594784353,
      "loss": 1.495,
      "theoretical_loss": 3.5883726844390096,
      "tokens_seen": 1305804800
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006105115346038115,
      "loss": 1.3656,
      "theoretical_loss": 3.5883565101744797,
      "tokens_seen": 1305870336
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006104914744232698,
      "loss": 1.3946,
      "theoretical_loss": 3.588340336948914,
      "tokens_seen": 1305935872
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006104714142427282,
      "loss": 1.3312,
      "theoretical_loss": 3.5883241647621933,
      "tokens_seen": 1306001408
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006104513540621865,
      "loss": 1.4638,
      "theoretical_loss": 3.5883079936141984,
      "tokens_seen": 1306066944
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000610431293881645,
      "loss": 1.4872,
      "theoretical_loss": 3.588291823504811,
      "tokens_seen": 1306132480
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006104112337011033,
      "loss": 1.3695,
      "theoretical_loss": 3.5882756544339123,
      "tokens_seen": 1306198016
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006103911735205617,
      "loss": 1.4246,
      "theoretical_loss": 3.588259486401383,
      "tokens_seen": 1306263552
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006103711133400201,
      "loss": 1.5573,
      "theoretical_loss": 3.5882433194071046,
      "tokens_seen": 1306329088
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006103510531594785,
      "loss": 1.5107,
      "theoretical_loss": 3.5882271534509584,
      "tokens_seen": 1306394624
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006103309929789368,
      "loss": 1.344,
      "theoretical_loss": 3.5882109885328255,
      "tokens_seen": 1306460160
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006103109327983953,
      "loss": 1.4422,
      "theoretical_loss": 3.5881948246525877,
      "tokens_seen": 1306525696
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006102908726178535,
      "loss": 1.4296,
      "theoretical_loss": 3.588178661810126,
      "tokens_seen": 1306591232
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000610270812437312,
      "loss": 1.4692,
      "theoretical_loss": 3.588162500005321,
      "tokens_seen": 1306656768
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006102507522567703,
      "loss": 1.5758,
      "theoretical_loss": 3.588146339238055,
      "tokens_seen": 1306722304
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006102306920762287,
      "loss": 1.3827,
      "theoretical_loss": 3.588130179508209,
      "tokens_seen": 1306787840
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006102106318956871,
      "loss": 1.4877,
      "theoretical_loss": 3.5881140208156643,
      "tokens_seen": 1306853376
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006101905717151454,
      "loss": 1.5539,
      "theoretical_loss": 3.5880978631603027,
      "tokens_seen": 1306918912
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006101705115346038,
      "loss": 1.4009,
      "theoretical_loss": 3.588081706542005,
      "tokens_seen": 1306984448
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006101504513540623,
      "loss": 1.4665,
      "theoretical_loss": 3.5880655509606534,
      "tokens_seen": 1307049984
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006101303911735205,
      "loss": 1.4446,
      "theoretical_loss": 3.5880493964161286,
      "tokens_seen": 1307115520
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000610110330992979,
      "loss": 1.527,
      "theoretical_loss": 3.5880332429083124,
      "tokens_seen": 1307181056
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006100902708124372,
      "loss": 1.6144,
      "theoretical_loss": 3.588017090437087,
      "tokens_seen": 1307246592
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006100702106318957,
      "loss": 1.4525,
      "theoretical_loss": 3.5880009390023324,
      "tokens_seen": 1307312128
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006100501504513541,
      "loss": 1.3768,
      "theoretical_loss": 3.587984788603931,
      "tokens_seen": 1307377664
    },
    {
      "epoch": 0.4,
      "objective/train/advantage_avg": 0.49331116676330566,
      "objective/train/docs_used": 478148,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4744012355804443,
      "objective/train/original_loss": 1.4744012355804443,
      "objective/train/theoretical_loss": 3.587968639241765,
      "objective/train/tokens_used": 1327903200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24383458495140076,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505707263946533,
      "objective/train/weighted_lm_loss": 1.5488524436950684,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.587968639241765,
      "tokens_seen": 1307443200
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006100300902708124,
      "loss": 1.4596,
      "theoretical_loss": 3.587968639241765,
      "tokens_seen": 1307443200
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006100100300902708,
      "loss": 1.5384,
      "theoretical_loss": 3.5879524909157148,
      "tokens_seen": 1307508736
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006099899699097293,
      "loss": 1.3564,
      "theoretical_loss": 3.5879363436256626,
      "tokens_seen": 1307574272
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006099699097291875,
      "loss": 1.4841,
      "theoretical_loss": 3.5879201973714903,
      "tokens_seen": 1307639808
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000609949849548646,
      "loss": 1.5025,
      "theoretical_loss": 3.5879040521530787,
      "tokens_seen": 1307705344
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006099297893681042,
      "loss": 1.533,
      "theoretical_loss": 3.58788790797031,
      "tokens_seen": 1307770880
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006099097291875627,
      "loss": 1.4269,
      "theoretical_loss": 3.587871764823066,
      "tokens_seen": 1307836416
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006098896690070211,
      "loss": 1.429,
      "theoretical_loss": 3.5878556227112277,
      "tokens_seen": 1307901952
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006098696088264794,
      "loss": 1.5446,
      "theoretical_loss": 3.5878394816346777,
      "tokens_seen": 1307967488
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006098495486459378,
      "loss": 1.5219,
      "theoretical_loss": 3.587823341593297,
      "tokens_seen": 1308033024
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006098294884653962,
      "loss": 1.5153,
      "theoretical_loss": 3.5878072025869683,
      "tokens_seen": 1308098560
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006098094282848545,
      "loss": 1.4853,
      "theoretical_loss": 3.5877910646155726,
      "tokens_seen": 1308164096
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000609789368104313,
      "loss": 1.4651,
      "theoretical_loss": 3.5877749276789914,
      "tokens_seen": 1308229632
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006097693079237712,
      "loss": 1.3761,
      "theoretical_loss": 3.587758791777107,
      "tokens_seen": 1308295168
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006097492477432297,
      "loss": 1.4058,
      "theoretical_loss": 3.5877426569098017,
      "tokens_seen": 1308360704
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000609729187562688,
      "loss": 1.4353,
      "theoretical_loss": 3.5877265230769564,
      "tokens_seen": 1308426240
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006097091273821464,
      "loss": 1.4377,
      "theoretical_loss": 3.5877103902784535,
      "tokens_seen": 1308491776
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006096890672016048,
      "loss": 1.3646,
      "theoretical_loss": 3.587694258514175,
      "tokens_seen": 1308557312
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006096690070210632,
      "loss": 1.4456,
      "theoretical_loss": 3.5876781277840024,
      "tokens_seen": 1308622848
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006096489468405215,
      "loss": 1.5819,
      "theoretical_loss": 3.5876619980878175,
      "tokens_seen": 1308688384
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00060962888665998,
      "loss": 1.3724,
      "theoretical_loss": 3.5876458694255033,
      "tokens_seen": 1308753920
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006096088264794384,
      "loss": 1.4428,
      "theoretical_loss": 3.58762974179694,
      "tokens_seen": 1308819456
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006095887662988967,
      "loss": 1.4616,
      "theoretical_loss": 3.5876136152020117,
      "tokens_seen": 1308884992
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006095687061183551,
      "loss": 1.4459,
      "theoretical_loss": 3.587597489640599,
      "tokens_seen": 1308950528
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006095486459378134,
      "loss": 1.4137,
      "theoretical_loss": 3.587581365112584,
      "tokens_seen": 1309016064
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006095285857572719,
      "loss": 1.4942,
      "theoretical_loss": 3.587565241617849,
      "tokens_seen": 1309081600
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006095085255767302,
      "loss": 1.5381,
      "theoretical_loss": 3.587549119156276,
      "tokens_seen": 1309147136
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006094884653961886,
      "loss": 1.5721,
      "theoretical_loss": 3.5875329977277475,
      "tokens_seen": 1309212672
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006094684052156469,
      "loss": 1.5649,
      "theoretical_loss": 3.5875168773321455,
      "tokens_seen": 1309278208
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006094483450351054,
      "loss": 1.5131,
      "theoretical_loss": 3.5875007579693508,
      "tokens_seen": 1309343744
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006094282848545637,
      "loss": 1.5175,
      "theoretical_loss": 3.5874846396392472,
      "tokens_seen": 1309409280
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006094082246740221,
      "loss": 1.5652,
      "theoretical_loss": 3.5874685223417164,
      "tokens_seen": 1309474816
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006093881644934805,
      "loss": 1.4678,
      "theoretical_loss": 3.5874524060766406,
      "tokens_seen": 1309540352
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006093681043129389,
      "loss": 1.4359,
      "theoretical_loss": 3.5874362908439013,
      "tokens_seen": 1309605888
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006093480441323972,
      "loss": 1.4496,
      "theoretical_loss": 3.5874201766433815,
      "tokens_seen": 1309671424
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006093279839518557,
      "loss": 1.4553,
      "theoretical_loss": 3.5874040634749633,
      "tokens_seen": 1309736960
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006093079237713139,
      "loss": 1.4861,
      "theoretical_loss": 3.5873879513385285,
      "tokens_seen": 1309802496
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006092878635907724,
      "loss": 1.5067,
      "theoretical_loss": 3.5873718402339603,
      "tokens_seen": 1309868032
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006092678034102307,
      "loss": 1.5395,
      "theoretical_loss": 3.58735573016114,
      "tokens_seen": 1309933568
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006092477432296891,
      "loss": 1.5147,
      "theoretical_loss": 3.58733962111995,
      "tokens_seen": 1309999104
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006092276830491475,
      "loss": 1.3346,
      "theoretical_loss": 3.587323513110274,
      "tokens_seen": 1310064640
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006092076228686058,
      "loss": 1.5365,
      "theoretical_loss": 3.587307406131992,
      "tokens_seen": 1310130176
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006091875626880642,
      "loss": 1.5703,
      "theoretical_loss": 3.5872913001849884,
      "tokens_seen": 1310195712
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006091675025075227,
      "loss": 1.3931,
      "theoretical_loss": 3.587275195269145,
      "tokens_seen": 1310261248
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006091474423269809,
      "loss": 1.3399,
      "theoretical_loss": 3.587259091384344,
      "tokens_seen": 1310326784
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006091273821464394,
      "loss": 1.4689,
      "theoretical_loss": 3.587242988530468,
      "tokens_seen": 1310392320
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006091073219658976,
      "loss": 1.5258,
      "theoretical_loss": 3.5872268867073993,
      "tokens_seen": 1310457856
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006090872617853561,
      "loss": 1.3836,
      "theoretical_loss": 3.58721078591502,
      "tokens_seen": 1310523392
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006090672016048145,
      "loss": 1.4451,
      "theoretical_loss": 3.5871946861532136,
      "tokens_seen": 1310588928
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006090471414242728,
      "loss": 1.3871,
      "theoretical_loss": 3.587178587421861,
      "tokens_seen": 1310654464
    },
    {
      "debugging/Compilability": 0.9,
      "debugging/distinct-1-grams": 0.7732995475821178,
      "debugging/entropy-1-grams": 5.431960387820077,
      "debugging/length": 501.4,
      "debugging/num_segments": 10,
      "debugging/raw_token_scores_avg": 0.009144442155957222,
      "debugging/raw_token_scores_std": 0.03945710137486458,
      "debugging/score": 0.007191963134632398,
      "debugging/score_std": 0.006508080924945895,
      "epoch": 0.4,
      "objective/train/advantage_avg": 0.4908607304096222,
      "objective/train/docs_used": 479375,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.447709083557129,
      "objective/train/original_loss": 1.4477088451385498,
      "objective/train/theoretical_loss": 3.587162489720847,
      "objective/train/tokens_used": 1331180000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24250146746635437,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050318956375122,
      "objective/train/weighted_lm_loss": 1.5207688808441162,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.587162489720847,
      "tokens_seen": 1310720000
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006090270812437312,
      "loss": 1.4967,
      "theoretical_loss": 3.587162489720847,
      "tokens_seen": 1310720000
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006090070210631897,
      "loss": 1.5214,
      "theoretical_loss": 3.5871463930500527,
      "tokens_seen": 1310785536
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006089869608826479,
      "loss": 1.3395,
      "theoretical_loss": 3.5871302974093604,
      "tokens_seen": 1310851072
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006089669007021064,
      "loss": 1.4327,
      "theoretical_loss": 3.587114202798653,
      "tokens_seen": 1310916608
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006089468405215646,
      "loss": 1.4543,
      "theoretical_loss": 3.5870981092178136,
      "tokens_seen": 1310982144
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006089267803410231,
      "loss": 1.4626,
      "theoretical_loss": 3.5870820166667245,
      "tokens_seen": 1311047680
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006089067201604815,
      "loss": 1.4673,
      "theoretical_loss": 3.5870659251452683,
      "tokens_seen": 1311113216
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006088866599799398,
      "loss": 1.457,
      "theoretical_loss": 3.5870498346533277,
      "tokens_seen": 1311178752
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006088665997993982,
      "loss": 1.528,
      "theoretical_loss": 3.5870337451907854,
      "tokens_seen": 1311244288
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006088465396188566,
      "loss": 1.5569,
      "theoretical_loss": 3.587017656757524,
      "tokens_seen": 1311309824
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006088264794383149,
      "loss": 1.5188,
      "theoretical_loss": 3.587001569353426,
      "tokens_seen": 1311375360
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006088064192577734,
      "loss": 1.3973,
      "theoretical_loss": 3.586985482978375,
      "tokens_seen": 1311440896
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006087863590772316,
      "loss": 1.4872,
      "theoretical_loss": 3.5869693976322523,
      "tokens_seen": 1311506432
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006087662988966901,
      "loss": 1.3441,
      "theoretical_loss": 3.5869533133149423,
      "tokens_seen": 1311571968
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006087462387161484,
      "loss": 1.4702,
      "theoretical_loss": 3.5869372300263267,
      "tokens_seen": 1311637504
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006087261785356068,
      "loss": 1.4659,
      "theoretical_loss": 3.5869211477662883,
      "tokens_seen": 1311703040
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006087061183550652,
      "loss": 1.3033,
      "theoretical_loss": 3.586905066534711,
      "tokens_seen": 1311768576
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006086860581745236,
      "loss": 1.3238,
      "theoretical_loss": 3.586888986331477,
      "tokens_seen": 1311834112
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006086659979939819,
      "loss": 1.3238,
      "theoretical_loss": 3.5868729071564687,
      "tokens_seen": 1311899648
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006086459378134404,
      "loss": 1.4317,
      "theoretical_loss": 3.5868568290095695,
      "tokens_seen": 1311965184
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006086258776328987,
      "loss": 1.3915,
      "theoretical_loss": 3.5868407518906618,
      "tokens_seen": 1312030720
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006086058174523571,
      "loss": 1.5067,
      "theoretical_loss": 3.5868246757996296,
      "tokens_seen": 1312096256
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006085857572718154,
      "loss": 1.4726,
      "theoretical_loss": 3.5868086007363544,
      "tokens_seen": 1312161792
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006085656970912738,
      "loss": 1.5533,
      "theoretical_loss": 3.586792526700721,
      "tokens_seen": 1312227328
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006085456369107322,
      "loss": 1.499,
      "theoretical_loss": 3.586776453692611,
      "tokens_seen": 1312292864
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006085255767301906,
      "loss": 1.4701,
      "theoretical_loss": 3.5867603817119074,
      "tokens_seen": 1312358400
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006085055165496489,
      "loss": 1.5319,
      "theoretical_loss": 3.5867443107584935,
      "tokens_seen": 1312423936
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006084854563691073,
      "loss": 1.4871,
      "theoretical_loss": 3.5867282408322527,
      "tokens_seen": 1312489472
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006084653961885657,
      "loss": 1.3257,
      "theoretical_loss": 3.5867121719330677,
      "tokens_seen": 1312555008
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006084453360080241,
      "loss": 1.5063,
      "theoretical_loss": 3.586696104060822,
      "tokens_seen": 1312620544
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006084252758274824,
      "loss": 1.4624,
      "theoretical_loss": 3.586680037215398,
      "tokens_seen": 1312686080
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006084052156469409,
      "loss": 1.4752,
      "theoretical_loss": 3.5866639713966793,
      "tokens_seen": 1312751616
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006083851554663992,
      "loss": 1.4184,
      "theoretical_loss": 3.586647906604549,
      "tokens_seen": 1312817152
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006083650952858576,
      "loss": 1.3782,
      "theoretical_loss": 3.58663184283889,
      "tokens_seen": 1312882688
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006083450351053159,
      "loss": 1.6428,
      "theoretical_loss": 3.586615780099586,
      "tokens_seen": 1312948224
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006083249749247743,
      "loss": 1.4478,
      "theoretical_loss": 3.5865997183865197,
      "tokens_seen": 1313013760
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006083049147442327,
      "loss": 1.3716,
      "theoretical_loss": 3.5865836576995744,
      "tokens_seen": 1313079296
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006082848545636911,
      "loss": 1.3522,
      "theoretical_loss": 3.5865675980386333,
      "tokens_seen": 1313144832
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006082647943831494,
      "loss": 1.4215,
      "theoretical_loss": 3.58655153940358,
      "tokens_seen": 1313210368
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006082447342026079,
      "loss": 1.6176,
      "theoretical_loss": 3.5865354817942974,
      "tokens_seen": 1313275904
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006082246740220661,
      "loss": 1.4282,
      "theoretical_loss": 3.5865194252106694,
      "tokens_seen": 1313341440
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006082046138415246,
      "loss": 1.5074,
      "theoretical_loss": 3.5865033696525783,
      "tokens_seen": 1313406976
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000608184553660983,
      "loss": 1.5373,
      "theoretical_loss": 3.5864873151199084,
      "tokens_seen": 1313472512
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006081644934804413,
      "loss": 1.4479,
      "theoretical_loss": 3.5864712616125427,
      "tokens_seen": 1313538048
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006081444332998997,
      "loss": 1.5122,
      "theoretical_loss": 3.586455209130364,
      "tokens_seen": 1313603584
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000608124373119358,
      "loss": 1.6029,
      "theoretical_loss": 3.5864391576732566,
      "tokens_seen": 1313669120
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006081043129388165,
      "loss": 1.4998,
      "theoretical_loss": 3.5864231072411035,
      "tokens_seen": 1313734656
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006080842527582749,
      "loss": 1.525,
      "theoretical_loss": 3.586407057833788,
      "tokens_seen": 1313800192
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006080641925777332,
      "loss": 1.4643,
      "theoretical_loss": 3.5863910094511935,
      "tokens_seen": 1313865728
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006080441323971916,
      "loss": 1.5604,
      "theoretical_loss": 3.586374962093204,
      "tokens_seen": 1313931264
    },
    {
      "epoch": 0.4,
      "objective/train/advantage_avg": 0.49490660429000854,
      "objective/train/docs_used": 480644,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4059205055236816,
      "objective/train/original_loss": 1.4059206247329712,
      "objective/train/theoretical_loss": 3.5863589157597024,
      "objective/train/tokens_used": 1334456800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24508771300315857,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050736665725708,
      "objective/train/weighted_lm_loss": 1.47727632522583,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5863589157597024,
      "tokens_seen": 1313996800
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006080240722166501,
      "loss": 1.478,
      "theoretical_loss": 3.5863589157597024,
      "tokens_seen": 1313996800
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006080040120361083,
      "loss": 1.3939,
      "theoretical_loss": 3.5863428704505727,
      "tokens_seen": 1314062336
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006079839518555668,
      "loss": 1.4325,
      "theoretical_loss": 3.586326826165698,
      "tokens_seen": 1314127872
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000607963891675025,
      "loss": 1.4412,
      "theoretical_loss": 3.586310782904962,
      "tokens_seen": 1314193408
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006079438314944835,
      "loss": 1.4008,
      "theoretical_loss": 3.586294740668248,
      "tokens_seen": 1314258944
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006079237713139419,
      "loss": 1.474,
      "theoretical_loss": 3.5862786994554403,
      "tokens_seen": 1314324480
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006079037111334002,
      "loss": 1.5713,
      "theoretical_loss": 3.5862626592664215,
      "tokens_seen": 1314390016
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006078836509528586,
      "loss": 1.5087,
      "theoretical_loss": 3.5862466201010763,
      "tokens_seen": 1314455552
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000607863590772317,
      "loss": 1.4944,
      "theoretical_loss": 3.5862305819592875,
      "tokens_seen": 1314521088
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006078435305917753,
      "loss": 1.3249,
      "theoretical_loss": 3.5862145448409395,
      "tokens_seen": 1314586624
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006078234704112338,
      "loss": 1.4527,
      "theoretical_loss": 3.586198508745915,
      "tokens_seen": 1314652160
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000607803410230692,
      "loss": 1.5452,
      "theoretical_loss": 3.586182473674098,
      "tokens_seen": 1314717696
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006077833500501505,
      "loss": 1.4373,
      "theoretical_loss": 3.586166439625373,
      "tokens_seen": 1314783232
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006077632898696088,
      "loss": 1.3894,
      "theoretical_loss": 3.5861504065996233,
      "tokens_seen": 1314848768
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006077432296890672,
      "loss": 1.432,
      "theoretical_loss": 3.586134374596732,
      "tokens_seen": 1314914304
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006077231695085256,
      "loss": 1.4749,
      "theoretical_loss": 3.5861183436165835,
      "tokens_seen": 1314979840
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000607703109327984,
      "loss": 1.4287,
      "theoretical_loss": 3.5861023136590617,
      "tokens_seen": 1315045376
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006076830491474423,
      "loss": 1.524,
      "theoretical_loss": 3.5860862847240496,
      "tokens_seen": 1315110912
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006076629889669008,
      "loss": 1.505,
      "theoretical_loss": 3.586070256811432,
      "tokens_seen": 1315176448
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006076429287863591,
      "loss": 1.3738,
      "theoretical_loss": 3.5860542299210927,
      "tokens_seen": 1315241984
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006076228686058175,
      "loss": 1.4226,
      "theoretical_loss": 3.5860382040529144,
      "tokens_seen": 1315307520
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006076028084252758,
      "loss": 1.3158,
      "theoretical_loss": 3.5860221792067826,
      "tokens_seen": 1315373056
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006075827482447342,
      "loss": 1.5796,
      "theoretical_loss": 3.58600615538258,
      "tokens_seen": 1315438592
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006075626880641926,
      "loss": 1.4961,
      "theoretical_loss": 3.585990132580191,
      "tokens_seen": 1315504128
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000607542627883651,
      "loss": 1.4686,
      "theoretical_loss": 3.5859741107994996,
      "tokens_seen": 1315569664
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006075225677031093,
      "loss": 1.5673,
      "theoretical_loss": 3.585958090040389,
      "tokens_seen": 1315635200
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006075025075225677,
      "loss": 1.5461,
      "theoretical_loss": 3.5859420703027447,
      "tokens_seen": 1315700736
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006074824473420261,
      "loss": 1.3928,
      "theoretical_loss": 3.5859260515864495,
      "tokens_seen": 1315766272
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006074623871614845,
      "loss": 1.5442,
      "theoretical_loss": 3.585910033891387,
      "tokens_seen": 1315831808
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006074423269809428,
      "loss": 1.4861,
      "theoretical_loss": 3.5858940172174427,
      "tokens_seen": 1315897344
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006074222668004013,
      "loss": 1.3421,
      "theoretical_loss": 3.5858780015644998,
      "tokens_seen": 1315962880
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006074022066198596,
      "loss": 1.6268,
      "theoretical_loss": 3.5858619869324424,
      "tokens_seen": 1316028416
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000607382146439318,
      "loss": 1.5434,
      "theoretical_loss": 3.5858459733211547,
      "tokens_seen": 1316093952
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006073620862587763,
      "loss": 1.4999,
      "theoretical_loss": 3.585829960730521,
      "tokens_seen": 1316159488
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006073420260782347,
      "loss": 1.4885,
      "theoretical_loss": 3.585813949160425,
      "tokens_seen": 1316225024
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006073219658976931,
      "loss": 1.3856,
      "theoretical_loss": 3.5857979386107512,
      "tokens_seen": 1316290560
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006073019057171515,
      "loss": 1.4235,
      "theoretical_loss": 3.5857819290813833,
      "tokens_seen": 1316356096
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006072818455366098,
      "loss": 1.5784,
      "theoretical_loss": 3.585765920572206,
      "tokens_seen": 1316421632
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006072617853560683,
      "loss": 1.3738,
      "theoretical_loss": 3.585749913083103,
      "tokens_seen": 1316487168
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006072417251755265,
      "loss": 1.5649,
      "theoretical_loss": 3.5857339066139593,
      "tokens_seen": 1316552704
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000607221664994985,
      "loss": 1.601,
      "theoretical_loss": 3.5857179011646587,
      "tokens_seen": 1316618240
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006072016048144433,
      "loss": 1.5604,
      "theoretical_loss": 3.585701896735085,
      "tokens_seen": 1316683776
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006071815446339017,
      "loss": 1.4542,
      "theoretical_loss": 3.5856858933251234,
      "tokens_seen": 1316749312
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006071614844533601,
      "loss": 1.5474,
      "theoretical_loss": 3.585669890934658,
      "tokens_seen": 1316814848
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006071414242728184,
      "loss": 1.4091,
      "theoretical_loss": 3.585653889563572,
      "tokens_seen": 1316880384
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006071213640922768,
      "loss": 1.5782,
      "theoretical_loss": 3.5856378892117506,
      "tokens_seen": 1316945920
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006071013039117353,
      "loss": 1.4398,
      "theoretical_loss": 3.5856218898790786,
      "tokens_seen": 1317011456
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006070812437311935,
      "loss": 1.4224,
      "theoretical_loss": 3.5856058915654403,
      "tokens_seen": 1317076992
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000607061183550652,
      "loss": 1.5366,
      "theoretical_loss": 3.585589894270719,
      "tokens_seen": 1317142528
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006070411233701104,
      "loss": 1.612,
      "theoretical_loss": 3.5855738979948004,
      "tokens_seen": 1317208064
    },
    {
      "epoch": 0.4,
      "objective/train/advantage_avg": 0.48856857419013977,
      "objective/train/docs_used": 481887,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5049604177474976,
      "objective/train/original_loss": 1.504960298538208,
      "objective/train/theoretical_loss": 3.585557902737568,
      "objective/train/tokens_used": 1337733600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2412024587392807,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500831604003906,
      "objective/train/weighted_lm_loss": 1.5808225870132446,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.585557902737568,
      "tokens_seen": 1317273600
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006070210631895687,
      "loss": 1.5288,
      "theoretical_loss": 3.585557902737568,
      "tokens_seen": 1317273600
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006070010030090271,
      "loss": 1.601,
      "theoretical_loss": 3.5855419084989064,
      "tokens_seen": 1317339136
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006069809428284854,
      "loss": 1.7366,
      "theoretical_loss": 3.5855259152787005,
      "tokens_seen": 1317404672
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006069608826479438,
      "loss": 1.5987,
      "theoretical_loss": 3.5855099230768346,
      "tokens_seen": 1317470208
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006069408224674023,
      "loss": 1.4122,
      "theoretical_loss": 3.5854939318931933,
      "tokens_seen": 1317535744
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006069207622868605,
      "loss": 1.467,
      "theoretical_loss": 3.585477941727661,
      "tokens_seen": 1317601280
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000606900702106319,
      "loss": 1.5625,
      "theoretical_loss": 3.5854619525801223,
      "tokens_seen": 1317666816
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006068806419257773,
      "loss": 1.3106,
      "theoretical_loss": 3.5854459644504617,
      "tokens_seen": 1317732352
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006068605817452357,
      "loss": 1.5903,
      "theoretical_loss": 3.5854299773385634,
      "tokens_seen": 1317797888
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006068405215646941,
      "loss": 1.4068,
      "theoretical_loss": 3.585413991244313,
      "tokens_seen": 1317863424
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006068204613841524,
      "loss": 1.5438,
      "theoretical_loss": 3.585398006167594,
      "tokens_seen": 1317928960
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006068004012036108,
      "loss": 1.5296,
      "theoretical_loss": 3.585382022108292,
      "tokens_seen": 1317994496
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006067803410230692,
      "loss": 1.4954,
      "theoretical_loss": 3.585366039066291,
      "tokens_seen": 1318060032
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006067602808425275,
      "loss": 1.4115,
      "theoretical_loss": 3.5853500570414756,
      "tokens_seen": 1318125568
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000606740220661986,
      "loss": 1.4746,
      "theoretical_loss": 3.5853340760337313,
      "tokens_seen": 1318191104
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006067201604814443,
      "loss": 1.5498,
      "theoretical_loss": 3.5853180960429425,
      "tokens_seen": 1318256640
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006067001003009027,
      "loss": 1.469,
      "theoretical_loss": 3.585302117068993,
      "tokens_seen": 1318322176
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006066800401203611,
      "loss": 1.4587,
      "theoretical_loss": 3.5852861391117687,
      "tokens_seen": 1318387712
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006066599799398195,
      "loss": 1.4935,
      "theoretical_loss": 3.5852701621711542,
      "tokens_seen": 1318453248
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006066399197592778,
      "loss": 1.4433,
      "theoretical_loss": 3.5852541862470337,
      "tokens_seen": 1318518784
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006066198595787362,
      "loss": 1.5469,
      "theoretical_loss": 3.585238211339292,
      "tokens_seen": 1318584320
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006065997993981947,
      "loss": 1.3261,
      "theoretical_loss": 3.585222237447815,
      "tokens_seen": 1318649856
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000606579739217653,
      "loss": 1.5822,
      "theoretical_loss": 3.5852062645724865,
      "tokens_seen": 1318715392
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006065596790371114,
      "loss": 1.4573,
      "theoretical_loss": 3.585190292713192,
      "tokens_seen": 1318780928
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006065396188565697,
      "loss": 1.4208,
      "theoretical_loss": 3.5851743218698156,
      "tokens_seen": 1318846464
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006065195586760281,
      "loss": 1.4881,
      "theoretical_loss": 3.585158352042243,
      "tokens_seen": 1318912000
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006064994984954865,
      "loss": 1.4187,
      "theoretical_loss": 3.585142383230359,
      "tokens_seen": 1318977536
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006064794383149449,
      "loss": 1.3802,
      "theoretical_loss": 3.585126415434048,
      "tokens_seen": 1319043072
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006064593781344032,
      "loss": 1.5821,
      "theoretical_loss": 3.585110448653195,
      "tokens_seen": 1319108608
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006064393179538617,
      "loss": 1.486,
      "theoretical_loss": 3.585094482887686,
      "tokens_seen": 1319174144
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00060641925777332,
      "loss": 1.45,
      "theoretical_loss": 3.5850785181374047,
      "tokens_seen": 1319239680
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006063991975927784,
      "loss": 1.3952,
      "theoretical_loss": 3.585062554402237,
      "tokens_seen": 1319305216
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006063791374122367,
      "loss": 1.4148,
      "theoretical_loss": 3.585046591682068,
      "tokens_seen": 1319370752
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006063590772316951,
      "loss": 1.4897,
      "theoretical_loss": 3.5850306299767816,
      "tokens_seen": 1319436288
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006063390170511535,
      "loss": 1.3789,
      "theoretical_loss": 3.585014669286264,
      "tokens_seen": 1319501824
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006063189568706119,
      "loss": 1.5434,
      "theoretical_loss": 3.5849987096104003,
      "tokens_seen": 1319567360
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006062988966900702,
      "loss": 1.5,
      "theoretical_loss": 3.5849827509490746,
      "tokens_seen": 1319632896
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006062788365095287,
      "loss": 1.4577,
      "theoretical_loss": 3.5849667933021734,
      "tokens_seen": 1319698432
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006062587763289869,
      "loss": 1.4538,
      "theoretical_loss": 3.584950836669581,
      "tokens_seen": 1319763968
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006062387161484454,
      "loss": 1.519,
      "theoretical_loss": 3.584934881051182,
      "tokens_seen": 1319829504
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006062186559679037,
      "loss": 1.4665,
      "theoretical_loss": 3.584918926446863,
      "tokens_seen": 1319895040
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006061985957873621,
      "loss": 1.5453,
      "theoretical_loss": 3.584902972856508,
      "tokens_seen": 1319960576
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006061785356068205,
      "loss": 1.3988,
      "theoretical_loss": 3.5848870202800027,
      "tokens_seen": 1320026112
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006061584754262788,
      "loss": 1.3519,
      "theoretical_loss": 3.5848710687172325,
      "tokens_seen": 1320091648
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006061384152457372,
      "loss": 1.5099,
      "theoretical_loss": 3.5848551181680826,
      "tokens_seen": 1320157184
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006061183550651957,
      "loss": 1.3721,
      "theoretical_loss": 3.5848391686324383,
      "tokens_seen": 1320222720
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006060982948846539,
      "loss": 1.4346,
      "theoretical_loss": 3.5848232201101844,
      "tokens_seen": 1320288256
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006060782347041124,
      "loss": 1.3813,
      "theoretical_loss": 3.584807272601207,
      "tokens_seen": 1320353792
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006060581745235708,
      "loss": 1.374,
      "theoretical_loss": 3.5847913261053908,
      "tokens_seen": 1320419328
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006060381143430291,
      "loss": 1.5569,
      "theoretical_loss": 3.584775380622621,
      "tokens_seen": 1320484864
    },
    {
      "epoch": 0.4,
      "objective/train/advantage_avg": 0.4927046000957489,
      "objective/train/docs_used": 483102,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3539386987686157,
      "objective/train/original_loss": 1.3539388179779053,
      "objective/train/theoretical_loss": 3.584759436152784,
      "objective/train/tokens_used": 1341010400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24326573312282562,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505071878433228,
      "objective/train/weighted_lm_loss": 1.4223684072494507,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.584759436152784,
      "tokens_seen": 1320550400
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006060180541624875,
      "loss": 1.4856,
      "theoretical_loss": 3.584759436152784,
      "tokens_seen": 1320550400
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006059979939819458,
      "loss": 1.5033,
      "theoretical_loss": 3.584743492695764,
      "tokens_seen": 1320615936
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006059779338014042,
      "loss": 1.3582,
      "theoretical_loss": 3.584727550251447,
      "tokens_seen": 1320681472
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006059578736208627,
      "loss": 1.4808,
      "theoretical_loss": 3.5847116088197186,
      "tokens_seen": 1320747008
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006059378134403209,
      "loss": 1.5581,
      "theoretical_loss": 3.584695668400464,
      "tokens_seen": 1320812544
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006059177532597794,
      "loss": 1.4448,
      "theoretical_loss": 3.584679728993569,
      "tokens_seen": 1320878080
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006058976930792377,
      "loss": 1.5621,
      "theoretical_loss": 3.5846637905989183,
      "tokens_seen": 1320943616
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006058776328986961,
      "loss": 1.5133,
      "theoretical_loss": 3.5846478532163975,
      "tokens_seen": 1321009152
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006058575727181545,
      "loss": 1.3537,
      "theoretical_loss": 3.5846319168458933,
      "tokens_seen": 1321074688
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006058375125376128,
      "loss": 1.4992,
      "theoretical_loss": 3.58461598148729,
      "tokens_seen": 1321140224
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006058174523570712,
      "loss": 1.4767,
      "theoretical_loss": 3.5846000471404738,
      "tokens_seen": 1321205760
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006057973921765296,
      "loss": 1.5572,
      "theoretical_loss": 3.584584113805329,
      "tokens_seen": 1321271296
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006057773319959879,
      "loss": 1.5398,
      "theoretical_loss": 3.5845681814817434,
      "tokens_seen": 1321336832
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006057572718154464,
      "loss": 1.678,
      "theoretical_loss": 3.584552250169601,
      "tokens_seen": 1321402368
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006057372116349047,
      "loss": 1.5228,
      "theoretical_loss": 3.5845363198687883,
      "tokens_seen": 1321467904
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006057171514543631,
      "loss": 1.5277,
      "theoretical_loss": 3.58452039057919,
      "tokens_seen": 1321533440
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006056970912738215,
      "loss": 1.5404,
      "theoretical_loss": 3.5845044623006928,
      "tokens_seen": 1321598976
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006056770310932799,
      "loss": 1.4862,
      "theoretical_loss": 3.584488535033181,
      "tokens_seen": 1321664512
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006056569709127382,
      "loss": 1.4234,
      "theoretical_loss": 3.584472608776542,
      "tokens_seen": 1321730048
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006056369107321966,
      "loss": 1.4974,
      "theoretical_loss": 3.584456683530661,
      "tokens_seen": 1321795584
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006056168505516549,
      "loss": 1.5711,
      "theoretical_loss": 3.584440759295423,
      "tokens_seen": 1321861120
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006055967903711134,
      "loss": 1.529,
      "theoretical_loss": 3.584424836070714,
      "tokens_seen": 1321926656
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006055767301905717,
      "loss": 1.4233,
      "theoretical_loss": 3.5844089138564197,
      "tokens_seen": 1321992192
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006055566700100301,
      "loss": 1.4406,
      "theoretical_loss": 3.584392992652427,
      "tokens_seen": 1322057728
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006055366098294884,
      "loss": 1.4403,
      "theoretical_loss": 3.5843770724586204,
      "tokens_seen": 1322123264
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006055165496489469,
      "loss": 1.5526,
      "theoretical_loss": 3.5843611532748865,
      "tokens_seen": 1322188800
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006054964894684052,
      "loss": 1.282,
      "theoretical_loss": 3.584345235101111,
      "tokens_seen": 1322254336
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006054764292878636,
      "loss": 1.491,
      "theoretical_loss": 3.584329317937179,
      "tokens_seen": 1322319872
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000605456369107322,
      "loss": 1.3582,
      "theoretical_loss": 3.5843134017829783,
      "tokens_seen": 1322385408
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006054363089267804,
      "loss": 1.5328,
      "theoretical_loss": 3.5842974866383925,
      "tokens_seen": 1322450944
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006054162487462387,
      "loss": 1.4475,
      "theoretical_loss": 3.584281572503309,
      "tokens_seen": 1322516480
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006053961885656971,
      "loss": 1.4769,
      "theoretical_loss": 3.584265659377613,
      "tokens_seen": 1322582016
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006053761283851554,
      "loss": 1.4876,
      "theoretical_loss": 3.5842497472611914,
      "tokens_seen": 1322647552
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006053560682046139,
      "loss": 1.5264,
      "theoretical_loss": 3.5842338361539294,
      "tokens_seen": 1322713088
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006053360080240722,
      "loss": 1.5019,
      "theoretical_loss": 3.584217926055713,
      "tokens_seen": 1322778624
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006053159478435306,
      "loss": 1.4834,
      "theoretical_loss": 3.5842020169664286,
      "tokens_seen": 1322844160
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000605295887662989,
      "loss": 1.4961,
      "theoretical_loss": 3.5841861088859623,
      "tokens_seen": 1322909696
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006052758274824473,
      "loss": 1.522,
      "theoretical_loss": 3.5841702018141994,
      "tokens_seen": 1322975232
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006052557673019057,
      "loss": 1.5227,
      "theoretical_loss": 3.584154295751027,
      "tokens_seen": 1323040768
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006052357071213642,
      "loss": 1.4323,
      "theoretical_loss": 3.5841383906963307,
      "tokens_seen": 1323106304
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006052156469408224,
      "loss": 1.4575,
      "theoretical_loss": 3.5841224866499966,
      "tokens_seen": 1323171840
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006051955867602809,
      "loss": 1.4651,
      "theoretical_loss": 3.58410658361191,
      "tokens_seen": 1323237376
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006051755265797391,
      "loss": 1.5461,
      "theoretical_loss": 3.5840906815819586,
      "tokens_seen": 1323302912
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006051554663991976,
      "loss": 1.3878,
      "theoretical_loss": 3.5840747805600284,
      "tokens_seen": 1323368448
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000605135406218656,
      "loss": 1.5013,
      "theoretical_loss": 3.584058880546004,
      "tokens_seen": 1323433984
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006051153460381143,
      "loss": 1.4778,
      "theoretical_loss": 3.584042981539773,
      "tokens_seen": 1323499520
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006050952858575728,
      "loss": 1.6121,
      "theoretical_loss": 3.584027083541222,
      "tokens_seen": 1323565056
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006050752256770312,
      "loss": 1.4584,
      "theoretical_loss": 3.5840111865502355,
      "tokens_seen": 1323630592
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006050551654964895,
      "loss": 1.512,
      "theoretical_loss": 3.583995290566701,
      "tokens_seen": 1323696128
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006050351053159479,
      "loss": 1.4902,
      "theoretical_loss": 3.5839793955905046,
      "tokens_seen": 1323761664
    },
    {
      "epoch": 0.4,
      "objective/train/advantage_avg": 0.4919940233230591,
      "objective/train/docs_used": 484332,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4468096494674683,
      "objective/train/original_loss": 1.4468095302581787,
      "objective/train/theoretical_loss": 3.583963501621533,
      "objective/train/tokens_used": 1344287200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24277126789093018,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504335165023804,
      "objective/train/weighted_lm_loss": 1.5197827816009521,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.583963501621533,
      "tokens_seen": 1323827200
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006050150451354062,
      "loss": 1.4892,
      "theoretical_loss": 3.583963501621533,
      "tokens_seen": 1323827200
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006049949849548646,
      "loss": 1.4163,
      "theoretical_loss": 3.5839476086596718,
      "tokens_seen": 1323892736
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006049749247743231,
      "loss": 1.464,
      "theoretical_loss": 3.5839317167048073,
      "tokens_seen": 1323958272
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006049548645937813,
      "loss": 1.4226,
      "theoretical_loss": 3.5839158257568267,
      "tokens_seen": 1324023808
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006049348044132398,
      "loss": 1.5894,
      "theoretical_loss": 3.583899935815616,
      "tokens_seen": 1324089344
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006049147442326981,
      "loss": 1.4187,
      "theoretical_loss": 3.583884046881061,
      "tokens_seen": 1324154880
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006048946840521565,
      "loss": 1.3635,
      "theoretical_loss": 3.5838681589530483,
      "tokens_seen": 1324220416
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006048746238716149,
      "loss": 1.6435,
      "theoretical_loss": 3.583852272031465,
      "tokens_seen": 1324285952
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006048545636910732,
      "loss": 1.4562,
      "theoretical_loss": 3.583836386116197,
      "tokens_seen": 1324351488
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006048345035105316,
      "loss": 1.4503,
      "theoretical_loss": 3.583820501207131,
      "tokens_seen": 1324417024
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00060481444332999,
      "loss": 1.4257,
      "theoretical_loss": 3.5838046173041533,
      "tokens_seen": 1324482560
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006047943831494483,
      "loss": 1.6003,
      "theoretical_loss": 3.5837887344071504,
      "tokens_seen": 1324548096
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006047743229689068,
      "loss": 1.4616,
      "theoretical_loss": 3.5837728525160086,
      "tokens_seen": 1324613632
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006047542627883651,
      "loss": 1.5543,
      "theoretical_loss": 3.5837569716306152,
      "tokens_seen": 1324679168
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006047342026078235,
      "loss": 1.5968,
      "theoretical_loss": 3.5837410917508556,
      "tokens_seen": 1324744704
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006047141424272819,
      "loss": 1.5427,
      "theoretical_loss": 3.583725212876618,
      "tokens_seen": 1324810240
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006046940822467403,
      "loss": 1.5926,
      "theoretical_loss": 3.5837093350077875,
      "tokens_seen": 1324875776
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006046740220661986,
      "loss": 1.4736,
      "theoretical_loss": 3.583693458144251,
      "tokens_seen": 1324941312
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000604653961885657,
      "loss": 1.4712,
      "theoretical_loss": 3.5836775822858957,
      "tokens_seen": 1325006848
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006046339017051153,
      "loss": 1.4551,
      "theoretical_loss": 3.5836617074326083,
      "tokens_seen": 1325072384
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006046138415245738,
      "loss": 1.4462,
      "theoretical_loss": 3.5836458335842747,
      "tokens_seen": 1325137920
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006045937813440321,
      "loss": 1.4899,
      "theoretical_loss": 3.583629960740782,
      "tokens_seen": 1325203456
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006045737211634905,
      "loss": 1.5021,
      "theoretical_loss": 3.583614088902017,
      "tokens_seen": 1325268992
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006045536609829488,
      "loss": 1.466,
      "theoretical_loss": 3.5835982180678663,
      "tokens_seen": 1325334528
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006045336008024073,
      "loss": 1.534,
      "theoretical_loss": 3.5835823482382163,
      "tokens_seen": 1325400064
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006045135406218656,
      "loss": 1.4165,
      "theoretical_loss": 3.583566479412955,
      "tokens_seen": 1325465600
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000604493480441324,
      "loss": 1.485,
      "theoretical_loss": 3.583550611591967,
      "tokens_seen": 1325531136
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006044734202607823,
      "loss": 1.3634,
      "theoretical_loss": 3.5835347447751413,
      "tokens_seen": 1325596672
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006044533600802408,
      "loss": 1.5891,
      "theoretical_loss": 3.583518878962364,
      "tokens_seen": 1325662208
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006044332998996991,
      "loss": 1.4226,
      "theoretical_loss": 3.5835030141535213,
      "tokens_seen": 1325727744
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006044132397191575,
      "loss": 1.5628,
      "theoretical_loss": 3.5834871503485,
      "tokens_seen": 1325793280
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006043931795386158,
      "loss": 1.4804,
      "theoretical_loss": 3.583471287547188,
      "tokens_seen": 1325858816
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006043731193580743,
      "loss": 1.3714,
      "theoretical_loss": 3.583455425749472,
      "tokens_seen": 1325924352
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006043530591775326,
      "loss": 1.4438,
      "theoretical_loss": 3.583439564955238,
      "tokens_seen": 1325989888
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000604332998996991,
      "loss": 1.5461,
      "theoretical_loss": 3.583423705164374,
      "tokens_seen": 1326055424
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006043129388164494,
      "loss": 1.4563,
      "theoretical_loss": 3.5834078463767653,
      "tokens_seen": 1326120960
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006042928786359077,
      "loss": 1.5303,
      "theoretical_loss": 3.583391988592301,
      "tokens_seen": 1326186496
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006042728184553661,
      "loss": 1.5311,
      "theoretical_loss": 3.5833761318108666,
      "tokens_seen": 1326252032
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006042527582748246,
      "loss": 1.5713,
      "theoretical_loss": 3.5833602760323497,
      "tokens_seen": 1326317568
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006042326980942828,
      "loss": 1.5531,
      "theoretical_loss": 3.5833444212566374,
      "tokens_seen": 1326383104
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006042126379137413,
      "loss": 1.504,
      "theoretical_loss": 3.5833285674836164,
      "tokens_seen": 1326448640
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006041925777331995,
      "loss": 1.5318,
      "theoretical_loss": 3.5833127147131734,
      "tokens_seen": 1326514176
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000604172517552658,
      "loss": 1.5225,
      "theoretical_loss": 3.583296862945196,
      "tokens_seen": 1326579712
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006041524573721164,
      "loss": 1.4981,
      "theoretical_loss": 3.583281012179572,
      "tokens_seen": 1326645248
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006041323971915747,
      "loss": 1.4187,
      "theoretical_loss": 3.583265162416187,
      "tokens_seen": 1326710784
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006041123370110331,
      "loss": 1.4246,
      "theoretical_loss": 3.583249313654929,
      "tokens_seen": 1326776320
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006040922768304916,
      "loss": 1.5064,
      "theoretical_loss": 3.583233465895685,
      "tokens_seen": 1326841856
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006040722166499498,
      "loss": 1.5019,
      "theoretical_loss": 3.5832176191383422,
      "tokens_seen": 1326907392
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006040521564694083,
      "loss": 1.5246,
      "theoretical_loss": 3.583201773382788,
      "tokens_seen": 1326972928
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006040320962888665,
      "loss": 1.4714,
      "theoretical_loss": 3.5831859286289083,
      "tokens_seen": 1327038464
    },
    {
      "epoch": 0.4,
      "objective/train/advantage_avg": 0.49329522252082825,
      "objective/train/docs_used": 485470,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4964334964752197,
      "objective/train/original_loss": 1.4964334964752197,
      "objective/train/theoretical_loss": 3.5831700848765924,
      "objective/train/tokens_used": 1347564000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24392816424369812,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505696535110474,
      "objective/train/weighted_lm_loss": 1.571977138519287,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5831700848765924,
      "tokens_seen": 1327104000
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000604012036108325,
      "loss": 1.6036,
      "theoretical_loss": 3.5831700848765924,
      "tokens_seen": 1327104000
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006039919759277834,
      "loss": 1.3856,
      "theoretical_loss": 3.583154242125726,
      "tokens_seen": 1327169536
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006039719157472417,
      "loss": 1.5519,
      "theoretical_loss": 3.583138400376197,
      "tokens_seen": 1327235072
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006039518555667001,
      "loss": 1.439,
      "theoretical_loss": 3.5831225596278924,
      "tokens_seen": 1327300608
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006039317953861585,
      "loss": 1.6162,
      "theoretical_loss": 3.5831067198807,
      "tokens_seen": 1327366144
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006039117352056168,
      "loss": 1.3857,
      "theoretical_loss": 3.583090881134506,
      "tokens_seen": 1327431680
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006038916750250753,
      "loss": 1.4099,
      "theoretical_loss": 3.583075043389199,
      "tokens_seen": 1327497216
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006038716148445335,
      "loss": 1.4972,
      "theoretical_loss": 3.5830592066446654,
      "tokens_seen": 1327562752
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000603851554663992,
      "loss": 1.5437,
      "theoretical_loss": 3.5830433709007927,
      "tokens_seen": 1327628288
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006038314944834503,
      "loss": 1.572,
      "theoretical_loss": 3.583027536157469,
      "tokens_seen": 1327693824
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006038114343029087,
      "loss": 1.4048,
      "theoretical_loss": 3.583011702414581,
      "tokens_seen": 1327759360
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006037913741223671,
      "loss": 1.4568,
      "theoretical_loss": 3.5829958696720166,
      "tokens_seen": 1327824896
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006037713139418255,
      "loss": 1.5162,
      "theoretical_loss": 3.582980037929663,
      "tokens_seen": 1327890432
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006037512537612838,
      "loss": 1.5354,
      "theoretical_loss": 3.5829642071874073,
      "tokens_seen": 1327955968
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006037311935807423,
      "loss": 1.5362,
      "theoretical_loss": 3.5829483774451374,
      "tokens_seen": 1328021504
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006037111334002005,
      "loss": 1.4935,
      "theoretical_loss": 3.5829325487027406,
      "tokens_seen": 1328087040
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000603691073219659,
      "loss": 1.441,
      "theoretical_loss": 3.582916720960105,
      "tokens_seen": 1328152576
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006036710130391173,
      "loss": 1.5241,
      "theoretical_loss": 3.5829008942171168,
      "tokens_seen": 1328218112
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006036509528585757,
      "loss": 1.4091,
      "theoretical_loss": 3.582885068473665,
      "tokens_seen": 1328283648
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006036308926780341,
      "loss": 1.4462,
      "theoretical_loss": 3.582869243729636,
      "tokens_seen": 1328349184
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006036108324974925,
      "loss": 1.5689,
      "theoretical_loss": 3.582853419984918,
      "tokens_seen": 1328414720
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006035907723169509,
      "loss": 1.4661,
      "theoretical_loss": 3.582837597239399,
      "tokens_seen": 1328480256
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006035707121364092,
      "loss": 1.4671,
      "theoretical_loss": 3.582821775492966,
      "tokens_seen": 1328545792
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006035506519558677,
      "loss": 1.4106,
      "theoretical_loss": 3.5828059547455062,
      "tokens_seen": 1328611328
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000603530591775326,
      "loss": 1.4326,
      "theoretical_loss": 3.5827901349969085,
      "tokens_seen": 1328676864
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006035105315947844,
      "loss": 1.4465,
      "theoretical_loss": 3.5827743162470593,
      "tokens_seen": 1328742400
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006034904714142427,
      "loss": 1.5594,
      "theoretical_loss": 3.5827584984958474,
      "tokens_seen": 1328807936
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006034704112337012,
      "loss": 1.4904,
      "theoretical_loss": 3.58274268174316,
      "tokens_seen": 1328873472
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006034503510531595,
      "loss": 1.4934,
      "theoretical_loss": 3.5827268659888842,
      "tokens_seen": 1328939008
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006034302908726179,
      "loss": 1.4721,
      "theoretical_loss": 3.582711051232909,
      "tokens_seen": 1329004544
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006034102306920762,
      "loss": 1.5814,
      "theoretical_loss": 3.582695237475121,
      "tokens_seen": 1329070080
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006033901705115347,
      "loss": 1.445,
      "theoretical_loss": 3.5826794247154092,
      "tokens_seen": 1329135616
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000603370110330993,
      "loss": 1.4378,
      "theoretical_loss": 3.58266361295366,
      "tokens_seen": 1329201152
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006033500501504514,
      "loss": 1.5874,
      "theoretical_loss": 3.5826478021897623,
      "tokens_seen": 1329266688
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006033299899699098,
      "loss": 1.3773,
      "theoretical_loss": 3.582631992423603,
      "tokens_seen": 1329332224
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006033099297893681,
      "loss": 1.4714,
      "theoretical_loss": 3.5826161836550714,
      "tokens_seen": 1329397760
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006032898696088265,
      "loss": 1.3587,
      "theoretical_loss": 3.582600375884054,
      "tokens_seen": 1329463296
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000603269809428285,
      "loss": 1.5441,
      "theoretical_loss": 3.582584569110439,
      "tokens_seen": 1329528832
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006032497492477432,
      "loss": 1.4571,
      "theoretical_loss": 3.582568763334115,
      "tokens_seen": 1329594368
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006032296890672017,
      "loss": 1.3912,
      "theoretical_loss": 3.582552958554969,
      "tokens_seen": 1329659904
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006032096288866599,
      "loss": 1.3791,
      "theoretical_loss": 3.582537154772889,
      "tokens_seen": 1329725440
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006031895687061184,
      "loss": 1.3652,
      "theoretical_loss": 3.582521351987764,
      "tokens_seen": 1329790976
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006031695085255768,
      "loss": 1.4899,
      "theoretical_loss": 3.582505550199481,
      "tokens_seen": 1329856512
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006031494483450351,
      "loss": 1.436,
      "theoretical_loss": 3.5824897494079284,
      "tokens_seen": 1329922048
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006031293881644935,
      "loss": 1.5627,
      "theoretical_loss": 3.582473949612994,
      "tokens_seen": 1329987584
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000603109327983952,
      "loss": 1.3945,
      "theoretical_loss": 3.582458150814566,
      "tokens_seen": 1330053120
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006030892678034102,
      "loss": 1.4201,
      "theoretical_loss": 3.5824423530125324,
      "tokens_seen": 1330118656
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006030692076228687,
      "loss": 1.5143,
      "theoretical_loss": 3.582426556206781,
      "tokens_seen": 1330184192
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006030491474423269,
      "loss": 1.4171,
      "theoretical_loss": 3.5824107603972006,
      "tokens_seen": 1330249728
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006030290872617854,
      "loss": 1.5291,
      "theoretical_loss": 3.5823949655836786,
      "tokens_seen": 1330315264
    },
    {
      "epoch": 0.4,
      "objective/train/advantage_avg": 0.49390482902526855,
      "objective/train/docs_used": 486514,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.412717580795288,
      "objective/train/original_loss": 1.4127174615859985,
      "objective/train/theoretical_loss": 3.582379171766104,
      "objective/train/tokens_used": 1350840800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2441677749156952,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506318807601929,
      "objective/train/weighted_lm_loss": 1.4841334819793701,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.582379171766104,
      "tokens_seen": 1330380800
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006030090270812438,
      "loss": 1.4264,
      "theoretical_loss": 3.582379171766104,
      "tokens_seen": 1330380800
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006029889669007021,
      "loss": 1.5771,
      "theoretical_loss": 3.5823633789443634,
      "tokens_seen": 1330446336
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006029689067201605,
      "loss": 1.5059,
      "theoretical_loss": 3.5823475871183463,
      "tokens_seen": 1330511872
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006029488465396189,
      "loss": 1.4353,
      "theoretical_loss": 3.5823317962879404,
      "tokens_seen": 1330577408
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006029287863590772,
      "loss": 1.4294,
      "theoretical_loss": 3.582316006453034,
      "tokens_seen": 1330642944
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006029087261785357,
      "loss": 1.524,
      "theoretical_loss": 3.5823002176135157,
      "tokens_seen": 1330708480
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006028886659979939,
      "loss": 1.5186,
      "theoretical_loss": 3.582284429769273,
      "tokens_seen": 1330774016
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006028686058174524,
      "loss": 1.4521,
      "theoretical_loss": 3.5822686429201944,
      "tokens_seen": 1330839552
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006028485456369107,
      "loss": 1.4354,
      "theoretical_loss": 3.5822528570661683,
      "tokens_seen": 1330905088
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006028284854563691,
      "loss": 1.5591,
      "theoretical_loss": 3.582237072207083,
      "tokens_seen": 1330970624
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006028084252758275,
      "loss": 1.5381,
      "theoretical_loss": 3.5822212883428266,
      "tokens_seen": 1331036160
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006027883650952859,
      "loss": 1.431,
      "theoretical_loss": 3.5822055054732873,
      "tokens_seen": 1331101696
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006027683049147442,
      "loss": 1.4883,
      "theoretical_loss": 3.582189723598354,
      "tokens_seen": 1331167232
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006027482447342027,
      "loss": 1.4973,
      "theoretical_loss": 3.5821739427179153,
      "tokens_seen": 1331232768
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000602728184553661,
      "loss": 1.4649,
      "theoretical_loss": 3.5821581628318584,
      "tokens_seen": 1331298304
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006027081243731194,
      "loss": 1.4336,
      "theoretical_loss": 3.5821423839400723,
      "tokens_seen": 1331363840
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006026880641925777,
      "loss": 1.5234,
      "theoretical_loss": 3.582126606042446,
      "tokens_seen": 1331429376
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006026680040120361,
      "loss": 1.5616,
      "theoretical_loss": 3.582110829138867,
      "tokens_seen": 1331494912
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006026479438314945,
      "loss": 1.482,
      "theoretical_loss": 3.582095053229224,
      "tokens_seen": 1331560448
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006026278836509529,
      "loss": 1.4318,
      "theoretical_loss": 3.5820792783134054,
      "tokens_seen": 1331625984
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006026078234704112,
      "loss": 1.4318,
      "theoretical_loss": 3.5820635043913005,
      "tokens_seen": 1331691520
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006025877632898696,
      "loss": 1.5181,
      "theoretical_loss": 3.582047731462797,
      "tokens_seen": 1331757056
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000602567703109328,
      "loss": 1.5557,
      "theoretical_loss": 3.582031959527783,
      "tokens_seen": 1331822592
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006025476429287864,
      "loss": 1.5203,
      "theoretical_loss": 3.582016188586149,
      "tokens_seen": 1331888128
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006025275827482447,
      "loss": 1.4129,
      "theoretical_loss": 3.582000418637781,
      "tokens_seen": 1331953664
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006025075225677031,
      "loss": 1.4111,
      "theoretical_loss": 3.5819846496825694,
      "tokens_seen": 1332019200
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006024874623871614,
      "loss": 1.3122,
      "theoretical_loss": 3.581968881720402,
      "tokens_seen": 1332084736
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006024674022066199,
      "loss": 1.4086,
      "theoretical_loss": 3.5819531147511676,
      "tokens_seen": 1332150272
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006024473420260782,
      "loss": 1.5127,
      "theoretical_loss": 3.581937348774755,
      "tokens_seen": 1332215808
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006024272818455366,
      "loss": 1.3593,
      "theoretical_loss": 3.5819215837910523,
      "tokens_seen": 1332281344
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000602407221664995,
      "loss": 1.6445,
      "theoretical_loss": 3.5819058197999487,
      "tokens_seen": 1332346880
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006023871614844534,
      "loss": 1.4623,
      "theoretical_loss": 3.581890056801333,
      "tokens_seen": 1332412416
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006023671013039117,
      "loss": 1.4763,
      "theoretical_loss": 3.5818742947950932,
      "tokens_seen": 1332477952
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006023470411233702,
      "loss": 1.554,
      "theoretical_loss": 3.5818585337811184,
      "tokens_seen": 1332543488
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006023269809428284,
      "loss": 1.4495,
      "theoretical_loss": 3.581842773759298,
      "tokens_seen": 1332609024
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006023069207622869,
      "loss": 1.4599,
      "theoretical_loss": 3.5818270147295195,
      "tokens_seen": 1332674560
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006022868605817452,
      "loss": 1.5127,
      "theoretical_loss": 3.5818112566916724,
      "tokens_seen": 1332740096
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006022668004012036,
      "loss": 1.4334,
      "theoretical_loss": 3.581795499645646,
      "tokens_seen": 1332805632
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000602246740220662,
      "loss": 1.3517,
      "theoretical_loss": 3.581779743591328,
      "tokens_seen": 1332871168
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006022266800401203,
      "loss": 1.5031,
      "theoretical_loss": 3.581763988528608,
      "tokens_seen": 1332936704
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006022066198595787,
      "loss": 1.5489,
      "theoretical_loss": 3.5817482344573746,
      "tokens_seen": 1333002240
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006021865596790372,
      "loss": 1.554,
      "theoretical_loss": 3.581732481377516,
      "tokens_seen": 1333067776
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006021664994984954,
      "loss": 1.4577,
      "theoretical_loss": 3.5817167292889223,
      "tokens_seen": 1333133312
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006021464393179539,
      "loss": 1.4219,
      "theoretical_loss": 3.581700978191482,
      "tokens_seen": 1333198848
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006021263791374121,
      "loss": 1.5554,
      "theoretical_loss": 3.5816852280850835,
      "tokens_seen": 1333264384
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006021063189568706,
      "loss": 1.3295,
      "theoretical_loss": 3.581669478969616,
      "tokens_seen": 1333329920
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006020862587763291,
      "loss": 1.5359,
      "theoretical_loss": 3.5816537308449687,
      "tokens_seen": 1333395456
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006020661985957873,
      "loss": 1.4653,
      "theoretical_loss": 3.5816379837110306,
      "tokens_seen": 1333460992
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006020461384152458,
      "loss": 1.472,
      "theoretical_loss": 3.5816222375676903,
      "tokens_seen": 1333526528
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006020260782347042,
      "loss": 1.4426,
      "theoretical_loss": 3.581606492414837,
      "tokens_seen": 1333592064
    },
    {
      "epoch": 0.4,
      "objective/train/advantage_avg": 0.4929308295249939,
      "objective/train/docs_used": 487665,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.362432599067688,
      "objective/train/original_loss": 1.362432599067688,
      "objective/train/theoretical_loss": 3.58159074825236,
      "objective/train/tokens_used": 1354117600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24362513422966003,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505316257476807,
      "objective/train/weighted_lm_loss": 1.4311251640319824,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.58159074825236,
      "tokens_seen": 1333657600
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006020060180541625,
      "loss": 1.4373,
      "theoretical_loss": 3.58159074825236,
      "tokens_seen": 1333657600
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006019859578736209,
      "loss": 1.5392,
      "theoretical_loss": 3.581575005080148,
      "tokens_seen": 1333723136
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006019658976930793,
      "loss": 1.5106,
      "theoretical_loss": 3.58155926289809,
      "tokens_seen": 1333788672
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006019458375125376,
      "loss": 1.4923,
      "theoretical_loss": 3.5815435217060747,
      "tokens_seen": 1333854208
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006019257773319961,
      "loss": 1.42,
      "theoretical_loss": 3.5815277815039925,
      "tokens_seen": 1333919744
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006019057171514543,
      "loss": 1.3708,
      "theoretical_loss": 3.5815120422917315,
      "tokens_seen": 1333985280
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006018856569709128,
      "loss": 1.3643,
      "theoretical_loss": 3.581496304069181,
      "tokens_seen": 1334050816
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006018655967903711,
      "loss": 1.4114,
      "theoretical_loss": 3.58148056683623,
      "tokens_seen": 1334116352
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006018455366098295,
      "loss": 1.5058,
      "theoretical_loss": 3.5814648305927683,
      "tokens_seen": 1334181888
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006018254764292879,
      "loss": 1.5807,
      "theoretical_loss": 3.581449095338684,
      "tokens_seen": 1334247424
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006018054162487463,
      "loss": 1.5539,
      "theoretical_loss": 3.5814333610738673,
      "tokens_seen": 1334312960
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006017853560682046,
      "loss": 1.5444,
      "theoretical_loss": 3.5814176277982073,
      "tokens_seen": 1334378496
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006017652958876631,
      "loss": 1.4302,
      "theoretical_loss": 3.581401895511593,
      "tokens_seen": 1334444032
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006017452357071213,
      "loss": 1.4281,
      "theoretical_loss": 3.5813861642139138,
      "tokens_seen": 1334509568
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006017251755265798,
      "loss": 1.5182,
      "theoretical_loss": 3.5813704339050583,
      "tokens_seen": 1334575104
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006017051153460381,
      "loss": 1.4771,
      "theoretical_loss": 3.581354704584917,
      "tokens_seen": 1334640640
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006016850551654965,
      "loss": 1.5351,
      "theoretical_loss": 3.5813389762533774,
      "tokens_seen": 1334706176
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006016649949849549,
      "loss": 1.5024,
      "theoretical_loss": 3.581323248910331,
      "tokens_seen": 1334771712
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006016449348044133,
      "loss": 1.3826,
      "theoretical_loss": 3.581307522555666,
      "tokens_seen": 1334837248
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006016248746238716,
      "loss": 1.376,
      "theoretical_loss": 3.5812917971892717,
      "tokens_seen": 1334902784
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00060160481444333,
      "loss": 1.4033,
      "theoretical_loss": 3.581276072811038,
      "tokens_seen": 1334968320
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006015847542627884,
      "loss": 1.4086,
      "theoretical_loss": 3.5812603494208535,
      "tokens_seen": 1335033856
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006015646940822468,
      "loss": 1.5161,
      "theoretical_loss": 3.5812446270186085,
      "tokens_seen": 1335099392
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006015446339017051,
      "loss": 1.4939,
      "theoretical_loss": 3.5812289056041915,
      "tokens_seen": 1335164928
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006015245737211635,
      "loss": 1.5703,
      "theoretical_loss": 3.581213185177493,
      "tokens_seen": 1335230464
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006015045135406218,
      "loss": 1.3679,
      "theoretical_loss": 3.581197465738401,
      "tokens_seen": 1335296000
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006014844533600803,
      "loss": 1.4009,
      "theoretical_loss": 3.5811817472868075,
      "tokens_seen": 1335361536
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006014643931795386,
      "loss": 1.4275,
      "theoretical_loss": 3.581166029822599,
      "tokens_seen": 1335427072
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000601444332998997,
      "loss": 1.4933,
      "theoretical_loss": 3.5811503133456672,
      "tokens_seen": 1335492608
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006014242728184554,
      "loss": 1.3478,
      "theoretical_loss": 3.5811345978559013,
      "tokens_seen": 1335558144
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006014042126379138,
      "loss": 1.5184,
      "theoretical_loss": 3.5811188833531897,
      "tokens_seen": 1335623680
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006013841524573721,
      "loss": 1.4056,
      "theoretical_loss": 3.581103169837423,
      "tokens_seen": 1335689216
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006013640922768306,
      "loss": 1.4153,
      "theoretical_loss": 3.5810874573084908,
      "tokens_seen": 1335754752
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006013440320962888,
      "loss": 1.3562,
      "theoretical_loss": 3.581071745766282,
      "tokens_seen": 1335820288
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006013239719157473,
      "loss": 1.4246,
      "theoretical_loss": 3.5810560352106866,
      "tokens_seen": 1335885824
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006013039117352056,
      "loss": 1.3975,
      "theoretical_loss": 3.5810403256415944,
      "tokens_seen": 1335951360
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000601283851554664,
      "loss": 1.3833,
      "theoretical_loss": 3.5810246170588957,
      "tokens_seen": 1336016896
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006012637913741224,
      "loss": 1.4699,
      "theoretical_loss": 3.581008909462479,
      "tokens_seen": 1336082432
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006012437311935807,
      "loss": 1.4018,
      "theoretical_loss": 3.580993202852234,
      "tokens_seen": 1336147968
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006012236710130391,
      "loss": 1.3494,
      "theoretical_loss": 3.5809774972280515,
      "tokens_seen": 1336213504
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006012036108324976,
      "loss": 1.3723,
      "theoretical_loss": 3.5809617925898207,
      "tokens_seen": 1336279040
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006011835506519558,
      "loss": 1.4946,
      "theoretical_loss": 3.5809460889374307,
      "tokens_seen": 1336344576
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006011634904714143,
      "loss": 1.3363,
      "theoretical_loss": 3.580930386270772,
      "tokens_seen": 1336410112
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0006011434302908725,
      "loss": 1.4267,
      "theoretical_loss": 3.580914684589735,
      "tokens_seen": 1336475648
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000601123370110331,
      "loss": 1.5366,
      "theoretical_loss": 3.5808989838942082,
      "tokens_seen": 1336541184
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006011033099297894,
      "loss": 1.4547,
      "theoretical_loss": 3.580883284184082,
      "tokens_seen": 1336606720
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006010832497492477,
      "loss": 1.4753,
      "theoretical_loss": 3.5808675854592464,
      "tokens_seen": 1336672256
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006010631895687061,
      "loss": 1.3822,
      "theoretical_loss": 3.580851887719591,
      "tokens_seen": 1336737792
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006010431293881646,
      "loss": 1.4616,
      "theoretical_loss": 3.580836190965006,
      "tokens_seen": 1336803328
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006010230692076228,
      "loss": 1.6567,
      "theoretical_loss": 3.580820495195381,
      "tokens_seen": 1336868864
    },
    {
      "epoch": 0.41,
      "objective/train/advantage_avg": 0.4916706383228302,
      "objective/train/docs_used": 488954,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2268729209899902,
      "objective/train/original_loss": 1.2268726825714111,
      "objective/train/theoretical_loss": 3.5808048004106054,
      "objective/train/tokens_used": 1357394400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24213974177837372,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503979921340942,
      "objective/train/weighted_lm_loss": 1.288705587387085,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5808048004106054,
      "tokens_seen": 1336934400
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006010030090270813,
      "loss": 1.4143,
      "theoretical_loss": 3.5808048004106054,
      "tokens_seen": 1336934400
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006009829488465395,
      "loss": 1.4499,
      "theoretical_loss": 3.5807891066105704,
      "tokens_seen": 1336999936
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000600962888665998,
      "loss": 1.4829,
      "theoretical_loss": 3.580773413795165,
      "tokens_seen": 1337065472
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006009428284854564,
      "loss": 1.4632,
      "theoretical_loss": 3.580757721964279,
      "tokens_seen": 1337131008
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006009227683049147,
      "loss": 1.4539,
      "theoretical_loss": 3.5807420311178033,
      "tokens_seen": 1337196544
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006009027081243731,
      "loss": 1.3859,
      "theoretical_loss": 3.5807263412556276,
      "tokens_seen": 1337262080
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006008826479438315,
      "loss": 1.3834,
      "theoretical_loss": 3.5807106523776415,
      "tokens_seen": 1337327616
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006008625877632898,
      "loss": 1.5577,
      "theoretical_loss": 3.580694964483735,
      "tokens_seen": 1337393152
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006008425275827483,
      "loss": 1.5276,
      "theoretical_loss": 3.580679277573799,
      "tokens_seen": 1337458688
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006008224674022066,
      "loss": 1.3819,
      "theoretical_loss": 3.5806635916477227,
      "tokens_seen": 1337524224
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000600802407221665,
      "loss": 1.5406,
      "theoretical_loss": 3.5806479067053965,
      "tokens_seen": 1337589760
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006007823470411234,
      "loss": 1.4928,
      "theoretical_loss": 3.5806322227467104,
      "tokens_seen": 1337655296
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006007622868605817,
      "loss": 1.4601,
      "theoretical_loss": 3.5806165397715546,
      "tokens_seen": 1337720832
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006007422266800401,
      "loss": 1.5153,
      "theoretical_loss": 3.58060085777982,
      "tokens_seen": 1337786368
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006007221664994985,
      "loss": 1.4392,
      "theoretical_loss": 3.5805851767713954,
      "tokens_seen": 1337851904
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006007021063189568,
      "loss": 1.4679,
      "theoretical_loss": 3.580569496746172,
      "tokens_seen": 1337917440
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006006820461384153,
      "loss": 1.4859,
      "theoretical_loss": 3.580553817704039,
      "tokens_seen": 1337982976
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006006619859578736,
      "loss": 1.3754,
      "theoretical_loss": 3.5805381396448874,
      "tokens_seen": 1338048512
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000600641925777332,
      "loss": 1.502,
      "theoretical_loss": 3.5805224625686076,
      "tokens_seen": 1338114048
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006006218655967903,
      "loss": 1.397,
      "theoretical_loss": 3.5805067864750892,
      "tokens_seen": 1338179584
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006006018054162488,
      "loss": 1.4616,
      "theoretical_loss": 3.580491111364223,
      "tokens_seen": 1338245120
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006005817452357072,
      "loss": 1.3706,
      "theoretical_loss": 3.580475437235899,
      "tokens_seen": 1338310656
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006005616850551655,
      "loss": 1.4427,
      "theoretical_loss": 3.5804597640900075,
      "tokens_seen": 1338376192
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000600541624874624,
      "loss": 1.4226,
      "theoretical_loss": 3.580444091926439,
      "tokens_seen": 1338441728
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006005215646940822,
      "loss": 1.4676,
      "theoretical_loss": 3.5804284207450836,
      "tokens_seen": 1338507264
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006005015045135407,
      "loss": 1.4658,
      "theoretical_loss": 3.5804127505458316,
      "tokens_seen": 1338572800
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000600481444332999,
      "loss": 1.3368,
      "theoretical_loss": 3.580397081328574,
      "tokens_seen": 1338638336
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006004613841524574,
      "loss": 1.3697,
      "theoretical_loss": 3.5803814130932006,
      "tokens_seen": 1338703872
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006004413239719158,
      "loss": 1.4419,
      "theoretical_loss": 3.580365745839602,
      "tokens_seen": 1338769408
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006004212637913742,
      "loss": 1.5096,
      "theoretical_loss": 3.5803500795676686,
      "tokens_seen": 1338834944
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006004012036108325,
      "loss": 1.4566,
      "theoretical_loss": 3.580334414277291,
      "tokens_seen": 1338900480
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000600381143430291,
      "loss": 1.4904,
      "theoretical_loss": 3.5803187499683586,
      "tokens_seen": 1338966016
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006003610832497492,
      "loss": 1.4479,
      "theoretical_loss": 3.5803030866407637,
      "tokens_seen": 1339031552
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006003410230692077,
      "loss": 1.4126,
      "theoretical_loss": 3.5802874242943954,
      "tokens_seen": 1339097088
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000600320962888666,
      "loss": 1.4424,
      "theoretical_loss": 3.580271762929145,
      "tokens_seen": 1339162624
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006003009027081244,
      "loss": 1.5049,
      "theoretical_loss": 3.5802561025449027,
      "tokens_seen": 1339228160
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006002808425275828,
      "loss": 1.4117,
      "theoretical_loss": 3.580240443141559,
      "tokens_seen": 1339293696
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006002607823470411,
      "loss": 1.485,
      "theoretical_loss": 3.580224784719004,
      "tokens_seen": 1339359232
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006002407221664995,
      "loss": 1.4858,
      "theoretical_loss": 3.5802091272771293,
      "tokens_seen": 1339424768
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000600220661985958,
      "loss": 1.49,
      "theoretical_loss": 3.580193470815825,
      "tokens_seen": 1339490304
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006002006018054162,
      "loss": 1.4305,
      "theoretical_loss": 3.5801778153349817,
      "tokens_seen": 1339555840
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006001805416248747,
      "loss": 1.5175,
      "theoretical_loss": 3.5801621608344902,
      "tokens_seen": 1339621376
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006001604814443329,
      "loss": 1.4185,
      "theoretical_loss": 3.5801465073142404,
      "tokens_seen": 1339686912
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006001404212637914,
      "loss": 1.5172,
      "theoretical_loss": 3.5801308547741244,
      "tokens_seen": 1339752448
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006001203610832498,
      "loss": 1.488,
      "theoretical_loss": 3.5801152032140315,
      "tokens_seen": 1339817984
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006001003009027081,
      "loss": 1.3128,
      "theoretical_loss": 3.580099552633853,
      "tokens_seen": 1339883520
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006000802407221665,
      "loss": 1.4138,
      "theoretical_loss": 3.58008390303348,
      "tokens_seen": 1339949056
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000600060180541625,
      "loss": 1.5519,
      "theoretical_loss": 3.5800682544128026,
      "tokens_seen": 1340014592
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006000401203610832,
      "loss": 1.4814,
      "theoretical_loss": 3.580052606771712,
      "tokens_seen": 1340080128
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006000200601805417,
      "loss": 1.3851,
      "theoretical_loss": 3.5800369601100988,
      "tokens_seen": 1340145664
    },
    {
      "epoch": 0.41,
      "objective/train/advantage_avg": 0.4960496723651886,
      "objective/train/docs_used": 490192,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3845856189727783,
      "objective/train/original_loss": 1.3845858573913574,
      "objective/train/theoretical_loss": 3.580021314427854,
      "objective/train/tokens_used": 1360671200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24621468782424927,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508568286895752,
      "objective/train/weighted_lm_loss": 1.4549685716629028,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.580021314427854,
      "tokens_seen": 1340211200
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0006,
      "loss": 1.4797,
      "theoretical_loss": 3.580021314427854,
      "tokens_seen": 1340211200
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005999799398194584,
      "loss": 1.5287,
      "theoretical_loss": 3.5800056697248674,
      "tokens_seen": 1340276736
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005999598796389168,
      "loss": 1.3435,
      "theoretical_loss": 3.579990026001031,
      "tokens_seen": 1340342272
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005999398194583751,
      "loss": 1.4418,
      "theoretical_loss": 3.5799743832562356,
      "tokens_seen": 1340407808
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005999197592778335,
      "loss": 1.4657,
      "theoretical_loss": 3.5799587414903717,
      "tokens_seen": 1340473344
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005998996990972919,
      "loss": 1.5375,
      "theoretical_loss": 3.57994310070333,
      "tokens_seen": 1340538880
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005998796389167502,
      "loss": 1.4678,
      "theoretical_loss": 3.579927460895002,
      "tokens_seen": 1340604416
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005998595787362087,
      "loss": 1.5166,
      "theoretical_loss": 3.5799118220652772,
      "tokens_seen": 1340669952
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000599839518555667,
      "loss": 1.5857,
      "theoretical_loss": 3.579896184214049,
      "tokens_seen": 1340735488
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005998194583751254,
      "loss": 1.5255,
      "theoretical_loss": 3.579880547341206,
      "tokens_seen": 1340801024
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005997993981945838,
      "loss": 1.4652,
      "theoretical_loss": 3.5798649114466405,
      "tokens_seen": 1340866560
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005997793380140421,
      "loss": 1.2171,
      "theoretical_loss": 3.5798492765302425,
      "tokens_seen": 1340932096
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005997592778335005,
      "loss": 1.5402,
      "theoretical_loss": 3.579833642591904,
      "tokens_seen": 1340997632
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005997392176529589,
      "loss": 1.5942,
      "theoretical_loss": 3.5798180096315155,
      "tokens_seen": 1341063168
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005997191574724172,
      "loss": 1.4368,
      "theoretical_loss": 3.579802377648969,
      "tokens_seen": 1341128704
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005996990972918757,
      "loss": 1.4678,
      "theoretical_loss": 3.5797867466441535,
      "tokens_seen": 1341194240
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000599679037111334,
      "loss": 1.4644,
      "theoretical_loss": 3.579771116616962,
      "tokens_seen": 1341259776
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005996589769307924,
      "loss": 1.4918,
      "theoretical_loss": 3.5797554875672843,
      "tokens_seen": 1341325312
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005996389167502507,
      "loss": 1.5556,
      "theoretical_loss": 3.579739859495013,
      "tokens_seen": 1341390848
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005996188565697092,
      "loss": 1.4555,
      "theoretical_loss": 3.5797242324000376,
      "tokens_seen": 1341456384
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005995987963891675,
      "loss": 1.376,
      "theoretical_loss": 3.57970860628225,
      "tokens_seen": 1341521920
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005995787362086259,
      "loss": 1.5515,
      "theoretical_loss": 3.579692981141541,
      "tokens_seen": 1341587456
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005995586760280842,
      "loss": 1.5534,
      "theoretical_loss": 3.5796773569778026,
      "tokens_seen": 1341652992
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005995386158475426,
      "loss": 1.4487,
      "theoretical_loss": 3.5796617337909256,
      "tokens_seen": 1341718528
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000599518555667001,
      "loss": 1.4359,
      "theoretical_loss": 3.579646111580801,
      "tokens_seen": 1341784064
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005994984954864594,
      "loss": 1.4772,
      "theoretical_loss": 3.57963049034732,
      "tokens_seen": 1341849600
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005994784353059177,
      "loss": 1.5254,
      "theoretical_loss": 3.579614870090374,
      "tokens_seen": 1341915136
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005994583751253762,
      "loss": 1.5666,
      "theoretical_loss": 3.579599250809854,
      "tokens_seen": 1341980672
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005994383149448345,
      "loss": 1.4731,
      "theoretical_loss": 3.579583632505652,
      "tokens_seen": 1342046208
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005994182547642929,
      "loss": 1.5753,
      "theoretical_loss": 3.5795680151776583,
      "tokens_seen": 1342111744
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005993981945837512,
      "loss": 1.5844,
      "theoretical_loss": 3.5795523988257654,
      "tokens_seen": 1342177280
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005993781344032096,
      "loss": 1.429,
      "theoretical_loss": 3.5795367834498637,
      "tokens_seen": 1342242816
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000599358074222668,
      "loss": 1.4428,
      "theoretical_loss": 3.579521169049845,
      "tokens_seen": 1342308352
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005993380140421264,
      "loss": 1.4983,
      "theoretical_loss": 3.5795055556256004,
      "tokens_seen": 1342373888
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005993179538615847,
      "loss": 1.5501,
      "theoretical_loss": 3.5794899431770215,
      "tokens_seen": 1342439424
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005992978936810432,
      "loss": 1.4892,
      "theoretical_loss": 3.579474331703999,
      "tokens_seen": 1342504960
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005992778335005014,
      "loss": 1.6235,
      "theoretical_loss": 3.579458721206426,
      "tokens_seen": 1342570496
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005992577733199599,
      "loss": 1.4235,
      "theoretical_loss": 3.579443111684192,
      "tokens_seen": 1342636032
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005992377131394183,
      "loss": 1.5073,
      "theoretical_loss": 3.5794275031371896,
      "tokens_seen": 1342701568
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005992176529588766,
      "loss": 1.4548,
      "theoretical_loss": 3.5794118955653103,
      "tokens_seen": 1342767104
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000599197592778335,
      "loss": 1.4435,
      "theoretical_loss": 3.5793962889684448,
      "tokens_seen": 1342832640
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005991775325977933,
      "loss": 1.4775,
      "theoretical_loss": 3.5793806833464856,
      "tokens_seen": 1342898176
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005991574724172517,
      "loss": 1.3256,
      "theoretical_loss": 3.579365078699323,
      "tokens_seen": 1342963712
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005991374122367102,
      "loss": 1.5708,
      "theoretical_loss": 3.57934947502685,
      "tokens_seen": 1343029248
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005991173520561684,
      "loss": 1.5005,
      "theoretical_loss": 3.579333872328957,
      "tokens_seen": 1343094784
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005990972918756269,
      "loss": 1.436,
      "theoretical_loss": 3.5793182706055364,
      "tokens_seen": 1343160320
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005990772316950854,
      "loss": 1.4178,
      "theoretical_loss": 3.579302669856479,
      "tokens_seen": 1343225856
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005990571715145436,
      "loss": 1.499,
      "theoretical_loss": 3.5792870700816772,
      "tokens_seen": 1343291392
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005990371113340021,
      "loss": 1.3743,
      "theoretical_loss": 3.5792714712810225,
      "tokens_seen": 1343356928
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005990170511534603,
      "loss": 1.423,
      "theoretical_loss": 3.5792558734544055,
      "tokens_seen": 1343422464
    },
    {
      "epoch": 0.41,
      "objective/train/advantage_avg": 0.4935036301612854,
      "objective/train/docs_used": 490931,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.530173897743225,
      "objective/train/original_loss": 1.5301740169525146,
      "objective/train/theoretical_loss": 3.5792402766017197,
      "objective/train/tokens_used": 1363948000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2439432293176651,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505903959274292,
      "objective/train/weighted_lm_loss": 1.6076054573059082,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5792402766017197,
      "tokens_seen": 1343488000
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005989969909729188,
      "loss": 1.4993,
      "theoretical_loss": 3.5792402766017197,
      "tokens_seen": 1343488000
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005989769307923772,
      "loss": 1.4103,
      "theoretical_loss": 3.579224680722855,
      "tokens_seen": 1343553536
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005989568706118355,
      "loss": 1.4476,
      "theoretical_loss": 3.579209085817704,
      "tokens_seen": 1343619072
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005989368104312939,
      "loss": 1.3929,
      "theoretical_loss": 3.5791934918861585,
      "tokens_seen": 1343684608
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005989167502507523,
      "loss": 1.4563,
      "theoretical_loss": 3.57917789892811,
      "tokens_seen": 1343750144
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005988966900702106,
      "loss": 1.4499,
      "theoretical_loss": 3.57916230694345,
      "tokens_seen": 1343815680
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005988766298896691,
      "loss": 1.3874,
      "theoretical_loss": 3.5791467159320707,
      "tokens_seen": 1343881216
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005988565697091274,
      "loss": 1.4721,
      "theoretical_loss": 3.5791311258938636,
      "tokens_seen": 1343946752
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005988365095285858,
      "loss": 1.3908,
      "theoretical_loss": 3.579115536828721,
      "tokens_seen": 1344012288
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005988164493480442,
      "loss": 1.4807,
      "theoretical_loss": 3.579099948736534,
      "tokens_seen": 1344077824
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005987963891675025,
      "loss": 1.441,
      "theoretical_loss": 3.579084361617195,
      "tokens_seen": 1344143360
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005987763289869609,
      "loss": 1.4247,
      "theoretical_loss": 3.5790687754705957,
      "tokens_seen": 1344208896
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005987562688064193,
      "loss": 1.4263,
      "theoretical_loss": 3.5790531902966274,
      "tokens_seen": 1344274432
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005987362086258776,
      "loss": 1.4757,
      "theoretical_loss": 3.579037606095183,
      "tokens_seen": 1344339968
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005987161484453361,
      "loss": 1.4457,
      "theoretical_loss": 3.5790220228661536,
      "tokens_seen": 1344405504
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005986960882647944,
      "loss": 1.4229,
      "theoretical_loss": 3.5790064406094317,
      "tokens_seen": 1344471040
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005986760280842528,
      "loss": 1.4122,
      "theoretical_loss": 3.578990859324909,
      "tokens_seen": 1344536576
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005986559679037111,
      "loss": 1.4469,
      "theoretical_loss": 3.5789752790124774,
      "tokens_seen": 1344602112
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005986359077231696,
      "loss": 1.4318,
      "theoretical_loss": 3.578959699672029,
      "tokens_seen": 1344667648
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005986158475426279,
      "loss": 1.4824,
      "theoretical_loss": 3.5789441213034556,
      "tokens_seen": 1344733184
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005985957873620863,
      "loss": 1.3768,
      "theoretical_loss": 3.5789285439066494,
      "tokens_seen": 1344798720
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005985757271815446,
      "loss": 1.5069,
      "theoretical_loss": 3.5789129674815023,
      "tokens_seen": 1344864256
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000598555667001003,
      "loss": 1.4452,
      "theoretical_loss": 3.5788973920279066,
      "tokens_seen": 1344929792
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005985356068204614,
      "loss": 1.4641,
      "theoretical_loss": 3.5788818175457537,
      "tokens_seen": 1344995328
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005985155466399198,
      "loss": 1.342,
      "theoretical_loss": 3.578866244034937,
      "tokens_seen": 1345060864
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005984954864593781,
      "loss": 1.3889,
      "theoretical_loss": 3.578850671495347,
      "tokens_seen": 1345126400
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005984754262788366,
      "loss": 1.5857,
      "theoretical_loss": 3.5788350999268768,
      "tokens_seen": 1345191936
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005984553660982949,
      "loss": 1.508,
      "theoretical_loss": 3.5788195293294183,
      "tokens_seen": 1345257472
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005984353059177533,
      "loss": 1.5518,
      "theoretical_loss": 3.5788039597028636,
      "tokens_seen": 1345323008
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005984152457372116,
      "loss": 1.486,
      "theoretical_loss": 3.578788391047105,
      "tokens_seen": 1345388544
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00059839518555667,
      "loss": 1.3261,
      "theoretical_loss": 3.5787728233620344,
      "tokens_seen": 1345454080
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005983751253761284,
      "loss": 1.444,
      "theoretical_loss": 3.5787572566475445,
      "tokens_seen": 1345519616
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005983550651955868,
      "loss": 1.3416,
      "theoretical_loss": 3.5787416909035272,
      "tokens_seen": 1345585152
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005983350050150451,
      "loss": 1.4871,
      "theoretical_loss": 3.578726126129875,
      "tokens_seen": 1345650688
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005983149448345036,
      "loss": 1.4787,
      "theoretical_loss": 3.578710562326479,
      "tokens_seen": 1345716224
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005982948846539618,
      "loss": 1.4785,
      "theoretical_loss": 3.5786949994932336,
      "tokens_seen": 1345781760
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005982748244734203,
      "loss": 1.382,
      "theoretical_loss": 3.578679437630029,
      "tokens_seen": 1345847296
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005982547642928787,
      "loss": 1.3936,
      "theoretical_loss": 3.5786638767367585,
      "tokens_seen": 1345912832
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000598234704112337,
      "loss": 1.5301,
      "theoretical_loss": 3.5786483168133145,
      "tokens_seen": 1345978368
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005982146439317954,
      "loss": 1.459,
      "theoretical_loss": 3.578632757859589,
      "tokens_seen": 1346043904
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005981945837512537,
      "loss": 1.5399,
      "theoretical_loss": 3.5786171998754748,
      "tokens_seen": 1346109440
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005981745235707121,
      "loss": 1.3484,
      "theoretical_loss": 3.5786016428608636,
      "tokens_seen": 1346174976
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005981544633901706,
      "loss": 1.5062,
      "theoretical_loss": 3.5785860868156485,
      "tokens_seen": 1346240512
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005981344032096288,
      "loss": 1.4242,
      "theoretical_loss": 3.578570531739721,
      "tokens_seen": 1346306048
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005981143430290873,
      "loss": 1.4566,
      "theoretical_loss": 3.5785549776329746,
      "tokens_seen": 1346371584
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005980942828485457,
      "loss": 1.5045,
      "theoretical_loss": 3.5785394244953013,
      "tokens_seen": 1346437120
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000598074222668004,
      "loss": 1.4406,
      "theoretical_loss": 3.578523872326593,
      "tokens_seen": 1346502656
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005980541624874624,
      "loss": 1.3648,
      "theoretical_loss": 3.578508321126743,
      "tokens_seen": 1346568192
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005980341023069207,
      "loss": 1.4239,
      "theoretical_loss": 3.578492770895643,
      "tokens_seen": 1346633728
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005980140421263791,
      "loss": 1.408,
      "theoretical_loss": 3.578477221633187,
      "tokens_seen": 1346699264
    },
    {
      "epoch": 0.41,
      "objective/train/advantage_avg": 0.49108660221099854,
      "objective/train/docs_used": 492127,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3430569171905518,
      "objective/train/original_loss": 1.3430569171905518,
      "objective/train/theoretical_loss": 3.5784616733392656,
      "objective/train/tokens_used": 1367224800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24174939095973969,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050337553024292,
      "objective/train/weighted_lm_loss": 1.410582184791565,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5784616733392656,
      "tokens_seen": 1346764800
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005979939819458376,
      "loss": 1.3805,
      "theoretical_loss": 3.5784616733392656,
      "tokens_seen": 1346764800
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005979739217652958,
      "loss": 1.4612,
      "theoretical_loss": 3.5784461260137728,
      "tokens_seen": 1346830336
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005979538615847543,
      "loss": 1.2875,
      "theoretical_loss": 3.5784305796566,
      "tokens_seen": 1346895872
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005979338014042126,
      "loss": 1.5017,
      "theoretical_loss": 3.578415034267641,
      "tokens_seen": 1346961408
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000597913741223671,
      "loss": 1.4663,
      "theoretical_loss": 3.5783994898467872,
      "tokens_seen": 1347026944
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005978936810431294,
      "loss": 1.3857,
      "theoretical_loss": 3.5783839463939326,
      "tokens_seen": 1347092480
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005978736208625878,
      "loss": 1.4824,
      "theoretical_loss": 3.5783684039089687,
      "tokens_seen": 1347158016
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005978535606820461,
      "loss": 1.3263,
      "theoretical_loss": 3.578352862391788,
      "tokens_seen": 1347223552
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005978335005015046,
      "loss": 1.4808,
      "theoretical_loss": 3.578337321842284,
      "tokens_seen": 1347289088
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005978134403209628,
      "loss": 1.5628,
      "theoretical_loss": 3.5783217822603492,
      "tokens_seen": 1347354624
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005977933801404213,
      "loss": 1.3108,
      "theoretical_loss": 3.578306243645876,
      "tokens_seen": 1347420160
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005977733199598796,
      "loss": 1.4393,
      "theoretical_loss": 3.578290705998757,
      "tokens_seen": 1347485696
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000597753259779338,
      "loss": 1.4289,
      "theoretical_loss": 3.5782751693188857,
      "tokens_seen": 1347551232
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005977331995987964,
      "loss": 1.3654,
      "theoretical_loss": 3.578259633606154,
      "tokens_seen": 1347616768
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005977131394182548,
      "loss": 1.413,
      "theoretical_loss": 3.5782440988604547,
      "tokens_seen": 1347682304
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005976930792377131,
      "loss": 1.4656,
      "theoretical_loss": 3.578228565081681,
      "tokens_seen": 1347747840
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005976730190571715,
      "loss": 1.4598,
      "theoretical_loss": 3.5782130322697263,
      "tokens_seen": 1347813376
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005976529588766298,
      "loss": 1.3707,
      "theoretical_loss": 3.578197500424482,
      "tokens_seen": 1347878912
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005976328986960883,
      "loss": 1.4801,
      "theoretical_loss": 3.5781819695458417,
      "tokens_seen": 1347944448
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005976128385155466,
      "loss": 1.4106,
      "theoretical_loss": 3.5781664396336987,
      "tokens_seen": 1348009984
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000597592778335005,
      "loss": 1.5447,
      "theoretical_loss": 3.578150910687945,
      "tokens_seen": 1348075520
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005975727181544634,
      "loss": 1.5338,
      "theoretical_loss": 3.578135382708474,
      "tokens_seen": 1348141056
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005975526579739218,
      "loss": 1.4156,
      "theoretical_loss": 3.578119855695178,
      "tokens_seen": 1348206592
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005975325977933802,
      "loss": 1.4334,
      "theoretical_loss": 3.578104329647951,
      "tokens_seen": 1348272128
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005975125376128385,
      "loss": 1.4252,
      "theoretical_loss": 3.578088804566685,
      "tokens_seen": 1348337664
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000597492477432297,
      "loss": 1.4229,
      "theoretical_loss": 3.578073280451273,
      "tokens_seen": 1348403200
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005974724172517553,
      "loss": 1.5054,
      "theoretical_loss": 3.5780577573016084,
      "tokens_seen": 1348468736
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005974523570712137,
      "loss": 1.4219,
      "theoretical_loss": 3.5780422351175845,
      "tokens_seen": 1348534272
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000597432296890672,
      "loss": 1.5377,
      "theoretical_loss": 3.5780267138990935,
      "tokens_seen": 1348599808
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005974122367101304,
      "loss": 1.5597,
      "theoretical_loss": 3.578011193646029,
      "tokens_seen": 1348665344
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005973921765295888,
      "loss": 1.4482,
      "theoretical_loss": 3.5779956743582835,
      "tokens_seen": 1348730880
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005973721163490472,
      "loss": 1.5378,
      "theoretical_loss": 3.5779801560357507,
      "tokens_seen": 1348796416
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005973520561685055,
      "loss": 1.3784,
      "theoretical_loss": 3.577964638678323,
      "tokens_seen": 1348861952
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000597331995987964,
      "loss": 1.4331,
      "theoretical_loss": 3.5779491222858937,
      "tokens_seen": 1348927488
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005973119358074222,
      "loss": 1.4641,
      "theoretical_loss": 3.5779336068583563,
      "tokens_seen": 1348993024
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005972918756268807,
      "loss": 1.5154,
      "theoretical_loss": 3.577918092395604,
      "tokens_seen": 1349058560
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005972718154463391,
      "loss": 1.4699,
      "theoretical_loss": 3.577902578897529,
      "tokens_seen": 1349124096
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005972517552657974,
      "loss": 1.4492,
      "theoretical_loss": 3.577887066364026,
      "tokens_seen": 1349189632
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005972316950852558,
      "loss": 1.4698,
      "theoretical_loss": 3.577871554794986,
      "tokens_seen": 1349255168
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005972116349047141,
      "loss": 1.3875,
      "theoretical_loss": 3.5778560441903045,
      "tokens_seen": 1349320704
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005971915747241725,
      "loss": 1.3684,
      "theoretical_loss": 3.577840534549873,
      "tokens_seen": 1349386240
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000597171514543631,
      "loss": 1.4018,
      "theoretical_loss": 3.5778250258735858,
      "tokens_seen": 1349451776
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005971514543630892,
      "loss": 1.4435,
      "theoretical_loss": 3.5778095181613354,
      "tokens_seen": 1349517312
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005971313941825477,
      "loss": 1.4949,
      "theoretical_loss": 3.577794011413016,
      "tokens_seen": 1349582848
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005971113340020061,
      "loss": 1.453,
      "theoretical_loss": 3.57777850562852,
      "tokens_seen": 1349648384
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005970912738214644,
      "loss": 1.4381,
      "theoretical_loss": 3.5777630008077406,
      "tokens_seen": 1349713920
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005970712136409228,
      "loss": 1.4871,
      "theoretical_loss": 3.577747496950572,
      "tokens_seen": 1349779456
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005970511534603811,
      "loss": 1.5578,
      "theoretical_loss": 3.577731994056906,
      "tokens_seen": 1349844992
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005970310932798395,
      "loss": 1.5049,
      "theoretical_loss": 3.577716492126638,
      "tokens_seen": 1349910528
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000597011033099298,
      "loss": 1.4758,
      "theoretical_loss": 3.5777009911596602,
      "tokens_seen": 1349976064
    },
    {
      "epoch": 0.41,
      "objective/train/advantage_avg": 0.49352627992630005,
      "objective/train/docs_used": 493399,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4361531734466553,
      "objective/train/original_loss": 1.4361531734466553,
      "objective/train/theoretical_loss": 3.577685491155866,
      "objective/train/tokens_used": 1370501600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24388453364372253,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505924224853516,
      "objective/train/weighted_lm_loss": 1.5086997747421265,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.577685491155866,
      "tokens_seen": 1350041600
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005969909729187562,
      "loss": 1.3927,
      "theoretical_loss": 3.577685491155866,
      "tokens_seen": 1350041600
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005969709127382147,
      "loss": 1.418,
      "theoretical_loss": 3.577669992115149,
      "tokens_seen": 1350107136
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000596950852557673,
      "loss": 1.4766,
      "theoretical_loss": 3.577654494037402,
      "tokens_seen": 1350172672
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005969307923771314,
      "loss": 1.5415,
      "theoretical_loss": 3.5776389969225195,
      "tokens_seen": 1350238208
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005969107321965898,
      "loss": 1.4503,
      "theoretical_loss": 3.577623500770394,
      "tokens_seen": 1350303744
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005968906720160482,
      "loss": 1.4419,
      "theoretical_loss": 3.5776080055809203,
      "tokens_seen": 1350369280
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005968706118355065,
      "loss": 1.4663,
      "theoretical_loss": 3.57759251135399,
      "tokens_seen": 1350434816
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000596850551654965,
      "loss": 1.4596,
      "theoretical_loss": 3.5775770180894977,
      "tokens_seen": 1350500352
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005968304914744232,
      "loss": 1.4467,
      "theoretical_loss": 3.577561525787337,
      "tokens_seen": 1350565888
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005968104312938817,
      "loss": 1.5112,
      "theoretical_loss": 3.5775460344474013,
      "tokens_seen": 1350631424
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00059679037111334,
      "loss": 1.3879,
      "theoretical_loss": 3.5775305440695835,
      "tokens_seen": 1350696960
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005967703109327984,
      "loss": 1.3093,
      "theoretical_loss": 3.5775150546537784,
      "tokens_seen": 1350762496
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005967502507522568,
      "loss": 1.5243,
      "theoretical_loss": 3.5774995661998785,
      "tokens_seen": 1350828032
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005967301905717152,
      "loss": 1.4807,
      "theoretical_loss": 3.577484078707778,
      "tokens_seen": 1350893568
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005967101303911735,
      "loss": 1.4472,
      "theoretical_loss": 3.5774685921773703,
      "tokens_seen": 1350959104
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005966900702106319,
      "loss": 1.4712,
      "theoretical_loss": 3.5774531066085493,
      "tokens_seen": 1351024640
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005966700100300902,
      "loss": 1.3904,
      "theoretical_loss": 3.5774376220012085,
      "tokens_seen": 1351090176
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005966499498495487,
      "loss": 1.4202,
      "theoretical_loss": 3.577422138355241,
      "tokens_seen": 1351155712
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000596629889669007,
      "loss": 1.5329,
      "theoretical_loss": 3.577406655670541,
      "tokens_seen": 1351221248
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005966098294884654,
      "loss": 1.5515,
      "theoretical_loss": 3.577391173947002,
      "tokens_seen": 1351286784
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005965897693079237,
      "loss": 1.5447,
      "theoretical_loss": 3.5773756931845186,
      "tokens_seen": 1351352320
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005965697091273822,
      "loss": 1.5996,
      "theoretical_loss": 3.577360213382984,
      "tokens_seen": 1351417856
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005965496489468405,
      "loss": 1.4016,
      "theoretical_loss": 3.577344734542291,
      "tokens_seen": 1351483392
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005965295887662989,
      "loss": 1.4668,
      "theoretical_loss": 3.5773292566623343,
      "tokens_seen": 1351548928
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005965095285857573,
      "loss": 1.3996,
      "theoretical_loss": 3.5773137797430077,
      "tokens_seen": 1351614464
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005964894684052157,
      "loss": 1.4793,
      "theoretical_loss": 3.5772983037842048,
      "tokens_seen": 1351680000
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000596469408224674,
      "loss": 1.4521,
      "theoretical_loss": 3.57728282878582,
      "tokens_seen": 1351745536
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005964493480441324,
      "loss": 1.5431,
      "theoretical_loss": 3.577267354747746,
      "tokens_seen": 1351811072
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005964292878635907,
      "loss": 1.4585,
      "theoretical_loss": 3.577251881669877,
      "tokens_seen": 1351876608
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005964092276830492,
      "loss": 1.499,
      "theoretical_loss": 3.577236409552108,
      "tokens_seen": 1351942144
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005963891675025075,
      "loss": 1.6232,
      "theoretical_loss": 3.5772209383943316,
      "tokens_seen": 1352007680
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005963691073219659,
      "loss": 1.5383,
      "theoretical_loss": 3.577205468196442,
      "tokens_seen": 1352073216
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005963490471414243,
      "loss": 1.5569,
      "theoretical_loss": 3.5771899989583336,
      "tokens_seen": 1352138752
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005963289869608826,
      "loss": 1.4347,
      "theoretical_loss": 3.5771745306798994,
      "tokens_seen": 1352204288
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000596308926780341,
      "loss": 1.4782,
      "theoretical_loss": 3.5771590633610346,
      "tokens_seen": 1352269824
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005962888665997995,
      "loss": 1.3826,
      "theoretical_loss": 3.5771435970016325,
      "tokens_seen": 1352335360
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005962688064192577,
      "loss": 1.4251,
      "theoretical_loss": 3.577128131601587,
      "tokens_seen": 1352400896
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005962487462387162,
      "loss": 1.4278,
      "theoretical_loss": 3.577112667160792,
      "tokens_seen": 1352466432
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005962286860581744,
      "loss": 1.4264,
      "theoretical_loss": 3.577097203679142,
      "tokens_seen": 1352531968
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005962086258776329,
      "loss": 1.3476,
      "theoretical_loss": 3.577081741156531,
      "tokens_seen": 1352597504
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005961885656970913,
      "loss": 1.3891,
      "theoretical_loss": 3.5770662795928527,
      "tokens_seen": 1352663040
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005961685055165496,
      "loss": 1.4386,
      "theoretical_loss": 3.577050818988001,
      "tokens_seen": 1352728576
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000596148445336008,
      "loss": 1.3663,
      "theoretical_loss": 3.5770353593418704,
      "tokens_seen": 1352794112
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005961283851554665,
      "loss": 1.3974,
      "theoretical_loss": 3.5770199006543555,
      "tokens_seen": 1352859648
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005961083249749247,
      "loss": 1.3923,
      "theoretical_loss": 3.5770044429253494,
      "tokens_seen": 1352925184
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005960882647943832,
      "loss": 1.4832,
      "theoretical_loss": 3.576988986154747,
      "tokens_seen": 1352990720
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005960682046138415,
      "loss": 1.4541,
      "theoretical_loss": 3.5769735303424417,
      "tokens_seen": 1353056256
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005960481444332999,
      "loss": 1.4685,
      "theoretical_loss": 3.576958075488328,
      "tokens_seen": 1353121792
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005960280842527584,
      "loss": 1.389,
      "theoretical_loss": 3.576942621592301,
      "tokens_seen": 1353187328
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005960080240722166,
      "loss": 1.4184,
      "theoretical_loss": 3.5769271686542536,
      "tokens_seen": 1353252864
    },
    {
      "epoch": 0.41,
      "objective/train/advantage_avg": 0.49008217453956604,
      "objective/train/docs_used": 494475,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3683398962020874,
      "objective/train/original_loss": 1.368340015411377,
      "objective/train/theoretical_loss": 3.5769117166740805,
      "objective/train/tokens_used": 1373778400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24086585640907288,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502325296401978,
      "objective/train/weighted_lm_loss": 1.4370107650756836,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5769117166740805,
      "tokens_seen": 1353318400
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005959879638916751,
      "loss": 1.4496,
      "theoretical_loss": 3.5769117166740805,
      "tokens_seen": 1353318400
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005959679037111334,
      "loss": 1.4528,
      "theoretical_loss": 3.5768962656516763,
      "tokens_seen": 1353383936
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005959478435305918,
      "loss": 1.5268,
      "theoretical_loss": 3.576880815586935,
      "tokens_seen": 1353449472
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005959277833500502,
      "loss": 1.4637,
      "theoretical_loss": 3.5768653664797503,
      "tokens_seen": 1353515008
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005959077231695086,
      "loss": 1.4706,
      "theoretical_loss": 3.5768499183300175,
      "tokens_seen": 1353580544
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005958876629889669,
      "loss": 1.4917,
      "theoretical_loss": 3.5768344711376305,
      "tokens_seen": 1353646080
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005958676028084254,
      "loss": 1.4553,
      "theoretical_loss": 3.576819024902483,
      "tokens_seen": 1353711616
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005958475426278836,
      "loss": 1.4631,
      "theoretical_loss": 3.5768035796244706,
      "tokens_seen": 1353777152
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005958274824473421,
      "loss": 1.4861,
      "theoretical_loss": 3.576788135303487,
      "tokens_seen": 1353842688
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005958074222668004,
      "loss": 1.6156,
      "theoretical_loss": 3.576772691939426,
      "tokens_seen": 1353908224
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005957873620862588,
      "loss": 1.5574,
      "theoretical_loss": 3.576757249532183,
      "tokens_seen": 1353973760
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005957673019057172,
      "loss": 1.4433,
      "theoretical_loss": 3.5767418080816515,
      "tokens_seen": 1354039296
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005957472417251756,
      "loss": 1.4818,
      "theoretical_loss": 3.576726367587727,
      "tokens_seen": 1354104832
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005957271815446339,
      "loss": 1.3348,
      "theoretical_loss": 3.5767109280503027,
      "tokens_seen": 1354170368
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005957071213640923,
      "loss": 1.4925,
      "theoretical_loss": 3.576695489469274,
      "tokens_seen": 1354235904
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005956870611835506,
      "loss": 1.5211,
      "theoretical_loss": 3.576680051844535,
      "tokens_seen": 1354301440
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005956670010030091,
      "loss": 1.4423,
      "theoretical_loss": 3.5766646151759804,
      "tokens_seen": 1354366976
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005956469408224674,
      "loss": 1.3748,
      "theoretical_loss": 3.5766491794635042,
      "tokens_seen": 1354432512
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005956268806419258,
      "loss": 1.5317,
      "theoretical_loss": 3.5766337447070016,
      "tokens_seen": 1354498048
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005956068204613841,
      "loss": 1.5492,
      "theoretical_loss": 3.576618310906367,
      "tokens_seen": 1354563584
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005955867602808426,
      "loss": 1.4417,
      "theoretical_loss": 3.5766028780614945,
      "tokens_seen": 1354629120
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005955667001003009,
      "loss": 1.4336,
      "theoretical_loss": 3.576587446172279,
      "tokens_seen": 1354694656
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005955466399197593,
      "loss": 1.4042,
      "theoretical_loss": 3.5765720152386153,
      "tokens_seen": 1354760192
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005955265797392177,
      "loss": 1.444,
      "theoretical_loss": 3.5765565852603975,
      "tokens_seen": 1354825728
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005955065195586761,
      "loss": 1.4895,
      "theoretical_loss": 3.5765411562375204,
      "tokens_seen": 1354891264
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005954864593781344,
      "loss": 1.4851,
      "theoretical_loss": 3.5765257281698792,
      "tokens_seen": 1354956800
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005954663991975928,
      "loss": 1.4046,
      "theoretical_loss": 3.5765103010573682,
      "tokens_seen": 1355022336
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005954463390170511,
      "loss": 1.4581,
      "theoretical_loss": 3.5764948748998817,
      "tokens_seen": 1355087872
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005954262788365096,
      "loss": 1.3961,
      "theoretical_loss": 3.576479449697315,
      "tokens_seen": 1355153408
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005954062186559679,
      "loss": 1.548,
      "theoretical_loss": 3.576464025449562,
      "tokens_seen": 1355218944
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005953861584754263,
      "loss": 1.457,
      "theoretical_loss": 3.576448602156518,
      "tokens_seen": 1355284480
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005953660982948847,
      "loss": 1.4829,
      "theoretical_loss": 3.576433179818078,
      "tokens_seen": 1355350016
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000595346038114343,
      "loss": 1.4633,
      "theoretical_loss": 3.576417758434136,
      "tokens_seen": 1355415552
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005953259779338014,
      "loss": 1.3767,
      "theoretical_loss": 3.5764023380045877,
      "tokens_seen": 1355481088
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005953059177532599,
      "loss": 1.3936,
      "theoretical_loss": 3.5763869185293276,
      "tokens_seen": 1355546624
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005952858575727181,
      "loss": 1.5517,
      "theoretical_loss": 3.57637150000825,
      "tokens_seen": 1355612160
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005952657973921766,
      "loss": 1.5058,
      "theoretical_loss": 3.57635608244125,
      "tokens_seen": 1355677696
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005952457372116348,
      "loss": 1.4533,
      "theoretical_loss": 3.5763406658282224,
      "tokens_seen": 1355743232
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005952256770310933,
      "loss": 1.4956,
      "theoretical_loss": 3.576325250169062,
      "tokens_seen": 1355808768
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005952056168505517,
      "loss": 1.5244,
      "theoretical_loss": 3.5763098354636647,
      "tokens_seen": 1355874304
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00059518555667001,
      "loss": 1.4119,
      "theoretical_loss": 3.576294421711924,
      "tokens_seen": 1355939840
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005951654964894684,
      "loss": 1.4765,
      "theoretical_loss": 3.576279008913735,
      "tokens_seen": 1356005376
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005951454363089269,
      "loss": 1.331,
      "theoretical_loss": 3.5762635970689933,
      "tokens_seen": 1356070912
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005951253761283851,
      "loss": 1.3257,
      "theoretical_loss": 3.576248186177594,
      "tokens_seen": 1356136448
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005951053159478436,
      "loss": 1.4523,
      "theoretical_loss": 3.576232776239431,
      "tokens_seen": 1356201984
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005950852557673018,
      "loss": 1.4843,
      "theoretical_loss": 3.5762173672544,
      "tokens_seen": 1356267520
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005950651955867603,
      "loss": 1.4098,
      "theoretical_loss": 3.576201959222396,
      "tokens_seen": 1356333056
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005950451354062187,
      "loss": 1.322,
      "theoretical_loss": 3.576186552143313,
      "tokens_seen": 1356398592
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000595025075225677,
      "loss": 1.4857,
      "theoretical_loss": 3.5761711460170478,
      "tokens_seen": 1356464128
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005950050150451354,
      "loss": 1.4651,
      "theoretical_loss": 3.576155740843495,
      "tokens_seen": 1356529664
    },
    {
      "epoch": 0.41,
      "objective/train/advantage_avg": 0.49025121331214905,
      "objective/train/docs_used": 495730,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5661019086837769,
      "objective/train/original_loss": 1.5661022663116455,
      "objective/train/theoretical_loss": 3.576140336622548,
      "objective/train/tokens_used": 1377055200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24227747321128845,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502567291259766,
      "objective/train/weighted_lm_loss": 1.645086407661438,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.576140336622548,
      "tokens_seen": 1356595200
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005949849548645938,
      "loss": 1.4875,
      "theoretical_loss": 3.576140336622548,
      "tokens_seen": 1356595200
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005949648946840521,
      "loss": 1.4854,
      "theoretical_loss": 3.5761249333541043,
      "tokens_seen": 1356660736
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005949448345035106,
      "loss": 1.5251,
      "theoretical_loss": 3.576109531038057,
      "tokens_seen": 1356726272
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005949247743229688,
      "loss": 1.5786,
      "theoretical_loss": 3.576094129674302,
      "tokens_seen": 1356791808
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005949047141424273,
      "loss": 1.4044,
      "theoretical_loss": 3.5760787292627345,
      "tokens_seen": 1356857344
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005948846539618856,
      "loss": 1.478,
      "theoretical_loss": 3.57606332980325,
      "tokens_seen": 1356922880
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000594864593781344,
      "loss": 1.5751,
      "theoretical_loss": 3.5760479312957427,
      "tokens_seen": 1356988416
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005948445336008024,
      "loss": 1.4761,
      "theoretical_loss": 3.5760325337401087,
      "tokens_seen": 1357053952
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005948244734202608,
      "loss": 1.4714,
      "theoretical_loss": 3.576017137136242,
      "tokens_seen": 1357119488
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005948044132397191,
      "loss": 1.4269,
      "theoretical_loss": 3.5760017414840393,
      "tokens_seen": 1357185024
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005947843530591776,
      "loss": 1.3914,
      "theoretical_loss": 3.575986346783395,
      "tokens_seen": 1357250560
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005947642928786359,
      "loss": 1.5089,
      "theoretical_loss": 3.5759709530342048,
      "tokens_seen": 1357316096
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005947442326980943,
      "loss": 1.5068,
      "theoretical_loss": 3.5759555602363635,
      "tokens_seen": 1357381632
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005947241725175526,
      "loss": 1.31,
      "theoretical_loss": 3.5759401683897662,
      "tokens_seen": 1357447168
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000594704112337011,
      "loss": 1.4808,
      "theoretical_loss": 3.575924777494309,
      "tokens_seen": 1357512704
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005946840521564694,
      "loss": 1.3067,
      "theoretical_loss": 3.575909387549886,
      "tokens_seen": 1357578240
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005946639919759278,
      "loss": 1.4866,
      "theoretical_loss": 3.5758939985563942,
      "tokens_seen": 1357643776
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005946439317953861,
      "loss": 1.5269,
      "theoretical_loss": 3.5758786105137275,
      "tokens_seen": 1357709312
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005946238716148445,
      "loss": 1.3744,
      "theoretical_loss": 3.5758632234217815,
      "tokens_seen": 1357774848
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005946038114343029,
      "loss": 1.4469,
      "theoretical_loss": 3.5758478372804525,
      "tokens_seen": 1357840384
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005945837512537613,
      "loss": 1.4992,
      "theoretical_loss": 3.5758324520896347,
      "tokens_seen": 1357905920
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005945636910732197,
      "loss": 1.4109,
      "theoretical_loss": 3.5758170678492243,
      "tokens_seen": 1357971456
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005945436308926781,
      "loss": 1.4354,
      "theoretical_loss": 3.5758016845591163,
      "tokens_seen": 1358036992
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005945235707121365,
      "loss": 1.569,
      "theoretical_loss": 3.5757863022192065,
      "tokens_seen": 1358102528
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005945035105315948,
      "loss": 1.4948,
      "theoretical_loss": 3.57577092082939,
      "tokens_seen": 1358168064
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005944834503510532,
      "loss": 1.5412,
      "theoretical_loss": 3.5757555403895624,
      "tokens_seen": 1358233600
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005944633901705115,
      "loss": 1.3712,
      "theoretical_loss": 3.5757401608996195,
      "tokens_seen": 1358299136
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00059444332998997,
      "loss": 1.4134,
      "theoretical_loss": 3.5757247823594565,
      "tokens_seen": 1358364672
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005944232698094283,
      "loss": 1.5239,
      "theoretical_loss": 3.5757094047689684,
      "tokens_seen": 1358430208
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005944032096288867,
      "loss": 1.4185,
      "theoretical_loss": 3.575694028128052,
      "tokens_seen": 1358495744
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005943831494483451,
      "loss": 1.4021,
      "theoretical_loss": 3.5756786524366015,
      "tokens_seen": 1358561280
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005943630892678034,
      "loss": 1.51,
      "theoretical_loss": 3.575663277694514,
      "tokens_seen": 1358626816
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005943430290872618,
      "loss": 1.3838,
      "theoretical_loss": 3.575647903901684,
      "tokens_seen": 1358692352
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005943229689067203,
      "loss": 1.4846,
      "theoretical_loss": 3.575632531058007,
      "tokens_seen": 1358757888
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005943029087261785,
      "loss": 1.5562,
      "theoretical_loss": 3.5756171591633796,
      "tokens_seen": 1358823424
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000594282848545637,
      "loss": 1.4042,
      "theoretical_loss": 3.5756017882176963,
      "tokens_seen": 1358888960
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005942627883650952,
      "loss": 1.4533,
      "theoretical_loss": 3.575586418220853,
      "tokens_seen": 1358954496
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005942427281845537,
      "loss": 1.4873,
      "theoretical_loss": 3.5755710491727464,
      "tokens_seen": 1359020032
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005942226680040121,
      "loss": 1.5013,
      "theoretical_loss": 3.575555681073271,
      "tokens_seen": 1359085568
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005942026078234704,
      "loss": 1.4593,
      "theoretical_loss": 3.575540313922323,
      "tokens_seen": 1359151104
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005941825476429288,
      "loss": 1.4576,
      "theoretical_loss": 3.5755249477197983,
      "tokens_seen": 1359216640
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005941624874623873,
      "loss": 1.4328,
      "theoretical_loss": 3.5755095824655925,
      "tokens_seen": 1359282176
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005941424272818455,
      "loss": 1.32,
      "theoretical_loss": 3.5754942181596006,
      "tokens_seen": 1359347712
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000594122367101304,
      "loss": 1.5008,
      "theoretical_loss": 3.5754788548017196,
      "tokens_seen": 1359413248
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005941023069207622,
      "loss": 1.4626,
      "theoretical_loss": 3.5754634923918447,
      "tokens_seen": 1359478784
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005940822467402207,
      "loss": 1.4658,
      "theoretical_loss": 3.575448130929872,
      "tokens_seen": 1359544320
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005940621865596791,
      "loss": 1.3161,
      "theoretical_loss": 3.5754327704156963,
      "tokens_seen": 1359609856
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005940421263791374,
      "loss": 1.5094,
      "theoretical_loss": 3.575417410849215,
      "tokens_seen": 1359675392
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005940220661985958,
      "loss": 1.428,
      "theoretical_loss": 3.5754020522303227,
      "tokens_seen": 1359740928
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005940020060180542,
      "loss": 1.4769,
      "theoretical_loss": 3.575386694558916,
      "tokens_seen": 1359806464
    },
    {
      "epoch": 0.41,
      "objective/train/advantage_avg": 0.49022120237350464,
      "objective/train/docs_used": 496861,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3403966426849365,
      "objective/train/original_loss": 1.3403964042663574,
      "objective/train/theoretical_loss": 3.5753713378348904,
      "objective/train/tokens_used": 1380332000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24074628949165344,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050246000289917,
      "objective/train/weighted_lm_loss": 1.4077306985855103,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.5753713378348904,
      "tokens_seen": 1359872000
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005939819458375125,
      "loss": 1.303,
      "theoretical_loss": 3.5753713378348904,
      "tokens_seen": 1359872000
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000593961885656971,
      "loss": 1.5318,
      "theoretical_loss": 3.5753559820581415,
      "tokens_seen": 1359937536
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005939418254764292,
      "loss": 1.4174,
      "theoretical_loss": 3.575340627228566,
      "tokens_seen": 1360003072
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005939217652958877,
      "loss": 1.5038,
      "theoretical_loss": 3.57532527334606,
      "tokens_seen": 1360068608
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000593901705115346,
      "loss": 1.475,
      "theoretical_loss": 3.5753099204105183,
      "tokens_seen": 1360134144
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005938816449348044,
      "loss": 1.3996,
      "theoretical_loss": 3.5752945684218376,
      "tokens_seen": 1360199680
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005938615847542628,
      "loss": 1.467,
      "theoretical_loss": 3.575279217379914,
      "tokens_seen": 1360265216
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005938415245737212,
      "loss": 1.504,
      "theoretical_loss": 3.5752638672846437,
      "tokens_seen": 1360330752
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005938214643931795,
      "loss": 1.6251,
      "theoretical_loss": 3.5752485181359215,
      "tokens_seen": 1360396288
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000593801404212638,
      "loss": 1.4138,
      "theoretical_loss": 3.575233169933645,
      "tokens_seen": 1360461824
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005937813440320963,
      "loss": 1.5338,
      "theoretical_loss": 3.575217822677709,
      "tokens_seen": 1360527360
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005937612838515547,
      "loss": 1.438,
      "theoretical_loss": 3.57520247636801,
      "tokens_seen": 1360592896
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000593741223671013,
      "loss": 1.4433,
      "theoretical_loss": 3.5751871310044443,
      "tokens_seen": 1360658432
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005937211634904714,
      "loss": 1.4611,
      "theoretical_loss": 3.575171786586908,
      "tokens_seen": 1360723968
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005937011033099298,
      "loss": 1.4169,
      "theoretical_loss": 3.575156443115297,
      "tokens_seen": 1360789504
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005936810431293882,
      "loss": 1.3839,
      "theoretical_loss": 3.5751411005895077,
      "tokens_seen": 1360855040
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005936609829488465,
      "loss": 1.4437,
      "theoretical_loss": 3.5751257590094365,
      "tokens_seen": 1360920576
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005936409227683049,
      "loss": 1.5586,
      "theoretical_loss": 3.5751104183749787,
      "tokens_seen": 1360986112
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005936208625877633,
      "loss": 1.3811,
      "theoretical_loss": 3.5750950786860307,
      "tokens_seen": 1361051648
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005936008024072217,
      "loss": 1.5,
      "theoretical_loss": 3.575079739942489,
      "tokens_seen": 1361117184
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00059358074222668,
      "loss": 1.5075,
      "theoretical_loss": 3.5750644021442506,
      "tokens_seen": 1361182720
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005935606820461385,
      "loss": 1.3875,
      "theoretical_loss": 3.5750490652912097,
      "tokens_seen": 1361248256
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005935406218655968,
      "loss": 1.4026,
      "theoretical_loss": 3.5750337293832644,
      "tokens_seen": 1361313792
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005935205616850552,
      "loss": 1.3472,
      "theoretical_loss": 3.57501839442031,
      "tokens_seen": 1361379328
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005935005015045135,
      "loss": 1.3868,
      "theoretical_loss": 3.5750030604022434,
      "tokens_seen": 1361444864
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005934804413239719,
      "loss": 1.5278,
      "theoretical_loss": 3.5749877273289608,
      "tokens_seen": 1361510400
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005934603811434303,
      "loss": 1.3424,
      "theoretical_loss": 3.5749723952003576,
      "tokens_seen": 1361575936
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005934403209628887,
      "loss": 1.6025,
      "theoretical_loss": 3.5749570640163313,
      "tokens_seen": 1361641472
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000593420260782347,
      "loss": 1.3646,
      "theoretical_loss": 3.574941733776778,
      "tokens_seen": 1361707008
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005934002006018055,
      "loss": 1.4982,
      "theoretical_loss": 3.5749264044815936,
      "tokens_seen": 1361772544
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005933801404212637,
      "loss": 1.6545,
      "theoretical_loss": 3.5749110761306744,
      "tokens_seen": 1361838080
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005933600802407222,
      "loss": 1.3717,
      "theoretical_loss": 3.5748957487239172,
      "tokens_seen": 1361903616
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005933400200601805,
      "loss": 1.4853,
      "theoretical_loss": 3.5748804222612187,
      "tokens_seen": 1361969152
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005933199598796389,
      "loss": 1.4439,
      "theoretical_loss": 3.574865096742475,
      "tokens_seen": 1362034688
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005932998996990973,
      "loss": 1.4196,
      "theoretical_loss": 3.5748497721675823,
      "tokens_seen": 1362100224
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005932798395185556,
      "loss": 1.4515,
      "theoretical_loss": 3.5748344485364374,
      "tokens_seen": 1362165760
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000593259779338014,
      "loss": 1.5917,
      "theoretical_loss": 3.5748191258489364,
      "tokens_seen": 1362231296
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005932397191574725,
      "loss": 1.413,
      "theoretical_loss": 3.574803804104976,
      "tokens_seen": 1362296832
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005932196589769307,
      "loss": 1.5059,
      "theoretical_loss": 3.574788483304453,
      "tokens_seen": 1362362368
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005931995987963892,
      "loss": 1.5401,
      "theoretical_loss": 3.574773163447264,
      "tokens_seen": 1362427904
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005931795386158476,
      "loss": 1.3626,
      "theoretical_loss": 3.5747578445333046,
      "tokens_seen": 1362493440
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005931594784353059,
      "loss": 1.4454,
      "theoretical_loss": 3.574742526562473,
      "tokens_seen": 1362558976
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005931394182547643,
      "loss": 1.4114,
      "theoretical_loss": 3.5747272095346636,
      "tokens_seen": 1362624512
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005931193580742226,
      "loss": 1.5757,
      "theoretical_loss": 3.5747118934497752,
      "tokens_seen": 1362690048
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000593099297893681,
      "loss": 1.5878,
      "theoretical_loss": 3.5746965783077025,
      "tokens_seen": 1362755584
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005930792377131395,
      "loss": 1.4001,
      "theoretical_loss": 3.5746812641083436,
      "tokens_seen": 1362821120
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005930591775325978,
      "loss": 1.3636,
      "theoretical_loss": 3.5746659508515943,
      "tokens_seen": 1362886656
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005930391173520562,
      "loss": 1.5594,
      "theoretical_loss": 3.574650638537352,
      "tokens_seen": 1362952192
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005930190571715146,
      "loss": 1.4172,
      "theoretical_loss": 3.5746353271655122,
      "tokens_seen": 1363017728
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005929989969909729,
      "loss": 1.4674,
      "theoretical_loss": 3.5746200167359725,
      "tokens_seen": 1363083264
    },
    {
      "epoch": 0.41,
      "objective/train/advantage_avg": 0.49429503083229065,
      "objective/train/docs_used": 498147,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3820762634277344,
      "objective/train/original_loss": 1.3820762634277344,
      "objective/train/theoretical_loss": 3.5746047072486293,
      "objective/train/tokens_used": 1383608800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2448446899652481,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506742000579834,
      "objective/train/weighted_lm_loss": 1.4521121978759766,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5746047072486293,
      "tokens_seen": 1363148800
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005929789368104314,
      "loss": 1.4349,
      "theoretical_loss": 3.5746047072486293,
      "tokens_seen": 1363148800
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005929588766298896,
      "loss": 1.3967,
      "theoretical_loss": 3.57458939870338,
      "tokens_seen": 1363214336
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005929388164493481,
      "loss": 1.3916,
      "theoretical_loss": 3.57457409110012,
      "tokens_seen": 1363279872
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005929187562688064,
      "loss": 1.4822,
      "theoretical_loss": 3.5745587844387474,
      "tokens_seen": 1363345408
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005928986960882648,
      "loss": 1.435,
      "theoretical_loss": 3.574543478719158,
      "tokens_seen": 1363410944
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005928786359077232,
      "loss": 1.4606,
      "theoretical_loss": 3.574528173941249,
      "tokens_seen": 1363476480
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005928585757271816,
      "loss": 1.3445,
      "theoretical_loss": 3.5745128701049174,
      "tokens_seen": 1363542016
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005928385155466399,
      "loss": 1.5608,
      "theoretical_loss": 3.5744975672100594,
      "tokens_seen": 1363607552
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005928184553660984,
      "loss": 1.3839,
      "theoretical_loss": 3.5744822652565724,
      "tokens_seen": 1363673088
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005927983951855567,
      "loss": 1.4741,
      "theoretical_loss": 3.5744669642443534,
      "tokens_seen": 1363738624
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005927783350050151,
      "loss": 1.5434,
      "theoretical_loss": 3.5744516641732984,
      "tokens_seen": 1363804160
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005927582748244734,
      "loss": 1.489,
      "theoretical_loss": 3.5744363650433053,
      "tokens_seen": 1363869696
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005927382146439318,
      "loss": 1.4983,
      "theoretical_loss": 3.5744210668542706,
      "tokens_seen": 1363935232
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005927181544633902,
      "loss": 1.5557,
      "theoretical_loss": 3.5744057696060905,
      "tokens_seen": 1364000768
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005926980942828486,
      "loss": 1.3946,
      "theoretical_loss": 3.5743904732986636,
      "tokens_seen": 1364066304
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005926780341023069,
      "loss": 1.5173,
      "theoretical_loss": 3.574375177931885,
      "tokens_seen": 1364131840
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005926579739217653,
      "loss": 1.492,
      "theoretical_loss": 3.574359883505653,
      "tokens_seen": 1364197376
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005926379137412237,
      "loss": 1.5282,
      "theoretical_loss": 3.574344590019864,
      "tokens_seen": 1364262912
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005926178535606821,
      "loss": 1.3975,
      "theoretical_loss": 3.574329297474415,
      "tokens_seen": 1364328448
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005925977933801404,
      "loss": 1.5998,
      "theoretical_loss": 3.574314005869203,
      "tokens_seen": 1364393984
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005925777331995989,
      "loss": 1.456,
      "theoretical_loss": 3.5742987152041255,
      "tokens_seen": 1364459520
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005925576730190572,
      "loss": 1.4965,
      "theoretical_loss": 3.574283425479079,
      "tokens_seen": 1364525056
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005925376128385156,
      "loss": 1.3858,
      "theoretical_loss": 3.574268136693961,
      "tokens_seen": 1364590592
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005925175526579739,
      "loss": 1.4283,
      "theoretical_loss": 3.5742528488486682,
      "tokens_seen": 1364656128
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005924974924774323,
      "loss": 1.3599,
      "theoretical_loss": 3.574237561943098,
      "tokens_seen": 1364721664
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005924774322968907,
      "loss": 1.4789,
      "theoretical_loss": 3.5742222759771467,
      "tokens_seen": 1364787200
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005924573721163491,
      "loss": 1.5298,
      "theoretical_loss": 3.574206990950713,
      "tokens_seen": 1364852736
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005924373119358074,
      "loss": 1.4084,
      "theoretical_loss": 3.574191706863693,
      "tokens_seen": 1364918272
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005924172517552659,
      "loss": 1.4508,
      "theoretical_loss": 3.5741764237159837,
      "tokens_seen": 1364983808
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005923971915747241,
      "loss": 1.4607,
      "theoretical_loss": 3.5741611415074823,
      "tokens_seen": 1365049344
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005923771313941826,
      "loss": 1.4103,
      "theoretical_loss": 3.5741458602380867,
      "tokens_seen": 1365114880
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000592357071213641,
      "loss": 1.6184,
      "theoretical_loss": 3.5741305799076932,
      "tokens_seen": 1365180416
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005923370110330993,
      "loss": 1.4715,
      "theoretical_loss": 3.5741153005162003,
      "tokens_seen": 1365245952
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005923169508525577,
      "loss": 1.5856,
      "theoretical_loss": 3.5741000220635035,
      "tokens_seen": 1365311488
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000592296890672016,
      "loss": 1.5107,
      "theoretical_loss": 3.574084744549502,
      "tokens_seen": 1365377024
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005922768304914744,
      "loss": 1.6099,
      "theoretical_loss": 3.574069467974091,
      "tokens_seen": 1365442560
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005922567703109329,
      "loss": 1.337,
      "theoretical_loss": 3.57405419233717,
      "tokens_seen": 1365508096
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005922367101303911,
      "loss": 1.4363,
      "theoretical_loss": 3.574038917638635,
      "tokens_seen": 1365573632
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005922166499498496,
      "loss": 1.449,
      "theoretical_loss": 3.574023643878383,
      "tokens_seen": 1365639168
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000592196589769308,
      "loss": 1.4581,
      "theoretical_loss": 3.574008371056312,
      "tokens_seen": 1365704704
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005921765295887663,
      "loss": 1.5259,
      "theoretical_loss": 3.5739930991723194,
      "tokens_seen": 1365770240
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005921564694082247,
      "loss": 1.5044,
      "theoretical_loss": 3.573977828226302,
      "tokens_seen": 1365835776
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000592136409227683,
      "loss": 1.4385,
      "theoretical_loss": 3.573962558218158,
      "tokens_seen": 1365901312
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005921163490471414,
      "loss": 1.59,
      "theoretical_loss": 3.5739472891477844,
      "tokens_seen": 1365966848
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005920962888665999,
      "loss": 1.5283,
      "theoretical_loss": 3.5739320210150787,
      "tokens_seen": 1366032384
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005920762286860581,
      "loss": 1.6417,
      "theoretical_loss": 3.5739167538199377,
      "tokens_seen": 1366097920
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005920561685055166,
      "loss": 1.3704,
      "theoretical_loss": 3.5739014875622597,
      "tokens_seen": 1366163456
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005920361083249749,
      "loss": 1.4178,
      "theoretical_loss": 3.573886222241942,
      "tokens_seen": 1366228992
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005920160481444333,
      "loss": 1.4467,
      "theoretical_loss": 3.5738709578588814,
      "tokens_seen": 1366294528
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005919959879638917,
      "loss": 1.4513,
      "theoretical_loss": 3.5738556944129765,
      "tokens_seen": 1366360064
    },
    {
      "epoch": 0.41,
      "objective/train/advantage_avg": 0.4917767643928528,
      "objective/train/docs_used": 499217,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3180235624313354,
      "objective/train/original_loss": 1.3180235624313354,
      "objective/train/theoretical_loss": 3.573840431904124,
      "objective/train/tokens_used": 1386885600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24224121868610382,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050409197807312,
      "objective/train/weighted_lm_loss": 1.3844918012619019,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.573840431904124,
      "tokens_seen": 1366425600
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00059197592778335,
      "loss": 1.3791,
      "theoretical_loss": 3.573840431904124,
      "tokens_seen": 1366425600
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005919558676028084,
      "loss": 1.4182,
      "theoretical_loss": 3.573825170332222,
      "tokens_seen": 1366491136
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005919358074222668,
      "loss": 1.408,
      "theoretical_loss": 3.573809909697167,
      "tokens_seen": 1366556672
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005919157472417251,
      "loss": 1.4885,
      "theoretical_loss": 3.5737946499988587,
      "tokens_seen": 1366622208
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005918956870611836,
      "loss": 1.3871,
      "theoretical_loss": 3.573779391237192,
      "tokens_seen": 1366687744
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005918756268806419,
      "loss": 1.43,
      "theoretical_loss": 3.5737641334120664,
      "tokens_seen": 1366753280
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005918555667001003,
      "loss": 1.4715,
      "theoretical_loss": 3.573748876523379,
      "tokens_seen": 1366818816
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005918355065195587,
      "loss": 1.4608,
      "theoretical_loss": 3.573733620571027,
      "tokens_seen": 1366884352
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005918154463390171,
      "loss": 1.4582,
      "theoretical_loss": 3.5737183655549085,
      "tokens_seen": 1366949888
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005917953861584754,
      "loss": 1.4739,
      "theoretical_loss": 3.573703111474921,
      "tokens_seen": 1367015424
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005917753259779338,
      "loss": 1.5163,
      "theoretical_loss": 3.5736878583309624,
      "tokens_seen": 1367080960
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005917552657973921,
      "loss": 1.5834,
      "theoretical_loss": 3.57367260612293,
      "tokens_seen": 1367146496
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005917352056168506,
      "loss": 1.3762,
      "theoretical_loss": 3.573657354850722,
      "tokens_seen": 1367212032
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005917151454363089,
      "loss": 1.4374,
      "theoretical_loss": 3.573642104514236,
      "tokens_seen": 1367277568
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005916950852557673,
      "loss": 1.4997,
      "theoretical_loss": 3.5736268551133694,
      "tokens_seen": 1367343104
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005916750250752256,
      "loss": 1.5234,
      "theoretical_loss": 3.5736116066480195,
      "tokens_seen": 1367408640
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005916549648946841,
      "loss": 1.4876,
      "theoretical_loss": 3.5735963591180857,
      "tokens_seen": 1367474176
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005916349047141424,
      "loss": 1.527,
      "theoretical_loss": 3.573581112523464,
      "tokens_seen": 1367539712
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005916148445336008,
      "loss": 1.432,
      "theoretical_loss": 3.5735658668640538,
      "tokens_seen": 1367605248
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005915947843530591,
      "loss": 1.462,
      "theoretical_loss": 3.573550622139752,
      "tokens_seen": 1367670784
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005915747241725176,
      "loss": 1.3839,
      "theoretical_loss": 3.5735353783504564,
      "tokens_seen": 1367736320
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000591554663991976,
      "loss": 1.4696,
      "theoretical_loss": 3.5735201354960653,
      "tokens_seen": 1367801856
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005915346038114343,
      "loss": 1.4385,
      "theoretical_loss": 3.573504893576476,
      "tokens_seen": 1367867392
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005915145436308927,
      "loss": 1.4485,
      "theoretical_loss": 3.5734896525915874,
      "tokens_seen": 1367932928
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005914944834503511,
      "loss": 1.3909,
      "theoretical_loss": 3.5734744125412963,
      "tokens_seen": 1367998464
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005914744232698095,
      "loss": 1.5118,
      "theoretical_loss": 3.573459173425501,
      "tokens_seen": 1368064000
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005914543630892678,
      "loss": 1.406,
      "theoretical_loss": 3.573443935244099,
      "tokens_seen": 1368129536
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005914343029087263,
      "loss": 1.4267,
      "theoretical_loss": 3.5734286979969894,
      "tokens_seen": 1368195072
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005914142427281845,
      "loss": 1.5628,
      "theoretical_loss": 3.573413461684069,
      "tokens_seen": 1368260608
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000591394182547643,
      "loss": 1.3324,
      "theoretical_loss": 3.573398226305237,
      "tokens_seen": 1368326144
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005913741223671013,
      "loss": 1.4745,
      "theoretical_loss": 3.5733829918603903,
      "tokens_seen": 1368391680
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005913540621865597,
      "loss": 1.5029,
      "theoretical_loss": 3.5733677583494274,
      "tokens_seen": 1368457216
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005913340020060181,
      "loss": 1.5013,
      "theoretical_loss": 3.573352525772246,
      "tokens_seen": 1368522752
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005913139418254764,
      "loss": 1.5252,
      "theoretical_loss": 3.5733372941287445,
      "tokens_seen": 1368588288
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005912938816449348,
      "loss": 1.4863,
      "theoretical_loss": 3.573322063418821,
      "tokens_seen": 1368653824
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005912738214643933,
      "loss": 1.5191,
      "theoretical_loss": 3.5733068336423734,
      "tokens_seen": 1368719360
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005912537612838515,
      "loss": 1.5513,
      "theoretical_loss": 3.5732916047992997,
      "tokens_seen": 1368784896
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00059123370110331,
      "loss": 1.3442,
      "theoretical_loss": 3.573276376889498,
      "tokens_seen": 1368850432
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005912136409227684,
      "loss": 1.4468,
      "theoretical_loss": 3.5732611499128666,
      "tokens_seen": 1368915968
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005911935807422267,
      "loss": 1.496,
      "theoretical_loss": 3.5732459238693037,
      "tokens_seen": 1368981504
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005911735205616851,
      "loss": 1.3689,
      "theoretical_loss": 3.5732306987587075,
      "tokens_seen": 1369047040
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005911534603811434,
      "loss": 1.4602,
      "theoretical_loss": 3.573215474580976,
      "tokens_seen": 1369112576
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005911334002006018,
      "loss": 1.4553,
      "theoretical_loss": 3.5732002513360075,
      "tokens_seen": 1369178112
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005911133400200603,
      "loss": 1.4239,
      "theoretical_loss": 3.5731850290237,
      "tokens_seen": 1369243648
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005910932798395185,
      "loss": 1.4518,
      "theoretical_loss": 3.5731698076439518,
      "tokens_seen": 1369309184
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.000591073219658977,
      "loss": 1.4207,
      "theoretical_loss": 3.5731545871966617,
      "tokens_seen": 1369374720
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0005910531594784353,
      "loss": 1.449,
      "theoretical_loss": 3.5731393676817267,
      "tokens_seen": 1369440256
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005910330992978937,
      "loss": 1.4034,
      "theoretical_loss": 3.573124149099046,
      "tokens_seen": 1369505792
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005910130391173521,
      "loss": 1.3881,
      "theoretical_loss": 3.5731089314485183,
      "tokens_seen": 1369571328
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005909929789368104,
      "loss": 1.5697,
      "theoretical_loss": 3.5730937147300406,
      "tokens_seen": 1369636864
    },
    {
      "epoch": 0.42,
      "objective/train/advantage_avg": 0.49026408791542053,
      "objective/train/docs_used": 500422,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4417644739151,
      "objective/train/original_loss": 1.4417643547058105,
      "objective/train/theoretical_loss": 3.573078498943513,
      "objective/train/tokens_used": 1390162400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24113242328166962,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502521991729736,
      "objective/train/weighted_lm_loss": 1.5143424272537231,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.573078498943513,
      "tokens_seen": 1369702400
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005909729187562688,
      "loss": 1.5118,
      "theoretical_loss": 3.573078498943513,
      "tokens_seen": 1369702400
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005909528585757272,
      "loss": 1.4999,
      "theoretical_loss": 3.573063284088832,
      "tokens_seen": 1369767936
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005909327983951855,
      "loss": 1.567,
      "theoretical_loss": 3.5730480701658967,
      "tokens_seen": 1369833472
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000590912738214644,
      "loss": 1.404,
      "theoretical_loss": 3.5730328571746055,
      "tokens_seen": 1369899008
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005908926780341023,
      "loss": 1.4311,
      "theoretical_loss": 3.5730176451148568,
      "tokens_seen": 1369964544
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005908726178535607,
      "loss": 1.5129,
      "theoretical_loss": 3.573002433986549,
      "tokens_seen": 1370030080
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005908525576730191,
      "loss": 1.4216,
      "theoretical_loss": 3.572987223789581,
      "tokens_seen": 1370095616
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005908324974924775,
      "loss": 1.4156,
      "theoretical_loss": 3.5729720145238506,
      "tokens_seen": 1370161152
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005908124373119358,
      "loss": 1.3926,
      "theoretical_loss": 3.572956806189256,
      "tokens_seen": 1370226688
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005907923771313942,
      "loss": 1.4616,
      "theoretical_loss": 3.5729415987856963,
      "tokens_seen": 1370292224
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005907723169508525,
      "loss": 1.5088,
      "theoretical_loss": 3.5729263923130694,
      "tokens_seen": 1370357760
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000590752256770311,
      "loss": 1.467,
      "theoretical_loss": 3.5729111867712744,
      "tokens_seen": 1370423296
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005907321965897693,
      "loss": 1.4141,
      "theoretical_loss": 3.5728959821602095,
      "tokens_seen": 1370488832
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005907121364092277,
      "loss": 1.383,
      "theoretical_loss": 3.572880778479773,
      "tokens_seen": 1370554368
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000590692076228686,
      "loss": 1.4503,
      "theoretical_loss": 3.572865575729864,
      "tokens_seen": 1370619904
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005906720160481445,
      "loss": 1.4815,
      "theoretical_loss": 3.5728503739103807,
      "tokens_seen": 1370685440
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005906519558676028,
      "loss": 1.4374,
      "theoretical_loss": 3.5728351730212218,
      "tokens_seen": 1370750976
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005906318956870612,
      "loss": 1.4574,
      "theoretical_loss": 3.572819973062286,
      "tokens_seen": 1370816512
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005906118355065195,
      "loss": 1.4718,
      "theoretical_loss": 3.5728047740334716,
      "tokens_seen": 1370882048
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000590591775325978,
      "loss": 1.514,
      "theoretical_loss": 3.572789575934677,
      "tokens_seen": 1370947584
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005905717151454363,
      "loss": 1.4045,
      "theoretical_loss": 3.5727743787658017,
      "tokens_seen": 1371013120
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005905516549648947,
      "loss": 1.5273,
      "theoretical_loss": 3.5727591825267435,
      "tokens_seen": 1371078656
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000590531594784353,
      "loss": 1.6749,
      "theoretical_loss": 3.572743987217402,
      "tokens_seen": 1371144192
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005905115346038115,
      "loss": 1.568,
      "theoretical_loss": 3.5727287928376747,
      "tokens_seen": 1371209728
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005904914744232698,
      "loss": 1.391,
      "theoretical_loss": 3.572713599387461,
      "tokens_seen": 1371275264
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005904714142427282,
      "loss": 1.4471,
      "theoretical_loss": 3.5726984068666594,
      "tokens_seen": 1371340800
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005904513540621866,
      "loss": 1.4461,
      "theoretical_loss": 3.572683215275169,
      "tokens_seen": 1371406336
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005904312938816449,
      "loss": 1.4876,
      "theoretical_loss": 3.5726680246128883,
      "tokens_seen": 1371471872
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005904112337011033,
      "loss": 1.3456,
      "theoretical_loss": 3.572652834879716,
      "tokens_seen": 1371537408
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005903911735205617,
      "loss": 1.4616,
      "theoretical_loss": 3.572637646075551,
      "tokens_seen": 1371602944
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00059037111334002,
      "loss": 1.3867,
      "theoretical_loss": 3.572622458200292,
      "tokens_seen": 1371668480
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005903510531594785,
      "loss": 1.5068,
      "theoretical_loss": 3.5726072712538386,
      "tokens_seen": 1371734016
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005903309929789367,
      "loss": 1.3447,
      "theoretical_loss": 3.5725920852360877,
      "tokens_seen": 1371799552
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005903109327983952,
      "loss": 1.465,
      "theoretical_loss": 3.57257690014694,
      "tokens_seen": 1371865088
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005902908726178536,
      "loss": 1.3482,
      "theoretical_loss": 3.5725617159862937,
      "tokens_seen": 1371930624
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005902708124373119,
      "loss": 1.458,
      "theoretical_loss": 3.572546532754047,
      "tokens_seen": 1371996160
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005902507522567703,
      "loss": 1.3837,
      "theoretical_loss": 3.5725313504501006,
      "tokens_seen": 1372061696
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005902306920762288,
      "loss": 1.4825,
      "theoretical_loss": 3.5725161690743517,
      "tokens_seen": 1372127232
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000590210631895687,
      "loss": 1.5645,
      "theoretical_loss": 3.5725009886267,
      "tokens_seen": 1372192768
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005901905717151455,
      "loss": 1.4919,
      "theoretical_loss": 3.572485809107044,
      "tokens_seen": 1372258304
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005901705115346037,
      "loss": 1.4319,
      "theoretical_loss": 3.5724706305152827,
      "tokens_seen": 1372323840
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005901504513540622,
      "loss": 1.3787,
      "theoretical_loss": 3.5724554528513157,
      "tokens_seen": 1372389376
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005901303911735206,
      "loss": 1.3144,
      "theoretical_loss": 3.5724402761150413,
      "tokens_seen": 1372454912
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005901103309929789,
      "loss": 1.4053,
      "theoretical_loss": 3.5724251003063587,
      "tokens_seen": 1372520448
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005900902708124373,
      "loss": 1.4056,
      "theoretical_loss": 3.572409925425167,
      "tokens_seen": 1372585984
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005900702106318957,
      "loss": 1.3832,
      "theoretical_loss": 3.5723947514713656,
      "tokens_seen": 1372651520
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005900501504513541,
      "loss": 1.5135,
      "theoretical_loss": 3.5723795784448527,
      "tokens_seen": 1372717056
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005900300902708125,
      "loss": 1.4908,
      "theoretical_loss": 3.572364406345528,
      "tokens_seen": 1372782592
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005900100300902708,
      "loss": 1.46,
      "theoretical_loss": 3.5723492351732906,
      "tokens_seen": 1372848128
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005899899699097292,
      "loss": 1.4263,
      "theoretical_loss": 3.5723340649280395,
      "tokens_seen": 1372913664
    },
    {
      "epoch": 0.42,
      "objective/train/advantage_avg": 0.4908875524997711,
      "objective/train/docs_used": 501474,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2542375326156616,
      "objective/train/original_loss": 1.254237413406372,
      "objective/train/theoretical_loss": 3.5723188956096736,
      "objective/train/tokens_used": 1393439200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24149060249328613,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050316333770752,
      "objective/train/weighted_lm_loss": 1.3174153566360474,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5723188956096736,
      "tokens_seen": 1372979200
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005899699097291876,
      "loss": 1.4002,
      "theoretical_loss": 3.5723188956096736,
      "tokens_seen": 1372979200
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005899498495486459,
      "loss": 1.4945,
      "theoretical_loss": 3.5723037272180918,
      "tokens_seen": 1373044736
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005899297893681044,
      "loss": 1.4612,
      "theoretical_loss": 3.572288559753194,
      "tokens_seen": 1373110272
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005899097291875627,
      "loss": 1.4181,
      "theoretical_loss": 3.572273393214879,
      "tokens_seen": 1373175808
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005898896690070211,
      "loss": 1.6634,
      "theoretical_loss": 3.5722582276030455,
      "tokens_seen": 1373241344
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005898696088264795,
      "loss": 1.6027,
      "theoretical_loss": 3.5722430629175936,
      "tokens_seen": 1373306880
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005898495486459379,
      "loss": 1.4745,
      "theoretical_loss": 3.5722278991584218,
      "tokens_seen": 1373372416
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005898294884653962,
      "loss": 1.4335,
      "theoretical_loss": 3.5722127363254295,
      "tokens_seen": 1373437952
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005898094282848546,
      "loss": 1.4875,
      "theoretical_loss": 3.5721975744185164,
      "tokens_seen": 1373503488
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005897893681043129,
      "loss": 1.5103,
      "theoretical_loss": 3.572182413437581,
      "tokens_seen": 1373569024
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005897693079237714,
      "loss": 1.407,
      "theoretical_loss": 3.572167253382523,
      "tokens_seen": 1373634560
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005897492477432297,
      "loss": 1.6087,
      "theoretical_loss": 3.5721520942532416,
      "tokens_seen": 1373700096
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005897291875626881,
      "loss": 1.5678,
      "theoretical_loss": 3.572136936049636,
      "tokens_seen": 1373765632
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005897091273821464,
      "loss": 1.4745,
      "theoretical_loss": 3.572121778771606,
      "tokens_seen": 1373831168
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005896890672016049,
      "loss": 1.4585,
      "theoretical_loss": 3.5721066224190503,
      "tokens_seen": 1373896704
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005896690070210632,
      "loss": 1.4832,
      "theoretical_loss": 3.5720914669918686,
      "tokens_seen": 1373962240
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005896489468405216,
      "loss": 1.4121,
      "theoretical_loss": 3.57207631248996,
      "tokens_seen": 1374027776
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00058962888665998,
      "loss": 1.462,
      "theoretical_loss": 3.5720611589132245,
      "tokens_seen": 1374093312
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005896088264794384,
      "loss": 1.422,
      "theoretical_loss": 3.572046006261561,
      "tokens_seen": 1374158848
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005895887662988967,
      "loss": 1.3785,
      "theoretical_loss": 3.5720308545348685,
      "tokens_seen": 1374224384
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005895687061183551,
      "loss": 1.4087,
      "theoretical_loss": 3.5720157037330473,
      "tokens_seen": 1374289920
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005895486459378134,
      "loss": 1.417,
      "theoretical_loss": 3.5720005538559967,
      "tokens_seen": 1374355456
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005895285857572719,
      "loss": 1.3825,
      "theoretical_loss": 3.5719854049036153,
      "tokens_seen": 1374420992
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005895085255767302,
      "loss": 1.4572,
      "theoretical_loss": 3.5719702568758036,
      "tokens_seen": 1374486528
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005894884653961886,
      "loss": 1.4318,
      "theoretical_loss": 3.5719551097724604,
      "tokens_seen": 1374552064
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000589468405215647,
      "loss": 1.4593,
      "theoretical_loss": 3.5719399635934863,
      "tokens_seen": 1374617600
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005894483450351053,
      "loss": 1.3705,
      "theoretical_loss": 3.571924818338779,
      "tokens_seen": 1374683136
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005894282848545637,
      "loss": 1.3251,
      "theoretical_loss": 3.57190967400824,
      "tokens_seen": 1374748672
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005894082246740221,
      "loss": 1.4751,
      "theoretical_loss": 3.5718945306017673,
      "tokens_seen": 1374814208
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005893881644934804,
      "loss": 1.3587,
      "theoretical_loss": 3.571879388119261,
      "tokens_seen": 1374879744
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005893681043129389,
      "loss": 1.4922,
      "theoretical_loss": 3.5718642465606214,
      "tokens_seen": 1374945280
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005893480441323971,
      "loss": 1.3737,
      "theoretical_loss": 3.5718491059257467,
      "tokens_seen": 1375010816
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005893279839518556,
      "loss": 1.4344,
      "theoretical_loss": 3.5718339662145375,
      "tokens_seen": 1375076352
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000589307923771314,
      "loss": 1.4333,
      "theoretical_loss": 3.5718188274268936,
      "tokens_seen": 1375141888
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005892878635907723,
      "loss": 1.4093,
      "theoretical_loss": 3.571803689562714,
      "tokens_seen": 1375207424
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005892678034102307,
      "loss": 1.5873,
      "theoretical_loss": 3.571788552621899,
      "tokens_seen": 1375272960
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005892477432296892,
      "loss": 1.408,
      "theoretical_loss": 3.5717734166043473,
      "tokens_seen": 1375338496
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005892276830491474,
      "loss": 1.4287,
      "theoretical_loss": 3.5717582815099598,
      "tokens_seen": 1375404032
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005892076228686059,
      "loss": 1.5241,
      "theoretical_loss": 3.571743147338635,
      "tokens_seen": 1375469568
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005891875626880641,
      "loss": 1.4405,
      "theoretical_loss": 3.571728014090274,
      "tokens_seen": 1375535104
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005891675025075226,
      "loss": 1.4406,
      "theoretical_loss": 3.571712881764775,
      "tokens_seen": 1375600640
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000589147442326981,
      "loss": 1.6176,
      "theoretical_loss": 3.5716977503620395,
      "tokens_seen": 1375666176
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005891273821464393,
      "loss": 1.421,
      "theoretical_loss": 3.5716826198819653,
      "tokens_seen": 1375731712
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005891073219658977,
      "loss": 1.39,
      "theoretical_loss": 3.571667490324454,
      "tokens_seen": 1375797248
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000589087261785356,
      "loss": 1.4986,
      "theoretical_loss": 3.571652361689404,
      "tokens_seen": 1375862784
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005890672016048144,
      "loss": 1.6455,
      "theoretical_loss": 3.571637233976716,
      "tokens_seen": 1375928320
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005890471414242729,
      "loss": 1.495,
      "theoretical_loss": 3.57162210718629,
      "tokens_seen": 1375993856
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005890270812437311,
      "loss": 1.4897,
      "theoretical_loss": 3.571606981318025,
      "tokens_seen": 1376059392
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005890070210631896,
      "loss": 1.4443,
      "theoretical_loss": 3.5715918563718216,
      "tokens_seen": 1376124928
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005889869608826479,
      "loss": 1.3995,
      "theoretical_loss": 3.5715767323475793,
      "tokens_seen": 1376190464
    },
    {
      "epoch": 0.42,
      "objective/train/advantage_avg": 0.4919305145740509,
      "objective/train/docs_used": 502650,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2530155181884766,
      "objective/train/original_loss": 1.2530152797698975,
      "objective/train/theoretical_loss": 3.5715616092451983,
      "objective/train/tokens_used": 1396716000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24257470667362213,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504261255264282,
      "objective/train/weighted_lm_loss": 1.3159908056259155,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5715616092451983,
      "tokens_seen": 1376256000
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005889669007021063,
      "loss": 1.3823,
      "theoretical_loss": 3.5715616092451983,
      "tokens_seen": 1376256000
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005889468405215647,
      "loss": 1.3376,
      "theoretical_loss": 3.5715464870645777,
      "tokens_seen": 1376321536
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005889267803410231,
      "loss": 1.5026,
      "theoretical_loss": 3.5715313658056185,
      "tokens_seen": 1376387072
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005889067201604814,
      "loss": 1.4086,
      "theoretical_loss": 3.57151624546822,
      "tokens_seen": 1376452608
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005888866599799399,
      "loss": 1.3187,
      "theoretical_loss": 3.5715011260522824,
      "tokens_seen": 1376518144
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005888665997993981,
      "loss": 1.4392,
      "theoretical_loss": 3.5714860075577057,
      "tokens_seen": 1376583680
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005888465396188566,
      "loss": 1.4497,
      "theoretical_loss": 3.57147088998439,
      "tokens_seen": 1376649216
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005888264794383149,
      "loss": 1.4073,
      "theoretical_loss": 3.571455773332234,
      "tokens_seen": 1376714752
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005888064192577733,
      "loss": 1.4986,
      "theoretical_loss": 3.57144065760114,
      "tokens_seen": 1376780288
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005887863590772317,
      "loss": 1.5534,
      "theoretical_loss": 3.5714255427910064,
      "tokens_seen": 1376845824
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005887662988966901,
      "loss": 1.4257,
      "theoretical_loss": 3.5714104289017343,
      "tokens_seen": 1376911360
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005887462387161484,
      "loss": 1.4304,
      "theoretical_loss": 3.571395315933223,
      "tokens_seen": 1376976896
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005887261785356068,
      "loss": 1.4423,
      "theoretical_loss": 3.5713802038853726,
      "tokens_seen": 1377042432
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005887061183550652,
      "loss": 1.3918,
      "theoretical_loss": 3.571365092758083,
      "tokens_seen": 1377107968
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005886860581745236,
      "loss": 1.3739,
      "theoretical_loss": 3.5713499825512556,
      "tokens_seen": 1377173504
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005886659979939819,
      "loss": 1.3591,
      "theoretical_loss": 3.571334873264789,
      "tokens_seen": 1377239040
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005886459378134403,
      "loss": 1.437,
      "theoretical_loss": 3.5713197648985844,
      "tokens_seen": 1377304576
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005886258776328986,
      "loss": 1.4817,
      "theoretical_loss": 3.5713046574525418,
      "tokens_seen": 1377370112
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005886058174523571,
      "loss": 1.4664,
      "theoretical_loss": 3.5712895509265605,
      "tokens_seen": 1377435648
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005885857572718155,
      "loss": 1.4045,
      "theoretical_loss": 3.5712744453205416,
      "tokens_seen": 1377501184
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005885656970912738,
      "loss": 1.6033,
      "theoretical_loss": 3.571259340634385,
      "tokens_seen": 1377566720
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005885456369107323,
      "loss": 1.3793,
      "theoretical_loss": 3.571244236867991,
      "tokens_seen": 1377632256
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005885255767301906,
      "loss": 1.4636,
      "theoretical_loss": 3.5712291340212605,
      "tokens_seen": 1377697792
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000588505516549649,
      "loss": 1.3581,
      "theoretical_loss": 3.571214032094092,
      "tokens_seen": 1377763328
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005884854563691074,
      "loss": 1.4023,
      "theoretical_loss": 3.5711989310863874,
      "tokens_seen": 1377828864
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005884653961885657,
      "loss": 1.2521,
      "theoretical_loss": 3.571183830998047,
      "tokens_seen": 1377894400
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005884453360080241,
      "loss": 1.4229,
      "theoretical_loss": 3.57116873182897,
      "tokens_seen": 1377959936
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005884252758274825,
      "loss": 1.3705,
      "theoretical_loss": 3.571153633579057,
      "tokens_seen": 1378025472
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005884052156469408,
      "loss": 1.4078,
      "theoretical_loss": 3.571138536248209,
      "tokens_seen": 1378091008
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005883851554663993,
      "loss": 1.5409,
      "theoretical_loss": 3.5711234398363256,
      "tokens_seen": 1378156544
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005883650952858575,
      "loss": 1.4453,
      "theoretical_loss": 3.5711083443433083,
      "tokens_seen": 1378222080
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000588345035105316,
      "loss": 1.4906,
      "theoretical_loss": 3.5710932497690555,
      "tokens_seen": 1378287616
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005883249749247744,
      "loss": 1.4051,
      "theoretical_loss": 3.57107815611347,
      "tokens_seen": 1378353152
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005883049147442327,
      "loss": 1.4738,
      "theoretical_loss": 3.5710630633764504,
      "tokens_seen": 1378418688
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005882848545636911,
      "loss": 1.5326,
      "theoretical_loss": 3.571047971557898,
      "tokens_seen": 1378484224
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005882647943831496,
      "loss": 1.4603,
      "theoretical_loss": 3.5710328806577127,
      "tokens_seen": 1378549760
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005882447342026078,
      "loss": 1.4606,
      "theoretical_loss": 3.571017790675796,
      "tokens_seen": 1378615296
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005882246740220663,
      "loss": 1.421,
      "theoretical_loss": 3.5710027016120467,
      "tokens_seen": 1378680832
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005882046138415245,
      "loss": 1.4832,
      "theoretical_loss": 3.570987613466367,
      "tokens_seen": 1378746368
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000588184553660983,
      "loss": 1.519,
      "theoretical_loss": 3.5709725262386565,
      "tokens_seen": 1378811904
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005881644934804414,
      "loss": 1.4656,
      "theoretical_loss": 3.570957439928815,
      "tokens_seen": 1378877440
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005881444332998997,
      "loss": 1.4787,
      "theoretical_loss": 3.570942354536745,
      "tokens_seen": 1378942976
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005881243731193581,
      "loss": 1.4849,
      "theoretical_loss": 3.5709272700623456,
      "tokens_seen": 1379008512
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005881043129388165,
      "loss": 1.3186,
      "theoretical_loss": 3.5709121865055176,
      "tokens_seen": 1379074048
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005880842527582748,
      "loss": 1.5567,
      "theoretical_loss": 3.5708971038661614,
      "tokens_seen": 1379139584
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005880641925777333,
      "loss": 1.5674,
      "theoretical_loss": 3.5708820221441786,
      "tokens_seen": 1379205120
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005880441323971915,
      "loss": 1.4257,
      "theoretical_loss": 3.570866941339469,
      "tokens_seen": 1379270656
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00058802407221665,
      "loss": 1.4901,
      "theoretical_loss": 3.570851861451933,
      "tokens_seen": 1379336192
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005880040120361083,
      "loss": 1.2338,
      "theoretical_loss": 3.5708367824814715,
      "tokens_seen": 1379401728
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005879839518555667,
      "loss": 1.4886,
      "theoretical_loss": 3.5708217044279857,
      "tokens_seen": 1379467264
    },
    {
      "epoch": 0.42,
      "objective/train/advantage_avg": 0.49669235944747925,
      "objective/train/docs_used": 503826,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4134360551834106,
      "objective/train/original_loss": 1.413435935974121,
      "objective/train/theoretical_loss": 3.5708066272913754,
      "objective/train/tokens_used": 1399992800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2467978447675705,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0509240627288818,
      "objective/train/weighted_lm_loss": 1.4853912591934204,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.5708066272913754,
      "tokens_seen": 1379532800
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005879638916750251,
      "loss": 1.475,
      "theoretical_loss": 3.5708066272913754,
      "tokens_seen": 1379532800
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005879438314944835,
      "loss": 1.3947,
      "theoretical_loss": 3.570791551071542,
      "tokens_seen": 1379598336
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005879237713139418,
      "loss": 1.4525,
      "theoretical_loss": 3.570776475768386,
      "tokens_seen": 1379663872
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005879037111334003,
      "loss": 1.5554,
      "theoretical_loss": 3.5707614013818083,
      "tokens_seen": 1379729408
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005878836509528585,
      "loss": 1.4932,
      "theoretical_loss": 3.570746327911709,
      "tokens_seen": 1379794944
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000587863590772317,
      "loss": 1.4791,
      "theoretical_loss": 3.5707312553579893,
      "tokens_seen": 1379860480
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005878435305917753,
      "loss": 1.3569,
      "theoretical_loss": 3.57071618372055,
      "tokens_seen": 1379926016
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005878234704112337,
      "loss": 1.4948,
      "theoretical_loss": 3.570701112999292,
      "tokens_seen": 1379991552
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005878034102306921,
      "loss": 1.51,
      "theoretical_loss": 3.5706860431941156,
      "tokens_seen": 1380057088
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005877833500501505,
      "loss": 1.5044,
      "theoretical_loss": 3.570670974304922,
      "tokens_seen": 1380122624
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005877632898696088,
      "loss": 1.4335,
      "theoretical_loss": 3.570655906331612,
      "tokens_seen": 1380188160
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005877432296890672,
      "loss": 1.4884,
      "theoretical_loss": 3.5706408392740867,
      "tokens_seen": 1380253696
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005877231695085256,
      "loss": 1.4286,
      "theoretical_loss": 3.570625773132246,
      "tokens_seen": 1380319232
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000587703109327984,
      "loss": 1.4091,
      "theoretical_loss": 3.570610707905992,
      "tokens_seen": 1380384768
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005876830491474423,
      "loss": 1.5198,
      "theoretical_loss": 3.570595643595225,
      "tokens_seen": 1380450304
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005876629889669007,
      "loss": 1.423,
      "theoretical_loss": 3.570580580199846,
      "tokens_seen": 1380515840
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000587642928786359,
      "loss": 1.4293,
      "theoretical_loss": 3.5705655177197557,
      "tokens_seen": 1380581376
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005876228686058175,
      "loss": 1.4587,
      "theoretical_loss": 3.570550456154855,
      "tokens_seen": 1380646912
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005876028084252758,
      "loss": 1.538,
      "theoretical_loss": 3.570535395505045,
      "tokens_seen": 1380712448
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005875827482447342,
      "loss": 1.4017,
      "theoretical_loss": 3.5705203357702278,
      "tokens_seen": 1380777984
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005875626880641926,
      "loss": 1.4771,
      "theoretical_loss": 3.5705052769503025,
      "tokens_seen": 1380843520
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000587542627883651,
      "loss": 1.2991,
      "theoretical_loss": 3.570490219045171,
      "tokens_seen": 1380909056
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005875225677031093,
      "loss": 1.4403,
      "theoretical_loss": 3.570475162054734,
      "tokens_seen": 1380974592
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005875025075225678,
      "loss": 1.4109,
      "theoretical_loss": 3.5704601059788934,
      "tokens_seen": 1381040128
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000587482447342026,
      "loss": 1.4179,
      "theoretical_loss": 3.5704450508175496,
      "tokens_seen": 1381105664
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005874623871614845,
      "loss": 1.4923,
      "theoretical_loss": 3.5704299965706037,
      "tokens_seen": 1381171200
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005874423269809428,
      "loss": 1.508,
      "theoretical_loss": 3.570414943237956,
      "tokens_seen": 1381236736
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005874222668004012,
      "loss": 1.4282,
      "theoretical_loss": 3.570399890819509,
      "tokens_seen": 1381302272
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005874022066198596,
      "loss": 1.4488,
      "theoretical_loss": 3.570384839315163,
      "tokens_seen": 1381367808
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005873821464393179,
      "loss": 1.3842,
      "theoretical_loss": 3.570369788724819,
      "tokens_seen": 1381433344
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005873620862587763,
      "loss": 1.5525,
      "theoretical_loss": 3.570354739048379,
      "tokens_seen": 1381498880
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005873420260782348,
      "loss": 1.4049,
      "theoretical_loss": 3.570339690285744,
      "tokens_seen": 1381564416
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000587321965897693,
      "loss": 1.4475,
      "theoretical_loss": 3.570324642436814,
      "tokens_seen": 1381629952
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005873019057171515,
      "loss": 1.4633,
      "theoretical_loss": 3.570309595501491,
      "tokens_seen": 1381695488
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005872818455366097,
      "loss": 1.4106,
      "theoretical_loss": 3.5702945494796765,
      "tokens_seen": 1381761024
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005872617853560682,
      "loss": 1.4733,
      "theoretical_loss": 3.570279504371271,
      "tokens_seen": 1381826560
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005872417251755266,
      "loss": 1.4652,
      "theoretical_loss": 3.5702644601761766,
      "tokens_seen": 1381892096
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005872216649949849,
      "loss": 1.3592,
      "theoretical_loss": 3.5702494168942938,
      "tokens_seen": 1381957632
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005872016048144433,
      "loss": 1.5356,
      "theoretical_loss": 3.5702343745255236,
      "tokens_seen": 1382023168
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005871815446339018,
      "loss": 1.5094,
      "theoretical_loss": 3.570219333069769,
      "tokens_seen": 1382088704
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00058716148445336,
      "loss": 1.3966,
      "theoretical_loss": 3.5702042925269293,
      "tokens_seen": 1382154240
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005871414242728185,
      "loss": 1.4793,
      "theoretical_loss": 3.570189252896906,
      "tokens_seen": 1382219776
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005871213640922767,
      "loss": 1.4334,
      "theoretical_loss": 3.5701742141796022,
      "tokens_seen": 1382285312
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005871013039117352,
      "loss": 1.4339,
      "theoretical_loss": 3.5701591763749176,
      "tokens_seen": 1382350848
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005870812437311937,
      "loss": 1.391,
      "theoretical_loss": 3.570144139482754,
      "tokens_seen": 1382416384
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005870611835506519,
      "loss": 1.4671,
      "theoretical_loss": 3.570129103503013,
      "tokens_seen": 1382481920
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005870411233701104,
      "loss": 1.3526,
      "theoretical_loss": 3.570114068435595,
      "tokens_seen": 1382547456
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005870210631895687,
      "loss": 1.4266,
      "theoretical_loss": 3.5700990342804033,
      "tokens_seen": 1382612992
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005870010030090271,
      "loss": 1.5675,
      "theoretical_loss": 3.570084001037338,
      "tokens_seen": 1382678528
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005869809428284855,
      "loss": 1.4363,
      "theoretical_loss": 3.5700689687063,
      "tokens_seen": 1382744064
    },
    {
      "epoch": 0.42,
      "objective/train/advantage_avg": 0.4883284866809845,
      "objective/train/docs_used": 505055,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.325297474861145,
      "objective/train/original_loss": 1.3252973556518555,
      "objective/train/theoretical_loss": 3.570053937287192,
      "objective/train/tokens_used": 1403269600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23912307620048523,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500483512878418,
      "objective/train/weighted_lm_loss": 1.3918594121932983,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.570053937287192,
      "tokens_seen": 1382809600
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005869608826479439,
      "loss": 1.4707,
      "theoretical_loss": 3.570053937287192,
      "tokens_seen": 1382809600
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005869408224674022,
      "loss": 1.396,
      "theoretical_loss": 3.570038906779915,
      "tokens_seen": 1382875136
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005869207622868607,
      "loss": 1.3958,
      "theoretical_loss": 3.57002387718437,
      "tokens_seen": 1382940672
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000586900702106319,
      "loss": 1.4595,
      "theoretical_loss": 3.5700088485004593,
      "tokens_seen": 1383006208
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005868806419257774,
      "loss": 1.5293,
      "theoretical_loss": 3.569993820728084,
      "tokens_seen": 1383071744
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005868605817452357,
      "loss": 1.4416,
      "theoretical_loss": 3.5699787938671452,
      "tokens_seen": 1383137280
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005868405215646941,
      "loss": 1.4508,
      "theoretical_loss": 3.569963767917545,
      "tokens_seen": 1383202816
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005868204613841525,
      "loss": 1.4171,
      "theoretical_loss": 3.569948742879186,
      "tokens_seen": 1383268352
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005868004012036109,
      "loss": 1.3439,
      "theoretical_loss": 3.569933718751967,
      "tokens_seen": 1383333888
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005867803410230692,
      "loss": 1.5342,
      "theoretical_loss": 3.569918695535792,
      "tokens_seen": 1383399424
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005867602808425276,
      "loss": 1.4765,
      "theoretical_loss": 3.569903673230562,
      "tokens_seen": 1383464960
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000586740220661986,
      "loss": 1.3943,
      "theoretical_loss": 3.5698886518361777,
      "tokens_seen": 1383530496
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005867201604814444,
      "loss": 1.4277,
      "theoretical_loss": 3.569873631352542,
      "tokens_seen": 1383596032
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005867001003009027,
      "loss": 1.4134,
      "theoretical_loss": 3.569858611779556,
      "tokens_seen": 1383661568
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005866800401203611,
      "loss": 1.4348,
      "theoretical_loss": 3.5698435931171213,
      "tokens_seen": 1383727104
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005866599799398194,
      "loss": 1.4679,
      "theoretical_loss": 3.5698285753651393,
      "tokens_seen": 1383792640
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005866399197592779,
      "loss": 1.5162,
      "theoretical_loss": 3.5698135585235122,
      "tokens_seen": 1383858176
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005866198595787362,
      "loss": 1.4236,
      "theoretical_loss": 3.5697985425921415,
      "tokens_seen": 1383923712
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005865997993981946,
      "loss": 1.4993,
      "theoretical_loss": 3.5697835275709293,
      "tokens_seen": 1383989248
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000586579739217653,
      "loss": 1.3229,
      "theoretical_loss": 3.5697685134597767,
      "tokens_seen": 1384054784
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005865596790371114,
      "loss": 1.3918,
      "theoretical_loss": 3.5697535002585856,
      "tokens_seen": 1384120320
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005865396188565697,
      "loss": 1.4034,
      "theoretical_loss": 3.569738487967258,
      "tokens_seen": 1384185856
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005865195586760282,
      "loss": 1.4187,
      "theoretical_loss": 3.5697234765856956,
      "tokens_seen": 1384251392
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005864994984954864,
      "loss": 1.4231,
      "theoretical_loss": 3.5697084661138003,
      "tokens_seen": 1384316928
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005864794383149449,
      "loss": 1.5541,
      "theoretical_loss": 3.569693456551474,
      "tokens_seen": 1384382464
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005864593781344032,
      "loss": 1.5065,
      "theoretical_loss": 3.569678447898618,
      "tokens_seen": 1384448000
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005864393179538616,
      "loss": 1.5522,
      "theoretical_loss": 3.5696634401551344,
      "tokens_seen": 1384513536
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00058641925777332,
      "loss": 1.4733,
      "theoretical_loss": 3.5696484333209253,
      "tokens_seen": 1384579072
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005863991975927783,
      "loss": 1.4058,
      "theoretical_loss": 3.5696334273958925,
      "tokens_seen": 1384644608
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005863791374122367,
      "loss": 1.5058,
      "theoretical_loss": 3.5696184223799374,
      "tokens_seen": 1384710144
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005863590772316952,
      "loss": 1.4288,
      "theoretical_loss": 3.569603418272963,
      "tokens_seen": 1384775680
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005863390170511534,
      "loss": 1.449,
      "theoretical_loss": 3.56958841507487,
      "tokens_seen": 1384841216
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005863189568706119,
      "loss": 1.384,
      "theoretical_loss": 3.569573412785561,
      "tokens_seen": 1384906752
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005862988966900701,
      "loss": 1.5221,
      "theoretical_loss": 3.5695584114049375,
      "tokens_seen": 1384972288
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005862788365095286,
      "loss": 1.4502,
      "theoretical_loss": 3.5695434109329023,
      "tokens_seen": 1385037824
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000586258776328987,
      "loss": 1.5506,
      "theoretical_loss": 3.569528411369357,
      "tokens_seen": 1385103360
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005862387161484453,
      "loss": 1.391,
      "theoretical_loss": 3.569513412714203,
      "tokens_seen": 1385168896
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005862186559679037,
      "loss": 1.394,
      "theoretical_loss": 3.569498414967343,
      "tokens_seen": 1385234432
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005861985957873622,
      "loss": 1.5711,
      "theoretical_loss": 3.5694834181286788,
      "tokens_seen": 1385299968
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005861785356068204,
      "loss": 1.5428,
      "theoretical_loss": 3.5694684221981126,
      "tokens_seen": 1385365504
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005861584754262789,
      "loss": 1.5104,
      "theoretical_loss": 3.569453427175546,
      "tokens_seen": 1385431040
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005861384152457371,
      "loss": 1.4035,
      "theoretical_loss": 3.5694384330608813,
      "tokens_seen": 1385496576
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005861183550651956,
      "loss": 1.4238,
      "theoretical_loss": 3.5694234398540208,
      "tokens_seen": 1385562112
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000586098294884654,
      "loss": 1.4489,
      "theoretical_loss": 3.5694084475548666,
      "tokens_seen": 1385627648
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005860782347041123,
      "loss": 1.3998,
      "theoretical_loss": 3.5693934561633203,
      "tokens_seen": 1385693184
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005860581745235707,
      "loss": 1.3792,
      "theoretical_loss": 3.569378465679285,
      "tokens_seen": 1385758720
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005860381143430291,
      "loss": 1.3647,
      "theoretical_loss": 3.5693634761026622,
      "tokens_seen": 1385824256
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005860180541624874,
      "loss": 1.487,
      "theoretical_loss": 3.569348487433354,
      "tokens_seen": 1385889792
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005859979939819459,
      "loss": 1.4779,
      "theoretical_loss": 3.5693334996712625,
      "tokens_seen": 1385955328
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005859779338014042,
      "loss": 1.4705,
      "theoretical_loss": 3.5693185128162903,
      "tokens_seen": 1386020864
    },
    {
      "epoch": 0.42,
      "objective/train/advantage_avg": 0.49301883578300476,
      "objective/train/docs_used": 506209,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.539597749710083,
      "objective/train/original_loss": 1.539597988128662,
      "objective/train/theoretical_loss": 3.5693035268683393,
      "objective/train/tokens_used": 1406546400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24396759271621704,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505421161651611,
      "objective/train/weighted_lm_loss": 1.6174156665802002,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5693035268683393,
      "tokens_seen": 1386086400
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005859578736208626,
      "loss": 1.4832,
      "theoretical_loss": 3.5693035268683393,
      "tokens_seen": 1386086400
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000585937813440321,
      "loss": 1.4655,
      "theoretical_loss": 3.569288541827312,
      "tokens_seen": 1386151936
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005859177532597793,
      "loss": 1.421,
      "theoretical_loss": 3.5692735576931103,
      "tokens_seen": 1386217472
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005858976930792377,
      "loss": 1.3541,
      "theoretical_loss": 3.569258574465637,
      "tokens_seen": 1386283008
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005858776328986961,
      "loss": 1.4242,
      "theoretical_loss": 3.569243592144794,
      "tokens_seen": 1386348544
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005858575727181544,
      "loss": 1.276,
      "theoretical_loss": 3.569228610730483,
      "tokens_seen": 1386414080
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005858375125376129,
      "loss": 1.5151,
      "theoretical_loss": 3.569213630222607,
      "tokens_seen": 1386479616
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005858174523570712,
      "loss": 1.4475,
      "theoretical_loss": 3.569198650621069,
      "tokens_seen": 1386545152
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005857973921765296,
      "loss": 1.478,
      "theoretical_loss": 3.56918367192577,
      "tokens_seen": 1386610688
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005857773319959879,
      "loss": 1.4404,
      "theoretical_loss": 3.5691686941366125,
      "tokens_seen": 1386676224
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005857572718154464,
      "loss": 1.4154,
      "theoretical_loss": 3.5691537172535,
      "tokens_seen": 1386741760
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005857372116349047,
      "loss": 1.4041,
      "theoretical_loss": 3.5691387412763333,
      "tokens_seen": 1386807296
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005857171514543631,
      "loss": 1.3647,
      "theoretical_loss": 3.569123766205016,
      "tokens_seen": 1386872832
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005856970912738214,
      "loss": 1.3969,
      "theoretical_loss": 3.5691087920394504,
      "tokens_seen": 1386938368
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005856770310932798,
      "loss": 1.3937,
      "theoretical_loss": 3.569093818779539,
      "tokens_seen": 1387003904
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005856569709127382,
      "loss": 1.4692,
      "theoretical_loss": 3.569078846425183,
      "tokens_seen": 1387069440
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005856369107321966,
      "loss": 1.4605,
      "theoretical_loss": 3.569063874976286,
      "tokens_seen": 1387134976
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005856168505516549,
      "loss": 1.5174,
      "theoretical_loss": 3.5690489044327505,
      "tokens_seen": 1387200512
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005855967903711134,
      "loss": 1.6352,
      "theoretical_loss": 3.5690339347944784,
      "tokens_seen": 1387266048
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005855767301905718,
      "loss": 1.4246,
      "theoretical_loss": 3.5690189660613725,
      "tokens_seen": 1387331584
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005855566700100301,
      "loss": 1.4413,
      "theoretical_loss": 3.5690039982333355,
      "tokens_seen": 1387397120
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005855366098294886,
      "loss": 1.4773,
      "theoretical_loss": 3.5689890313102692,
      "tokens_seen": 1387462656
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005855165496489468,
      "loss": 1.2733,
      "theoretical_loss": 3.568974065292077,
      "tokens_seen": 1387528192
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005854964894684053,
      "loss": 1.4719,
      "theoretical_loss": 3.5689591001786614,
      "tokens_seen": 1387593728
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005854764292878636,
      "loss": 1.7115,
      "theoretical_loss": 3.5689441359699243,
      "tokens_seen": 1387659264
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000585456369107322,
      "loss": 1.4599,
      "theoretical_loss": 3.5689291726657686,
      "tokens_seen": 1387724800
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005854363089267804,
      "loss": 1.3353,
      "theoretical_loss": 3.5689142102660973,
      "tokens_seen": 1387790336
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005854162487462387,
      "loss": 1.4549,
      "theoretical_loss": 3.568899248770813,
      "tokens_seen": 1387855872
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005853961885656971,
      "loss": 1.3592,
      "theoretical_loss": 3.5688842881798175,
      "tokens_seen": 1387921408
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005853761283851556,
      "loss": 1.4417,
      "theoretical_loss": 3.5688693284930135,
      "tokens_seen": 1387986944
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005853560682046138,
      "loss": 1.3554,
      "theoretical_loss": 3.568854369710305,
      "tokens_seen": 1388052480
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005853360080240723,
      "loss": 1.2826,
      "theoretical_loss": 3.5688394118315934,
      "tokens_seen": 1388118016
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005853159478435305,
      "loss": 1.502,
      "theoretical_loss": 3.5688244548567822,
      "tokens_seen": 1388183552
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000585295887662989,
      "loss": 1.375,
      "theoretical_loss": 3.5688094987857735,
      "tokens_seen": 1388249088
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005852758274824474,
      "loss": 1.3813,
      "theoretical_loss": 3.5687945436184703,
      "tokens_seen": 1388314624
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005852557673019057,
      "loss": 1.4861,
      "theoretical_loss": 3.568779589354775,
      "tokens_seen": 1388380160
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005852357071213641,
      "loss": 1.3308,
      "theoretical_loss": 3.568764635994591,
      "tokens_seen": 1388445696
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005852156469408226,
      "loss": 1.5353,
      "theoretical_loss": 3.5687496835378205,
      "tokens_seen": 1388511232
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005851955867602808,
      "loss": 1.4331,
      "theoretical_loss": 3.5687347319843665,
      "tokens_seen": 1388576768
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005851755265797393,
      "loss": 1.4657,
      "theoretical_loss": 3.5687197813341323,
      "tokens_seen": 1388642304
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005851554663991975,
      "loss": 1.3764,
      "theoretical_loss": 3.5687048315870196,
      "tokens_seen": 1388707840
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000585135406218656,
      "loss": 1.4969,
      "theoretical_loss": 3.568689882742932,
      "tokens_seen": 1388773376
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005851153460381144,
      "loss": 1.4988,
      "theoretical_loss": 3.5686749348017726,
      "tokens_seen": 1388838912
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005850952858575727,
      "loss": 1.4246,
      "theoretical_loss": 3.568659987763443,
      "tokens_seen": 1388904448
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005850752256770311,
      "loss": 1.4505,
      "theoretical_loss": 3.5686450416278475,
      "tokens_seen": 1388969984
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005850551654964895,
      "loss": 1.4672,
      "theoretical_loss": 3.5686300963948883,
      "tokens_seen": 1389035520
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005850351053159478,
      "loss": 1.5818,
      "theoretical_loss": 3.5686151520644684,
      "tokens_seen": 1389101056
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005850150451354063,
      "loss": 1.4593,
      "theoretical_loss": 3.5686002086364903,
      "tokens_seen": 1389166592
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005849949849548646,
      "loss": 1.392,
      "theoretical_loss": 3.568585266110858,
      "tokens_seen": 1389232128
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000584974924774323,
      "loss": 1.583,
      "theoretical_loss": 3.5685703244874736,
      "tokens_seen": 1389297664
    },
    {
      "epoch": 0.42,
      "objective/train/advantage_avg": 0.49430257081985474,
      "objective/train/docs_used": 507487,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.35800302028656,
      "objective/train/original_loss": 1.3580026626586914,
      "objective/train/theoretical_loss": 3.56855538376624,
      "objective/train/tokens_used": 1409823200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.244618222117424,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506738424301147,
      "objective/train/weighted_lm_loss": 1.4270060062408447,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.56855538376624,
      "tokens_seen": 1389363200
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005849548645937814,
      "loss": 1.3839,
      "theoretical_loss": 3.56855538376624,
      "tokens_seen": 1389363200
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005849348044132397,
      "loss": 1.4434,
      "theoretical_loss": 3.568540443947061,
      "tokens_seen": 1389428736
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005849147442326981,
      "loss": 1.427,
      "theoretical_loss": 3.5685255050298386,
      "tokens_seen": 1389494272
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005848946840521565,
      "loss": 1.3504,
      "theoretical_loss": 3.568510567014476,
      "tokens_seen": 1389559808
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005848746238716148,
      "loss": 1.4136,
      "theoretical_loss": 3.568495629900877,
      "tokens_seen": 1389625344
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005848545636910733,
      "loss": 1.5688,
      "theoretical_loss": 3.568480693688944,
      "tokens_seen": 1389690880
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005848345035105316,
      "loss": 1.5591,
      "theoretical_loss": 3.56846575837858,
      "tokens_seen": 1389756416
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00058481444332999,
      "loss": 1.4611,
      "theoretical_loss": 3.5684508239696884,
      "tokens_seen": 1389821952
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005847943831494483,
      "loss": 1.4342,
      "theoretical_loss": 3.5684358904621725,
      "tokens_seen": 1389887488
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005847743229689068,
      "loss": 1.3333,
      "theoretical_loss": 3.5684209578559343,
      "tokens_seen": 1389953024
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005847542627883651,
      "loss": 1.4426,
      "theoretical_loss": 3.5684060261508783,
      "tokens_seen": 1390018560
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005847342026078235,
      "loss": 1.3483,
      "theoretical_loss": 3.5683910953469065,
      "tokens_seen": 1390084096
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005847141424272818,
      "loss": 1.4481,
      "theoretical_loss": 3.5683761654439223,
      "tokens_seen": 1390149632
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005846940822467402,
      "loss": 1.2911,
      "theoretical_loss": 3.5683612364418296,
      "tokens_seen": 1390215168
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005846740220661986,
      "loss": 1.3799,
      "theoretical_loss": 3.568346308340531,
      "tokens_seen": 1390280704
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000584653961885657,
      "loss": 1.3669,
      "theoretical_loss": 3.568331381139929,
      "tokens_seen": 1390346240
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005846339017051153,
      "loss": 1.4862,
      "theoretical_loss": 3.5683164548399287,
      "tokens_seen": 1390411776
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005846138415245738,
      "loss": 1.5367,
      "theoretical_loss": 3.5683015294404314,
      "tokens_seen": 1390477312
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005845937813440321,
      "loss": 1.4291,
      "theoretical_loss": 3.568286604941341,
      "tokens_seen": 1390542848
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005845737211634905,
      "loss": 1.5088,
      "theoretical_loss": 3.5682716813425612,
      "tokens_seen": 1390608384
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005845536609829488,
      "loss": 1.4725,
      "theoretical_loss": 3.568256758643995,
      "tokens_seen": 1390673920
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005845336008024072,
      "loss": 1.4257,
      "theoretical_loss": 3.5682418368455453,
      "tokens_seen": 1390739456
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005845135406218656,
      "loss": 1.2942,
      "theoretical_loss": 3.5682269159471156,
      "tokens_seen": 1390804992
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000584493480441324,
      "loss": 1.3764,
      "theoretical_loss": 3.5682119959486096,
      "tokens_seen": 1390870528
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005844734202607823,
      "loss": 1.4425,
      "theoretical_loss": 3.5681970768499305,
      "tokens_seen": 1390936064
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005844533600802408,
      "loss": 1.388,
      "theoretical_loss": 3.568182158650981,
      "tokens_seen": 1391001600
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000584433299899699,
      "loss": 1.4648,
      "theoretical_loss": 3.5681672413516647,
      "tokens_seen": 1391067136
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005844132397191575,
      "loss": 1.4095,
      "theoretical_loss": 3.568152324951886,
      "tokens_seen": 1391132672
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005843931795386159,
      "loss": 1.3596,
      "theoretical_loss": 3.5681374094515466,
      "tokens_seen": 1391198208
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005843731193580742,
      "loss": 1.3506,
      "theoretical_loss": 3.568122494850551,
      "tokens_seen": 1391263744
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005843530591775326,
      "loss": 1.4616,
      "theoretical_loss": 3.5681075811488023,
      "tokens_seen": 1391329280
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005843329989969909,
      "loss": 1.4259,
      "theoretical_loss": 3.568092668346204,
      "tokens_seen": 1391394816
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005843129388164493,
      "loss": 1.5251,
      "theoretical_loss": 3.5680777564426602,
      "tokens_seen": 1391460352
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005842928786359078,
      "loss": 1.3384,
      "theoretical_loss": 3.5680628454380727,
      "tokens_seen": 1391525888
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000584272818455366,
      "loss": 1.4275,
      "theoretical_loss": 3.5680479353323467,
      "tokens_seen": 1391591424
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005842527582748245,
      "loss": 1.2883,
      "theoretical_loss": 3.5680330261253843,
      "tokens_seen": 1391656960
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005842326980942829,
      "loss": 1.4503,
      "theoretical_loss": 3.56801811781709,
      "tokens_seen": 1391722496
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005842126379137412,
      "loss": 1.4603,
      "theoretical_loss": 3.568003210407367,
      "tokens_seen": 1391788032
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005841925777331996,
      "loss": 1.358,
      "theoretical_loss": 3.567988303896119,
      "tokens_seen": 1391853568
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000584172517552658,
      "loss": 1.3889,
      "theoretical_loss": 3.5679733982832493,
      "tokens_seen": 1391919104
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005841524573721163,
      "loss": 1.4694,
      "theoretical_loss": 3.5679584935686615,
      "tokens_seen": 1391984640
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005841323971915748,
      "loss": 1.4364,
      "theoretical_loss": 3.5679435897522587,
      "tokens_seen": 1392050176
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000584112337011033,
      "loss": 1.4031,
      "theoretical_loss": 3.5679286868339455,
      "tokens_seen": 1392115712
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005840922768304915,
      "loss": 1.362,
      "theoretical_loss": 3.5679137848136246,
      "tokens_seen": 1392181248
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005840722166499499,
      "loss": 1.4335,
      "theoretical_loss": 3.5678988836912007,
      "tokens_seen": 1392246784
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005840521564694082,
      "loss": 1.438,
      "theoretical_loss": 3.567883983466576,
      "tokens_seen": 1392312320
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005840320962888667,
      "loss": 1.456,
      "theoretical_loss": 3.567869084139655,
      "tokens_seen": 1392377856
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000584012036108325,
      "loss": 1.3463,
      "theoretical_loss": 3.567854185710342,
      "tokens_seen": 1392443392
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005839919759277834,
      "loss": 1.4948,
      "theoretical_loss": 3.567839288178539,
      "tokens_seen": 1392508928
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005839719157472418,
      "loss": 1.496,
      "theoretical_loss": 3.5678243915441508,
      "tokens_seen": 1392574464
    },
    {
      "epoch": 0.42,
      "objective/train/advantage_avg": 0.4923599362373352,
      "objective/train/docs_used": 508733,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2537610530853271,
      "objective/train/original_loss": 1.2537609338760376,
      "objective/train/theoretical_loss": 3.567809495807081,
      "objective/train/tokens_used": 1413100000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2427421361207962,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050469994544983,
      "objective/train/weighted_lm_loss": 1.316812515258789,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0216598510742188,
      "theoretical_loss": 3.567809495807081,
      "tokens_seen": 1392640000
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005839518555667001,
      "loss": 1.356,
      "theoretical_loss": 3.567809495807081,
      "tokens_seen": 1392640000
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005839317953861585,
      "loss": 1.497,
      "theoretical_loss": 3.5677946009672343,
      "tokens_seen": 1392705536
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005839117352056169,
      "loss": 1.4662,
      "theoretical_loss": 3.5677797070245125,
      "tokens_seen": 1392771072
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005838916750250752,
      "loss": 1.4662,
      "theoretical_loss": 3.56776481397882,
      "tokens_seen": 1392836608
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005838716148445337,
      "loss": 1.458,
      "theoretical_loss": 3.5677499218300612,
      "tokens_seen": 1392902144
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000583851554663992,
      "loss": 1.3505,
      "theoretical_loss": 3.56773503057814,
      "tokens_seen": 1392967680
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005838314944834504,
      "loss": 1.4311,
      "theoretical_loss": 3.567720140222959,
      "tokens_seen": 1393033216
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005838114343029087,
      "loss": 1.4464,
      "theoretical_loss": 3.567705250764423,
      "tokens_seen": 1393098752
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005837913741223672,
      "loss": 1.4074,
      "theoretical_loss": 3.567690362202436,
      "tokens_seen": 1393164288
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005837713139418255,
      "loss": 1.5427,
      "theoretical_loss": 3.567675474536901,
      "tokens_seen": 1393229824
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005837512537612839,
      "loss": 1.504,
      "theoretical_loss": 3.567660587767722,
      "tokens_seen": 1393295360
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005837311935807422,
      "loss": 1.3857,
      "theoretical_loss": 3.5676457018948033,
      "tokens_seen": 1393360896
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005837111334002006,
      "loss": 1.4131,
      "theoretical_loss": 3.567630816918049,
      "tokens_seen": 1393426432
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000583691073219659,
      "loss": 1.3928,
      "theoretical_loss": 3.567615932837363,
      "tokens_seen": 1393491968
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005836710130391174,
      "loss": 1.54,
      "theoretical_loss": 3.567601049652648,
      "tokens_seen": 1393557504
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005836509528585757,
      "loss": 1.4505,
      "theoretical_loss": 3.5675861673638094,
      "tokens_seen": 1393623040
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005836308926780342,
      "loss": 1.4172,
      "theoretical_loss": 3.5675712859707502,
      "tokens_seen": 1393688576
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005836108324974925,
      "loss": 1.4369,
      "theoretical_loss": 3.5675564054733746,
      "tokens_seen": 1393754112
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005835907723169509,
      "loss": 1.4089,
      "theoretical_loss": 3.567541525871587,
      "tokens_seen": 1393819648
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005835707121364092,
      "loss": 1.535,
      "theoretical_loss": 3.567526647165291,
      "tokens_seen": 1393885184
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005835506519558676,
      "loss": 1.3514,
      "theoretical_loss": 3.5675117693543905,
      "tokens_seen": 1393950720
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000583530591775326,
      "loss": 1.51,
      "theoretical_loss": 3.5674968924387898,
      "tokens_seen": 1394016256
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005835105315947844,
      "loss": 1.3172,
      "theoretical_loss": 3.5674820164183934,
      "tokens_seen": 1394081792
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005834904714142427,
      "loss": 1.5111,
      "theoretical_loss": 3.567467141293104,
      "tokens_seen": 1394147328
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005834704112337012,
      "loss": 1.5045,
      "theoretical_loss": 3.5674522670628273,
      "tokens_seen": 1394212864
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005834503510531594,
      "loss": 1.4044,
      "theoretical_loss": 3.5674373937274657,
      "tokens_seen": 1394278400
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005834302908726179,
      "loss": 1.5173,
      "theoretical_loss": 3.567422521286925,
      "tokens_seen": 1394343936
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005834102306920763,
      "loss": 1.5075,
      "theoretical_loss": 3.567407649741108,
      "tokens_seen": 1394409472
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005833901705115346,
      "loss": 1.3993,
      "theoretical_loss": 3.567392779089919,
      "tokens_seen": 1394475008
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000583370110330993,
      "loss": 1.4631,
      "theoretical_loss": 3.567377909333263,
      "tokens_seen": 1394540544
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005833500501504513,
      "loss": 1.5188,
      "theoretical_loss": 3.5673630404710432,
      "tokens_seen": 1394606080
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005833299899699097,
      "loss": 1.3751,
      "theoretical_loss": 3.567348172503164,
      "tokens_seen": 1394671616
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005833099297893682,
      "loss": 1.5286,
      "theoretical_loss": 3.56733330542953,
      "tokens_seen": 1394737152
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005832898696088264,
      "loss": 1.4316,
      "theoretical_loss": 3.5673184392500454,
      "tokens_seen": 1394802688
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005832698094282849,
      "loss": 1.4805,
      "theoretical_loss": 3.567303573964614,
      "tokens_seen": 1394868224
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005832497492477433,
      "loss": 1.55,
      "theoretical_loss": 3.5672887095731403,
      "tokens_seen": 1394933760
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005832296890672016,
      "loss": 1.376,
      "theoretical_loss": 3.567273846075528,
      "tokens_seen": 1394999296
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00058320962888666,
      "loss": 1.5191,
      "theoretical_loss": 3.5672589834716817,
      "tokens_seen": 1395064832
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005831895687061183,
      "loss": 1.3674,
      "theoretical_loss": 3.5672441217615063,
      "tokens_seen": 1395130368
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005831695085255767,
      "loss": 1.4857,
      "theoretical_loss": 3.567229260944905,
      "tokens_seen": 1395195904
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005831494483450352,
      "loss": 1.4685,
      "theoretical_loss": 3.567214401021783,
      "tokens_seen": 1395261440
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005831293881644934,
      "loss": 1.4219,
      "theoretical_loss": 3.5671995419920446,
      "tokens_seen": 1395326976
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005831093279839519,
      "loss": 1.6682,
      "theoretical_loss": 3.5671846838555936,
      "tokens_seen": 1395392512
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005830892678034102,
      "loss": 1.4162,
      "theoretical_loss": 3.567169826612334,
      "tokens_seen": 1395458048
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005830692076228686,
      "loss": 1.4619,
      "theoretical_loss": 3.5671549702621714,
      "tokens_seen": 1395523584
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000583049147442327,
      "loss": 1.4855,
      "theoretical_loss": 3.5671401148050093,
      "tokens_seen": 1395589120
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005830290872617854,
      "loss": 1.4561,
      "theoretical_loss": 3.567125260240752,
      "tokens_seen": 1395654656
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005830090270812437,
      "loss": 1.4586,
      "theoretical_loss": 3.5671104065693044,
      "tokens_seen": 1395720192
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005829889669007022,
      "loss": 1.3456,
      "theoretical_loss": 3.567095553790571,
      "tokens_seen": 1395785728
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005829689067201604,
      "loss": 1.45,
      "theoretical_loss": 3.5670807019044553,
      "tokens_seen": 1395851264
    },
    {
      "epoch": 0.42,
      "objective/train/advantage_avg": 0.49091631174087524,
      "objective/train/docs_used": 509897,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2668709754943848,
      "objective/train/original_loss": 1.2668712139129639,
      "objective/train/theoretical_loss": 3.5670658509108626,
      "objective/train/tokens_used": 1416376800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24198754131793976,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503218173980713,
      "objective/train/weighted_lm_loss": 1.3307123184204102,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.5670658509108626,
      "tokens_seen": 1395916800
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005829488465396189,
      "loss": 1.3941,
      "theoretical_loss": 3.5670658509108626,
      "tokens_seen": 1395916800
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005829287863590772,
      "loss": 1.5917,
      "theoretical_loss": 3.5670510008096974,
      "tokens_seen": 1395982336
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005829087261785356,
      "loss": 1.5042,
      "theoretical_loss": 3.5670361516008637,
      "tokens_seen": 1396047872
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000582888665997994,
      "loss": 1.3852,
      "theoretical_loss": 3.5670213032842666,
      "tokens_seen": 1396113408
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005828686058174524,
      "loss": 1.5274,
      "theoretical_loss": 3.5670064558598096,
      "tokens_seen": 1396178944
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005828485456369107,
      "loss": 1.3984,
      "theoretical_loss": 3.5669916093273986,
      "tokens_seen": 1396244480
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005828284854563691,
      "loss": 1.4495,
      "theoretical_loss": 3.566976763686937,
      "tokens_seen": 1396310016
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005828084252758274,
      "loss": 1.5588,
      "theoretical_loss": 3.56696191893833,
      "tokens_seen": 1396375552
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005827883650952859,
      "loss": 1.3947,
      "theoretical_loss": 3.5669470750814813,
      "tokens_seen": 1396441088
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005827683049147442,
      "loss": 1.4863,
      "theoretical_loss": 3.566932232116297,
      "tokens_seen": 1396506624
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005827482447342026,
      "loss": 1.4083,
      "theoretical_loss": 3.56691739004268,
      "tokens_seen": 1396572160
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005827281845536609,
      "loss": 1.5423,
      "theoretical_loss": 3.5669025488605364,
      "tokens_seen": 1396637696
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005827081243731194,
      "loss": 1.4198,
      "theoretical_loss": 3.5668877085697694,
      "tokens_seen": 1396703232
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005826880641925777,
      "loss": 1.4935,
      "theoretical_loss": 3.566872869170285,
      "tokens_seen": 1396768768
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005826680040120361,
      "loss": 1.5335,
      "theoretical_loss": 3.5668580306619875,
      "tokens_seen": 1396834304
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005826479438314945,
      "loss": 1.48,
      "theoretical_loss": 3.5668431930447806,
      "tokens_seen": 1396899840
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005826278836509529,
      "loss": 1.5093,
      "theoretical_loss": 3.5668283563185703,
      "tokens_seen": 1396965376
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005826078234704112,
      "loss": 1.3484,
      "theoretical_loss": 3.5668135204832603,
      "tokens_seen": 1397030912
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005825877632898696,
      "loss": 1.3792,
      "theoretical_loss": 3.566798685538756,
      "tokens_seen": 1397096448
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000582567703109328,
      "loss": 1.589,
      "theoretical_loss": 3.5667838514849617,
      "tokens_seen": 1397161984
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005825476429287864,
      "loss": 1.4142,
      "theoretical_loss": 3.566769018321782,
      "tokens_seen": 1397227520
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005825275827482448,
      "loss": 1.4867,
      "theoretical_loss": 3.5667541860491223,
      "tokens_seen": 1397293056
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005825075225677031,
      "loss": 1.5171,
      "theoretical_loss": 3.566739354666887,
      "tokens_seen": 1397358592
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005824874623871616,
      "loss": 1.6021,
      "theoretical_loss": 3.566724524174981,
      "tokens_seen": 1397424128
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005824674022066198,
      "loss": 1.4372,
      "theoretical_loss": 3.5667096945733086,
      "tokens_seen": 1397489664
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005824473420260783,
      "loss": 1.3924,
      "theoretical_loss": 3.566694865861775,
      "tokens_seen": 1397555200
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005824272818455367,
      "loss": 1.4367,
      "theoretical_loss": 3.5666800380402854,
      "tokens_seen": 1397620736
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000582407221664995,
      "loss": 1.4439,
      "theoretical_loss": 3.5666652111087447,
      "tokens_seen": 1397686272
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005823871614844534,
      "loss": 1.4635,
      "theoretical_loss": 3.566650385067057,
      "tokens_seen": 1397751808
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005823671013039117,
      "loss": 1.4255,
      "theoretical_loss": 3.5666355599151274,
      "tokens_seen": 1397817344
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005823470411233701,
      "loss": 1.548,
      "theoretical_loss": 3.566620735652861,
      "tokens_seen": 1397882880
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005823269809428286,
      "loss": 1.4199,
      "theoretical_loss": 3.5666059122801625,
      "tokens_seen": 1397948416
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005823069207622868,
      "loss": 1.3757,
      "theoretical_loss": 3.5665910897969377,
      "tokens_seen": 1398013952
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005822868605817453,
      "loss": 1.446,
      "theoretical_loss": 3.56657626820309,
      "tokens_seen": 1398079488
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005822668004012037,
      "loss": 1.4498,
      "theoretical_loss": 3.5665614474985254,
      "tokens_seen": 1398145024
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000582246740220662,
      "loss": 1.5569,
      "theoretical_loss": 3.566546627683149,
      "tokens_seen": 1398210560
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005822266800401204,
      "loss": 1.513,
      "theoretical_loss": 3.5665318087568645,
      "tokens_seen": 1398276096
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005822066198595787,
      "loss": 1.5051,
      "theoretical_loss": 3.5665169907195784,
      "tokens_seen": 1398341632
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005821865596790371,
      "loss": 1.427,
      "theoretical_loss": 3.5665021735711946,
      "tokens_seen": 1398407168
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005821664994984956,
      "loss": 1.4952,
      "theoretical_loss": 3.566487357311619,
      "tokens_seen": 1398472704
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005821464393179538,
      "loss": 1.4328,
      "theoretical_loss": 3.5664725419407564,
      "tokens_seen": 1398538240
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005821263791374123,
      "loss": 1.5028,
      "theoretical_loss": 3.566457727458512,
      "tokens_seen": 1398603776
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005821063189568706,
      "loss": 1.4275,
      "theoretical_loss": 3.5664429138647895,
      "tokens_seen": 1398669312
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000582086258776329,
      "loss": 1.3987,
      "theoretical_loss": 3.566428101159496,
      "tokens_seen": 1398734848
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005820661985957874,
      "loss": 1.4942,
      "theoretical_loss": 3.566413289342535,
      "tokens_seen": 1398800384
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005820461384152458,
      "loss": 1.4392,
      "theoretical_loss": 3.566398478413813,
      "tokens_seen": 1398865920
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005820260782347041,
      "loss": 1.5099,
      "theoretical_loss": 3.5663836683732333,
      "tokens_seen": 1398931456
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005820060180541626,
      "loss": 1.364,
      "theoretical_loss": 3.5663688592207032,
      "tokens_seen": 1398996992
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005819859578736208,
      "loss": 1.4814,
      "theoretical_loss": 3.566354050956126,
      "tokens_seen": 1399062528
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005819658976930793,
      "loss": 1.3986,
      "theoretical_loss": 3.5663392435794083,
      "tokens_seen": 1399128064
    },
    {
      "epoch": 0.42,
      "objective/train/advantage_avg": 0.4922877550125122,
      "objective/train/docs_used": 511124,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.389640212059021,
      "objective/train/original_loss": 1.3896400928497314,
      "objective/train/theoretical_loss": 3.5663244370904543,
      "objective/train/tokens_used": 1419653600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24263879656791687,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050462245941162,
      "objective/train/weighted_lm_loss": 1.4596641063690186,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.5663244370904543,
      "tokens_seen": 1399193600
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005819458375125376,
      "loss": 1.4952,
      "theoretical_loss": 3.5663244370904543,
      "tokens_seen": 1399193600
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000581925777331996,
      "loss": 1.5068,
      "theoretical_loss": 3.566309631489169,
      "tokens_seen": 1399259136
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005819057171514544,
      "loss": 1.5421,
      "theoretical_loss": 3.566294826775459,
      "tokens_seen": 1399324672
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005818856569709128,
      "loss": 1.3134,
      "theoretical_loss": 3.566280022949228,
      "tokens_seen": 1399390208
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005818655967903711,
      "loss": 1.4867,
      "theoretical_loss": 3.566265220010382,
      "tokens_seen": 1399455744
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005818455366098295,
      "loss": 1.4002,
      "theoretical_loss": 3.5662504179588264,
      "tokens_seen": 1399521280
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005818254764292878,
      "loss": 1.4272,
      "theoretical_loss": 3.566235616794466,
      "tokens_seen": 1399586816
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005818054162487463,
      "loss": 1.4582,
      "theoretical_loss": 3.5662208165172067,
      "tokens_seen": 1399652352
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005817853560682046,
      "loss": 1.368,
      "theoretical_loss": 3.566206017126953,
      "tokens_seen": 1399717888
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000581765295887663,
      "loss": 1.4699,
      "theoretical_loss": 3.5661912186236107,
      "tokens_seen": 1399783424
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005817452357071213,
      "loss": 1.4522,
      "theoretical_loss": 3.566176421007085,
      "tokens_seen": 1399848960
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005817251755265798,
      "loss": 1.4571,
      "theoretical_loss": 3.5661616242772816,
      "tokens_seen": 1399914496
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005817051153460381,
      "loss": 1.4873,
      "theoretical_loss": 3.566146828434105,
      "tokens_seen": 1399980032
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005816850551654965,
      "loss": 1.3391,
      "theoretical_loss": 3.5661320334774613,
      "tokens_seen": 1400045568
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005816649949849549,
      "loss": 1.4206,
      "theoretical_loss": 3.566117239407256,
      "tokens_seen": 1400111104
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005816449348044133,
      "loss": 1.4773,
      "theoretical_loss": 3.566102446223394,
      "tokens_seen": 1400176640
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005816248746238716,
      "loss": 1.4924,
      "theoretical_loss": 3.566087653925781,
      "tokens_seen": 1400242176
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00058160481444333,
      "loss": 1.4984,
      "theoretical_loss": 3.5660728625143223,
      "tokens_seen": 1400307712
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005815847542627883,
      "loss": 1.4729,
      "theoretical_loss": 3.5660580719889237,
      "tokens_seen": 1400373248
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005815646940822468,
      "loss": 1.4828,
      "theoretical_loss": 3.56604328234949,
      "tokens_seen": 1400438784
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005815446339017051,
      "loss": 1.4355,
      "theoretical_loss": 3.566028493595927,
      "tokens_seen": 1400504320
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005815245737211635,
      "loss": 1.4749,
      "theoretical_loss": 3.5660137057281402,
      "tokens_seen": 1400569856
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005815045135406219,
      "loss": 1.6225,
      "theoretical_loss": 3.5659989187460353,
      "tokens_seen": 1400635392
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005814844533600802,
      "loss": 1.5072,
      "theoretical_loss": 3.5659841326495174,
      "tokens_seen": 1400700928
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005814643931795386,
      "loss": 1.564,
      "theoretical_loss": 3.5659693474384926,
      "tokens_seen": 1400766464
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005814443329989971,
      "loss": 1.6002,
      "theoretical_loss": 3.5659545631128657,
      "tokens_seen": 1400832000
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005814242728184553,
      "loss": 1.4549,
      "theoretical_loss": 3.5659397796725427,
      "tokens_seen": 1400897536
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005814042126379138,
      "loss": 1.4833,
      "theoretical_loss": 3.5659249971174294,
      "tokens_seen": 1400963072
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000581384152457372,
      "loss": 1.3624,
      "theoretical_loss": 3.565910215447431,
      "tokens_seen": 1401028608
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005813640922768305,
      "loss": 1.3408,
      "theoretical_loss": 3.565895434662454,
      "tokens_seen": 1401094144
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005813440320962889,
      "loss": 1.3437,
      "theoretical_loss": 3.565880654762402,
      "tokens_seen": 1401159680
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005813239719157472,
      "loss": 1.4558,
      "theoretical_loss": 3.5658658757471824,
      "tokens_seen": 1401225216
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005813039117352056,
      "loss": 1.4715,
      "theoretical_loss": 3.5658510976167004,
      "tokens_seen": 1401290752
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005812838515546641,
      "loss": 1.3538,
      "theoretical_loss": 3.5658363203708614,
      "tokens_seen": 1401356288
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005812637913741223,
      "loss": 1.3738,
      "theoretical_loss": 3.5658215440095717,
      "tokens_seen": 1401421824
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005812437311935808,
      "loss": 1.4233,
      "theoretical_loss": 3.5658067685327364,
      "tokens_seen": 1401487360
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000581223671013039,
      "loss": 1.4576,
      "theoretical_loss": 3.5657919939402616,
      "tokens_seen": 1401552896
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005812036108324975,
      "loss": 1.477,
      "theoretical_loss": 3.565777220232052,
      "tokens_seen": 1401618432
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005811835506519559,
      "loss": 1.4074,
      "theoretical_loss": 3.565762447408015,
      "tokens_seen": 1401683968
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005811634904714142,
      "loss": 1.435,
      "theoretical_loss": 3.565747675468055,
      "tokens_seen": 1401749504
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005811434302908726,
      "loss": 1.4865,
      "theoretical_loss": 3.5657329044120782,
      "tokens_seen": 1401815040
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000581123370110331,
      "loss": 1.5538,
      "theoretical_loss": 3.56571813423999,
      "tokens_seen": 1401880576
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005811033099297893,
      "loss": 1.4828,
      "theoretical_loss": 3.5657033649516974,
      "tokens_seen": 1401946112
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005810832497492478,
      "loss": 1.4809,
      "theoretical_loss": 3.565688596547105,
      "tokens_seen": 1402011648
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005810631895687062,
      "loss": 1.4387,
      "theoretical_loss": 3.565673829026119,
      "tokens_seen": 1402077184
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005810431293881645,
      "loss": 1.3263,
      "theoretical_loss": 3.5656590623886455,
      "tokens_seen": 1402142720
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000581023069207623,
      "loss": 1.4758,
      "theoretical_loss": 3.5656442966345905,
      "tokens_seen": 1402208256
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005810030090270812,
      "loss": 1.4037,
      "theoretical_loss": 3.5656295317638587,
      "tokens_seen": 1402273792
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005809829488465397,
      "loss": 1.5104,
      "theoretical_loss": 3.565614767776357,
      "tokens_seen": 1402339328
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000580962888665998,
      "loss": 1.6331,
      "theoretical_loss": 3.5656000046719916,
      "tokens_seen": 1402404864
    },
    {
      "epoch": 0.42,
      "objective/train/advantage_avg": 0.4945162236690521,
      "objective/train/docs_used": 512300,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.513278603553772,
      "objective/train/original_loss": 1.5132784843444824,
      "objective/train/theoretical_loss": 3.565585242450667,
      "objective/train/tokens_used": 1422930400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.244964599609375,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506969690322876,
      "objective/train/weighted_lm_loss": 1.5900007486343384,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.565585242450667,
      "tokens_seen": 1402470400
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0005809428284854564,
      "loss": 1.5836,
      "theoretical_loss": 3.565585242450667,
      "tokens_seen": 1402470400
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005809227683049148,
      "loss": 1.4586,
      "theoretical_loss": 3.5655704811122906,
      "tokens_seen": 1402535936
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005809027081243732,
      "loss": 1.3637,
      "theoretical_loss": 3.5655557206567674,
      "tokens_seen": 1402601472
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005808826479438315,
      "loss": 1.5372,
      "theoretical_loss": 3.565540961084004,
      "tokens_seen": 1402667008
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005808625877632899,
      "loss": 1.5706,
      "theoretical_loss": 3.5655262023939054,
      "tokens_seen": 1402732544
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005808425275827482,
      "loss": 1.3491,
      "theoretical_loss": 3.5655114445863783,
      "tokens_seen": 1402798080
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005808224674022067,
      "loss": 1.5642,
      "theoretical_loss": 3.565496687661329,
      "tokens_seen": 1402863616
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000580802407221665,
      "loss": 1.3514,
      "theoretical_loss": 3.5654819316186632,
      "tokens_seen": 1402929152
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005807823470411234,
      "loss": 1.394,
      "theoretical_loss": 3.5654671764582866,
      "tokens_seen": 1402994688
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005807622868605817,
      "loss": 1.4636,
      "theoretical_loss": 3.5654524221801056,
      "tokens_seen": 1403060224
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005807422266800402,
      "loss": 1.4468,
      "theoretical_loss": 3.5654376687840257,
      "tokens_seen": 1403125760
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005807221664994985,
      "loss": 1.4627,
      "theoretical_loss": 3.565422916269954,
      "tokens_seen": 1403191296
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005807021063189569,
      "loss": 1.4338,
      "theoretical_loss": 3.5654081646377955,
      "tokens_seen": 1403256832
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005806820461384153,
      "loss": 1.4394,
      "theoretical_loss": 3.565393413887457,
      "tokens_seen": 1403322368
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005806619859578737,
      "loss": 1.494,
      "theoretical_loss": 3.5653786640188443,
      "tokens_seen": 1403387904
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000580641925777332,
      "loss": 1.388,
      "theoretical_loss": 3.565363915031864,
      "tokens_seen": 1403453440
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005806218655967904,
      "loss": 1.4892,
      "theoretical_loss": 3.5653491669264215,
      "tokens_seen": 1403518976
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005806018054162487,
      "loss": 1.4348,
      "theoretical_loss": 3.565334419702423,
      "tokens_seen": 1403584512
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005805817452357072,
      "loss": 1.5337,
      "theoretical_loss": 3.5653196733597756,
      "tokens_seen": 1403650048
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005805616850551655,
      "loss": 1.3832,
      "theoretical_loss": 3.565304927898384,
      "tokens_seen": 1403715584
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005805416248746239,
      "loss": 1.5221,
      "theoretical_loss": 3.565290183318156,
      "tokens_seen": 1403781120
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005805215646940823,
      "loss": 1.4868,
      "theoretical_loss": 3.5652754396189965,
      "tokens_seen": 1403846656
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005805015045135406,
      "loss": 1.4545,
      "theoretical_loss": 3.5652606968008125,
      "tokens_seen": 1403912192
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000580481444332999,
      "loss": 1.4863,
      "theoretical_loss": 3.56524595486351,
      "tokens_seen": 1403977728
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005804613841524575,
      "loss": 1.3995,
      "theoretical_loss": 3.565231213806995,
      "tokens_seen": 1404043264
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005804413239719157,
      "loss": 1.3972,
      "theoretical_loss": 3.5652164736311747,
      "tokens_seen": 1404108800
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005804212637913742,
      "loss": 1.4173,
      "theoretical_loss": 3.565201734335954,
      "tokens_seen": 1404174336
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005804012036108324,
      "loss": 1.4388,
      "theoretical_loss": 3.5651869959212403,
      "tokens_seen": 1404239872
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005803811434302909,
      "loss": 1.4495,
      "theoretical_loss": 3.5651722583869394,
      "tokens_seen": 1404305408
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005803610832497493,
      "loss": 1.4775,
      "theoretical_loss": 3.5651575217329574,
      "tokens_seen": 1404370944
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005803410230692076,
      "loss": 1.5152,
      "theoretical_loss": 3.5651427859592015,
      "tokens_seen": 1404436480
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000580320962888666,
      "loss": 1.4693,
      "theoretical_loss": 3.565128051065577,
      "tokens_seen": 1404502016
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005803009027081245,
      "loss": 1.5103,
      "theoretical_loss": 3.565113317051991,
      "tokens_seen": 1404567552
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005802808425275827,
      "loss": 1.4628,
      "theoretical_loss": 3.56509858391835,
      "tokens_seen": 1404633088
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005802607823470412,
      "loss": 1.4219,
      "theoretical_loss": 3.5650838516645598,
      "tokens_seen": 1404698624
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005802407221664994,
      "loss": 1.372,
      "theoretical_loss": 3.565069120290527,
      "tokens_seen": 1404764160
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005802206619859579,
      "loss": 1.5156,
      "theoretical_loss": 3.5650543897961584,
      "tokens_seen": 1404829696
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005802006018054163,
      "loss": 1.4595,
      "theoretical_loss": 3.56503966018136,
      "tokens_seen": 1404895232
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005801805416248746,
      "loss": 1.3663,
      "theoretical_loss": 3.5650249314460383,
      "tokens_seen": 1404960768
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000580160481444333,
      "loss": 1.54,
      "theoretical_loss": 3.5650102035901,
      "tokens_seen": 1405026304
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005801404212637914,
      "loss": 1.5996,
      "theoretical_loss": 3.5649954766134515,
      "tokens_seen": 1405091840
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005801203610832497,
      "loss": 1.5611,
      "theoretical_loss": 3.564980750515999,
      "tokens_seen": 1405157376
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005801003009027082,
      "loss": 1.4371,
      "theoretical_loss": 3.5649660252976494,
      "tokens_seen": 1405222912
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005800802407221664,
      "loss": 1.5265,
      "theoretical_loss": 3.5649513009583087,
      "tokens_seen": 1405288448
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005800601805416249,
      "loss": 1.4516,
      "theoretical_loss": 3.5649365774978845,
      "tokens_seen": 1405353984
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005800401203610832,
      "loss": 1.4831,
      "theoretical_loss": 3.5649218549162818,
      "tokens_seen": 1405419520
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005800200601805416,
      "loss": 1.4867,
      "theoretical_loss": 3.564907133213409,
      "tokens_seen": 1405485056
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00058,
      "loss": 1.4033,
      "theoretical_loss": 3.564892412389171,
      "tokens_seen": 1405550592
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005799799398194584,
      "loss": 1.3901,
      "theoretical_loss": 3.5648776924434755,
      "tokens_seen": 1405616128
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005799598796389167,
      "loss": 1.6335,
      "theoretical_loss": 3.5648629733762283,
      "tokens_seen": 1405681664
    },
    {
      "epoch": 0.43,
      "objective/train/advantage_avg": 0.49203693866729736,
      "objective/train/docs_used": 513528,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3456885814666748,
      "objective/train/original_loss": 1.3456885814666748,
      "objective/train/theoretical_loss": 3.5648482551873366,
      "objective/train/tokens_used": 1426207200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24253125488758087,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504366159439087,
      "objective/train/weighted_lm_loss": 1.4134820699691772,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5648482551873366,
      "tokens_seen": 1405747200
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005799398194583752,
      "loss": 1.4601,
      "theoretical_loss": 3.5648482551873366,
      "tokens_seen": 1405747200
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005799197592778335,
      "loss": 1.5433,
      "theoretical_loss": 3.5648335378767078,
      "tokens_seen": 1405812736
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005798996990972919,
      "loss": 1.3582,
      "theoretical_loss": 3.5648188214442467,
      "tokens_seen": 1405878272
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005798796389167502,
      "loss": 1.5025,
      "theoretical_loss": 3.564804105889861,
      "tokens_seen": 1405943808
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005798595787362086,
      "loss": 1.4896,
      "theoretical_loss": 3.5647893912134574,
      "tokens_seen": 1406009344
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000579839518555667,
      "loss": 1.5129,
      "theoretical_loss": 3.5647746774149427,
      "tokens_seen": 1406074880
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005798194583751254,
      "loss": 1.541,
      "theoretical_loss": 3.5647599644942227,
      "tokens_seen": 1406140416
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005797993981945837,
      "loss": 1.2881,
      "theoretical_loss": 3.5647452524512055,
      "tokens_seen": 1406205952
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005797793380140421,
      "loss": 1.5706,
      "theoretical_loss": 3.564730541285797,
      "tokens_seen": 1406271488
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005797592778335005,
      "loss": 1.5366,
      "theoretical_loss": 3.564715830997904,
      "tokens_seen": 1406337024
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005797392176529589,
      "loss": 1.4759,
      "theoretical_loss": 3.564701121587434,
      "tokens_seen": 1406402560
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005797191574724172,
      "loss": 1.4488,
      "theoretical_loss": 3.564686413054292,
      "tokens_seen": 1406468096
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005796990972918757,
      "loss": 1.3495,
      "theoretical_loss": 3.5646717053983865,
      "tokens_seen": 1406533632
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000579679037111334,
      "loss": 1.3975,
      "theoretical_loss": 3.5646569986196246,
      "tokens_seen": 1406599168
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005796589769307924,
      "loss": 1.3212,
      "theoretical_loss": 3.5646422927179113,
      "tokens_seen": 1406664704
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005796389167502507,
      "loss": 1.4491,
      "theoretical_loss": 3.5646275876931544,
      "tokens_seen": 1406730240
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005796188565697091,
      "loss": 1.3665,
      "theoretical_loss": 3.5646128835452613,
      "tokens_seen": 1406795776
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005795987963891675,
      "loss": 1.5313,
      "theoretical_loss": 3.564598180274138,
      "tokens_seen": 1406861312
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005795787362086259,
      "loss": 1.3914,
      "theoretical_loss": 3.564583477879692,
      "tokens_seen": 1406926848
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005795586760280843,
      "loss": 1.4499,
      "theoretical_loss": 3.5645687763618303,
      "tokens_seen": 1406992384
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005795386158475427,
      "loss": 1.4993,
      "theoretical_loss": 3.5645540757204586,
      "tokens_seen": 1407057920
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000579518555667001,
      "loss": 1.4681,
      "theoretical_loss": 3.5645393759554853,
      "tokens_seen": 1407123456
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005794984954864594,
      "loss": 1.4066,
      "theoretical_loss": 3.5645246770668164,
      "tokens_seen": 1407188992
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005794784353059179,
      "loss": 1.4741,
      "theoretical_loss": 3.564509979054359,
      "tokens_seen": 1407254528
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005794583751253761,
      "loss": 1.5372,
      "theoretical_loss": 3.5644952819180205,
      "tokens_seen": 1407320064
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005794383149448346,
      "loss": 1.4449,
      "theoretical_loss": 3.5644805856577078,
      "tokens_seen": 1407385600
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005794182547642928,
      "loss": 1.4496,
      "theoretical_loss": 3.5644658902733273,
      "tokens_seen": 1407451136
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005793981945837513,
      "loss": 1.6061,
      "theoretical_loss": 3.5644511957647866,
      "tokens_seen": 1407516672
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005793781344032097,
      "loss": 1.3316,
      "theoretical_loss": 3.5644365021319926,
      "tokens_seen": 1407582208
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000579358074222668,
      "loss": 1.5168,
      "theoretical_loss": 3.564421809374852,
      "tokens_seen": 1407647744
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005793380140421264,
      "loss": 1.4264,
      "theoretical_loss": 3.564407117493272,
      "tokens_seen": 1407713280
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005793179538615849,
      "loss": 1.3774,
      "theoretical_loss": 3.56439242648716,
      "tokens_seen": 1407778816
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005792978936810431,
      "loss": 1.4317,
      "theoretical_loss": 3.5643777363564233,
      "tokens_seen": 1407844352
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005792778335005016,
      "loss": 1.4557,
      "theoretical_loss": 3.5643630471009677,
      "tokens_seen": 1407909888
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005792577733199598,
      "loss": 1.4636,
      "theoretical_loss": 3.564348358720702,
      "tokens_seen": 1407975424
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005792377131394183,
      "loss": 1.4685,
      "theoretical_loss": 3.5643336712155325,
      "tokens_seen": 1408040960
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005792176529588767,
      "loss": 1.6001,
      "theoretical_loss": 3.564318984585366,
      "tokens_seen": 1408106496
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000579197592778335,
      "loss": 1.4226,
      "theoretical_loss": 3.56430429883011,
      "tokens_seen": 1408172032
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005791775325977934,
      "loss": 1.5666,
      "theoretical_loss": 3.564289613949671,
      "tokens_seen": 1408237568
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005791574724172518,
      "loss": 1.421,
      "theoretical_loss": 3.5642749299439576,
      "tokens_seen": 1408303104
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005791374122367101,
      "loss": 1.5574,
      "theoretical_loss": 3.5642602468128763,
      "tokens_seen": 1408368640
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005791173520561686,
      "loss": 1.5254,
      "theoretical_loss": 3.5642455645563342,
      "tokens_seen": 1408434176
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005790972918756268,
      "loss": 1.4225,
      "theoretical_loss": 3.5642308831742384,
      "tokens_seen": 1408499712
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005790772316950853,
      "loss": 1.5672,
      "theoretical_loss": 3.564216202666496,
      "tokens_seen": 1408565248
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005790571715145436,
      "loss": 1.3923,
      "theoretical_loss": 3.564201523033015,
      "tokens_seen": 1408630784
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000579037111334002,
      "loss": 1.4921,
      "theoretical_loss": 3.564186844273702,
      "tokens_seen": 1408696320
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005790170511534604,
      "loss": 1.3831,
      "theoretical_loss": 3.564172166388465,
      "tokens_seen": 1408761856
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005789969909729188,
      "loss": 1.4728,
      "theoretical_loss": 3.5641574893772097,
      "tokens_seen": 1408827392
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005789769307923771,
      "loss": 1.3497,
      "theoretical_loss": 3.564142813239845,
      "tokens_seen": 1408892928
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005789568706118356,
      "loss": 1.4318,
      "theoretical_loss": 3.5641281379762777,
      "tokens_seen": 1408958464
    },
    {
      "epoch": 0.43,
      "objective/train/advantage_avg": 0.49372395873069763,
      "objective/train/docs_used": 514827,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5697238445281982,
      "objective/train/original_loss": 1.5697238445281982,
      "objective/train/theoretical_loss": 3.5641134635864153,
      "objective/train/tokens_used": 1429484000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24396269023418427,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506128072738647,
      "objective/train/weighted_lm_loss": 1.6490398645401,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0100501775741577,
      "theoretical_loss": 3.5641134635864153,
      "tokens_seen": 1409024000
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005789368104312939,
      "loss": 1.5819,
      "theoretical_loss": 3.5641134635864153,
      "tokens_seen": 1409024000
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005789167502507523,
      "loss": 1.4348,
      "theoretical_loss": 3.5640987900701653,
      "tokens_seen": 1409089536
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005788966900702106,
      "loss": 1.4633,
      "theoretical_loss": 3.5640841174274343,
      "tokens_seen": 1409155072
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000578876629889669,
      "loss": 1.4048,
      "theoretical_loss": 3.56406944565813,
      "tokens_seen": 1409220608
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005788565697091274,
      "loss": 1.4194,
      "theoretical_loss": 3.56405477476216,
      "tokens_seen": 1409286144
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005788365095285858,
      "loss": 1.4911,
      "theoretical_loss": 3.5640401047394317,
      "tokens_seen": 1409351680
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005788164493480441,
      "loss": 1.4808,
      "theoretical_loss": 3.5640254355898526,
      "tokens_seen": 1409417216
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005787963891675025,
      "loss": 1.3952,
      "theoretical_loss": 3.56401076731333,
      "tokens_seen": 1409482752
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005787763289869609,
      "loss": 1.5065,
      "theoretical_loss": 3.563996099909771,
      "tokens_seen": 1409548288
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005787562688064193,
      "loss": 1.4698,
      "theoretical_loss": 3.563981433379084,
      "tokens_seen": 1409613824
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005787362086258776,
      "loss": 1.4959,
      "theoretical_loss": 3.5639667677211753,
      "tokens_seen": 1409679360
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000578716148445336,
      "loss": 1.3926,
      "theoretical_loss": 3.563952102935953,
      "tokens_seen": 1409744896
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005786960882647944,
      "loss": 1.6432,
      "theoretical_loss": 3.5639374390233245,
      "tokens_seen": 1409810432
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005786760280842528,
      "loss": 1.4375,
      "theoretical_loss": 3.5639227759831975,
      "tokens_seen": 1409875968
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005786559679037111,
      "loss": 1.5064,
      "theoretical_loss": 3.5639081138154793,
      "tokens_seen": 1409941504
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005786359077231695,
      "loss": 1.3938,
      "theoretical_loss": 3.563893452520078,
      "tokens_seen": 1410007040
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005786158475426279,
      "loss": 1.4843,
      "theoretical_loss": 3.563878792096901,
      "tokens_seen": 1410072576
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005785957873620863,
      "loss": 1.392,
      "theoretical_loss": 3.563864132545855,
      "tokens_seen": 1410138112
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005785757271815446,
      "loss": 1.4522,
      "theoretical_loss": 3.5638494738668482,
      "tokens_seen": 1410203648
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005785556670010031,
      "loss": 1.5532,
      "theoretical_loss": 3.563834816059788,
      "tokens_seen": 1410269184
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005785356068204613,
      "loss": 1.4362,
      "theoretical_loss": 3.5638201591245826,
      "tokens_seen": 1410334720
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005785155466399198,
      "loss": 1.451,
      "theoretical_loss": 3.563805503061139,
      "tokens_seen": 1410400256
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005784954864593781,
      "loss": 1.3773,
      "theoretical_loss": 3.563790847869365,
      "tokens_seen": 1410465792
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005784754262788365,
      "loss": 1.4432,
      "theoretical_loss": 3.5637761935491685,
      "tokens_seen": 1410531328
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005784553660982949,
      "loss": 1.4703,
      "theoretical_loss": 3.563761540100457,
      "tokens_seen": 1410596864
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005784353059177532,
      "loss": 1.6067,
      "theoretical_loss": 3.5637468875231386,
      "tokens_seen": 1410662400
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005784152457372116,
      "loss": 1.4731,
      "theoretical_loss": 3.5637322358171204,
      "tokens_seen": 1410727936
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005783951855566701,
      "loss": 1.4953,
      "theoretical_loss": 3.56371758498231,
      "tokens_seen": 1410793472
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005783751253761283,
      "loss": 1.3853,
      "theoretical_loss": 3.5637029350186156,
      "tokens_seen": 1410859008
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005783550651955868,
      "loss": 1.5292,
      "theoretical_loss": 3.563688285925945,
      "tokens_seen": 1410924544
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005783350050150452,
      "loss": 1.5417,
      "theoretical_loss": 3.5636736377042055,
      "tokens_seen": 1410990080
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005783149448345035,
      "loss": 1.3872,
      "theoretical_loss": 3.5636589903533054,
      "tokens_seen": 1411055616
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005782948846539619,
      "loss": 1.4531,
      "theoretical_loss": 3.563644343873152,
      "tokens_seen": 1411121152
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005782748244734202,
      "loss": 1.5289,
      "theoretical_loss": 3.5636296982636537,
      "tokens_seen": 1411186688
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005782547642928786,
      "loss": 1.4571,
      "theoretical_loss": 3.5636150535247175,
      "tokens_seen": 1411252224
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005782347041123371,
      "loss": 1.5861,
      "theoretical_loss": 3.5636004096562512,
      "tokens_seen": 1411317760
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005782146439317953,
      "loss": 1.5481,
      "theoretical_loss": 3.5635857666581643,
      "tokens_seen": 1411383296
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005781945837512538,
      "loss": 1.502,
      "theoretical_loss": 3.5635711245303625,
      "tokens_seen": 1411448832
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000578174523570712,
      "loss": 1.4765,
      "theoretical_loss": 3.5635564832727553,
      "tokens_seen": 1411514368
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005781544633901705,
      "loss": 1.5307,
      "theoretical_loss": 3.5635418428852494,
      "tokens_seen": 1411579904
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005781344032096289,
      "loss": 1.4599,
      "theoretical_loss": 3.5635272033677534,
      "tokens_seen": 1411645440
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005781143430290872,
      "loss": 1.4674,
      "theoretical_loss": 3.563512564720175,
      "tokens_seen": 1411710976
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005780942828485456,
      "loss": 1.7263,
      "theoretical_loss": 3.5634979269424223,
      "tokens_seen": 1411776512
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000578074222668004,
      "loss": 1.447,
      "theoretical_loss": 3.563483290034403,
      "tokens_seen": 1411842048
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005780541624874624,
      "loss": 1.4327,
      "theoretical_loss": 3.5634686539960247,
      "tokens_seen": 1411907584
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005780341023069208,
      "loss": 1.4683,
      "theoretical_loss": 3.563454018827197,
      "tokens_seen": 1411973120
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005780140421263792,
      "loss": 1.5809,
      "theoretical_loss": 3.5634393845278254,
      "tokens_seen": 1412038656
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005779939819458375,
      "loss": 1.5102,
      "theoretical_loss": 3.56342475109782,
      "tokens_seen": 1412104192
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000577973921765296,
      "loss": 1.4274,
      "theoretical_loss": 3.5634101185370874,
      "tokens_seen": 1412169728
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005779538615847543,
      "loss": 1.3563,
      "theoretical_loss": 3.5633954868455366,
      "tokens_seen": 1412235264
    },
    {
      "epoch": 0.43,
      "objective/train/advantage_avg": 0.49203649163246155,
      "objective/train/docs_used": 516049,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.447616696357727,
      "objective/train/original_loss": 1.4476168155670166,
      "objective/train/theoretical_loss": 3.563380856023075,
      "objective/train/tokens_used": 1432760800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24242514371871948,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504359006881714,
      "objective/train/weighted_lm_loss": 1.5205994844436646,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.563380856023075,
      "tokens_seen": 1412300800
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005779338014042127,
      "loss": 1.4794,
      "theoretical_loss": 3.563380856023075,
      "tokens_seen": 1412300800
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000577913741223671,
      "loss": 1.5181,
      "theoretical_loss": 3.5633662260696113,
      "tokens_seen": 1412366336
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005778936810431294,
      "loss": 1.4964,
      "theoretical_loss": 3.5633515969850533,
      "tokens_seen": 1412431872
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005778736208625878,
      "loss": 1.4714,
      "theoretical_loss": 3.563336968769309,
      "tokens_seen": 1412497408
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005778535606820462,
      "loss": 1.4524,
      "theoretical_loss": 3.563322341422286,
      "tokens_seen": 1412562944
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005778335005015045,
      "loss": 1.3953,
      "theoretical_loss": 3.5633077149438925,
      "tokens_seen": 1412628480
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005778134403209629,
      "loss": 1.5673,
      "theoretical_loss": 3.5632930893340378,
      "tokens_seen": 1412694016
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005777933801404213,
      "loss": 1.4732,
      "theoretical_loss": 3.5632784645926288,
      "tokens_seen": 1412759552
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005777733199598797,
      "loss": 1.5347,
      "theoretical_loss": 3.563263840719575,
      "tokens_seen": 1412825088
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000577753259779338,
      "loss": 1.4398,
      "theoretical_loss": 3.563249217714782,
      "tokens_seen": 1412890624
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005777331995987965,
      "loss": 1.5018,
      "theoretical_loss": 3.5632345955781606,
      "tokens_seen": 1412956160
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005777131394182548,
      "loss": 1.4515,
      "theoretical_loss": 3.563219974309618,
      "tokens_seen": 1413021696
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005776930792377132,
      "loss": 1.469,
      "theoretical_loss": 3.563205353909063,
      "tokens_seen": 1413087232
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005776730190571715,
      "loss": 1.4785,
      "theoretical_loss": 3.5631907343764024,
      "tokens_seen": 1413152768
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005776529588766299,
      "loss": 1.3558,
      "theoretical_loss": 3.5631761157115456,
      "tokens_seen": 1413218304
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005776328986960883,
      "loss": 1.4878,
      "theoretical_loss": 3.5631614979144004,
      "tokens_seen": 1413283840
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005776128385155467,
      "loss": 1.3162,
      "theoretical_loss": 3.563146880984875,
      "tokens_seen": 1413349376
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000577592778335005,
      "loss": 1.5159,
      "theoretical_loss": 3.563132264922878,
      "tokens_seen": 1413414912
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005775727181544635,
      "loss": 1.4086,
      "theoretical_loss": 3.5631176497283175,
      "tokens_seen": 1413480448
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005775526579739217,
      "loss": 1.4658,
      "theoretical_loss": 3.5631030354011024,
      "tokens_seen": 1413545984
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005775325977933802,
      "loss": 1.427,
      "theoretical_loss": 3.56308842194114,
      "tokens_seen": 1413611520
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005775125376128385,
      "loss": 1.5152,
      "theoretical_loss": 3.563073809348339,
      "tokens_seen": 1413677056
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005774924774322969,
      "loss": 1.5668,
      "theoretical_loss": 3.563059197622608,
      "tokens_seen": 1413742592
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005774724172517553,
      "loss": 1.5643,
      "theoretical_loss": 3.5630445867638554,
      "tokens_seen": 1413808128
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005774523570712136,
      "loss": 1.5179,
      "theoretical_loss": 3.5630299767719893,
      "tokens_seen": 1413873664
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000577432296890672,
      "loss": 1.5103,
      "theoretical_loss": 3.5630153676469183,
      "tokens_seen": 1413939200
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005774122367101305,
      "loss": 1.3629,
      "theoretical_loss": 3.56300075938855,
      "tokens_seen": 1414004736
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005773921765295887,
      "loss": 1.5108,
      "theoretical_loss": 3.5629861519967942,
      "tokens_seen": 1414070272
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005773721163490472,
      "loss": 1.3964,
      "theoretical_loss": 3.5629715454715587,
      "tokens_seen": 1414135808
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005773520561685056,
      "loss": 1.4997,
      "theoretical_loss": 3.5629569398127514,
      "tokens_seen": 1414201344
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005773319959879639,
      "loss": 1.5422,
      "theoretical_loss": 3.5629423350202813,
      "tokens_seen": 1414266880
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005773119358074223,
      "loss": 1.5763,
      "theoretical_loss": 3.562927731094057,
      "tokens_seen": 1414332416
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005772918756268806,
      "loss": 1.4302,
      "theoretical_loss": 3.562913128033987,
      "tokens_seen": 1414397952
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000577271815446339,
      "loss": 1.3822,
      "theoretical_loss": 3.5628985258399792,
      "tokens_seen": 1414463488
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005772517552657975,
      "loss": 1.4561,
      "theoretical_loss": 3.5628839245119424,
      "tokens_seen": 1414529024
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005772316950852557,
      "loss": 1.5569,
      "theoretical_loss": 3.5628693240497853,
      "tokens_seen": 1414594560
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005772116349047142,
      "loss": 1.5803,
      "theoretical_loss": 3.5628547244534166,
      "tokens_seen": 1414660096
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005771915747241725,
      "loss": 1.5216,
      "theoretical_loss": 3.5628401257227447,
      "tokens_seen": 1414725632
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005771715145436309,
      "loss": 1.4238,
      "theoretical_loss": 3.5628255278576777,
      "tokens_seen": 1414791168
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005771514543630893,
      "loss": 1.5147,
      "theoretical_loss": 3.5628109308581246,
      "tokens_seen": 1414856704
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005771313941825476,
      "loss": 1.4387,
      "theoretical_loss": 3.562796334723994,
      "tokens_seen": 1414922240
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000577111334002006,
      "loss": 1.3763,
      "theoretical_loss": 3.5627817394551946,
      "tokens_seen": 1414987776
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005770912738214644,
      "loss": 1.624,
      "theoretical_loss": 3.5627671450516347,
      "tokens_seen": 1415053312
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005770712136409227,
      "loss": 1.4091,
      "theoretical_loss": 3.5627525515132232,
      "tokens_seen": 1415118848
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005770511534603812,
      "loss": 1.4783,
      "theoretical_loss": 3.562737958839869,
      "tokens_seen": 1415184384
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005770310932798395,
      "loss": 1.5482,
      "theoretical_loss": 3.56272336703148,
      "tokens_seen": 1415249920
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005770110330992979,
      "loss": 1.5445,
      "theoretical_loss": 3.5627087760879657,
      "tokens_seen": 1415315456
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005769909729187563,
      "loss": 1.4579,
      "theoretical_loss": 3.5626941860092343,
      "tokens_seen": 1415380992
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005769709127382147,
      "loss": 1.4713,
      "theoretical_loss": 3.5626795967951947,
      "tokens_seen": 1415446528
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000576950852557673,
      "loss": 1.5633,
      "theoretical_loss": 3.562665008445755,
      "tokens_seen": 1415512064
    },
    {
      "epoch": 0.43,
      "objective/train/advantage_avg": 0.49415871500968933,
      "objective/train/docs_used": 517252,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4566549062728882,
      "objective/train/original_loss": 1.4566549062728882,
      "objective/train/theoretical_loss": 3.5626504209608254,
      "objective/train/tokens_used": 1436037600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24447205662727356,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506587028503418,
      "objective/train/weighted_lm_loss": 1.5303364992141724,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5626504209608254,
      "tokens_seen": 1415577600
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005769307923771314,
      "loss": 1.5125,
      "theoretical_loss": 3.5626504209608254,
      "tokens_seen": 1415577600
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005769107321965897,
      "loss": 1.31,
      "theoretical_loss": 3.5626358343403135,
      "tokens_seen": 1415643136
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005768906720160482,
      "loss": 1.3653,
      "theoretical_loss": 3.562621248584128,
      "tokens_seen": 1415708672
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005768706118355065,
      "loss": 1.463,
      "theoretical_loss": 3.5626066636921783,
      "tokens_seen": 1415774208
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005768505516549649,
      "loss": 1.4763,
      "theoretical_loss": 3.5625920796643724,
      "tokens_seen": 1415839744
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005768304914744232,
      "loss": 1.3948,
      "theoretical_loss": 3.56257749650062,
      "tokens_seen": 1415905280
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005768104312938817,
      "loss": 1.6824,
      "theoretical_loss": 3.5625629142008295,
      "tokens_seen": 1415970816
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00057679037111334,
      "loss": 1.5222,
      "theoretical_loss": 3.562548332764909,
      "tokens_seen": 1416036352
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005767703109327984,
      "loss": 1.5925,
      "theoretical_loss": 3.5625337521927687,
      "tokens_seen": 1416101888
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005767502507522567,
      "loss": 1.4904,
      "theoretical_loss": 3.5625191724843166,
      "tokens_seen": 1416167424
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005767301905717152,
      "loss": 1.3727,
      "theoretical_loss": 3.562504593639462,
      "tokens_seen": 1416232960
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005767101303911735,
      "loss": 1.5902,
      "theoretical_loss": 3.5624900156581134,
      "tokens_seen": 1416298496
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005767101303911735,
      "loss": 1.6206,
      "theoretical_loss": 3.5624754385401802,
      "tokens_seen": 1416364032
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005766900702106319,
      "loss": 1.4706,
      "theoretical_loss": 3.5624608622855707,
      "tokens_seen": 1416429568
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005766700100300902,
      "loss": 1.5945,
      "theoretical_loss": 3.5624462868941946,
      "tokens_seen": 1416495104
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005766499498495487,
      "loss": 1.3765,
      "theoretical_loss": 3.56243171236596,
      "tokens_seen": 1416560640
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000576629889669007,
      "loss": 1.4039,
      "theoretical_loss": 3.562417138700776,
      "tokens_seen": 1416626176
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005766098294884654,
      "loss": 1.564,
      "theoretical_loss": 3.5624025658985525,
      "tokens_seen": 1416691712
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005765897693079238,
      "loss": 1.5447,
      "theoretical_loss": 3.5623879939591974,
      "tokens_seen": 1416757248
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005765697091273821,
      "loss": 1.4429,
      "theoretical_loss": 3.5623734228826196,
      "tokens_seen": 1416822784
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005765496489468406,
      "loss": 1.4862,
      "theoretical_loss": 3.5623588526687295,
      "tokens_seen": 1416888320
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000576529588766299,
      "loss": 1.3117,
      "theoretical_loss": 3.562344283317435,
      "tokens_seen": 1416953856
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005765095285857573,
      "loss": 1.5007,
      "theoretical_loss": 3.562329714828645,
      "tokens_seen": 1417019392
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005764894684052157,
      "loss": 1.3809,
      "theoretical_loss": 3.562315147202269,
      "tokens_seen": 1417084928
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000576469408224674,
      "loss": 1.4683,
      "theoretical_loss": 3.562300580438216,
      "tokens_seen": 1417150464
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005764493480441324,
      "loss": 1.5059,
      "theoretical_loss": 3.5622860145363955,
      "tokens_seen": 1417216000
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005764292878635909,
      "loss": 1.3708,
      "theoretical_loss": 3.5622714494967154,
      "tokens_seen": 1417281536
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005764092276830491,
      "loss": 1.6744,
      "theoretical_loss": 3.562256885319086,
      "tokens_seen": 1417347072
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005763891675025076,
      "loss": 1.5052,
      "theoretical_loss": 3.562242322003416,
      "tokens_seen": 1417412608
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000576369107321966,
      "loss": 1.4195,
      "theoretical_loss": 3.5622277595496143,
      "tokens_seen": 1417478144
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005763490471414243,
      "loss": 1.5005,
      "theoretical_loss": 3.5622131979575906,
      "tokens_seen": 1417543680
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005763289869608827,
      "loss": 1.4305,
      "theoretical_loss": 3.5621986372272536,
      "tokens_seen": 1417609216
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000576308926780341,
      "loss": 1.5413,
      "theoretical_loss": 3.5621840773585127,
      "tokens_seen": 1417674752
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005762888665997994,
      "loss": 1.4641,
      "theoretical_loss": 3.5621695183512765,
      "tokens_seen": 1417740288
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005762688064192579,
      "loss": 1.55,
      "theoretical_loss": 3.562154960205455,
      "tokens_seen": 1417805824
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005762487462387161,
      "loss": 1.4148,
      "theoretical_loss": 3.5621404029209573,
      "tokens_seen": 1417871360
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005762286860581746,
      "loss": 1.3197,
      "theoretical_loss": 3.5621258464976924,
      "tokens_seen": 1417936896
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005762086258776329,
      "loss": 1.383,
      "theoretical_loss": 3.5621112909355697,
      "tokens_seen": 1418002432
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005761885656970913,
      "loss": 1.4747,
      "theoretical_loss": 3.5620967362344986,
      "tokens_seen": 1418067968
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005761685055165497,
      "loss": 1.481,
      "theoretical_loss": 3.562082182394388,
      "tokens_seen": 1418133504
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000576148445336008,
      "loss": 1.3738,
      "theoretical_loss": 3.5620676294151465,
      "tokens_seen": 1418199040
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005761283851554664,
      "loss": 1.4513,
      "theoretical_loss": 3.562053077296685,
      "tokens_seen": 1418264576
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005761083249749248,
      "loss": 1.3719,
      "theoretical_loss": 3.5620385260389122,
      "tokens_seen": 1418330112
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005760882647943831,
      "loss": 1.4425,
      "theoretical_loss": 3.5620239756417367,
      "tokens_seen": 1418395648
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005760682046138416,
      "loss": 1.5059,
      "theoretical_loss": 3.562009426105069,
      "tokens_seen": 1418461184
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005760481444332999,
      "loss": 1.5384,
      "theoretical_loss": 3.5619948774288175,
      "tokens_seen": 1418526720
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005760280842527583,
      "loss": 1.4806,
      "theoretical_loss": 3.561980329612892,
      "tokens_seen": 1418592256
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005760080240722167,
      "loss": 1.4182,
      "theoretical_loss": 3.561965782657202,
      "tokens_seen": 1418657792
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000575987963891675,
      "loss": 1.4752,
      "theoretical_loss": 3.561951236561656,
      "tokens_seen": 1418723328
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005759679037111334,
      "loss": 1.5232,
      "theoretical_loss": 3.5619366913261654,
      "tokens_seen": 1418788864
    },
    {
      "epoch": 0.43,
      "objective/train/advantage_avg": 0.49392983317375183,
      "objective/train/docs_used": 518371,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3717641830444336,
      "objective/train/original_loss": 1.3717644214630127,
      "objective/train/theoretical_loss": 3.5619221469506375,
      "objective/train/tokens_used": 1439314400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24428243935108185,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506349802017212,
      "objective/train/weighted_lm_loss": 1.4410974979400635,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5619221469506375,
      "tokens_seen": 1418854400
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005759478435305918,
      "loss": 1.4517,
      "theoretical_loss": 3.5619221469506375,
      "tokens_seen": 1418854400
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005759277833500501,
      "loss": 1.4199,
      "theoretical_loss": 3.5619076034349826,
      "tokens_seen": 1418919936
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005759077231695086,
      "loss": 1.6237,
      "theoretical_loss": 3.56189306077911,
      "tokens_seen": 1418985472
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005758876629889669,
      "loss": 1.4513,
      "theoretical_loss": 3.5618785189829296,
      "tokens_seen": 1419051008
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005758676028084253,
      "loss": 1.5169,
      "theoretical_loss": 3.561863978046351,
      "tokens_seen": 1419116544
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005758475426278836,
      "loss": 1.4745,
      "theoretical_loss": 3.5618494379692827,
      "tokens_seen": 1419182080
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005758274824473421,
      "loss": 1.514,
      "theoretical_loss": 3.561834898751635,
      "tokens_seen": 1419247616
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005758074222668004,
      "loss": 1.516,
      "theoretical_loss": 3.5618203603933174,
      "tokens_seen": 1419313152
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005757873620862588,
      "loss": 1.3845,
      "theoretical_loss": 3.5618058228942386,
      "tokens_seen": 1419378688
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005757673019057171,
      "loss": 1.4636,
      "theoretical_loss": 3.5617912862543095,
      "tokens_seen": 1419444224
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005757472417251756,
      "loss": 1.4811,
      "theoretical_loss": 3.561776750473439,
      "tokens_seen": 1419509760
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005757271815446339,
      "loss": 1.5354,
      "theoretical_loss": 3.5617622155515365,
      "tokens_seen": 1419575296
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005757071213640923,
      "loss": 1.4203,
      "theoretical_loss": 3.5617476814885114,
      "tokens_seen": 1419640832
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005756870611835506,
      "loss": 1.6149,
      "theoretical_loss": 3.561733148284274,
      "tokens_seen": 1419706368
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005756670010030091,
      "loss": 1.49,
      "theoretical_loss": 3.561718615938733,
      "tokens_seen": 1419771904
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005756469408224674,
      "loss": 1.5258,
      "theoretical_loss": 3.561704084451799,
      "tokens_seen": 1419837440
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005756268806419258,
      "loss": 1.4722,
      "theoretical_loss": 3.5616895538233813,
      "tokens_seen": 1419902976
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005756068204613842,
      "loss": 1.4395,
      "theoretical_loss": 3.5616750240533888,
      "tokens_seen": 1419968512
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005755867602808425,
      "loss": 1.4034,
      "theoretical_loss": 3.5616604951417328,
      "tokens_seen": 1420034048
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005755667001003009,
      "loss": 1.4972,
      "theoretical_loss": 3.5616459670883214,
      "tokens_seen": 1420099584
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005755466399197593,
      "loss": 1.4354,
      "theoretical_loss": 3.5616314398930653,
      "tokens_seen": 1420165120
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005755265797392176,
      "loss": 1.4285,
      "theoretical_loss": 3.5616169135558735,
      "tokens_seen": 1420230656
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005755065195586761,
      "loss": 1.4957,
      "theoretical_loss": 3.561602388076656,
      "tokens_seen": 1420296192
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005754864593781343,
      "loss": 1.4682,
      "theoretical_loss": 3.5615878634553226,
      "tokens_seen": 1420361728
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005754663991975928,
      "loss": 1.4124,
      "theoretical_loss": 3.5615733396917837,
      "tokens_seen": 1420427264
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005754463390170512,
      "loss": 1.4531,
      "theoretical_loss": 3.561558816785948,
      "tokens_seen": 1420492800
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005754262788365095,
      "loss": 1.444,
      "theoretical_loss": 3.5615442947377254,
      "tokens_seen": 1420558336
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005754062186559679,
      "loss": 1.4898,
      "theoretical_loss": 3.5615297735470266,
      "tokens_seen": 1420623872
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005753861584754264,
      "loss": 1.5295,
      "theoretical_loss": 3.56151525321376,
      "tokens_seen": 1420689408
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005753660982948846,
      "loss": 1.4703,
      "theoretical_loss": 3.5615007337378373,
      "tokens_seen": 1420754944
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005753460381143431,
      "loss": 1.4658,
      "theoretical_loss": 3.5614862151191664,
      "tokens_seen": 1420820480
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005753259779338013,
      "loss": 1.504,
      "theoretical_loss": 3.561471697357659,
      "tokens_seen": 1420886016
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005753059177532598,
      "loss": 1.4593,
      "theoretical_loss": 3.561457180453223,
      "tokens_seen": 1420951552
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005752858575727182,
      "loss": 1.5023,
      "theoretical_loss": 3.5614426644057695,
      "tokens_seen": 1421017088
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005752657973921765,
      "loss": 1.4741,
      "theoretical_loss": 3.561428149215208,
      "tokens_seen": 1421082624
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005752457372116349,
      "loss": 1.3789,
      "theoretical_loss": 3.5614136348814487,
      "tokens_seen": 1421148160
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005752256770310933,
      "loss": 1.4747,
      "theoretical_loss": 3.561399121404402,
      "tokens_seen": 1421213696
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005752056168505516,
      "loss": 1.4965,
      "theoretical_loss": 3.5613846087839764,
      "tokens_seen": 1421279232
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005751855566700101,
      "loss": 1.365,
      "theoretical_loss": 3.561370097020083,
      "tokens_seen": 1421344768
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005751654964894683,
      "loss": 1.5128,
      "theoretical_loss": 3.561355586112631,
      "tokens_seen": 1421410304
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005751454363089268,
      "loss": 1.5582,
      "theoretical_loss": 3.561341076061531,
      "tokens_seen": 1421475840
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005751253761283851,
      "loss": 1.4445,
      "theoretical_loss": 3.5613265668666925,
      "tokens_seen": 1421541376
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005751053159478435,
      "loss": 1.3535,
      "theoretical_loss": 3.561312058528026,
      "tokens_seen": 1421606912
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005750852557673019,
      "loss": 1.4011,
      "theoretical_loss": 3.5612975510454414,
      "tokens_seen": 1421672448
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005750651955867603,
      "loss": 1.434,
      "theoretical_loss": 3.5612830444188486,
      "tokens_seen": 1421737984
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005750451354062187,
      "loss": 1.4592,
      "theoretical_loss": 3.5612685386481573,
      "tokens_seen": 1421803520
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005750250752256771,
      "loss": 1.5762,
      "theoretical_loss": 3.5612540337332783,
      "tokens_seen": 1421869056
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005750050150451355,
      "loss": 1.533,
      "theoretical_loss": 3.561239529674121,
      "tokens_seen": 1421934592
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005749849548645938,
      "loss": 1.4129,
      "theoretical_loss": 3.561225026470596,
      "tokens_seen": 1422000128
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005749648946840522,
      "loss": 1.4722,
      "theoretical_loss": 3.561210524122613,
      "tokens_seen": 1422065664
    },
    {
      "epoch": 0.43,
      "objective/train/advantage_avg": 0.4940011203289032,
      "objective/train/docs_used": 519513,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5365597009658813,
      "objective/train/original_loss": 1.5365595817565918,
      "objective/train/theoretical_loss": 3.5611960226300816,
      "objective/train/tokens_used": 1442591200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24428236484527588,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506420135498047,
      "objective/train/weighted_lm_loss": 1.6143674850463867,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 3.5611960226300816,
      "tokens_seen": 1422131200
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005749448345035105,
      "loss": 1.5504,
      "theoretical_loss": 3.5611960226300816,
      "tokens_seen": 1422131200
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000574924774322969,
      "loss": 1.4659,
      "theoretical_loss": 3.5611815219929133,
      "tokens_seen": 1422196736
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005749047141424273,
      "loss": 1.4608,
      "theoretical_loss": 3.561167022211017,
      "tokens_seen": 1422262272
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005748846539618857,
      "loss": 1.3958,
      "theoretical_loss": 3.5611525232843038,
      "tokens_seen": 1422327808
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000574864593781344,
      "loss": 1.5533,
      "theoretical_loss": 3.561138025212683,
      "tokens_seen": 1422393344
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005748445336008025,
      "loss": 1.435,
      "theoretical_loss": 3.5611235279960654,
      "tokens_seen": 1422458880
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005748244734202608,
      "loss": 1.408,
      "theoretical_loss": 3.561109031634361,
      "tokens_seen": 1422524416
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005748044132397192,
      "loss": 1.4073,
      "theoretical_loss": 3.56109453612748,
      "tokens_seen": 1422589952
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005747843530591775,
      "loss": 1.5178,
      "theoretical_loss": 3.561080041475332,
      "tokens_seen": 1422655488
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000574764292878636,
      "loss": 1.542,
      "theoretical_loss": 3.5610655476778286,
      "tokens_seen": 1422721024
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005747442326980943,
      "loss": 1.3798,
      "theoretical_loss": 3.561051054734879,
      "tokens_seen": 1422786560
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005747241725175527,
      "loss": 1.385,
      "theoretical_loss": 3.561036562646393,
      "tokens_seen": 1422852096
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000574704112337011,
      "loss": 1.4094,
      "theoretical_loss": 3.5610220714122827,
      "tokens_seen": 1422917632
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005746840521564695,
      "loss": 1.4265,
      "theoretical_loss": 3.561007581032457,
      "tokens_seen": 1422983168
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005746639919759278,
      "loss": 1.4239,
      "theoretical_loss": 3.560993091506826,
      "tokens_seen": 1423048704
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005746439317953862,
      "loss": 1.4634,
      "theoretical_loss": 3.5609786028353003,
      "tokens_seen": 1423114240
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005746238716148446,
      "loss": 1.4357,
      "theoretical_loss": 3.560964115017791,
      "tokens_seen": 1423179776
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005746038114343029,
      "loss": 1.4966,
      "theoretical_loss": 3.5609496280542077,
      "tokens_seen": 1423245312
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005745837512537613,
      "loss": 1.3585,
      "theoretical_loss": 3.5609351419444604,
      "tokens_seen": 1423310848
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005745636910732197,
      "loss": 1.5057,
      "theoretical_loss": 3.5609206566884604,
      "tokens_seen": 1423376384
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000574543630892678,
      "loss": 1.3981,
      "theoretical_loss": 3.560906172286118,
      "tokens_seen": 1423441920
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005745235707121365,
      "loss": 1.4538,
      "theoretical_loss": 3.5608916887373425,
      "tokens_seen": 1423507456
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005745035105315947,
      "loss": 1.3853,
      "theoretical_loss": 3.5608772060420453,
      "tokens_seen": 1423572992
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005744834503510532,
      "loss": 1.5708,
      "theoretical_loss": 3.5608627242001365,
      "tokens_seen": 1423638528
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005744633901705116,
      "loss": 1.3772,
      "theoretical_loss": 3.5608482432115265,
      "tokens_seen": 1423704064
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005744433299899699,
      "loss": 1.4729,
      "theoretical_loss": 3.5608337630761264,
      "tokens_seen": 1423769600
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005744232698094283,
      "loss": 1.3085,
      "theoretical_loss": 3.5608192837938457,
      "tokens_seen": 1423835136
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005744032096288868,
      "loss": 1.4665,
      "theoretical_loss": 3.560804805364595,
      "tokens_seen": 1423900672
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000574383149448345,
      "loss": 1.509,
      "theoretical_loss": 3.5607903277882853,
      "tokens_seen": 1423966208
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005743630892678035,
      "loss": 1.5695,
      "theoretical_loss": 3.5607758510648266,
      "tokens_seen": 1424031744
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005743430290872617,
      "loss": 1.4701,
      "theoretical_loss": 3.5607613751941303,
      "tokens_seen": 1424097280
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005743229689067202,
      "loss": 1.4109,
      "theoretical_loss": 3.5607469001761056,
      "tokens_seen": 1424162816
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005743029087261786,
      "loss": 1.4127,
      "theoretical_loss": 3.560732426010664,
      "tokens_seen": 1424228352
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005742828485456369,
      "loss": 1.4097,
      "theoretical_loss": 3.5607179526977157,
      "tokens_seen": 1424293888
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005742627883650953,
      "loss": 1.5199,
      "theoretical_loss": 3.560703480237171,
      "tokens_seen": 1424359424
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005742427281845537,
      "loss": 1.5146,
      "theoretical_loss": 3.560689008628942,
      "tokens_seen": 1424424960
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000574222668004012,
      "loss": 1.5692,
      "theoretical_loss": 3.560674537872937,
      "tokens_seen": 1424490496
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005742026078234705,
      "loss": 1.4676,
      "theoretical_loss": 3.560660067969068,
      "tokens_seen": 1424556032
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005741825476429287,
      "loss": 1.4581,
      "theoretical_loss": 3.5606455989172456,
      "tokens_seen": 1424621568
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005741624874623872,
      "loss": 1.4005,
      "theoretical_loss": 3.56063113071738,
      "tokens_seen": 1424687104
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005741424272818455,
      "loss": 1.4186,
      "theoretical_loss": 3.560616663369382,
      "tokens_seen": 1424752640
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005741223671013039,
      "loss": 1.4132,
      "theoretical_loss": 3.5606021968731625,
      "tokens_seen": 1424818176
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005741023069207623,
      "loss": 1.4412,
      "theoretical_loss": 3.560587731228632,
      "tokens_seen": 1424883712
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005740822467402207,
      "loss": 1.4098,
      "theoretical_loss": 3.560573266435701,
      "tokens_seen": 1424949248
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000574062186559679,
      "loss": 1.5508,
      "theoretical_loss": 3.5605588024942803,
      "tokens_seen": 1425014784
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005740421263791375,
      "loss": 1.477,
      "theoretical_loss": 3.560544339404281,
      "tokens_seen": 1425080320
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005740220661985957,
      "loss": 1.4179,
      "theoretical_loss": 3.560529877165613,
      "tokens_seen": 1425145856
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005740020060180542,
      "loss": 1.4866,
      "theoretical_loss": 3.5605154157781875,
      "tokens_seen": 1425211392
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005739819458375125,
      "loss": 1.4527,
      "theoretical_loss": 3.560500955241916,
      "tokens_seen": 1425276928
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005739618856569709,
      "loss": 1.5691,
      "theoretical_loss": 3.5604864955567086,
      "tokens_seen": 1425342464
    },
    {
      "epoch": 0.43,
      "objective/train/advantage_avg": 0.493914395570755,
      "objective/train/docs_used": 520731,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.282238483428955,
      "objective/train/original_loss": 1.282238483428955,
      "objective/train/theoretical_loss": 3.5604720367224756,
      "objective/train/tokens_used": 1445868000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24452559649944305,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050634503364563,
      "objective/train/weighted_lm_loss": 1.3469407558441162,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5604720367224756,
      "tokens_seen": 1425408000
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005739418254764293,
      "loss": 1.3479,
      "theoretical_loss": 3.5604720367224756,
      "tokens_seen": 1425408000
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005739217652958877,
      "loss": 1.439,
      "theoretical_loss": 3.560457578739128,
      "tokens_seen": 1425473536
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000573901705115346,
      "loss": 1.5291,
      "theoretical_loss": 3.5604431216065775,
      "tokens_seen": 1425539072
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005738816449348044,
      "loss": 1.4456,
      "theoretical_loss": 3.560428665324734,
      "tokens_seen": 1425604608
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005738615847542628,
      "loss": 1.4651,
      "theoretical_loss": 3.560414209893509,
      "tokens_seen": 1425670144
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005738415245737212,
      "loss": 1.3731,
      "theoretical_loss": 3.5603997553128126,
      "tokens_seen": 1425735680
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005738214643931795,
      "loss": 1.5245,
      "theoretical_loss": 3.5603853015825564,
      "tokens_seen": 1425801216
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000573801404212638,
      "loss": 1.4268,
      "theoretical_loss": 3.56037084870265,
      "tokens_seen": 1425866752
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005737813440320962,
      "loss": 1.4345,
      "theoretical_loss": 3.560356396673006,
      "tokens_seen": 1425932288
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005737612838515547,
      "loss": 1.4963,
      "theoretical_loss": 3.5603419454935348,
      "tokens_seen": 1425997824
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000573741223671013,
      "loss": 1.3719,
      "theoretical_loss": 3.560327495164147,
      "tokens_seen": 1426063360
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005737211634904714,
      "loss": 1.511,
      "theoretical_loss": 3.560313045684753,
      "tokens_seen": 1426128896
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005737011033099298,
      "loss": 1.3752,
      "theoretical_loss": 3.5602985970552647,
      "tokens_seen": 1426194432
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005736810431293882,
      "loss": 1.3741,
      "theoretical_loss": 3.560284149275592,
      "tokens_seen": 1426259968
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005736609829488465,
      "loss": 1.3658,
      "theoretical_loss": 3.5602697023456473,
      "tokens_seen": 1426325504
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000573640922768305,
      "loss": 1.4578,
      "theoretical_loss": 3.560255256265341,
      "tokens_seen": 1426391040
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005736208625877632,
      "loss": 1.3973,
      "theoretical_loss": 3.560240811034584,
      "tokens_seen": 1426456576
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005736008024072217,
      "loss": 1.487,
      "theoretical_loss": 3.5602263666532865,
      "tokens_seen": 1426522112
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00057358074222668,
      "loss": 1.4857,
      "theoretical_loss": 3.5602119231213605,
      "tokens_seen": 1426587648
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005735606820461384,
      "loss": 1.3489,
      "theoretical_loss": 3.560197480438717,
      "tokens_seen": 1426653184
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005735406218655969,
      "loss": 1.4423,
      "theoretical_loss": 3.560183038605267,
      "tokens_seen": 1426718720
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005735205616850551,
      "loss": 1.5145,
      "theoretical_loss": 3.5601685976209207,
      "tokens_seen": 1426784256
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005735005015045136,
      "loss": 1.5062,
      "theoretical_loss": 3.5601541574855906,
      "tokens_seen": 1426849792
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000573480441323972,
      "loss": 1.4541,
      "theoretical_loss": 3.5601397181991867,
      "tokens_seen": 1426915328
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005734603811434303,
      "loss": 1.4024,
      "theoretical_loss": 3.560125279761621,
      "tokens_seen": 1426980864
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005734403209628887,
      "loss": 1.5022,
      "theoretical_loss": 3.560110842172804,
      "tokens_seen": 1427046400
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005734202607823472,
      "loss": 1.5055,
      "theoretical_loss": 3.5600964054326463,
      "tokens_seen": 1427111936
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005734002006018054,
      "loss": 1.4805,
      "theoretical_loss": 3.5600819695410606,
      "tokens_seen": 1427177472
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005733801404212639,
      "loss": 1.5204,
      "theoretical_loss": 3.5600675344979567,
      "tokens_seen": 1427243008
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005733600802407221,
      "loss": 1.4015,
      "theoretical_loss": 3.5600531003032465,
      "tokens_seen": 1427308544
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005733400200601806,
      "loss": 1.5402,
      "theoretical_loss": 3.5600386669568405,
      "tokens_seen": 1427374080
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000573319959879639,
      "loss": 1.575,
      "theoretical_loss": 3.5600242344586506,
      "tokens_seen": 1427439616
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005732998996990973,
      "loss": 1.6265,
      "theoretical_loss": 3.5600098028085876,
      "tokens_seen": 1427505152
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005732798395185557,
      "loss": 1.6042,
      "theoretical_loss": 3.559995372006563,
      "tokens_seen": 1427570688
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000573259779338014,
      "loss": 1.4038,
      "theoretical_loss": 3.559980942052488,
      "tokens_seen": 1427636224
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005732397191574724,
      "loss": 1.414,
      "theoretical_loss": 3.559966512946273,
      "tokens_seen": 1427701760
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005732196589769309,
      "loss": 1.4159,
      "theoretical_loss": 3.559952084687831,
      "tokens_seen": 1427767296
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005731995987963891,
      "loss": 1.4919,
      "theoretical_loss": 3.5599376572770716,
      "tokens_seen": 1427832832
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005731795386158476,
      "loss": 1.4336,
      "theoretical_loss": 3.559923230713907,
      "tokens_seen": 1427898368
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005731594784353059,
      "loss": 1.4775,
      "theoretical_loss": 3.5599088049982486,
      "tokens_seen": 1427963904
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005731394182547643,
      "loss": 1.3716,
      "theoretical_loss": 3.5598943801300074,
      "tokens_seen": 1428029440
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005731193580742227,
      "loss": 1.5327,
      "theoretical_loss": 3.5598799561090946,
      "tokens_seen": 1428094976
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005730992978936811,
      "loss": 1.4267,
      "theoretical_loss": 3.5598655329354214,
      "tokens_seen": 1428160512
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005730792377131394,
      "loss": 1.6317,
      "theoretical_loss": 3.5598511106089,
      "tokens_seen": 1428226048
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005730591775325979,
      "loss": 1.4498,
      "theoretical_loss": 3.5598366891294404,
      "tokens_seen": 1428291584
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005730391173520561,
      "loss": 1.4561,
      "theoretical_loss": 3.559822268496956,
      "tokens_seen": 1428357120
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005730190571715146,
      "loss": 1.4237,
      "theoretical_loss": 3.5598078487113556,
      "tokens_seen": 1428422656
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005729989969909729,
      "loss": 1.4574,
      "theoretical_loss": 3.5597934297725526,
      "tokens_seen": 1428488192
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005729789368104313,
      "loss": 1.5318,
      "theoretical_loss": 3.559779011680458,
      "tokens_seen": 1428553728
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005729588766298897,
      "loss": 1.5673,
      "theoretical_loss": 3.5597645944349834,
      "tokens_seen": 1428619264
    },
    {
      "epoch": 0.43,
      "objective/train/advantage_avg": 0.49411433935165405,
      "objective/train/docs_used": 521864,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.395490050315857,
      "objective/train/original_loss": 1.3954899311065674,
      "objective/train/theoretical_loss": 3.55975017803604,
      "objective/train/tokens_used": 1449144800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24449215829372406,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050654411315918,
      "objective/train/weighted_lm_loss": 1.4660526514053345,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.55975017803604,
      "tokens_seen": 1428684800
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005729388164493481,
      "loss": 1.4801,
      "theoretical_loss": 3.55975017803604,
      "tokens_seen": 1428684800
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005729187562688064,
      "loss": 1.5067,
      "theoretical_loss": 3.559735762483539,
      "tokens_seen": 1428750336
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005728986960882648,
      "loss": 1.4988,
      "theoretical_loss": 3.5597213477773915,
      "tokens_seen": 1428815872
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005728786359077232,
      "loss": 1.5771,
      "theoretical_loss": 3.5597069339175103,
      "tokens_seen": 1428881408
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005728585757271816,
      "loss": 1.6187,
      "theoretical_loss": 3.5596925209038055,
      "tokens_seen": 1428946944
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005728385155466399,
      "loss": 1.5569,
      "theoretical_loss": 3.5596781087361897,
      "tokens_seen": 1429012480
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005728184553660983,
      "loss": 1.3895,
      "theoretical_loss": 3.5596636974145746,
      "tokens_seen": 1429078016
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005727983951855566,
      "loss": 1.5482,
      "theoretical_loss": 3.5596492869388703,
      "tokens_seen": 1429143552
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005727783350050151,
      "loss": 1.4571,
      "theoretical_loss": 3.55963487730899,
      "tokens_seen": 1429209088
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005727582748244734,
      "loss": 1.4266,
      "theoretical_loss": 3.559620468524844,
      "tokens_seen": 1429274624
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005727382146439318,
      "loss": 1.3459,
      "theoretical_loss": 3.5596060605863444,
      "tokens_seen": 1429340160
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005727181544633902,
      "loss": 1.4744,
      "theoretical_loss": 3.559591653493403,
      "tokens_seen": 1429405696
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005726980942828486,
      "loss": 1.516,
      "theoretical_loss": 3.5595772472459313,
      "tokens_seen": 1429471232
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005726780341023069,
      "loss": 1.5051,
      "theoretical_loss": 3.5595628418438405,
      "tokens_seen": 1429536768
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005726579739217654,
      "loss": 1.4749,
      "theoretical_loss": 3.559548437287043,
      "tokens_seen": 1429602304
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005726379137412236,
      "loss": 1.4394,
      "theoretical_loss": 3.5595340335754497,
      "tokens_seen": 1429667840
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005726178535606821,
      "loss": 1.4072,
      "theoretical_loss": 3.559519630708973,
      "tokens_seen": 1429733376
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005725977933801404,
      "loss": 1.5085,
      "theoretical_loss": 3.5595052286875237,
      "tokens_seen": 1429798912
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005725777331995988,
      "loss": 1.4358,
      "theoretical_loss": 3.559490827511014,
      "tokens_seen": 1429864448
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005725576730190572,
      "loss": 1.5013,
      "theoretical_loss": 3.559476427179356,
      "tokens_seen": 1429929984
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005725376128385155,
      "loss": 1.4287,
      "theoretical_loss": 3.5594620276924607,
      "tokens_seen": 1429995520
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005725175526579739,
      "loss": 1.3575,
      "theoretical_loss": 3.5594476290502404,
      "tokens_seen": 1430061056
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005724974924774324,
      "loss": 1.495,
      "theoretical_loss": 3.5594332312526062,
      "tokens_seen": 1430126592
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005724774322968906,
      "loss": 1.5112,
      "theoretical_loss": 3.5594188342994704,
      "tokens_seen": 1430192128
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005724573721163491,
      "loss": 1.4328,
      "theoretical_loss": 3.559404438190745,
      "tokens_seen": 1430257664
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005724373119358073,
      "loss": 1.429,
      "theoretical_loss": 3.5593900429263408,
      "tokens_seen": 1430323200
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005724172517552658,
      "loss": 1.387,
      "theoretical_loss": 3.5593756485061707,
      "tokens_seen": 1430388736
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005723971915747242,
      "loss": 1.4774,
      "theoretical_loss": 3.5593612549301454,
      "tokens_seen": 1430454272
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005723771313941825,
      "loss": 1.5275,
      "theoretical_loss": 3.559346862198178,
      "tokens_seen": 1430519808
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005723570712136409,
      "loss": 1.4189,
      "theoretical_loss": 3.559332470310179,
      "tokens_seen": 1430585344
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005723370110330994,
      "loss": 1.5842,
      "theoretical_loss": 3.5593180792660615,
      "tokens_seen": 1430650880
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005723169508525576,
      "loss": 1.5337,
      "theoretical_loss": 3.5593036890657364,
      "tokens_seen": 1430716416
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005722968906720161,
      "loss": 1.3867,
      "theoretical_loss": 3.559289299709116,
      "tokens_seen": 1430781952
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005722768304914743,
      "loss": 1.3543,
      "theoretical_loss": 3.559274911196112,
      "tokens_seen": 1430847488
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005722567703109328,
      "loss": 1.4948,
      "theoretical_loss": 3.559260523526637,
      "tokens_seen": 1430913024
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005722367101303912,
      "loss": 1.4664,
      "theoretical_loss": 3.559246136700602,
      "tokens_seen": 1430978560
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005722166499498495,
      "loss": 1.5356,
      "theoretical_loss": 3.5592317507179194,
      "tokens_seen": 1431044096
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005721965897693079,
      "loss": 1.4491,
      "theoretical_loss": 3.559217365578501,
      "tokens_seen": 1431109632
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005721765295887663,
      "loss": 1.531,
      "theoretical_loss": 3.559202981282259,
      "tokens_seen": 1431175168
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005721564694082246,
      "loss": 1.4802,
      "theoretical_loss": 3.559188597829105,
      "tokens_seen": 1431240704
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005721364092276831,
      "loss": 1.4834,
      "theoretical_loss": 3.5591742152189507,
      "tokens_seen": 1431306240
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005721163490471414,
      "loss": 1.4611,
      "theoretical_loss": 3.5591598334517096,
      "tokens_seen": 1431371776
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005720962888665998,
      "loss": 1.352,
      "theoretical_loss": 3.559145452527292,
      "tokens_seen": 1431437312
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005720762286860582,
      "loss": 1.4658,
      "theoretical_loss": 3.5591310724456102,
      "tokens_seen": 1431502848
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005720561685055165,
      "loss": 1.4357,
      "theoretical_loss": 3.559116693206577,
      "tokens_seen": 1431568384
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000572036108324975,
      "loss": 1.5144,
      "theoretical_loss": 3.5591023148101044,
      "tokens_seen": 1431633920
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005720160481444333,
      "loss": 1.5456,
      "theoretical_loss": 3.559087937256104,
      "tokens_seen": 1431699456
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005719959879638917,
      "loss": 1.5002,
      "theoretical_loss": 3.559073560544488,
      "tokens_seen": 1431764992
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005719759277833501,
      "loss": 1.5077,
      "theoretical_loss": 3.5590591846751685,
      "tokens_seen": 1431830528
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005719558676028085,
      "loss": 1.3931,
      "theoretical_loss": 3.5590448096480576,
      "tokens_seen": 1431896064
    },
    {
      "epoch": 0.43,
      "objective/train/advantage_avg": 0.49437186121940613,
      "objective/train/docs_used": 522998,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2693687677383423,
      "objective/train/original_loss": 1.2693686485290527,
      "objective/train/theoretical_loss": 3.5590304354630673,
      "objective/train/tokens_used": 1452421600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2446669638156891,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506811141967773,
      "objective/train/weighted_lm_loss": 1.3337132930755615,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5590304354630673,
      "tokens_seen": 1431961600
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005719358074222668,
      "loss": 1.4425,
      "theoretical_loss": 3.5590304354630673,
      "tokens_seen": 1431961600
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005719157472417252,
      "loss": 1.4131,
      "theoretical_loss": 3.55901606212011,
      "tokens_seen": 1432027136
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005718956870611836,
      "loss": 1.5281,
      "theoretical_loss": 3.5590016896190977,
      "tokens_seen": 1432092672
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000571875626880642,
      "loss": 1.5369,
      "theoretical_loss": 3.558987317959942,
      "tokens_seen": 1432158208
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005718555667001003,
      "loss": 1.4023,
      "theoretical_loss": 3.558972947142556,
      "tokens_seen": 1432223744
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005718355065195587,
      "loss": 1.3892,
      "theoretical_loss": 3.558958577166852,
      "tokens_seen": 1432289280
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000571815446339017,
      "loss": 1.4475,
      "theoretical_loss": 3.5589442080327416,
      "tokens_seen": 1432354816
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005717953861584755,
      "loss": 1.41,
      "theoretical_loss": 3.558929839740137,
      "tokens_seen": 1432420352
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005717753259779338,
      "loss": 1.4511,
      "theoretical_loss": 3.55891547228895,
      "tokens_seen": 1432485888
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005717552657973922,
      "loss": 1.4708,
      "theoretical_loss": 3.5589011056790936,
      "tokens_seen": 1432551424
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005717352056168506,
      "loss": 1.5276,
      "theoretical_loss": 3.5588867399104798,
      "tokens_seen": 1432616960
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000571715145436309,
      "loss": 1.5117,
      "theoretical_loss": 3.5588723749830216,
      "tokens_seen": 1432682496
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005716950852557673,
      "loss": 1.5518,
      "theoretical_loss": 3.55885801089663,
      "tokens_seen": 1432748032
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005716750250752258,
      "loss": 1.4887,
      "theoretical_loss": 3.5588436476512175,
      "tokens_seen": 1432813568
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000571654964894684,
      "loss": 1.388,
      "theoretical_loss": 3.5588292852466967,
      "tokens_seen": 1432879104
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005716349047141425,
      "loss": 1.3032,
      "theoretical_loss": 3.5588149236829807,
      "tokens_seen": 1432944640
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005716148445336008,
      "loss": 1.4081,
      "theoretical_loss": 3.5588005629599806,
      "tokens_seen": 1433010176
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005715947843530592,
      "loss": 1.4079,
      "theoretical_loss": 3.5587862030776094,
      "tokens_seen": 1433075712
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005715747241725176,
      "loss": 1.4086,
      "theoretical_loss": 3.558771844035779,
      "tokens_seen": 1433141248
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005715546639919759,
      "loss": 1.3835,
      "theoretical_loss": 3.558757485834402,
      "tokens_seen": 1433206784
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005715346038114343,
      "loss": 1.5688,
      "theoretical_loss": 3.558743128473391,
      "tokens_seen": 1433272320
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005715145436308928,
      "loss": 1.376,
      "theoretical_loss": 3.558728771952658,
      "tokens_seen": 1433337856
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000571494483450351,
      "loss": 1.3062,
      "theoretical_loss": 3.5587144162721156,
      "tokens_seen": 1433403392
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005714744232698095,
      "loss": 1.4759,
      "theoretical_loss": 3.5587000614316766,
      "tokens_seen": 1433468928
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005714543630892677,
      "loss": 1.286,
      "theoretical_loss": 3.558685707431253,
      "tokens_seen": 1433534464
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005714343029087262,
      "loss": 1.4129,
      "theoretical_loss": 3.558671354270757,
      "tokens_seen": 1433600000
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005714142427281846,
      "loss": 1.4976,
      "theoretical_loss": 3.5586570019501016,
      "tokens_seen": 1433665536
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005713941825476429,
      "loss": 1.5197,
      "theoretical_loss": 3.5586426504691993,
      "tokens_seen": 1433731072
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005713741223671013,
      "loss": 1.5158,
      "theoretical_loss": 3.5586282998279617,
      "tokens_seen": 1433796608
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005713540621865598,
      "loss": 1.5522,
      "theoretical_loss": 3.5586139500263023,
      "tokens_seen": 1433862144
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000571334002006018,
      "loss": 1.5112,
      "theoretical_loss": 3.558599601064133,
      "tokens_seen": 1433927680
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005713139418254765,
      "loss": 1.4737,
      "theoretical_loss": 3.5585852529413664,
      "tokens_seen": 1433993216
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005712938816449347,
      "loss": 1.5491,
      "theoretical_loss": 3.558570905657916,
      "tokens_seen": 1434058752
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005712738214643932,
      "loss": 1.4666,
      "theoretical_loss": 3.5585565592136925,
      "tokens_seen": 1434124288
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005712537612838516,
      "loss": 1.4485,
      "theoretical_loss": 3.5585422136086104,
      "tokens_seen": 1434189824
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005712337011033099,
      "loss": 1.505,
      "theoretical_loss": 3.5585278688425808,
      "tokens_seen": 1434255360
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005712136409227683,
      "loss": 1.452,
      "theoretical_loss": 3.558513524915517,
      "tokens_seen": 1434320896
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005711935807422267,
      "loss": 1.5645,
      "theoretical_loss": 3.5584991818273317,
      "tokens_seen": 1434386432
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000571173520561685,
      "loss": 1.4122,
      "theoretical_loss": 3.558484839577937,
      "tokens_seen": 1434451968
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005711534603811435,
      "loss": 1.504,
      "theoretical_loss": 3.558470498167246,
      "tokens_seen": 1434517504
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005711334002006018,
      "loss": 1.4872,
      "theoretical_loss": 3.558456157595171,
      "tokens_seen": 1434583040
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005711133400200602,
      "loss": 1.4604,
      "theoretical_loss": 3.558441817861625,
      "tokens_seen": 1434648576
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005710932798395186,
      "loss": 1.411,
      "theoretical_loss": 3.55842747896652,
      "tokens_seen": 1434714112
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000571073219658977,
      "loss": 1.4424,
      "theoretical_loss": 3.55841314090977,
      "tokens_seen": 1434779648
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005710531594784353,
      "loss": 1.4144,
      "theoretical_loss": 3.558398803691286,
      "tokens_seen": 1434845184
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005710330992978937,
      "loss": 1.4235,
      "theoretical_loss": 3.558384467310982,
      "tokens_seen": 1434910720
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000571013039117352,
      "loss": 1.3058,
      "theoretical_loss": 3.55837013176877,
      "tokens_seen": 1434976256
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005709929789368105,
      "loss": 1.2987,
      "theoretical_loss": 3.5583557970645634,
      "tokens_seen": 1435041792
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005709729187562688,
      "loss": 1.4841,
      "theoretical_loss": 3.558341463198274,
      "tokens_seen": 1435107328
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005709528585757272,
      "loss": 1.4522,
      "theoretical_loss": 3.5583271301698156,
      "tokens_seen": 1435172864
    },
    {
      "epoch": 0.43,
      "objective/train/advantage_avg": 0.49529004096984863,
      "objective/train/docs_used": 524207,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4512397050857544,
      "objective/train/original_loss": 1.4512395858764648,
      "objective/train/theoretical_loss": 3.5583127979791005,
      "objective/train/tokens_used": 1455698400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2455739974975586,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050777554512024,
      "objective/train/weighted_lm_loss": 1.5249494314193726,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5583127979791005,
      "tokens_seen": 1435238400
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005709327983951855,
      "loss": 1.515,
      "theoretical_loss": 3.5583127979791005,
      "tokens_seen": 1435238400
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.000570912738214644,
      "loss": 1.5147,
      "theoretical_loss": 3.5582984666260415,
      "tokens_seen": 1435303936
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005708926780341023,
      "loss": 1.4825,
      "theoretical_loss": 3.558284136110551,
      "tokens_seen": 1435369472
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0005708726178535607,
      "loss": 1.472,
      "theoretical_loss": 3.558269806432542,
      "tokens_seen": 1435435008
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000570852557673019,
      "loss": 1.4157,
      "theoretical_loss": 3.558255477591928,
      "tokens_seen": 1435500544
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005708324974924774,
      "loss": 1.4506,
      "theoretical_loss": 3.5582411495886213,
      "tokens_seen": 1435566080
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005708124373119358,
      "loss": 1.3956,
      "theoretical_loss": 3.558226822422535,
      "tokens_seen": 1435631616
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005707923771313942,
      "loss": 1.3785,
      "theoretical_loss": 3.5582124960935815,
      "tokens_seen": 1435697152
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005707723169508525,
      "loss": 1.5107,
      "theoretical_loss": 3.558198170601674,
      "tokens_seen": 1435762688
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000570752256770311,
      "loss": 1.4995,
      "theoretical_loss": 3.558183845946725,
      "tokens_seen": 1435828224
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005707321965897693,
      "loss": 1.3916,
      "theoretical_loss": 3.5581695221286482,
      "tokens_seen": 1435893760
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005707121364092277,
      "loss": 1.391,
      "theoretical_loss": 3.5581551991473557,
      "tokens_seen": 1435959296
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000570692076228686,
      "loss": 1.4596,
      "theoretical_loss": 3.558140877002761,
      "tokens_seen": 1436024832
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005706720160481444,
      "loss": 1.4873,
      "theoretical_loss": 3.558126555694777,
      "tokens_seen": 1436090368
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005706519558676028,
      "loss": 1.4255,
      "theoretical_loss": 3.558112235223316,
      "tokens_seen": 1436155904
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005706318956870612,
      "loss": 1.514,
      "theoretical_loss": 3.5580979155882915,
      "tokens_seen": 1436221440
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005706118355065195,
      "loss": 1.5773,
      "theoretical_loss": 3.5580835967896167,
      "tokens_seen": 1436286976
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000570591775325978,
      "loss": 1.4063,
      "theoretical_loss": 3.558069278827204,
      "tokens_seen": 1436352512
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005705717151454362,
      "loss": 1.4481,
      "theoretical_loss": 3.558054961700967,
      "tokens_seen": 1436418048
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005705516549648947,
      "loss": 1.3864,
      "theoretical_loss": 3.558040645410818,
      "tokens_seen": 1436483584
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005705315947843532,
      "loss": 1.5129,
      "theoretical_loss": 3.5580263299566712,
      "tokens_seen": 1436549120
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005705115346038114,
      "loss": 1.544,
      "theoretical_loss": 3.5580120153384387,
      "tokens_seen": 1436614656
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005704914744232699,
      "loss": 1.3294,
      "theoretical_loss": 3.557997701556033,
      "tokens_seen": 1436680192
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005704714142427281,
      "loss": 1.6587,
      "theoretical_loss": 3.557983388609369,
      "tokens_seen": 1436745728
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005704513540621866,
      "loss": 1.4739,
      "theoretical_loss": 3.5579690764983587,
      "tokens_seen": 1436811264
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000570431293881645,
      "loss": 1.4073,
      "theoretical_loss": 3.557954765222915,
      "tokens_seen": 1436876800
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005704112337011033,
      "loss": 1.3023,
      "theoretical_loss": 3.5579404547829503,
      "tokens_seen": 1436942336
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005703911735205617,
      "loss": 1.4521,
      "theoretical_loss": 3.5579261451783797,
      "tokens_seen": 1437007872
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005703711133400202,
      "loss": 1.3648,
      "theoretical_loss": 3.557911836409115,
      "tokens_seen": 1437073408
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005703510531594784,
      "loss": 1.3445,
      "theoretical_loss": 3.5578975284750696,
      "tokens_seen": 1437138944
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005703309929789369,
      "loss": 1.4447,
      "theoretical_loss": 3.557883221376157,
      "tokens_seen": 1437204480
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005703109327983951,
      "loss": 1.521,
      "theoretical_loss": 3.55786891511229,
      "tokens_seen": 1437270016
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005702908726178536,
      "loss": 1.4039,
      "theoretical_loss": 3.557854609683382,
      "tokens_seen": 1437335552
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000570270812437312,
      "loss": 1.4227,
      "theoretical_loss": 3.557840305089346,
      "tokens_seen": 1437401088
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005702507522567703,
      "loss": 1.4424,
      "theoretical_loss": 3.5578260013300955,
      "tokens_seen": 1437466624
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005702306920762287,
      "loss": 1.4784,
      "theoretical_loss": 3.5578116984055432,
      "tokens_seen": 1437532160
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005702106318956871,
      "loss": 1.3736,
      "theoretical_loss": 3.5577973963156024,
      "tokens_seen": 1437597696
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005701905717151454,
      "loss": 1.3439,
      "theoretical_loss": 3.557783095060187,
      "tokens_seen": 1437663232
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005701705115346039,
      "loss": 1.4475,
      "theoretical_loss": 3.55776879463921,
      "tokens_seen": 1437728768
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005701504513540622,
      "loss": 1.4278,
      "theoretical_loss": 3.557754495052585,
      "tokens_seen": 1437794304
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005701303911735206,
      "loss": 1.4744,
      "theoretical_loss": 3.557740196300224,
      "tokens_seen": 1437859840
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000570110330992979,
      "loss": 1.5345,
      "theoretical_loss": 3.5577258983820412,
      "tokens_seen": 1437925376
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005700902708124373,
      "loss": 1.5624,
      "theoretical_loss": 3.5577116012979504,
      "tokens_seen": 1437990912
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005700702106318957,
      "loss": 1.4986,
      "theoretical_loss": 3.557697305047864,
      "tokens_seen": 1438056448
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005700501504513541,
      "loss": 1.4038,
      "theoretical_loss": 3.5576830096316963,
      "tokens_seen": 1438121984
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005700300902708124,
      "loss": 1.4179,
      "theoretical_loss": 3.5576687150493598,
      "tokens_seen": 1438187520
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005700100300902709,
      "loss": 1.5122,
      "theoretical_loss": 3.557654421300768,
      "tokens_seen": 1438253056
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005699899699097292,
      "loss": 1.4635,
      "theoretical_loss": 3.5576401283858345,
      "tokens_seen": 1438318592
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005699699097291876,
      "loss": 1.5012,
      "theoretical_loss": 3.5576258363044735,
      "tokens_seen": 1438384128
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005699498495486459,
      "loss": 1.5529,
      "theoretical_loss": 3.557611545056597,
      "tokens_seen": 1438449664
    },
    {
      "epoch": 0.44,
      "objective/train/advantage_avg": 0.4948044419288635,
      "objective/train/docs_used": 525418,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3803629875183105,
      "objective/train/original_loss": 1.3803627490997314,
      "objective/train/theoretical_loss": 3.5575972546421184,
      "objective/train/tokens_used": 1458975200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24508336186408997,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507265329360962,
      "objective/train/weighted_lm_loss": 1.4504742622375488,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.5575972546421184,
      "tokens_seen": 1438515200
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005699297893681044,
      "loss": 1.3849,
      "theoretical_loss": 3.5575972546421184,
      "tokens_seen": 1438515200
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005699097291875627,
      "loss": 1.5566,
      "theoretical_loss": 3.5575829650609525,
      "tokens_seen": 1438580736
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005698896690070211,
      "loss": 1.5175,
      "theoretical_loss": 3.5575686763130117,
      "tokens_seen": 1438646272
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005698696088264794,
      "loss": 1.5845,
      "theoretical_loss": 3.55755438839821,
      "tokens_seen": 1438711808
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005698495486459378,
      "loss": 1.4663,
      "theoretical_loss": 3.5575401013164605,
      "tokens_seen": 1438777344
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005698294884653962,
      "loss": 1.416,
      "theoretical_loss": 3.557525815067677,
      "tokens_seen": 1438842880
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005698094282848546,
      "loss": 1.4504,
      "theoretical_loss": 3.5575115296517725,
      "tokens_seen": 1438908416
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005697893681043129,
      "loss": 1.3861,
      "theoretical_loss": 3.557497245068661,
      "tokens_seen": 1438973952
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005697693079237714,
      "loss": 1.455,
      "theoretical_loss": 3.557482961318256,
      "tokens_seen": 1439039488
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005697492477432297,
      "loss": 1.4784,
      "theoretical_loss": 3.5574686784004705,
      "tokens_seen": 1439105024
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005697291875626881,
      "loss": 1.5025,
      "theoretical_loss": 3.5574543963152188,
      "tokens_seen": 1439170560
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005697091273821464,
      "loss": 1.4206,
      "theoretical_loss": 3.5574401150624135,
      "tokens_seen": 1439236096
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005696890672016048,
      "loss": 1.4044,
      "theoretical_loss": 3.5574258346419696,
      "tokens_seen": 1439301632
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005696690070210632,
      "loss": 1.3517,
      "theoretical_loss": 3.5574115550537995,
      "tokens_seen": 1439367168
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005696489468405216,
      "loss": 1.4982,
      "theoretical_loss": 3.5573972762978174,
      "tokens_seen": 1439432704
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005696288866599799,
      "loss": 1.5586,
      "theoretical_loss": 3.5573829983739365,
      "tokens_seen": 1439498240
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005696088264794384,
      "loss": 1.5262,
      "theoretical_loss": 3.557368721282071,
      "tokens_seen": 1439563776
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005695887662988966,
      "loss": 1.5208,
      "theoretical_loss": 3.5573544450221335,
      "tokens_seen": 1439629312
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005695687061183551,
      "loss": 1.5217,
      "theoretical_loss": 3.5573401695940383,
      "tokens_seen": 1439694848
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005695486459378135,
      "loss": 1.3913,
      "theoretical_loss": 3.5573258949976996,
      "tokens_seen": 1439760384
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005695285857572718,
      "loss": 1.4308,
      "theoretical_loss": 3.5573116212330307,
      "tokens_seen": 1439825920
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005695085255767302,
      "loss": 1.504,
      "theoretical_loss": 3.5572973482999446,
      "tokens_seen": 1439891456
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005694884653961885,
      "loss": 1.5002,
      "theoretical_loss": 3.557283076198356,
      "tokens_seen": 1439956992
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005694684052156469,
      "loss": 1.5207,
      "theoretical_loss": 3.557268804928178,
      "tokens_seen": 1440022528
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005694483450351054,
      "loss": 1.4387,
      "theoretical_loss": 3.5572545344893243,
      "tokens_seen": 1440088064
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005694282848545636,
      "loss": 1.4897,
      "theoretical_loss": 3.5572402648817096,
      "tokens_seen": 1440153600
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005694082246740221,
      "loss": 1.5433,
      "theoretical_loss": 3.557225996105246,
      "tokens_seen": 1440219136
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005693881644934805,
      "loss": 1.4308,
      "theoretical_loss": 3.557211728159849,
      "tokens_seen": 1440284672
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005693681043129388,
      "loss": 1.6531,
      "theoretical_loss": 3.5571974610454316,
      "tokens_seen": 1440350208
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005693480441323972,
      "loss": 1.4963,
      "theoretical_loss": 3.557183194761907,
      "tokens_seen": 1440415744
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005693279839518555,
      "loss": 1.4771,
      "theoretical_loss": 3.5571689293091895,
      "tokens_seen": 1440481280
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005693079237713139,
      "loss": 1.4303,
      "theoretical_loss": 3.5571546646871934,
      "tokens_seen": 1440546816
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005692878635907724,
      "loss": 1.4899,
      "theoretical_loss": 3.5571404008958325,
      "tokens_seen": 1440612352
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005692678034102306,
      "loss": 1.4033,
      "theoretical_loss": 3.5571261379350196,
      "tokens_seen": 1440677888
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005692477432296891,
      "loss": 1.4202,
      "theoretical_loss": 3.5571118758046696,
      "tokens_seen": 1440743424
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005692276830491474,
      "loss": 1.2946,
      "theoretical_loss": 3.5570976145046957,
      "tokens_seen": 1440808960
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005692076228686058,
      "loss": 1.426,
      "theoretical_loss": 3.5570833540350124,
      "tokens_seen": 1440874496
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005691875626880642,
      "loss": 1.5525,
      "theoretical_loss": 3.5570690943955334,
      "tokens_seen": 1440940032
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005691675025075226,
      "loss": 1.3685,
      "theoretical_loss": 3.5570548355861726,
      "tokens_seen": 1441005568
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005691474423269809,
      "loss": 1.5487,
      "theoretical_loss": 3.5570405776068434,
      "tokens_seen": 1441071104
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005691273821464394,
      "loss": 1.4597,
      "theoretical_loss": 3.5570263204574606,
      "tokens_seen": 1441136640
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005691073219658976,
      "loss": 1.5453,
      "theoretical_loss": 3.557012064137938,
      "tokens_seen": 1441202176
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005690872617853561,
      "loss": 1.5089,
      "theoretical_loss": 3.5569978086481884,
      "tokens_seen": 1441267712
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005690672016048145,
      "loss": 1.4579,
      "theoretical_loss": 3.556983553988127,
      "tokens_seen": 1441333248
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005690471414242728,
      "loss": 1.3538,
      "theoretical_loss": 3.556969300157668,
      "tokens_seen": 1441398784
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005690270812437313,
      "loss": 1.459,
      "theoretical_loss": 3.556955047156724,
      "tokens_seen": 1441464320
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005690070210631896,
      "loss": 1.367,
      "theoretical_loss": 3.556940794985211,
      "tokens_seen": 1441529856
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000568986960882648,
      "loss": 1.4714,
      "theoretical_loss": 3.556926543643041,
      "tokens_seen": 1441595392
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005689669007021063,
      "loss": 1.5029,
      "theoretical_loss": 3.5569122931301296,
      "tokens_seen": 1441660928
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005689468405215648,
      "loss": 1.4003,
      "theoretical_loss": 3.55689804344639,
      "tokens_seen": 1441726464
    },
    {
      "epoch": 0.44,
      "objective/train/advantage_avg": 0.49439483880996704,
      "objective/train/docs_used": 526671,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2947261333465576,
      "objective/train/original_loss": 1.2947262525558472,
      "objective/train/theoretical_loss": 3.5568837945917364,
      "objective/train/tokens_used": 1462252000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24474790692329407,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506837368011475,
      "objective/train/weighted_lm_loss": 1.3603771924972534,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5568837945917364,
      "tokens_seen": 1441792000
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005689267803410231,
      "loss": 1.4227,
      "theoretical_loss": 3.5568837945917364,
      "tokens_seen": 1441792000
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005689067201604815,
      "loss": 1.6565,
      "theoretical_loss": 3.556869546566083,
      "tokens_seen": 1441857536
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005688866599799398,
      "loss": 1.5577,
      "theoretical_loss": 3.556855299369343,
      "tokens_seen": 1441923072
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005688665997993982,
      "loss": 1.4283,
      "theoretical_loss": 3.556841053001433,
      "tokens_seen": 1441988608
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005688465396188566,
      "loss": 1.4518,
      "theoretical_loss": 3.556826807462264,
      "tokens_seen": 1442054144
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000568826479438315,
      "loss": 1.3993,
      "theoretical_loss": 3.5568125627517526,
      "tokens_seen": 1442119680
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005688064192577733,
      "loss": 1.6016,
      "theoretical_loss": 3.556798318869811,
      "tokens_seen": 1442185216
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005687863590772318,
      "loss": 1.4621,
      "theoretical_loss": 3.5567840758163554,
      "tokens_seen": 1442250752
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005687662988966901,
      "loss": 1.4152,
      "theoretical_loss": 3.5567698335912983,
      "tokens_seen": 1442316288
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005687462387161485,
      "loss": 1.415,
      "theoretical_loss": 3.556755592194554,
      "tokens_seen": 1442381824
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005687261785356068,
      "loss": 1.5306,
      "theoretical_loss": 3.5567413516260378,
      "tokens_seen": 1442447360
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005687061183550652,
      "loss": 1.5585,
      "theoretical_loss": 3.556727111885663,
      "tokens_seen": 1442512896
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005686860581745236,
      "loss": 1.4829,
      "theoretical_loss": 3.5567128729733444,
      "tokens_seen": 1442578432
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000568665997993982,
      "loss": 1.5341,
      "theoretical_loss": 3.5566986348889955,
      "tokens_seen": 1442643968
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005686459378134403,
      "loss": 1.4886,
      "theoretical_loss": 3.556684397632531,
      "tokens_seen": 1442709504
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005686258776328988,
      "loss": 1.4253,
      "theoretical_loss": 3.556670161203865,
      "tokens_seen": 1442775040
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000568605817452357,
      "loss": 1.3564,
      "theoretical_loss": 3.5566559256029118,
      "tokens_seen": 1442840576
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005685857572718155,
      "loss": 1.3793,
      "theoretical_loss": 3.5566416908295864,
      "tokens_seen": 1442906112
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005685656970912739,
      "loss": 1.4691,
      "theoretical_loss": 3.5566274568838017,
      "tokens_seen": 1442971648
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005685456369107322,
      "loss": 1.4093,
      "theoretical_loss": 3.5566132237654733,
      "tokens_seen": 1443037184
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005685255767301906,
      "loss": 1.4101,
      "theoretical_loss": 3.556598991474515,
      "tokens_seen": 1443102720
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005685055165496489,
      "loss": 1.3285,
      "theoretical_loss": 3.556584760010841,
      "tokens_seen": 1443168256
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005684854563691073,
      "loss": 1.5818,
      "theoretical_loss": 3.5565705293743655,
      "tokens_seen": 1443233792
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005684653961885658,
      "loss": 1.4944,
      "theoretical_loss": 3.5565562995650035,
      "tokens_seen": 1443299328
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000568445336008024,
      "loss": 1.4525,
      "theoretical_loss": 3.556542070582669,
      "tokens_seen": 1443364864
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005684252758274825,
      "loss": 1.5227,
      "theoretical_loss": 3.5565278424272755,
      "tokens_seen": 1443430400
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005684052156469409,
      "loss": 1.4046,
      "theoretical_loss": 3.556513615098739,
      "tokens_seen": 1443495936
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005683851554663992,
      "loss": 1.3962,
      "theoretical_loss": 3.5564993885969733,
      "tokens_seen": 1443561472
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005683650952858576,
      "loss": 1.4966,
      "theoretical_loss": 3.5564851629218928,
      "tokens_seen": 1443627008
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000568345035105316,
      "loss": 1.4205,
      "theoretical_loss": 3.556470938073412,
      "tokens_seen": 1443692544
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005683249749247743,
      "loss": 1.3917,
      "theoretical_loss": 3.5564567140514445,
      "tokens_seen": 1443758080
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005683049147442328,
      "loss": 1.4573,
      "theoretical_loss": 3.556442490855906,
      "tokens_seen": 1443823616
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000568284854563691,
      "loss": 1.3944,
      "theoretical_loss": 3.55642826848671,
      "tokens_seen": 1443889152
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005682647943831495,
      "loss": 1.4549,
      "theoretical_loss": 3.5564140469437717,
      "tokens_seen": 1443954688
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005682447342026078,
      "loss": 1.4008,
      "theoretical_loss": 3.556399826227006,
      "tokens_seen": 1444020224
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005682246740220662,
      "loss": 1.3077,
      "theoretical_loss": 3.5563856063363257,
      "tokens_seen": 1444085760
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005682046138415246,
      "loss": 1.4575,
      "theoretical_loss": 3.5563713872716467,
      "tokens_seen": 1444151296
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000568184553660983,
      "loss": 1.4543,
      "theoretical_loss": 3.5563571690328835,
      "tokens_seen": 1444216832
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005681644934804413,
      "loss": 1.4666,
      "theoretical_loss": 3.55634295161995,
      "tokens_seen": 1444282368
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005681444332998998,
      "loss": 1.4724,
      "theoretical_loss": 3.5563287350327615,
      "tokens_seen": 1444347904
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000568124373119358,
      "loss": 1.5364,
      "theoretical_loss": 3.5563145192712318,
      "tokens_seen": 1444413440
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005681043129388165,
      "loss": 1.4846,
      "theoretical_loss": 3.5563003043352763,
      "tokens_seen": 1444478976
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005680842527582748,
      "loss": 1.5743,
      "theoretical_loss": 3.556286090224809,
      "tokens_seen": 1444544512
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005680641925777332,
      "loss": 1.3591,
      "theoretical_loss": 3.5562718769397446,
      "tokens_seen": 1444610048
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005680441323971916,
      "loss": 1.5567,
      "theoretical_loss": 3.5562576644799977,
      "tokens_seen": 1444675584
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00056802407221665,
      "loss": 1.4924,
      "theoretical_loss": 3.556243452845483,
      "tokens_seen": 1444741120
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005680040120361083,
      "loss": 1.4372,
      "theoretical_loss": 3.556229242036116,
      "tokens_seen": 1444806656
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005679839518555667,
      "loss": 1.5229,
      "theoretical_loss": 3.55621503205181,
      "tokens_seen": 1444872192
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000567963891675025,
      "loss": 1.4753,
      "theoretical_loss": 3.5562008228924804,
      "tokens_seen": 1444937728
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005679438314944835,
      "loss": 1.4594,
      "theoretical_loss": 3.556186614558042,
      "tokens_seen": 1445003264
    },
    {
      "epoch": 0.44,
      "objective/train/advantage_avg": 0.4924595057964325,
      "objective/train/docs_used": 528002,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.31443190574646,
      "objective/train/original_loss": 1.31443190574646,
      "objective/train/theoretical_loss": 3.556172407048409,
      "objective/train/tokens_used": 1465528800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24305583536624908,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504815578460693,
      "objective/train/weighted_lm_loss": 1.380705714225769,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.556172407048409,
      "tokens_seen": 1445068800
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005679237713139418,
      "loss": 1.3701,
      "theoretical_loss": 3.556172407048409,
      "tokens_seen": 1445068800
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005679037111334002,
      "loss": 1.3504,
      "theoretical_loss": 3.5561582003634964,
      "tokens_seen": 1445134336
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005678836509528585,
      "loss": 1.6032,
      "theoretical_loss": 3.556143994503219,
      "tokens_seen": 1445199872
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000567863590772317,
      "loss": 1.3273,
      "theoretical_loss": 3.5561297894674917,
      "tokens_seen": 1445265408
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005678435305917753,
      "loss": 1.4324,
      "theoretical_loss": 3.5561155852562285,
      "tokens_seen": 1445330944
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005678234704112337,
      "loss": 1.2814,
      "theoretical_loss": 3.556101381869345,
      "tokens_seen": 1445396480
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000567803410230692,
      "loss": 1.5118,
      "theoretical_loss": 3.5560871793067554,
      "tokens_seen": 1445462016
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005677833500501505,
      "loss": 1.2376,
      "theoretical_loss": 3.556072977568375,
      "tokens_seen": 1445527552
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005677632898696088,
      "loss": 1.5299,
      "theoretical_loss": 3.556058776654118,
      "tokens_seen": 1445593088
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005677432296890672,
      "loss": 1.5464,
      "theoretical_loss": 3.5560445765638997,
      "tokens_seen": 1445658624
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005677231695085255,
      "loss": 1.4263,
      "theoretical_loss": 3.556030377297635,
      "tokens_seen": 1445724160
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000567703109327984,
      "loss": 1.5084,
      "theoretical_loss": 3.5560161788552387,
      "tokens_seen": 1445789696
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005676830491474423,
      "loss": 1.53,
      "theoretical_loss": 3.5560019812366255,
      "tokens_seen": 1445855232
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005676629889669007,
      "loss": 1.4319,
      "theoretical_loss": 3.55598778444171,
      "tokens_seen": 1445920768
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005676429287863591,
      "loss": 1.4267,
      "theoretical_loss": 3.5559735884704073,
      "tokens_seen": 1445986304
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005676228686058174,
      "loss": 1.4429,
      "theoretical_loss": 3.5559593933226323,
      "tokens_seen": 1446051840
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005676028084252758,
      "loss": 1.4861,
      "theoretical_loss": 3.5559451989983,
      "tokens_seen": 1446117376
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005675827482447343,
      "loss": 1.4788,
      "theoretical_loss": 3.5559310054973254,
      "tokens_seen": 1446182912
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005675626880641926,
      "loss": 1.4099,
      "theoretical_loss": 3.5559168128196235,
      "tokens_seen": 1446248448
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000567542627883651,
      "loss": 1.5165,
      "theoretical_loss": 3.555902620965109,
      "tokens_seen": 1446313984
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005675225677031093,
      "loss": 1.4691,
      "theoretical_loss": 3.5558884299336966,
      "tokens_seen": 1446379520
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005675025075225677,
      "loss": 1.4319,
      "theoretical_loss": 3.555874239725302,
      "tokens_seen": 1446445056
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005674824473420262,
      "loss": 1.3864,
      "theoretical_loss": 3.555860050339839,
      "tokens_seen": 1446510592
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005674623871614844,
      "loss": 1.4373,
      "theoretical_loss": 3.5558458617772244,
      "tokens_seen": 1446576128
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005674423269809429,
      "loss": 1.5403,
      "theoretical_loss": 3.555831674037371,
      "tokens_seen": 1446641664
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005674222668004013,
      "loss": 1.4209,
      "theoretical_loss": 3.555817487120196,
      "tokens_seen": 1446707200
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005674022066198596,
      "loss": 1.4308,
      "theoretical_loss": 3.555803301025613,
      "tokens_seen": 1446772736
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000567382146439318,
      "loss": 1.547,
      "theoretical_loss": 3.5557891157535373,
      "tokens_seen": 1446838272
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005673620862587763,
      "loss": 1.3592,
      "theoretical_loss": 3.5557749313038842,
      "tokens_seen": 1446903808
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005673420260782347,
      "loss": 1.4354,
      "theoretical_loss": 3.555760747676569,
      "tokens_seen": 1446969344
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005673219658976932,
      "loss": 1.4208,
      "theoretical_loss": 3.555746564871506,
      "tokens_seen": 1447034880
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005673019057171514,
      "loss": 1.3532,
      "theoretical_loss": 3.555732382888611,
      "tokens_seen": 1447100416
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005672818455366099,
      "loss": 1.4203,
      "theoretical_loss": 3.5557182017277986,
      "tokens_seen": 1447165952
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005672617853560682,
      "loss": 1.371,
      "theoretical_loss": 3.555704021388985,
      "tokens_seen": 1447231488
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005672417251755266,
      "loss": 1.5026,
      "theoretical_loss": 3.5556898418720837,
      "tokens_seen": 1447297024
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000567221664994985,
      "loss": 1.4441,
      "theoretical_loss": 3.555675663177011,
      "tokens_seen": 1447362560
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005672016048144434,
      "loss": 1.3104,
      "theoretical_loss": 3.555661485303682,
      "tokens_seen": 1447428096
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005671815446339017,
      "loss": 1.461,
      "theoretical_loss": 3.555647308252011,
      "tokens_seen": 1447493632
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005671614844533602,
      "loss": 1.3752,
      "theoretical_loss": 3.555633132021914,
      "tokens_seen": 1447559168
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005671414242728184,
      "loss": 1.4517,
      "theoretical_loss": 3.5556189566133063,
      "tokens_seen": 1447624704
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005671213640922769,
      "loss": 1.4944,
      "theoretical_loss": 3.5556047820261023,
      "tokens_seen": 1447690240
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005671013039117352,
      "loss": 1.4606,
      "theoretical_loss": 3.555590608260218,
      "tokens_seen": 1447755776
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005670812437311936,
      "loss": 1.4011,
      "theoretical_loss": 3.5555764353155688,
      "tokens_seen": 1447821312
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000567061183550652,
      "loss": 1.5005,
      "theoretical_loss": 3.555562263192069,
      "tokens_seen": 1447886848
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005670411233701104,
      "loss": 1.4465,
      "theoretical_loss": 3.5555480918896345,
      "tokens_seen": 1447952384
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005670210631895687,
      "loss": 1.441,
      "theoretical_loss": 3.55553392140818,
      "tokens_seen": 1448017920
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005670010030090271,
      "loss": 1.41,
      "theoretical_loss": 3.555519751747622,
      "tokens_seen": 1448083456
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005669809428284854,
      "loss": 1.4933,
      "theoretical_loss": 3.555505582907874,
      "tokens_seen": 1448148992
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005669608826479439,
      "loss": 1.5193,
      "theoretical_loss": 3.555491414888853,
      "tokens_seen": 1448214528
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005669408224674022,
      "loss": 1.4126,
      "theoretical_loss": 3.555477247690474,
      "tokens_seen": 1448280064
    },
    {
      "epoch": 0.44,
      "objective/train/advantage_avg": 0.4920799136161804,
      "objective/train/docs_used": 529166,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.415010929107666,
      "objective/train/original_loss": 1.415010929107666,
      "objective/train/theoretical_loss": 3.5554630813126513,
      "objective/train/tokens_used": 1468805600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.243229478597641,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050444483757019,
      "objective/train/weighted_lm_loss": 1.4863362312316895,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5554630813126513,
      "tokens_seen": 1448345600
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005669207622868606,
      "loss": 1.4972,
      "theoretical_loss": 3.5554630813126513,
      "tokens_seen": 1448345600
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005669007021063189,
      "loss": 1.4139,
      "theoretical_loss": 3.555448915755301,
      "tokens_seen": 1448411136
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005668806419257774,
      "loss": 1.4918,
      "theoretical_loss": 3.555434751018339,
      "tokens_seen": 1448476672
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005668605817452357,
      "loss": 1.5807,
      "theoretical_loss": 3.5554205871016795,
      "tokens_seen": 1448542208
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005668405215646941,
      "loss": 1.486,
      "theoretical_loss": 3.5554064240052385,
      "tokens_seen": 1448607744
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005668204613841525,
      "loss": 1.437,
      "theoretical_loss": 3.5553922617289313,
      "tokens_seen": 1448673280
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005668004012036109,
      "loss": 1.5019,
      "theoretical_loss": 3.555378100272674,
      "tokens_seen": 1448738816
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005667803410230692,
      "loss": 1.5149,
      "theoretical_loss": 3.5553639396363805,
      "tokens_seen": 1448804352
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005667602808425276,
      "loss": 1.3511,
      "theoretical_loss": 3.5553497798199674,
      "tokens_seen": 1448869888
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005667402206619859,
      "loss": 1.4448,
      "theoretical_loss": 3.5553356208233504,
      "tokens_seen": 1448935424
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005667201604814444,
      "loss": 1.4673,
      "theoretical_loss": 3.555321462646444,
      "tokens_seen": 1449000960
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005667001003009027,
      "loss": 1.5171,
      "theoretical_loss": 3.5553073052891637,
      "tokens_seen": 1449066496
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005666800401203611,
      "loss": 1.6098,
      "theoretical_loss": 3.555293148751426,
      "tokens_seen": 1449132032
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005666599799398195,
      "loss": 1.383,
      "theoretical_loss": 3.5552789930331454,
      "tokens_seen": 1449197568
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005666399197592778,
      "loss": 1.4814,
      "theoretical_loss": 3.5552648381342378,
      "tokens_seen": 1449263104
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005666198595787362,
      "loss": 1.548,
      "theoretical_loss": 3.555250684054619,
      "tokens_seen": 1449328640
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005665997993981947,
      "loss": 1.6049,
      "theoretical_loss": 3.555236530794204,
      "tokens_seen": 1449394176
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005665797392176529,
      "loss": 1.4352,
      "theoretical_loss": 3.5552223783529087,
      "tokens_seen": 1449459712
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005665596790371114,
      "loss": 1.4483,
      "theoretical_loss": 3.555208226730649,
      "tokens_seen": 1449525248
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005665396188565696,
      "loss": 1.5088,
      "theoretical_loss": 3.5551940759273393,
      "tokens_seen": 1449590784
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005665195586760281,
      "loss": 1.4046,
      "theoretical_loss": 3.5551799259428964,
      "tokens_seen": 1449656320
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005664994984954865,
      "loss": 1.5455,
      "theoretical_loss": 3.555165776777235,
      "tokens_seen": 1449721856
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005664794383149448,
      "loss": 1.2971,
      "theoretical_loss": 3.5551516284302718,
      "tokens_seen": 1449787392
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005664593781344032,
      "loss": 1.5294,
      "theoretical_loss": 3.555137480901921,
      "tokens_seen": 1449852928
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005664393179538617,
      "loss": 1.4075,
      "theoretical_loss": 3.5551233341920994,
      "tokens_seen": 1449918464
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005664192577733199,
      "loss": 1.4612,
      "theoretical_loss": 3.5551091883007224,
      "tokens_seen": 1449984000
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005663991975927784,
      "loss": 1.5204,
      "theoretical_loss": 3.555095043227705,
      "tokens_seen": 1450049536
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005663791374122366,
      "loss": 1.5573,
      "theoretical_loss": 3.555080898972964,
      "tokens_seen": 1450115072
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005663590772316951,
      "loss": 1.4135,
      "theoretical_loss": 3.555066755536414,
      "tokens_seen": 1450180608
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005663390170511535,
      "loss": 1.5188,
      "theoretical_loss": 3.555052612917971,
      "tokens_seen": 1450246144
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005663189568706118,
      "loss": 1.5322,
      "theoretical_loss": 3.5550384711175513,
      "tokens_seen": 1450311680
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005662988966900702,
      "loss": 1.5717,
      "theoretical_loss": 3.55502433013507,
      "tokens_seen": 1450377216
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005662788365095286,
      "loss": 1.4816,
      "theoretical_loss": 3.555010189970443,
      "tokens_seen": 1450442752
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005662587763289869,
      "loss": 1.3778,
      "theoretical_loss": 3.5549960506235863,
      "tokens_seen": 1450508288
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005662387161484454,
      "loss": 1.4058,
      "theoretical_loss": 3.554981912094415,
      "tokens_seen": 1450573824
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005662186559679036,
      "loss": 1.3522,
      "theoretical_loss": 3.554967774382846,
      "tokens_seen": 1450639360
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005661985957873621,
      "loss": 1.5083,
      "theoretical_loss": 3.5549536374887936,
      "tokens_seen": 1450704896
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005661785356068204,
      "loss": 1.502,
      "theoretical_loss": 3.554939501412175,
      "tokens_seen": 1450770432
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005661584754262788,
      "loss": 1.4194,
      "theoretical_loss": 3.554925366152905,
      "tokens_seen": 1450835968
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005661384152457372,
      "loss": 1.436,
      "theoretical_loss": 3.5549112317109,
      "tokens_seen": 1450901504
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005661183550651956,
      "loss": 1.5153,
      "theoretical_loss": 3.5548970980860757,
      "tokens_seen": 1450967040
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005660982948846539,
      "loss": 1.4028,
      "theoretical_loss": 3.554882965278348,
      "tokens_seen": 1451032576
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005660782347041124,
      "loss": 1.4434,
      "theoretical_loss": 3.5548688332876326,
      "tokens_seen": 1451098112
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005660581745235708,
      "loss": 1.4672,
      "theoretical_loss": 3.554854702113845,
      "tokens_seen": 1451163648
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005660381143430291,
      "loss": 1.4516,
      "theoretical_loss": 3.5548405717569023,
      "tokens_seen": 1451229184
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005660180541624875,
      "loss": 1.4684,
      "theoretical_loss": 3.5548264422167195,
      "tokens_seen": 1451294720
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005659979939819458,
      "loss": 1.5415,
      "theoretical_loss": 3.554812313493212,
      "tokens_seen": 1451360256
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005659779338014043,
      "loss": 1.4106,
      "theoretical_loss": 3.554798185586297,
      "tokens_seen": 1451425792
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005659578736208626,
      "loss": 1.5315,
      "theoretical_loss": 3.5547840584958896,
      "tokens_seen": 1451491328
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000565937813440321,
      "loss": 1.5428,
      "theoretical_loss": 3.554769932221906,
      "tokens_seen": 1451556864
    },
    {
      "epoch": 0.44,
      "objective/train/advantage_avg": 0.4878488779067993,
      "objective/train/docs_used": 530451,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5102205276489258,
      "objective/train/original_loss": 1.5102205276489258,
      "objective/train/theoretical_loss": 3.5547558067642617,
      "objective/train/tokens_used": 1472082400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23977050185203552,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500037670135498,
      "objective/train/weighted_lm_loss": 1.5864441394805908,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 3.5547558067642617,
      "tokens_seen": 1451622400
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005659177532597793,
      "loss": 1.6108,
      "theoretical_loss": 3.5547558067642617,
      "tokens_seen": 1451622400
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005658976930792378,
      "loss": 1.4138,
      "theoretical_loss": 3.5547416821228737,
      "tokens_seen": 1451687936
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005658776328986961,
      "loss": 1.4228,
      "theoretical_loss": 3.554727558297657,
      "tokens_seen": 1451753472
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005658575727181545,
      "loss": 1.3975,
      "theoretical_loss": 3.5547134352885283,
      "tokens_seen": 1451819008
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005658375125376129,
      "loss": 1.5203,
      "theoretical_loss": 3.5546993130954028,
      "tokens_seen": 1451884544
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005658174523570713,
      "loss": 1.403,
      "theoretical_loss": 3.5546851917181974,
      "tokens_seen": 1451950080
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005657973921765296,
      "loss": 1.4676,
      "theoretical_loss": 3.554671071156828,
      "tokens_seen": 1452015616
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000565777331995988,
      "loss": 1.4174,
      "theoretical_loss": 3.5546569514112094,
      "tokens_seen": 1452081152
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005657572718154463,
      "loss": 1.5067,
      "theoretical_loss": 3.5546428324812593,
      "tokens_seen": 1452146688
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005657372116349048,
      "loss": 1.5404,
      "theoretical_loss": 3.554628714366893,
      "tokens_seen": 1452212224
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005657171514543631,
      "loss": 1.3902,
      "theoretical_loss": 3.554614597068027,
      "tokens_seen": 1452277760
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005656970912738215,
      "loss": 1.5004,
      "theoretical_loss": 3.5546004805845763,
      "tokens_seen": 1452343296
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005656770310932799,
      "loss": 1.4433,
      "theoretical_loss": 3.5545863649164584,
      "tokens_seen": 1452408832
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005656569709127382,
      "loss": 1.4223,
      "theoretical_loss": 3.554572250063589,
      "tokens_seen": 1452474368
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005656369107321966,
      "loss": 1.5384,
      "theoretical_loss": 3.554558136025884,
      "tokens_seen": 1452539904
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000565616850551655,
      "loss": 1.549,
      "theoretical_loss": 3.554544022803259,
      "tokens_seen": 1452605440
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005655967903711133,
      "loss": 1.4854,
      "theoretical_loss": 3.5545299103956314,
      "tokens_seen": 1452670976
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005655767301905718,
      "loss": 1.4856,
      "theoretical_loss": 3.5545157988029166,
      "tokens_seen": 1452736512
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00056555667001003,
      "loss": 1.4518,
      "theoretical_loss": 3.554501688025031,
      "tokens_seen": 1452802048
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005655366098294885,
      "loss": 1.423,
      "theoretical_loss": 3.5544875780618903,
      "tokens_seen": 1452867584
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005655165496489469,
      "loss": 1.4436,
      "theoretical_loss": 3.5544734689134114,
      "tokens_seen": 1452933120
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005654964894684052,
      "loss": 1.4342,
      "theoretical_loss": 3.5544593605795103,
      "tokens_seen": 1452998656
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005654764292878636,
      "loss": 1.5306,
      "theoretical_loss": 3.554445253060103,
      "tokens_seen": 1453064192
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005654563691073221,
      "loss": 1.4338,
      "theoretical_loss": 3.554431146355106,
      "tokens_seen": 1453129728
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005654363089267803,
      "loss": 1.479,
      "theoretical_loss": 3.5544170404644357,
      "tokens_seen": 1453195264
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005654162487462388,
      "loss": 1.4451,
      "theoretical_loss": 3.5544029353880084,
      "tokens_seen": 1453260800
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000565396188565697,
      "loss": 1.4357,
      "theoretical_loss": 3.5543888311257397,
      "tokens_seen": 1453326336
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005653761283851555,
      "loss": 1.4807,
      "theoretical_loss": 3.554374727677546,
      "tokens_seen": 1453391872
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005653560682046139,
      "loss": 1.4046,
      "theoretical_loss": 3.5543606250433446,
      "tokens_seen": 1453457408
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005653360080240722,
      "loss": 1.4951,
      "theoretical_loss": 3.554346523223051,
      "tokens_seen": 1453522944
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005653159478435306,
      "loss": 1.5401,
      "theoretical_loss": 3.5543324222165813,
      "tokens_seen": 1453588480
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000565295887662989,
      "loss": 1.4567,
      "theoretical_loss": 3.5543183220238523,
      "tokens_seen": 1453654016
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005652758274824473,
      "loss": 1.3384,
      "theoretical_loss": 3.554304222644781,
      "tokens_seen": 1453719552
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005652557673019058,
      "loss": 1.4351,
      "theoretical_loss": 3.554290124079283,
      "tokens_seen": 1453785088
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000565235707121364,
      "loss": 1.5093,
      "theoretical_loss": 3.554276026327274,
      "tokens_seen": 1453850624
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005652156469408225,
      "loss": 1.4167,
      "theoretical_loss": 3.5542619293886712,
      "tokens_seen": 1453916160
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005651955867602808,
      "loss": 1.5197,
      "theoretical_loss": 3.5542478332633918,
      "tokens_seen": 1453981696
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005651755265797392,
      "loss": 1.3606,
      "theoretical_loss": 3.5542337379513507,
      "tokens_seen": 1454047232
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005651554663991976,
      "loss": 1.62,
      "theoretical_loss": 3.5542196434524653,
      "tokens_seen": 1454112768
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000565135406218656,
      "loss": 1.3371,
      "theoretical_loss": 3.5542055497666514,
      "tokens_seen": 1454178304
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005651153460381143,
      "loss": 1.4577,
      "theoretical_loss": 3.554191456893826,
      "tokens_seen": 1454243840
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005650952858575728,
      "loss": 1.6272,
      "theoretical_loss": 3.5541773648339055,
      "tokens_seen": 1454309376
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000565075225677031,
      "loss": 1.5361,
      "theoretical_loss": 3.554163273586806,
      "tokens_seen": 1454374912
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005650551654964895,
      "loss": 1.3795,
      "theoretical_loss": 3.5541491831524445,
      "tokens_seen": 1454440448
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005650351053159478,
      "loss": 1.4409,
      "theoretical_loss": 3.5541350935307374,
      "tokens_seen": 1454505984
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005650150451354062,
      "loss": 1.5133,
      "theoretical_loss": 3.5541210047216003,
      "tokens_seen": 1454571520
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005649949849548646,
      "loss": 1.4307,
      "theoretical_loss": 3.554106916724951,
      "tokens_seen": 1454637056
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000564974924774323,
      "loss": 1.4626,
      "theoretical_loss": 3.554092829540706,
      "tokens_seen": 1454702592
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005649548645937813,
      "loss": 1.4325,
      "theoretical_loss": 3.5540787431687813,
      "tokens_seen": 1454768128
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005649348044132397,
      "loss": 1.4677,
      "theoretical_loss": 3.554064657609093,
      "tokens_seen": 1454833664
    },
    {
      "epoch": 0.44,
      "objective/train/advantage_avg": 0.49466994404792786,
      "objective/train/docs_used": 531642,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4578299522399902,
      "objective/train/original_loss": 1.4578298330307007,
      "objective/train/theoretical_loss": 3.5540505728615583,
      "objective/train/tokens_used": 1475359200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24525608122348785,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507137775421143,
      "objective/train/weighted_lm_loss": 1.53154456615448,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.5540505728615583,
      "tokens_seen": 1454899200
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005649147442326981,
      "loss": 1.414,
      "theoretical_loss": 3.5540505728615583,
      "tokens_seen": 1454899200
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005648946840521565,
      "loss": 1.4714,
      "theoretical_loss": 3.5540364889260943,
      "tokens_seen": 1454964736
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005648746238716148,
      "loss": 1.3651,
      "theoretical_loss": 3.5540224058026166,
      "tokens_seen": 1455030272
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005648545636910733,
      "loss": 1.5073,
      "theoretical_loss": 3.5540083234910425,
      "tokens_seen": 1455095808
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005648345035105316,
      "loss": 1.4838,
      "theoretical_loss": 3.5539942419912878,
      "tokens_seen": 1455161344
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00056481444332999,
      "loss": 1.4016,
      "theoretical_loss": 3.5539801613032704,
      "tokens_seen": 1455226880
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005647943831494483,
      "loss": 1.3785,
      "theoretical_loss": 3.553966081426906,
      "tokens_seen": 1455292416
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005647743229689067,
      "loss": 1.4109,
      "theoretical_loss": 3.5539520023621116,
      "tokens_seen": 1455357952
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005647542627883651,
      "loss": 1.6293,
      "theoretical_loss": 3.553937924108804,
      "tokens_seen": 1455423488
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005647342026078235,
      "loss": 1.2834,
      "theoretical_loss": 3.5539238466668994,
      "tokens_seen": 1455489024
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005647141424272818,
      "loss": 1.5203,
      "theoretical_loss": 3.553909770036315,
      "tokens_seen": 1455554560
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005646940822467403,
      "loss": 1.4228,
      "theoretical_loss": 3.5538956942169673,
      "tokens_seen": 1455620096
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005646740220661985,
      "loss": 1.4462,
      "theoretical_loss": 3.5538816192087728,
      "tokens_seen": 1455685632
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000564653961885657,
      "loss": 1.4518,
      "theoretical_loss": 3.553867545011649,
      "tokens_seen": 1455751168
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005646339017051153,
      "loss": 1.4027,
      "theoretical_loss": 3.553853471625512,
      "tokens_seen": 1455816704
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005646138415245737,
      "loss": 1.4482,
      "theoretical_loss": 3.553839399050279,
      "tokens_seen": 1455882240
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005645937813440321,
      "loss": 1.4359,
      "theoretical_loss": 3.5538253272858658,
      "tokens_seen": 1455947776
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005645737211634904,
      "loss": 1.5687,
      "theoretical_loss": 3.55381125633219,
      "tokens_seen": 1456013312
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005645536609829489,
      "loss": 1.4178,
      "theoretical_loss": 3.553797186189169,
      "tokens_seen": 1456078848
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005645336008024073,
      "loss": 1.4056,
      "theoretical_loss": 3.553783116856718,
      "tokens_seen": 1456144384
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005645135406218656,
      "loss": 1.3869,
      "theoretical_loss": 3.5537690483347557,
      "tokens_seen": 1456209920
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000564493480441324,
      "loss": 1.395,
      "theoretical_loss": 3.5537549806231974,
      "tokens_seen": 1456275456
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005644734202607825,
      "loss": 1.4415,
      "theoretical_loss": 3.5537409137219607,
      "tokens_seen": 1456340992
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005644533600802407,
      "loss": 1.3527,
      "theoretical_loss": 3.553726847630962,
      "tokens_seen": 1456406528
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005644332998996992,
      "loss": 1.4305,
      "theoretical_loss": 3.5537127823501184,
      "tokens_seen": 1456472064
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005644132397191574,
      "loss": 1.5121,
      "theoretical_loss": 3.553698717879347,
      "tokens_seen": 1456537600
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005643931795386159,
      "loss": 1.4215,
      "theoretical_loss": 3.5536846542185647,
      "tokens_seen": 1456603136
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005643731193580743,
      "loss": 1.3898,
      "theoretical_loss": 3.5536705913676885,
      "tokens_seen": 1456668672
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005643530591775326,
      "loss": 1.414,
      "theoretical_loss": 3.5536565293266342,
      "tokens_seen": 1456734208
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000564332998996991,
      "loss": 1.5496,
      "theoretical_loss": 3.5536424680953203,
      "tokens_seen": 1456799744
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005643129388164494,
      "loss": 1.624,
      "theoretical_loss": 3.5536284076736626,
      "tokens_seen": 1456865280
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005642928786359077,
      "loss": 1.484,
      "theoretical_loss": 3.553614348061579,
      "tokens_seen": 1456930816
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005642728184553662,
      "loss": 1.5714,
      "theoretical_loss": 3.553600289258986,
      "tokens_seen": 1456996352
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005642527582748244,
      "loss": 1.4164,
      "theoretical_loss": 3.5535862312658004,
      "tokens_seen": 1457061888
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005642326980942829,
      "loss": 1.3587,
      "theoretical_loss": 3.553572174081939,
      "tokens_seen": 1457127424
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005642126379137412,
      "loss": 1.5883,
      "theoretical_loss": 3.5535581177073197,
      "tokens_seen": 1457192960
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005641925777331996,
      "loss": 1.5255,
      "theoretical_loss": 3.553544062141859,
      "tokens_seen": 1457258496
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000564172517552658,
      "loss": 1.5278,
      "theoretical_loss": 3.553530007385474,
      "tokens_seen": 1457324032
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005641524573721164,
      "loss": 1.3974,
      "theoretical_loss": 3.5535159534380814,
      "tokens_seen": 1457389568
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005641323971915747,
      "loss": 1.3822,
      "theoretical_loss": 3.5535019002995982,
      "tokens_seen": 1457455104
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005641123370110332,
      "loss": 1.5126,
      "theoretical_loss": 3.5534878479699423,
      "tokens_seen": 1457520640
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005640922768304915,
      "loss": 1.4261,
      "theoretical_loss": 3.5534737964490306,
      "tokens_seen": 1457586176
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005640722166499499,
      "loss": 1.5034,
      "theoretical_loss": 3.553459745736779,
      "tokens_seen": 1457651712
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005640521564694082,
      "loss": 1.5053,
      "theoretical_loss": 3.553445695833106,
      "tokens_seen": 1457717248
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005640320962888666,
      "loss": 1.5999,
      "theoretical_loss": 3.5534316467379288,
      "tokens_seen": 1457782784
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000564012036108325,
      "loss": 1.5841,
      "theoretical_loss": 3.553417598451163,
      "tokens_seen": 1457848320
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005639919759277834,
      "loss": 1.5327,
      "theoretical_loss": 3.5534035509727273,
      "tokens_seen": 1457913856
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005639719157472417,
      "loss": 1.4896,
      "theoretical_loss": 3.553389504302538,
      "tokens_seen": 1457979392
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005639518555667001,
      "loss": 1.3913,
      "theoretical_loss": 3.5533754584405126,
      "tokens_seen": 1458044928
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005639317953861585,
      "loss": 1.4974,
      "theoretical_loss": 3.5533614133865683,
      "tokens_seen": 1458110464
    },
    {
      "epoch": 0.44,
      "objective/train/advantage_avg": 0.4931744933128357,
      "objective/train/docs_used": 532886,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3077855110168457,
      "objective/train/original_loss": 1.3077855110168457,
      "objective/train/theoretical_loss": 3.553347369140622,
      "objective/train/tokens_used": 1478636000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24354708194732666,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505554676055908,
      "objective/train/weighted_lm_loss": 1.3738805055618286,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.553347369140622,
      "tokens_seen": 1458176000
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005639117352056169,
      "loss": 1.4544,
      "theoretical_loss": 3.553347369140622,
      "tokens_seen": 1458176000
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005638916750250752,
      "loss": 1.4821,
      "theoretical_loss": 3.5533333257025914,
      "tokens_seen": 1458241536
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005638716148445337,
      "loss": 1.3153,
      "theoretical_loss": 3.553319283072393,
      "tokens_seen": 1458307072
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000563851554663992,
      "loss": 1.4785,
      "theoretical_loss": 3.553305241249945,
      "tokens_seen": 1458372608
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005638314944834504,
      "loss": 1.5813,
      "theoretical_loss": 3.5532912002351633,
      "tokens_seen": 1458438144
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005638114343029087,
      "loss": 1.548,
      "theoretical_loss": 3.5532771600279665,
      "tokens_seen": 1458503680
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005637913741223671,
      "loss": 1.3891,
      "theoretical_loss": 3.553263120628271,
      "tokens_seen": 1458569216
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005637713139418255,
      "loss": 1.5121,
      "theoretical_loss": 3.553249082035995,
      "tokens_seen": 1458634752
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005637512537612839,
      "loss": 1.3813,
      "theoretical_loss": 3.553235044251055,
      "tokens_seen": 1458700288
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005637311935807422,
      "loss": 1.5152,
      "theoretical_loss": 3.5532210072733683,
      "tokens_seen": 1458765824
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005637111334002007,
      "loss": 1.4396,
      "theoretical_loss": 3.553206971102852,
      "tokens_seen": 1458831360
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005636910732196589,
      "loss": 1.4448,
      "theoretical_loss": 3.5531929357394247,
      "tokens_seen": 1458896896
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005636710130391174,
      "loss": 1.4123,
      "theoretical_loss": 3.553178901183002,
      "tokens_seen": 1458962432
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005636509528585757,
      "loss": 1.4694,
      "theoretical_loss": 3.5531648674335026,
      "tokens_seen": 1459027968
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005636308926780341,
      "loss": 1.4377,
      "theoretical_loss": 3.5531508344908436,
      "tokens_seen": 1459093504
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005636108324974925,
      "loss": 1.5219,
      "theoretical_loss": 3.553136802354942,
      "tokens_seen": 1459159040
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005635907723169508,
      "loss": 1.3763,
      "theoretical_loss": 3.553122771025715,
      "tokens_seen": 1459224576
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005635707121364092,
      "loss": 1.358,
      "theoretical_loss": 3.5531087405030806,
      "tokens_seen": 1459290112
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005635506519558677,
      "loss": 1.6151,
      "theoretical_loss": 3.5530947107869557,
      "tokens_seen": 1459355648
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005635305917753259,
      "loss": 1.3999,
      "theoretical_loss": 3.5530806818772582,
      "tokens_seen": 1459421184
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005635105315947844,
      "loss": 1.5919,
      "theoretical_loss": 3.5530666537739055,
      "tokens_seen": 1459486720
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005634904714142428,
      "loss": 1.5455,
      "theoretical_loss": 3.5530526264768145,
      "tokens_seen": 1459552256
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005634704112337011,
      "loss": 1.4177,
      "theoretical_loss": 3.5530385999859035,
      "tokens_seen": 1459617792
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005634503510531595,
      "loss": 1.4246,
      "theoretical_loss": 3.553024574301089,
      "tokens_seen": 1459683328
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005634302908726178,
      "loss": 1.5497,
      "theoretical_loss": 3.5530105494222894,
      "tokens_seen": 1459748864
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005634102306920762,
      "loss": 1.5011,
      "theoretical_loss": 3.552996525349421,
      "tokens_seen": 1459814400
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005633901705115347,
      "loss": 1.3755,
      "theoretical_loss": 3.5529825020824033,
      "tokens_seen": 1459879936
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005633701103309929,
      "loss": 1.5157,
      "theoretical_loss": 3.5529684796211516,
      "tokens_seen": 1459945472
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005633500501504514,
      "loss": 1.3551,
      "theoretical_loss": 3.552954457965585,
      "tokens_seen": 1460011008
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005633299899699097,
      "loss": 1.5209,
      "theoretical_loss": 3.5529404371156206,
      "tokens_seen": 1460076544
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005633099297893681,
      "loss": 1.4943,
      "theoretical_loss": 3.5529264170711756,
      "tokens_seen": 1460142080
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005632898696088265,
      "loss": 1.3646,
      "theoretical_loss": 3.5529123978321677,
      "tokens_seen": 1460207616
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005632698094282848,
      "loss": 1.5189,
      "theoretical_loss": 3.552898379398515,
      "tokens_seen": 1460273152
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005632497492477432,
      "loss": 1.4973,
      "theoretical_loss": 3.5528843617701344,
      "tokens_seen": 1460338688
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005632296890672016,
      "loss": 1.5224,
      "theoretical_loss": 3.552870344946944,
      "tokens_seen": 1460404224
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005632096288866599,
      "loss": 1.4561,
      "theoretical_loss": 3.552856328928861,
      "tokens_seen": 1460469760
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005631895687061184,
      "loss": 1.3929,
      "theoretical_loss": 3.552842313715803,
      "tokens_seen": 1460535296
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005631695085255767,
      "loss": 1.4871,
      "theoretical_loss": 3.5528282993076887,
      "tokens_seen": 1460600832
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005631494483450351,
      "loss": 1.5483,
      "theoretical_loss": 3.5528142857044345,
      "tokens_seen": 1460666368
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005631293881644935,
      "loss": 1.433,
      "theoretical_loss": 3.5528002729059587,
      "tokens_seen": 1460731904
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005631093279839519,
      "loss": 1.3644,
      "theoretical_loss": 3.5527862609121783,
      "tokens_seen": 1460797440
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005630892678034102,
      "loss": 1.4279,
      "theoretical_loss": 3.552772249723012,
      "tokens_seen": 1460862976
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005630692076228686,
      "loss": 1.6055,
      "theoretical_loss": 3.5527582393383765,
      "tokens_seen": 1460928512
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000563049147442327,
      "loss": 1.5201,
      "theoretical_loss": 3.552744229758191,
      "tokens_seen": 1460994048
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005630290872617854,
      "loss": 1.5478,
      "theoretical_loss": 3.5527302209823715,
      "tokens_seen": 1461059584
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005630090270812438,
      "loss": 1.5239,
      "theoretical_loss": 3.5527162130108363,
      "tokens_seen": 1461125120
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005629889669007021,
      "loss": 1.4287,
      "theoretical_loss": 3.5527022058435036,
      "tokens_seen": 1461190656
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005629689067201605,
      "loss": 1.3729,
      "theoretical_loss": 3.552688199480291,
      "tokens_seen": 1461256192
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005629488465396189,
      "loss": 1.5365,
      "theoretical_loss": 3.552674193921116,
      "tokens_seen": 1461321728
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005629287863590773,
      "loss": 1.4008,
      "theoretical_loss": 3.5526601891658967,
      "tokens_seen": 1461387264
    },
    {
      "epoch": 0.44,
      "objective/train/advantage_avg": 0.4928682744503021,
      "objective/train/docs_used": 534190,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3752458095550537,
      "objective/train/original_loss": 1.3752458095550537,
      "objective/train/theoretical_loss": 3.552646185214551,
      "objective/train/tokens_used": 1481912800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2432597577571869,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505235195159912,
      "objective/train/weighted_lm_loss": 1.4446871280670166,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.552646185214551,
      "tokens_seen": 1461452800
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005629087261785356,
      "loss": 1.4842,
      "theoretical_loss": 3.552646185214551,
      "tokens_seen": 1461452800
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000562888665997994,
      "loss": 1.4738,
      "theoretical_loss": 3.552632182066996,
      "tokens_seen": 1461518336
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005628686058174524,
      "loss": 1.4611,
      "theoretical_loss": 3.5526181797231504,
      "tokens_seen": 1461583872
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005628485456369108,
      "loss": 1.4113,
      "theoretical_loss": 3.552604178182931,
      "tokens_seen": 1461649408
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005628284854563691,
      "loss": 1.4073,
      "theoretical_loss": 3.552590177446257,
      "tokens_seen": 1461714944
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005628084252758275,
      "loss": 1.42,
      "theoretical_loss": 3.5525761775130453,
      "tokens_seen": 1461780480
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005627883650952859,
      "loss": 1.4226,
      "theoretical_loss": 3.552562178383214,
      "tokens_seen": 1461846016
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005627683049147443,
      "loss": 1.4556,
      "theoretical_loss": 3.5525481800566814,
      "tokens_seen": 1461911552
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005627482447342026,
      "loss": 1.4891,
      "theoretical_loss": 3.5525341825333645,
      "tokens_seen": 1461977088
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005627281845536611,
      "loss": 1.5011,
      "theoretical_loss": 3.552520185813182,
      "tokens_seen": 1462042624
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005627081243731193,
      "loss": 1.5,
      "theoretical_loss": 3.552506189896051,
      "tokens_seen": 1462108160
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005626880641925778,
      "loss": 1.5023,
      "theoretical_loss": 3.552492194781891,
      "tokens_seen": 1462173696
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005626680040120361,
      "loss": 1.5097,
      "theoretical_loss": 3.552478200470618,
      "tokens_seen": 1462239232
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005626479438314945,
      "loss": 1.4684,
      "theoretical_loss": 3.552464206962151,
      "tokens_seen": 1462304768
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005626278836509529,
      "loss": 1.4309,
      "theoretical_loss": 3.5524502142564085,
      "tokens_seen": 1462370304
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005626078234704112,
      "loss": 1.5701,
      "theoretical_loss": 3.5524362223533075,
      "tokens_seen": 1462435840
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005625877632898696,
      "loss": 1.5233,
      "theoretical_loss": 3.552422231252766,
      "tokens_seen": 1462501376
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005625677031093281,
      "loss": 1.3313,
      "theoretical_loss": 3.552408240954702,
      "tokens_seen": 1462566912
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005625476429287863,
      "loss": 1.46,
      "theoretical_loss": 3.5523942514590345,
      "tokens_seen": 1462632448
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005625275827482448,
      "loss": 1.5413,
      "theoretical_loss": 3.5523802627656806,
      "tokens_seen": 1462697984
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005625075225677032,
      "loss": 1.4653,
      "theoretical_loss": 3.552366274874559,
      "tokens_seen": 1462763520
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005624874623871615,
      "loss": 1.5669,
      "theoretical_loss": 3.552352287785587,
      "tokens_seen": 1462829056
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005624674022066199,
      "loss": 1.4601,
      "theoretical_loss": 3.5523383014986827,
      "tokens_seen": 1462894592
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005624473420260782,
      "loss": 1.45,
      "theoretical_loss": 3.552324316013765,
      "tokens_seen": 1462960128
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005624272818455366,
      "loss": 1.4163,
      "theoretical_loss": 3.5523103313307516,
      "tokens_seen": 1463025664
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005624072216649951,
      "loss": 1.5252,
      "theoretical_loss": 3.5522963474495604,
      "tokens_seen": 1463091200
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005623871614844533,
      "loss": 1.5623,
      "theoretical_loss": 3.5522823643701096,
      "tokens_seen": 1463156736
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005623671013039118,
      "loss": 1.4799,
      "theoretical_loss": 3.552268382092317,
      "tokens_seen": 1463222272
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00056234704112337,
      "loss": 1.4782,
      "theoretical_loss": 3.5522544006161016,
      "tokens_seen": 1463287808
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005623269809428285,
      "loss": 1.4373,
      "theoretical_loss": 3.5522404199413806,
      "tokens_seen": 1463353344
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005623069207622869,
      "loss": 1.4127,
      "theoretical_loss": 3.5522264400680728,
      "tokens_seen": 1463418880
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005622868605817452,
      "loss": 1.4873,
      "theoretical_loss": 3.5522124609960963,
      "tokens_seen": 1463484416
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005622668004012036,
      "loss": 1.5027,
      "theoretical_loss": 3.5521984827253688,
      "tokens_seen": 1463549952
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000562246740220662,
      "loss": 1.4322,
      "theoretical_loss": 3.552184505255809,
      "tokens_seen": 1463615488
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005622266800401203,
      "loss": 1.4198,
      "theoretical_loss": 3.552170528587335,
      "tokens_seen": 1463681024
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005622066198595788,
      "loss": 1.444,
      "theoretical_loss": 3.5521565527198655,
      "tokens_seen": 1463746560
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005621865596790371,
      "loss": 1.3911,
      "theoretical_loss": 3.5521425776533175,
      "tokens_seen": 1463812096
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005621664994984955,
      "loss": 1.4109,
      "theoretical_loss": 3.55212860338761,
      "tokens_seen": 1463877632
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005621464393179539,
      "loss": 1.4712,
      "theoretical_loss": 3.5521146299226616,
      "tokens_seen": 1463943168
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005621263791374123,
      "loss": 1.4645,
      "theoretical_loss": 3.55210065725839,
      "tokens_seen": 1464008704
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005621063189568706,
      "loss": 1.4423,
      "theoretical_loss": 3.5520866853947135,
      "tokens_seen": 1464074240
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000562086258776329,
      "loss": 1.5594,
      "theoretical_loss": 3.552072714331551,
      "tokens_seen": 1464139776
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005620661985957873,
      "loss": 1.3724,
      "theoretical_loss": 3.55205874406882,
      "tokens_seen": 1464205312
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005620461384152458,
      "loss": 1.3552,
      "theoretical_loss": 3.5520447746064394,
      "tokens_seen": 1464270848
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005620260782347041,
      "loss": 1.4682,
      "theoretical_loss": 3.5520308059443275,
      "tokens_seen": 1464336384
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005620060180541625,
      "loss": 1.474,
      "theoretical_loss": 3.5520168380824018,
      "tokens_seen": 1464401920
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005619859578736208,
      "loss": 1.516,
      "theoretical_loss": 3.5520028710205818,
      "tokens_seen": 1464467456
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005619658976930793,
      "loss": 1.4625,
      "theoretical_loss": 3.551988904758786,
      "tokens_seen": 1464532992
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005619458375125376,
      "loss": 1.5336,
      "theoretical_loss": 3.5519749392969313,
      "tokens_seen": 1464598528
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000561925777331996,
      "loss": 1.5186,
      "theoretical_loss": 3.5519609746349365,
      "tokens_seen": 1464664064
    },
    {
      "epoch": 0.44,
      "objective/train/advantage_avg": 0.4921218156814575,
      "objective/train/docs_used": 534938,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3705878257751465,
      "objective/train/original_loss": 1.3705878257751465,
      "objective/train/theoretical_loss": 3.5519470107727216,
      "objective/train/tokens_used": 1485189600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2429831326007843,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504473447799683,
      "objective/train/weighted_lm_loss": 1.4397681951522827,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5519470107727216,
      "tokens_seen": 1464729600
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005619057171514543,
      "loss": 1.4344,
      "theoretical_loss": 3.5519470107727216,
      "tokens_seen": 1464729600
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005618856569709127,
      "loss": 1.4964,
      "theoretical_loss": 3.5519330477102034,
      "tokens_seen": 1464795136
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005618655967903711,
      "loss": 1.4668,
      "theoretical_loss": 3.5519190854473006,
      "tokens_seen": 1464860672
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005618455366098295,
      "loss": 1.3976,
      "theoretical_loss": 3.551905123983932,
      "tokens_seen": 1464926208
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005618254764292878,
      "loss": 1.5122,
      "theoretical_loss": 3.551891163320016,
      "tokens_seen": 1464991744
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005618054162487463,
      "loss": 1.4852,
      "theoretical_loss": 3.551877203455471,
      "tokens_seen": 1465057280
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005617853560682046,
      "loss": 1.4904,
      "theoretical_loss": 3.5518632443902156,
      "tokens_seen": 1465122816
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000561765295887663,
      "loss": 1.3685,
      "theoretical_loss": 3.5518492861241677,
      "tokens_seen": 1465188352
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005617452357071214,
      "loss": 1.5159,
      "theoretical_loss": 3.551835328657246,
      "tokens_seen": 1465253888
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005617251755265797,
      "loss": 1.5448,
      "theoretical_loss": 3.5518213719893694,
      "tokens_seen": 1465319424
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005617051153460381,
      "loss": 1.3915,
      "theoretical_loss": 3.5518074161204565,
      "tokens_seen": 1465384960
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005616850551654965,
      "loss": 1.5434,
      "theoretical_loss": 3.5517934610504254,
      "tokens_seen": 1465450496
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005616649949849548,
      "loss": 1.5319,
      "theoretical_loss": 3.5517795067791953,
      "tokens_seen": 1465516032
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005616449348044133,
      "loss": 1.5758,
      "theoretical_loss": 3.5517655533066836,
      "tokens_seen": 1465581568
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005616248746238715,
      "loss": 1.4144,
      "theoretical_loss": 3.5517516006328096,
      "tokens_seen": 1465647104
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00056160481444333,
      "loss": 1.4125,
      "theoretical_loss": 3.5517376487574923,
      "tokens_seen": 1465712640
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005615847542627884,
      "loss": 1.5107,
      "theoretical_loss": 3.551723697680649,
      "tokens_seen": 1465778176
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005615646940822467,
      "loss": 1.3855,
      "theoretical_loss": 3.5517097474022004,
      "tokens_seen": 1465843712
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005615446339017052,
      "loss": 1.5661,
      "theoretical_loss": 3.5516957979220627,
      "tokens_seen": 1465909248
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005615245737211636,
      "loss": 1.5633,
      "theoretical_loss": 3.551681849240156,
      "tokens_seen": 1465974784
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005615045135406219,
      "loss": 1.4224,
      "theoretical_loss": 3.551667901356399,
      "tokens_seen": 1466040320
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005614844533600803,
      "loss": 1.5114,
      "theoretical_loss": 3.5516539542707095,
      "tokens_seen": 1466105856
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005614643931795386,
      "loss": 1.3896,
      "theoretical_loss": 3.551640007983007,
      "tokens_seen": 1466171392
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000561444332998997,
      "loss": 1.3414,
      "theoretical_loss": 3.5516260624932094,
      "tokens_seen": 1466236928
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005614242728184555,
      "loss": 1.4689,
      "theoretical_loss": 3.551612117801236,
      "tokens_seen": 1466302464
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005614042126379137,
      "loss": 1.4458,
      "theoretical_loss": 3.5515981739070055,
      "tokens_seen": 1466368000
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005613841524573722,
      "loss": 1.397,
      "theoretical_loss": 3.551584230810436,
      "tokens_seen": 1466433536
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005613640922768305,
      "loss": 1.4979,
      "theoretical_loss": 3.5515702885114466,
      "tokens_seen": 1466499072
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005613440320962889,
      "loss": 1.5068,
      "theoretical_loss": 3.551556347009957,
      "tokens_seen": 1466564608
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005613239719157473,
      "loss": 1.3693,
      "theoretical_loss": 3.551542406305884,
      "tokens_seen": 1466630144
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005613039117352056,
      "loss": 1.4991,
      "theoretical_loss": 3.551528466399148,
      "tokens_seen": 1466695680
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000561283851554664,
      "loss": 1.3836,
      "theoretical_loss": 3.551514527289667,
      "tokens_seen": 1466761216
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005612637913741224,
      "loss": 1.5765,
      "theoretical_loss": 3.5515005889773597,
      "tokens_seen": 1466826752
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005612437311935807,
      "loss": 1.5564,
      "theoretical_loss": 3.5514866514621453,
      "tokens_seen": 1466892288
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005612236710130392,
      "loss": 1.3807,
      "theoretical_loss": 3.551472714743942,
      "tokens_seen": 1466957824
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005612036108324975,
      "loss": 1.3882,
      "theoretical_loss": 3.5514587788226697,
      "tokens_seen": 1467023360
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005611835506519559,
      "loss": 1.4933,
      "theoretical_loss": 3.5514448436982464,
      "tokens_seen": 1467088896
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005611634904714143,
      "loss": 1.3816,
      "theoretical_loss": 3.551430909370591,
      "tokens_seen": 1467154432
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005611434302908727,
      "loss": 1.4507,
      "theoretical_loss": 3.551416975839623,
      "tokens_seen": 1467219968
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000561123370110331,
      "loss": 1.4304,
      "theoretical_loss": 3.5514030431052603,
      "tokens_seen": 1467285504
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005611033099297894,
      "loss": 1.4426,
      "theoretical_loss": 3.5513891111674223,
      "tokens_seen": 1467351040
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005610832497492477,
      "loss": 1.4793,
      "theoretical_loss": 3.551375180026028,
      "tokens_seen": 1467416576
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005610631895687062,
      "loss": 1.4262,
      "theoretical_loss": 3.551361249680996,
      "tokens_seen": 1467482112
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005610431293881645,
      "loss": 1.4446,
      "theoretical_loss": 3.551347320132245,
      "tokens_seen": 1467547648
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005610230692076229,
      "loss": 1.3925,
      "theoretical_loss": 3.5513333913796945,
      "tokens_seen": 1467613184
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005610030090270812,
      "loss": 1.4037,
      "theoretical_loss": 3.5513194634232637,
      "tokens_seen": 1467678720
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005609829488465397,
      "loss": 1.4923,
      "theoretical_loss": 3.5513055362628707,
      "tokens_seen": 1467744256
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000560962888665998,
      "loss": 1.5782,
      "theoretical_loss": 3.551291609898435,
      "tokens_seen": 1467809792
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005609428284854564,
      "loss": 1.5133,
      "theoretical_loss": 3.5512776843298752,
      "tokens_seen": 1467875328
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005609227683049147,
      "loss": 1.4509,
      "theoretical_loss": 3.55126375955711,
      "tokens_seen": 1467940864
    },
    {
      "epoch": 0.44,
      "objective/train/advantage_avg": 0.49261385202407837,
      "objective/train/docs_used": 535989,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.349312663078308,
      "objective/train/original_loss": 1.3493125438690186,
      "objective/train/theoretical_loss": 3.5512498355800597,
      "objective/train/tokens_used": 1488466400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24329595267772675,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504982471466064,
      "objective/train/weighted_lm_loss": 1.417466640472412,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5512498355800597,
      "tokens_seen": 1468006400
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005609027081243731,
      "loss": 1.4688,
      "theoretical_loss": 3.5512498355800597,
      "tokens_seen": 1468006400
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005608826479438315,
      "loss": 1.4477,
      "theoretical_loss": 3.551235912398642,
      "tokens_seen": 1468071936
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005608625877632899,
      "loss": 1.4202,
      "theoretical_loss": 3.5512219900127766,
      "tokens_seen": 1468137472
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005608425275827482,
      "loss": 1.4198,
      "theoretical_loss": 3.551208068422383,
      "tokens_seen": 1468203008
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005608224674022067,
      "loss": 1.4739,
      "theoretical_loss": 3.5511941476273785,
      "tokens_seen": 1468268544
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.000560802407221665,
      "loss": 1.4137,
      "theoretical_loss": 3.5511802276276843,
      "tokens_seen": 1468334080
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005607823470411234,
      "loss": 1.4978,
      "theoretical_loss": 3.5511663084232175,
      "tokens_seen": 1468399616
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0005607622868605818,
      "loss": 1.395,
      "theoretical_loss": 3.5511523900138986,
      "tokens_seen": 1468465152
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005607422266800401,
      "loss": 1.3712,
      "theoretical_loss": 3.551138472399646,
      "tokens_seen": 1468530688
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005607221664994985,
      "loss": 1.4649,
      "theoretical_loss": 3.5511245555803788,
      "tokens_seen": 1468596224
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000560702106318957,
      "loss": 1.4358,
      "theoretical_loss": 3.551110639556017,
      "tokens_seen": 1468661760
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005606820461384152,
      "loss": 1.4785,
      "theoretical_loss": 3.5510967243264786,
      "tokens_seen": 1468727296
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005606619859578737,
      "loss": 1.4482,
      "theoretical_loss": 3.5510828098916836,
      "tokens_seen": 1468792832
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005606419257773319,
      "loss": 1.4123,
      "theoretical_loss": 3.55106889625155,
      "tokens_seen": 1468858368
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005606218655967904,
      "loss": 1.5565,
      "theoretical_loss": 3.551054983405998,
      "tokens_seen": 1468923904
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005606018054162488,
      "loss": 1.5352,
      "theoretical_loss": 3.551041071354947,
      "tokens_seen": 1468989440
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005605817452357071,
      "loss": 1.3462,
      "theoretical_loss": 3.5510271600983154,
      "tokens_seen": 1469054976
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005605616850551655,
      "loss": 1.5274,
      "theoretical_loss": 3.5510132496360223,
      "tokens_seen": 1469120512
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000560541624874624,
      "loss": 1.506,
      "theoretical_loss": 3.550999339967987,
      "tokens_seen": 1469186048
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005605215646940822,
      "loss": 1.391,
      "theoretical_loss": 3.55098543109413,
      "tokens_seen": 1469251584
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005605015045135407,
      "loss": 1.3746,
      "theoretical_loss": 3.5509715230143692,
      "tokens_seen": 1469317120
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005604814443329989,
      "loss": 1.3888,
      "theoretical_loss": 3.550957615728624,
      "tokens_seen": 1469382656
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005604613841524574,
      "loss": 1.5406,
      "theoretical_loss": 3.5509437092368135,
      "tokens_seen": 1469448192
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005604413239719158,
      "loss": 1.452,
      "theoretical_loss": 3.5509298035388577,
      "tokens_seen": 1469513728
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005604212637913741,
      "loss": 1.414,
      "theoretical_loss": 3.550915898634676,
      "tokens_seen": 1469579264
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005604012036108325,
      "loss": 1.3036,
      "theoretical_loss": 3.5509019945241866,
      "tokens_seen": 1469644800
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005603811434302909,
      "loss": 1.5561,
      "theoretical_loss": 3.5508880912073093,
      "tokens_seen": 1469710336
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005603610832497492,
      "loss": 1.3147,
      "theoretical_loss": 3.550874188683964,
      "tokens_seen": 1469775872
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005603410230692077,
      "loss": 1.4481,
      "theoretical_loss": 3.550860286954069,
      "tokens_seen": 1469841408
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005603209628886659,
      "loss": 1.3756,
      "theoretical_loss": 3.5508463860175445,
      "tokens_seen": 1469906944
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005603009027081244,
      "loss": 1.4444,
      "theoretical_loss": 3.5508324858743094,
      "tokens_seen": 1469972480
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005602808425275827,
      "loss": 1.4531,
      "theoretical_loss": 3.5508185865242834,
      "tokens_seen": 1470038016
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005602607823470411,
      "loss": 1.3526,
      "theoretical_loss": 3.5508046879673856,
      "tokens_seen": 1470103552
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005602407221664995,
      "loss": 1.5848,
      "theoretical_loss": 3.5507907902035356,
      "tokens_seen": 1470169088
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005602206619859579,
      "loss": 1.5324,
      "theoretical_loss": 3.550776893232652,
      "tokens_seen": 1470234624
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005602006018054162,
      "loss": 1.4605,
      "theoretical_loss": 3.5507629970546555,
      "tokens_seen": 1470300160
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005601805416248747,
      "loss": 1.4038,
      "theoretical_loss": 3.550749101669465,
      "tokens_seen": 1470365696
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005601604814443329,
      "loss": 1.3602,
      "theoretical_loss": 3.5507352070769995,
      "tokens_seen": 1470431232
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005601404212637914,
      "loss": 1.4571,
      "theoretical_loss": 3.5507213132771787,
      "tokens_seen": 1470496768
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005601203610832497,
      "loss": 1.3834,
      "theoretical_loss": 3.550707420269923,
      "tokens_seen": 1470562304
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005601003009027081,
      "loss": 1.3818,
      "theoretical_loss": 3.5506935280551497,
      "tokens_seen": 1470627840
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005600802407221665,
      "loss": 1.4221,
      "theoretical_loss": 3.5506796366327804,
      "tokens_seen": 1470693376
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005600601805416249,
      "loss": 1.4714,
      "theoretical_loss": 3.5506657460027338,
      "tokens_seen": 1470758912
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005600401203610833,
      "loss": 1.4866,
      "theoretical_loss": 3.5506518561649294,
      "tokens_seen": 1470824448
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005600200601805416,
      "loss": 1.5054,
      "theoretical_loss": 3.5506379671192865,
      "tokens_seen": 1470889984
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005600000000000001,
      "loss": 1.5205,
      "theoretical_loss": 3.5506240788657246,
      "tokens_seen": 1470955520
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005599799398194584,
      "loss": 1.4562,
      "theoretical_loss": 3.5506101914041643,
      "tokens_seen": 1471021056
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005599598796389168,
      "loss": 1.5006,
      "theoretical_loss": 3.5505963047345235,
      "tokens_seen": 1471086592
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005599398194583751,
      "loss": 1.3926,
      "theoretical_loss": 3.550582418856723,
      "tokens_seen": 1471152128
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005599197592778335,
      "loss": 1.3739,
      "theoretical_loss": 3.550568533770682,
      "tokens_seen": 1471217664
    },
    {
      "epoch": 0.45,
      "objective/train/advantage_avg": 0.4925982654094696,
      "objective/train/docs_used": 537210,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3364008665084839,
      "objective/train/original_loss": 1.3364008665084839,
      "objective/train/theoretical_loss": 3.55055464947632,
      "objective/train/tokens_used": 1491743200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24316617846488953,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504961013793945,
      "objective/train/weighted_lm_loss": 1.4037643671035767,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.55055464947632,
      "tokens_seen": 1471283200
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005598996990972919,
      "loss": 1.4286,
      "theoretical_loss": 3.55055464947632,
      "tokens_seen": 1471283200
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005598796389167503,
      "loss": 1.4412,
      "theoretical_loss": 3.5505407659735564,
      "tokens_seen": 1471348736
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005598595787362086,
      "loss": 1.4719,
      "theoretical_loss": 3.550526883262312,
      "tokens_seen": 1471414272
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005598395185556671,
      "loss": 1.5401,
      "theoretical_loss": 3.5505130013425044,
      "tokens_seen": 1471479808
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005598194583751254,
      "loss": 1.4347,
      "theoretical_loss": 3.550499120214055,
      "tokens_seen": 1471545344
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005597993981945838,
      "loss": 1.5263,
      "theoretical_loss": 3.5504852398768825,
      "tokens_seen": 1471610880
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005597793380140422,
      "loss": 1.3773,
      "theoretical_loss": 3.550471360330907,
      "tokens_seen": 1471676416
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005597592778335005,
      "loss": 1.4611,
      "theoretical_loss": 3.550457481576048,
      "tokens_seen": 1471741952
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005597392176529589,
      "loss": 1.3793,
      "theoretical_loss": 3.550443603612225,
      "tokens_seen": 1471807488
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005597191574724173,
      "loss": 1.4998,
      "theoretical_loss": 3.550429726439358,
      "tokens_seen": 1471873024
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005596990972918756,
      "loss": 1.3685,
      "theoretical_loss": 3.550415850057367,
      "tokens_seen": 1471938560
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005596790371113341,
      "loss": 1.4942,
      "theoretical_loss": 3.550401974466171,
      "tokens_seen": 1472004096
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005596589769307923,
      "loss": 1.4638,
      "theoretical_loss": 3.55038809966569,
      "tokens_seen": 1472069632
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005596389167502508,
      "loss": 1.4195,
      "theoretical_loss": 3.550374225655844,
      "tokens_seen": 1472135168
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005596188565697092,
      "loss": 1.3921,
      "theoretical_loss": 3.5503603524365523,
      "tokens_seen": 1472200704
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005595987963891675,
      "loss": 1.4704,
      "theoretical_loss": 3.5503464800077347,
      "tokens_seen": 1472266240
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005595787362086259,
      "loss": 1.5115,
      "theoretical_loss": 3.5503326083693114,
      "tokens_seen": 1472331776
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005595586760280844,
      "loss": 1.5134,
      "theoretical_loss": 3.550318737521202,
      "tokens_seen": 1472397312
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005595386158475426,
      "loss": 1.5901,
      "theoretical_loss": 3.5503048674633266,
      "tokens_seen": 1472462848
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005595185556670011,
      "loss": 1.4189,
      "theoretical_loss": 3.550290998195604,
      "tokens_seen": 1472528384
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005594984954864593,
      "loss": 1.5061,
      "theoretical_loss": 3.550277129717955,
      "tokens_seen": 1472593920
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005594784353059178,
      "loss": 1.4028,
      "theoretical_loss": 3.5502632620302994,
      "tokens_seen": 1472659456
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005594583751253762,
      "loss": 1.4111,
      "theoretical_loss": 3.5502493951325564,
      "tokens_seen": 1472724992
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005594383149448345,
      "loss": 1.5097,
      "theoretical_loss": 3.5502355290246466,
      "tokens_seen": 1472790528
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005594182547642929,
      "loss": 1.5296,
      "theoretical_loss": 3.5502216637064894,
      "tokens_seen": 1472856064
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005593981945837513,
      "loss": 1.5708,
      "theoretical_loss": 3.550207799178005,
      "tokens_seen": 1472921600
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005593781344032096,
      "loss": 1.4949,
      "theoretical_loss": 3.550193935439112,
      "tokens_seen": 1472987136
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005593580742226681,
      "loss": 1.4809,
      "theoretical_loss": 3.5501800724897326,
      "tokens_seen": 1473052672
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005593380140421263,
      "loss": 1.4215,
      "theoretical_loss": 3.550166210329785,
      "tokens_seen": 1473118208
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005593179538615848,
      "loss": 1.4413,
      "theoretical_loss": 3.55015234895919,
      "tokens_seen": 1473183744
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005592978936810431,
      "loss": 1.4095,
      "theoretical_loss": 3.5501384883778666,
      "tokens_seen": 1473249280
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005592778335005015,
      "loss": 1.5101,
      "theoretical_loss": 3.5501246285857357,
      "tokens_seen": 1473314816
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005592577733199599,
      "loss": 1.4107,
      "theoretical_loss": 3.550110769582717,
      "tokens_seen": 1473380352
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005592377131394183,
      "loss": 1.6197,
      "theoretical_loss": 3.5500969113687306,
      "tokens_seen": 1473445888
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005592176529588766,
      "loss": 1.4792,
      "theoretical_loss": 3.5500830539436956,
      "tokens_seen": 1473511424
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005591975927783351,
      "loss": 1.5191,
      "theoretical_loss": 3.5500691973075327,
      "tokens_seen": 1473576960
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005591775325977933,
      "loss": 1.3795,
      "theoretical_loss": 3.550055341460162,
      "tokens_seen": 1473642496
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005591574724172518,
      "loss": 1.4241,
      "theoretical_loss": 3.5500414864015033,
      "tokens_seen": 1473708032
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005591374122367101,
      "loss": 1.411,
      "theoretical_loss": 3.550027632131477,
      "tokens_seen": 1473773568
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005591173520561685,
      "loss": 1.4937,
      "theoretical_loss": 3.550013778650003,
      "tokens_seen": 1473839104
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005590972918756269,
      "loss": 1.4409,
      "theoretical_loss": 3.549999925957001,
      "tokens_seen": 1473904640
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005590772316950853,
      "loss": 1.4504,
      "theoretical_loss": 3.5499860740523914,
      "tokens_seen": 1473970176
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005590571715145436,
      "loss": 1.4743,
      "theoretical_loss": 3.549972222936094,
      "tokens_seen": 1474035712
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000559037111334002,
      "loss": 1.4256,
      "theoretical_loss": 3.549958372608029,
      "tokens_seen": 1474101248
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005590170511534604,
      "loss": 1.4652,
      "theoretical_loss": 3.549944523068117,
      "tokens_seen": 1474166784
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005589969909729188,
      "loss": 1.426,
      "theoretical_loss": 3.5499306743162764,
      "tokens_seen": 1474232320
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005589769307923771,
      "loss": 1.4345,
      "theoretical_loss": 3.5499168263524297,
      "tokens_seen": 1474297856
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005589568706118355,
      "loss": 1.4294,
      "theoretical_loss": 3.5499029791764958,
      "tokens_seen": 1474363392
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005589368104312938,
      "loss": 1.5316,
      "theoretical_loss": 3.549889132788395,
      "tokens_seen": 1474428928
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005589167502507523,
      "loss": 1.5719,
      "theoretical_loss": 3.5498752871880472,
      "tokens_seen": 1474494464
    },
    {
      "epoch": 0.45,
      "objective/train/advantage_avg": 0.4951992928981781,
      "objective/train/docs_used": 538469,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3503663539886475,
      "objective/train/original_loss": 1.350366473197937,
      "objective/train/theoretical_loss": 3.549861442375373,
      "objective/train/tokens_used": 1495020000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24547205865383148,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507678985595703,
      "objective/train/weighted_lm_loss": 1.4189212322235107,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.549861442375373,
      "tokens_seen": 1474560000
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005588966900702106,
      "loss": 1.3945,
      "theoretical_loss": 3.549861442375373,
      "tokens_seen": 1474560000
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000558876629889669,
      "loss": 1.3264,
      "theoretical_loss": 3.5498475983502926,
      "tokens_seen": 1474625536
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005588565697091274,
      "loss": 1.4377,
      "theoretical_loss": 3.5498337551127257,
      "tokens_seen": 1474691072
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005588365095285858,
      "loss": 1.334,
      "theoretical_loss": 3.549819912662593,
      "tokens_seen": 1474756608
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005588164493480441,
      "loss": 1.5128,
      "theoretical_loss": 3.5498060709998143,
      "tokens_seen": 1474822144
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005587963891675026,
      "loss": 1.472,
      "theoretical_loss": 3.54979223012431,
      "tokens_seen": 1474887680
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005587763289869608,
      "loss": 1.5069,
      "theoretical_loss": 3.5497783900360007,
      "tokens_seen": 1474953216
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005587562688064193,
      "loss": 1.4019,
      "theoretical_loss": 3.5497645507348063,
      "tokens_seen": 1475018752
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005587362086258776,
      "loss": 1.496,
      "theoretical_loss": 3.5497507122206473,
      "tokens_seen": 1475084288
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000558716148445336,
      "loss": 1.5141,
      "theoretical_loss": 3.549736874493443,
      "tokens_seen": 1475149824
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005586960882647944,
      "loss": 1.3937,
      "theoretical_loss": 3.5497230375531155,
      "tokens_seen": 1475215360
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005586760280842527,
      "loss": 1.4003,
      "theoretical_loss": 3.5497092013995837,
      "tokens_seen": 1475280896
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005586559679037111,
      "loss": 1.5103,
      "theoretical_loss": 3.5496953660327684,
      "tokens_seen": 1475346432
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005586359077231696,
      "loss": 1.5778,
      "theoretical_loss": 3.5496815314525896,
      "tokens_seen": 1475411968
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005586158475426278,
      "loss": 1.4885,
      "theoretical_loss": 3.5496676976589683,
      "tokens_seen": 1475477504
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005585957873620863,
      "loss": 1.3861,
      "theoretical_loss": 3.549653864651824,
      "tokens_seen": 1475543040
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005585757271815445,
      "loss": 1.3845,
      "theoretical_loss": 3.5496400324310775,
      "tokens_seen": 1475608576
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000558555667001003,
      "loss": 1.4791,
      "theoretical_loss": 3.549626200996649,
      "tokens_seen": 1475674112
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005585356068204615,
      "loss": 1.4586,
      "theoretical_loss": 3.5496123703484592,
      "tokens_seen": 1475739648
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005585155466399197,
      "loss": 1.3073,
      "theoretical_loss": 3.5495985404864285,
      "tokens_seen": 1475805184
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005584954864593782,
      "loss": 1.2746,
      "theoretical_loss": 3.5495847114104766,
      "tokens_seen": 1475870720
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005584754262788366,
      "loss": 1.5388,
      "theoretical_loss": 3.5495708831205253,
      "tokens_seen": 1475936256
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005584553660982949,
      "loss": 1.4043,
      "theoretical_loss": 3.5495570556164937,
      "tokens_seen": 1476001792
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005584353059177533,
      "loss": 1.5609,
      "theoretical_loss": 3.549543228898303,
      "tokens_seen": 1476067328
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005584152457372117,
      "loss": 1.4544,
      "theoretical_loss": 3.549529402965873,
      "tokens_seen": 1476132864
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00055839518555667,
      "loss": 1.4833,
      "theoretical_loss": 3.5495155778191245,
      "tokens_seen": 1476198400
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005583751253761285,
      "loss": 1.4449,
      "theoretical_loss": 3.549501753457978,
      "tokens_seen": 1476263936
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005583550651955867,
      "loss": 1.4304,
      "theoretical_loss": 3.5494879298823543,
      "tokens_seen": 1476329472
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005583350050150452,
      "loss": 1.4236,
      "theoretical_loss": 3.549474107092173,
      "tokens_seen": 1476395008
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005583149448345035,
      "loss": 1.4117,
      "theoretical_loss": 3.5494602850873553,
      "tokens_seen": 1476460544
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005582948846539619,
      "loss": 1.5491,
      "theoretical_loss": 3.549446463867822,
      "tokens_seen": 1476526080
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005582748244734203,
      "loss": 1.5231,
      "theoretical_loss": 3.549432643433493,
      "tokens_seen": 1476591616
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005582547642928787,
      "loss": 1.5832,
      "theoretical_loss": 3.5494188237842894,
      "tokens_seen": 1476657152
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000558234704112337,
      "loss": 1.5423,
      "theoretical_loss": 3.549405004920131,
      "tokens_seen": 1476722688
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005582146439317955,
      "loss": 1.4309,
      "theoretical_loss": 3.549391186840939,
      "tokens_seen": 1476788224
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005581945837512537,
      "loss": 1.5097,
      "theoretical_loss": 3.549377369546634,
      "tokens_seen": 1476853760
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005581745235707122,
      "loss": 1.4291,
      "theoretical_loss": 3.5493635530371366,
      "tokens_seen": 1476919296
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005581544633901705,
      "loss": 1.4292,
      "theoretical_loss": 3.549349737312367,
      "tokens_seen": 1476984832
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005581344032096289,
      "loss": 1.3729,
      "theoretical_loss": 3.5493359223722454,
      "tokens_seen": 1477050368
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005581143430290873,
      "loss": 1.4334,
      "theoretical_loss": 3.5493221082166935,
      "tokens_seen": 1477115904
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005580942828485457,
      "loss": 1.5064,
      "theoretical_loss": 3.5493082948456314,
      "tokens_seen": 1477181440
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000558074222668004,
      "loss": 1.528,
      "theoretical_loss": 3.54929448225898,
      "tokens_seen": 1477246976
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005580541624874624,
      "loss": 1.4623,
      "theoretical_loss": 3.5492806704566595,
      "tokens_seen": 1477312512
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005580341023069208,
      "loss": 1.5274,
      "theoretical_loss": 3.549266859438591,
      "tokens_seen": 1477378048
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005580140421263792,
      "loss": 1.4497,
      "theoretical_loss": 3.549253049204695,
      "tokens_seen": 1477443584
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005579939819458375,
      "loss": 1.3909,
      "theoretical_loss": 3.5492392397548924,
      "tokens_seen": 1477509120
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000557973921765296,
      "loss": 1.4715,
      "theoretical_loss": 3.549225431089104,
      "tokens_seen": 1477574656
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005579538615847542,
      "loss": 1.424,
      "theoretical_loss": 3.549211623207249,
      "tokens_seen": 1477640192
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005579338014042127,
      "loss": 1.3477,
      "theoretical_loss": 3.549197816109251,
      "tokens_seen": 1477705728
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000557913741223671,
      "loss": 1.453,
      "theoretical_loss": 3.549184009795028,
      "tokens_seen": 1477771264
    },
    {
      "epoch": 0.45,
      "objective/train/advantage_avg": 0.49081602692604065,
      "objective/train/docs_used": 539614,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3338030576705933,
      "objective/train/original_loss": 1.3338030576705933,
      "objective/train/theoretical_loss": 3.549170204264502,
      "objective/train/tokens_used": 1498296800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24170133471488953,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503103733062744,
      "objective/train/weighted_lm_loss": 1.400624394416809,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.549170204264502,
      "tokens_seen": 1477836800
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005578936810431294,
      "loss": 1.3732,
      "theoretical_loss": 3.549170204264502,
      "tokens_seen": 1477836800
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005578736208625878,
      "loss": 1.4225,
      "theoretical_loss": 3.5491563995175937,
      "tokens_seen": 1477902336
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005578535606820462,
      "loss": 1.4626,
      "theoretical_loss": 3.549142595554224,
      "tokens_seen": 1477967872
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005578335005015045,
      "loss": 1.5128,
      "theoretical_loss": 3.5491287923743133,
      "tokens_seen": 1478033408
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000557813440320963,
      "loss": 1.4952,
      "theoretical_loss": 3.5491149899777827,
      "tokens_seen": 1478098944
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005577933801404212,
      "loss": 1.4877,
      "theoretical_loss": 3.5491011883645527,
      "tokens_seen": 1478164480
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005577733199598797,
      "loss": 1.461,
      "theoretical_loss": 3.5490873875345446,
      "tokens_seen": 1478230016
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000557753259779338,
      "loss": 1.3651,
      "theoretical_loss": 3.549073587487679,
      "tokens_seen": 1478295552
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005577331995987964,
      "loss": 1.417,
      "theoretical_loss": 3.549059788223876,
      "tokens_seen": 1478361088
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005577131394182548,
      "loss": 1.4589,
      "theoretical_loss": 3.5490459897430577,
      "tokens_seen": 1478426624
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005576930792377131,
      "loss": 1.4556,
      "theoretical_loss": 3.5490321920451446,
      "tokens_seen": 1478492160
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005576730190571715,
      "loss": 1.4721,
      "theoretical_loss": 3.5490183951300573,
      "tokens_seen": 1478557696
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00055765295887663,
      "loss": 1.4043,
      "theoretical_loss": 3.549004598997717,
      "tokens_seen": 1478623232
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005576328986960882,
      "loss": 1.4926,
      "theoretical_loss": 3.548990803648044,
      "tokens_seen": 1478688768
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005576128385155467,
      "loss": 1.4478,
      "theoretical_loss": 3.54897700908096,
      "tokens_seen": 1478754304
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005575927783350049,
      "loss": 1.3621,
      "theoretical_loss": 3.548963215296385,
      "tokens_seen": 1478819840
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005575727181544634,
      "loss": 1.5447,
      "theoretical_loss": 3.548949422294241,
      "tokens_seen": 1478885376
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005575526579739218,
      "loss": 1.4689,
      "theoretical_loss": 3.548935630074448,
      "tokens_seen": 1478950912
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005575325977933801,
      "loss": 1.6113,
      "theoretical_loss": 3.548921838636927,
      "tokens_seen": 1479016448
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005575125376128385,
      "loss": 1.4111,
      "theoretical_loss": 3.5489080479816,
      "tokens_seen": 1479081984
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000557492477432297,
      "loss": 1.4824,
      "theoretical_loss": 3.548894258108387,
      "tokens_seen": 1479147520
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005574724172517552,
      "loss": 1.56,
      "theoretical_loss": 3.5488804690172095,
      "tokens_seen": 1479213056
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005574523570712137,
      "loss": 1.5501,
      "theoretical_loss": 3.5488666807079885,
      "tokens_seen": 1479278592
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005574322968906719,
      "loss": 1.4228,
      "theoretical_loss": 3.5488528931806447,
      "tokens_seen": 1479344128
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005574122367101304,
      "loss": 1.4696,
      "theoretical_loss": 3.548839106435099,
      "tokens_seen": 1479409664
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005573921765295888,
      "loss": 1.4324,
      "theoretical_loss": 3.548825320471273,
      "tokens_seen": 1479475200
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005573721163490471,
      "loss": 1.4821,
      "theoretical_loss": 3.5488115352890874,
      "tokens_seen": 1479540736
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005573520561685055,
      "loss": 1.3158,
      "theoretical_loss": 3.5487977508884634,
      "tokens_seen": 1479606272
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005573319959879639,
      "loss": 1.4866,
      "theoretical_loss": 3.5487839672693218,
      "tokens_seen": 1479671808
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005573119358074222,
      "loss": 1.4801,
      "theoretical_loss": 3.548770184431584,
      "tokens_seen": 1479737344
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005572918756268807,
      "loss": 1.5693,
      "theoretical_loss": 3.5487564023751714,
      "tokens_seen": 1479802880
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000557271815446339,
      "loss": 1.5083,
      "theoretical_loss": 3.548742621100004,
      "tokens_seen": 1479868416
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005572517552657974,
      "loss": 1.488,
      "theoretical_loss": 3.548728840606004,
      "tokens_seen": 1479933952
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005572316950852558,
      "loss": 1.4199,
      "theoretical_loss": 3.548715060893092,
      "tokens_seen": 1479999488
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005572116349047141,
      "loss": 1.497,
      "theoretical_loss": 3.5487012819611894,
      "tokens_seen": 1480065024
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005571915747241725,
      "loss": 1.5769,
      "theoretical_loss": 3.5486875038102172,
      "tokens_seen": 1480130560
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005571715145436309,
      "loss": 1.4783,
      "theoretical_loss": 3.5486737264400965,
      "tokens_seen": 1480196096
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005571514543630892,
      "loss": 1.4174,
      "theoretical_loss": 3.548659949850749,
      "tokens_seen": 1480261632
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005571313941825477,
      "loss": 1.4351,
      "theoretical_loss": 3.548646174042095,
      "tokens_seen": 1480327168
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000557111334002006,
      "loss": 1.4875,
      "theoretical_loss": 3.548632399014056,
      "tokens_seen": 1480392704
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005570912738214644,
      "loss": 1.4669,
      "theoretical_loss": 3.548618624766554,
      "tokens_seen": 1480458240
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005570712136409227,
      "loss": 1.6272,
      "theoretical_loss": 3.548604851299509,
      "tokens_seen": 1480523776
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005570511534603812,
      "loss": 1.4225,
      "theoretical_loss": 3.5485910786128434,
      "tokens_seen": 1480589312
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005570310932798396,
      "loss": 1.493,
      "theoretical_loss": 3.5485773067064774,
      "tokens_seen": 1480654848
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005570110330992979,
      "loss": 1.4817,
      "theoretical_loss": 3.548563535580333,
      "tokens_seen": 1480720384
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005569909729187563,
      "loss": 1.4687,
      "theoretical_loss": 3.5485497652343314,
      "tokens_seen": 1480785920
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005569709127382146,
      "loss": 1.5227,
      "theoretical_loss": 3.5485359956683933,
      "tokens_seen": 1480851456
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005569508525576731,
      "loss": 1.4131,
      "theoretical_loss": 3.5485222268824406,
      "tokens_seen": 1480916992
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005569307923771314,
      "loss": 1.531,
      "theoretical_loss": 3.5485084588763947,
      "tokens_seen": 1480982528
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005569107321965898,
      "loss": 1.4891,
      "theoretical_loss": 3.548494691650176,
      "tokens_seen": 1481048064
    },
    {
      "epoch": 0.45,
      "objective/train/advantage_avg": 0.48901039361953735,
      "objective/train/docs_used": 540822,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4360733032226562,
      "objective/train/original_loss": 1.4360733032226562,
      "objective/train/theoretical_loss": 3.548480925203706,
      "objective/train/tokens_used": 1501573600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2399754524230957,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501209497451782,
      "objective/train/weighted_lm_loss": 1.5082576274871826,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.548480925203706,
      "tokens_seen": 1481113600
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005568906720160482,
      "loss": 1.5434,
      "theoretical_loss": 3.548480925203706,
      "tokens_seen": 1481113600
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005568706118355066,
      "loss": 1.4017,
      "theoretical_loss": 3.5484671595369077,
      "tokens_seen": 1481179136
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005568505516549649,
      "loss": 1.5846,
      "theoretical_loss": 3.5484533946497008,
      "tokens_seen": 1481244672
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005568304914744234,
      "loss": 1.4085,
      "theoretical_loss": 3.5484396305420063,
      "tokens_seen": 1481310208
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005568104312938816,
      "loss": 1.4808,
      "theoretical_loss": 3.548425867213747,
      "tokens_seen": 1481375744
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005567903711133401,
      "loss": 1.5065,
      "theoretical_loss": 3.548412104664844,
      "tokens_seen": 1481441280
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005567703109327984,
      "loss": 1.471,
      "theoretical_loss": 3.5483983428952177,
      "tokens_seen": 1481506816
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005567502507522568,
      "loss": 1.4463,
      "theoretical_loss": 3.5483845819047906,
      "tokens_seen": 1481572352
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005567301905717152,
      "loss": 1.4032,
      "theoretical_loss": 3.5483708216934833,
      "tokens_seen": 1481637888
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005567101303911735,
      "loss": 1.5205,
      "theoretical_loss": 3.5483570622612177,
      "tokens_seen": 1481703424
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005566900702106319,
      "loss": 1.4383,
      "theoretical_loss": 3.548343303607915,
      "tokens_seen": 1481768960
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005566700100300904,
      "loss": 1.493,
      "theoretical_loss": 3.5483295457334965,
      "tokens_seen": 1481834496
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005566499498495486,
      "loss": 1.38,
      "theoretical_loss": 3.5483157886378844,
      "tokens_seen": 1481900032
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005566298896690071,
      "loss": 1.5573,
      "theoretical_loss": 3.5483020323209997,
      "tokens_seen": 1481965568
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005566098294884653,
      "loss": 1.5108,
      "theoretical_loss": 3.5482882767827637,
      "tokens_seen": 1482031104
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005565897693079238,
      "loss": 1.3899,
      "theoretical_loss": 3.5482745220230982,
      "tokens_seen": 1482096640
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005565697091273822,
      "loss": 1.531,
      "theoretical_loss": 3.5482607680419243,
      "tokens_seen": 1482162176
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005565496489468405,
      "loss": 1.44,
      "theoretical_loss": 3.5482470148391645,
      "tokens_seen": 1482227712
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005565295887662989,
      "loss": 1.49,
      "theoretical_loss": 3.5482332624147386,
      "tokens_seen": 1482293248
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005565095285857574,
      "loss": 1.5221,
      "theoretical_loss": 3.54821951076857,
      "tokens_seen": 1482358784
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005564894684052156,
      "loss": 1.3907,
      "theoretical_loss": 3.548205759900579,
      "tokens_seen": 1482424320
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005564694082246741,
      "loss": 1.4175,
      "theoretical_loss": 3.548192009810688,
      "tokens_seen": 1482489856
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005564493480441323,
      "loss": 1.5426,
      "theoretical_loss": 3.548178260498818,
      "tokens_seen": 1482555392
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005564292878635908,
      "loss": 1.3948,
      "theoretical_loss": 3.548164511964891,
      "tokens_seen": 1482620928
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005564092276830492,
      "loss": 1.4862,
      "theoretical_loss": 3.548150764208828,
      "tokens_seen": 1482686464
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005563891675025075,
      "loss": 1.4085,
      "theoretical_loss": 3.548137017230551,
      "tokens_seen": 1482752000
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005563691073219659,
      "loss": 1.4419,
      "theoretical_loss": 3.548123271029982,
      "tokens_seen": 1482817536
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005563490471414243,
      "loss": 1.4264,
      "theoretical_loss": 3.5481095256070416,
      "tokens_seen": 1482883072
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005563289869608826,
      "loss": 1.5014,
      "theoretical_loss": 3.5480957809616527,
      "tokens_seen": 1482948608
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005563089267803411,
      "loss": 1.4762,
      "theoretical_loss": 3.548082037093736,
      "tokens_seen": 1483014144
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005562888665997994,
      "loss": 1.4746,
      "theoretical_loss": 3.5480682940032136,
      "tokens_seen": 1483079680
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005562688064192578,
      "loss": 1.4967,
      "theoretical_loss": 3.548054551690007,
      "tokens_seen": 1483145216
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005562487462387162,
      "loss": 1.5506,
      "theoretical_loss": 3.548040810154038,
      "tokens_seen": 1483210752
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005562286860581745,
      "loss": 1.4486,
      "theoretical_loss": 3.5480270693952285,
      "tokens_seen": 1483276288
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005562086258776329,
      "loss": 1.6495,
      "theoretical_loss": 3.5480133294134992,
      "tokens_seen": 1483341824
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005561885656970913,
      "loss": 1.5039,
      "theoretical_loss": 3.5479995902087733,
      "tokens_seen": 1483407360
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005561685055165496,
      "loss": 1.5218,
      "theoretical_loss": 3.5479858517809717,
      "tokens_seen": 1483472896
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005561484453360081,
      "loss": 1.5172,
      "theoretical_loss": 3.5479721141300162,
      "tokens_seen": 1483538432
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005561283851554664,
      "loss": 1.4448,
      "theoretical_loss": 3.5479583772558287,
      "tokens_seen": 1483603968
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005561083249749248,
      "loss": 1.5451,
      "theoretical_loss": 3.547944641158331,
      "tokens_seen": 1483669504
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005560882647943831,
      "loss": 1.4811,
      "theoretical_loss": 3.547930905837445,
      "tokens_seen": 1483735040
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005560682046138416,
      "loss": 1.5679,
      "theoretical_loss": 3.547917171293092,
      "tokens_seen": 1483800576
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005560481444332999,
      "loss": 1.683,
      "theoretical_loss": 3.5479034375251945,
      "tokens_seen": 1483866112
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005560280842527583,
      "loss": 1.5014,
      "theoretical_loss": 3.5478897045336732,
      "tokens_seen": 1483931648
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005560080240722166,
      "loss": 1.4256,
      "theoretical_loss": 3.547875972318451,
      "tokens_seen": 1483997184
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000555987963891675,
      "loss": 1.4747,
      "theoretical_loss": 3.54786224087945,
      "tokens_seen": 1484062720
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005559679037111334,
      "loss": 1.35,
      "theoretical_loss": 3.5478485102165904,
      "tokens_seen": 1484128256
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005559478435305918,
      "loss": 1.4133,
      "theoretical_loss": 3.5478347803297954,
      "tokens_seen": 1484193792
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005559277833500501,
      "loss": 1.4729,
      "theoretical_loss": 3.547821051218987,
      "tokens_seen": 1484259328
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005559077231695086,
      "loss": 1.5404,
      "theoretical_loss": 3.5478073228840863,
      "tokens_seen": 1484324864
    },
    {
      "epoch": 0.45,
      "objective/train/advantage_avg": 0.4934280216693878,
      "objective/train/docs_used": 542078,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.387536644935608,
      "objective/train/original_loss": 1.387536644935608,
      "objective/train/theoretical_loss": 3.5477935953250155,
      "objective/train/tokens_used": 1504850400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24421967566013336,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050584316253662,
      "objective/train/weighted_lm_loss": 1.457723617553711,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.5477935953250155,
      "tokens_seen": 1484390400
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005558876629889669,
      "loss": 1.4714,
      "theoretical_loss": 3.5477935953250155,
      "tokens_seen": 1484390400
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005558676028084253,
      "loss": 1.4397,
      "theoretical_loss": 3.5477798685416966,
      "tokens_seen": 1484455936
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005558475426278836,
      "loss": 1.4234,
      "theoretical_loss": 3.5477661425340514,
      "tokens_seen": 1484521472
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000555827482447342,
      "loss": 1.4742,
      "theoretical_loss": 3.547752417302002,
      "tokens_seen": 1484587008
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005558074222668004,
      "loss": 1.4614,
      "theoretical_loss": 3.5477386928454697,
      "tokens_seen": 1484652544
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005557873620862588,
      "loss": 1.4682,
      "theoretical_loss": 3.5477249691643777,
      "tokens_seen": 1484718080
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005557673019057171,
      "loss": 1.5288,
      "theoretical_loss": 3.547711246258647,
      "tokens_seen": 1484783616
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005557472417251756,
      "loss": 1.561,
      "theoretical_loss": 3.5476975241282,
      "tokens_seen": 1484849152
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005557271815446338,
      "loss": 1.416,
      "theoretical_loss": 3.547683802772958,
      "tokens_seen": 1484914688
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005557071213640923,
      "loss": 1.4778,
      "theoretical_loss": 3.5476700821928437,
      "tokens_seen": 1484980224
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005556870611835507,
      "loss": 1.5796,
      "theoretical_loss": 3.547656362387779,
      "tokens_seen": 1485045760
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000555667001003009,
      "loss": 1.5626,
      "theoretical_loss": 3.547642643357686,
      "tokens_seen": 1485111296
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005556469408224674,
      "loss": 1.5215,
      "theoretical_loss": 3.5476289251024857,
      "tokens_seen": 1485176832
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005556268806419257,
      "loss": 1.5179,
      "theoretical_loss": 3.547615207622102,
      "tokens_seen": 1485242368
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005556068204613841,
      "loss": 1.4503,
      "theoretical_loss": 3.5476014909164553,
      "tokens_seen": 1485307904
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005555867602808426,
      "loss": 1.365,
      "theoretical_loss": 3.5475877749854687,
      "tokens_seen": 1485373440
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005555667001003008,
      "loss": 1.3979,
      "theoretical_loss": 3.5475740598290635,
      "tokens_seen": 1485438976
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005555466399197593,
      "loss": 1.4789,
      "theoretical_loss": 3.5475603454471623,
      "tokens_seen": 1485504512
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005555265797392178,
      "loss": 1.497,
      "theoretical_loss": 3.547546631839687,
      "tokens_seen": 1485570048
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000555506519558676,
      "loss": 1.4839,
      "theoretical_loss": 3.54753291900656,
      "tokens_seen": 1485635584
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005554864593781345,
      "loss": 1.5224,
      "theoretical_loss": 3.547519206947703,
      "tokens_seen": 1485701120
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005554663991975927,
      "loss": 1.4453,
      "theoretical_loss": 3.5475054956630387,
      "tokens_seen": 1485766656
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005554463390170512,
      "loss": 1.3418,
      "theoretical_loss": 3.547491785152488,
      "tokens_seen": 1485832192
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005554262788365096,
      "loss": 1.5089,
      "theoretical_loss": 3.5474780754159747,
      "tokens_seen": 1485897728
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005554062186559679,
      "loss": 1.4925,
      "theoretical_loss": 3.54746436645342,
      "tokens_seen": 1485963264
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005553861584754263,
      "loss": 1.4623,
      "theoretical_loss": 3.5474506582647463,
      "tokens_seen": 1486028800
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005553660982948847,
      "loss": 1.4518,
      "theoretical_loss": 3.5474369508498755,
      "tokens_seen": 1486094336
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000555346038114343,
      "loss": 1.4743,
      "theoretical_loss": 3.54742324420873,
      "tokens_seen": 1486159872
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005553259779338015,
      "loss": 1.4905,
      "theoretical_loss": 3.547409538341232,
      "tokens_seen": 1486225408
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005553059177532598,
      "loss": 1.4199,
      "theoretical_loss": 3.5473958332473043,
      "tokens_seen": 1486290944
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005552858575727182,
      "loss": 1.5409,
      "theoretical_loss": 3.547382128926868,
      "tokens_seen": 1486356480
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005552657973921766,
      "loss": 1.3893,
      "theoretical_loss": 3.5473684253798465,
      "tokens_seen": 1486422016
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000555245737211635,
      "loss": 1.4767,
      "theoretical_loss": 3.547354722606161,
      "tokens_seen": 1486487552
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005552256770310933,
      "loss": 1.4579,
      "theoretical_loss": 3.5473410206057348,
      "tokens_seen": 1486553088
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005552056168505517,
      "loss": 1.4472,
      "theoretical_loss": 3.5473273193784896,
      "tokens_seen": 1486618624
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00055518555667001,
      "loss": 1.4914,
      "theoretical_loss": 3.5473136189243473,
      "tokens_seen": 1486684160
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005551654964894685,
      "loss": 1.5266,
      "theoretical_loss": 3.5472999192432306,
      "tokens_seen": 1486749696
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005551454363089268,
      "loss": 1.3762,
      "theoretical_loss": 3.5472862203350624,
      "tokens_seen": 1486815232
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005551253761283852,
      "loss": 1.368,
      "theoretical_loss": 3.547272522199764,
      "tokens_seen": 1486880768
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005551053159478435,
      "loss": 1.4468,
      "theoretical_loss": 3.5472588248372583,
      "tokens_seen": 1486946304
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000555085255767302,
      "loss": 1.4957,
      "theoretical_loss": 3.547245128247468,
      "tokens_seen": 1487011840
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005550651955867603,
      "loss": 1.557,
      "theoretical_loss": 3.5472314324303147,
      "tokens_seen": 1487077376
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005550451354062187,
      "loss": 1.4495,
      "theoretical_loss": 3.5472177373857208,
      "tokens_seen": 1487142912
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000555025075225677,
      "loss": 1.4512,
      "theoretical_loss": 3.5472040431136094,
      "tokens_seen": 1487208448
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005550050150451354,
      "loss": 1.4537,
      "theoretical_loss": 3.547190349613902,
      "tokens_seen": 1487273984
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005549849548645938,
      "loss": 1.3009,
      "theoretical_loss": 3.5471766568865215,
      "tokens_seen": 1487339520
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005549648946840522,
      "loss": 1.4309,
      "theoretical_loss": 3.5471629649313905,
      "tokens_seen": 1487405056
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005549448345035105,
      "loss": 1.4301,
      "theoretical_loss": 3.547149273748431,
      "tokens_seen": 1487470592
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000554924774322969,
      "loss": 1.4661,
      "theoretical_loss": 3.5471355833375657,
      "tokens_seen": 1487536128
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005549047141424273,
      "loss": 1.3596,
      "theoretical_loss": 3.547121893698717,
      "tokens_seen": 1487601664
    },
    {
      "epoch": 0.45,
      "objective/train/advantage_avg": 0.48960351943969727,
      "objective/train/docs_used": 543335,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3424406051635742,
      "objective/train/original_loss": 1.3424406051635742,
      "objective/train/theoretical_loss": 3.547108204831807,
      "objective/train/tokens_used": 1508127200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24117901921272278,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050186276435852,
      "objective/train/weighted_lm_loss": 1.4096623659133911,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.547108204831807,
      "tokens_seen": 1487667200
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005548846539618857,
      "loss": 1.4357,
      "theoretical_loss": 3.547108204831807,
      "tokens_seen": 1487667200
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000554864593781344,
      "loss": 1.4572,
      "theoretical_loss": 3.5470945167367587,
      "tokens_seen": 1487732736
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005548445336008024,
      "loss": 1.403,
      "theoretical_loss": 3.5470808294134946,
      "tokens_seen": 1487798272
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005548244734202608,
      "loss": 1.5385,
      "theoretical_loss": 3.5470671428619367,
      "tokens_seen": 1487863808
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005548044132397192,
      "loss": 1.6151,
      "theoretical_loss": 3.547053457082008,
      "tokens_seen": 1487929344
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005547843530591775,
      "loss": 1.4561,
      "theoretical_loss": 3.5470397720736306,
      "tokens_seen": 1487994880
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000554764292878636,
      "loss": 1.5729,
      "theoretical_loss": 3.547026087836727,
      "tokens_seen": 1488060416
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005547442326980942,
      "loss": 1.4823,
      "theoretical_loss": 3.5470124043712197,
      "tokens_seen": 1488125952
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005547241725175527,
      "loss": 1.4251,
      "theoretical_loss": 3.5469987216770322,
      "tokens_seen": 1488191488
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000554704112337011,
      "loss": 1.542,
      "theoretical_loss": 3.5469850397540856,
      "tokens_seen": 1488257024
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005546840521564694,
      "loss": 1.4842,
      "theoretical_loss": 3.546971358602304,
      "tokens_seen": 1488322560
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005546639919759278,
      "loss": 1.4631,
      "theoretical_loss": 3.5469576782216086,
      "tokens_seen": 1488388096
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005546439317953861,
      "loss": 1.4911,
      "theoretical_loss": 3.5469439986119227,
      "tokens_seen": 1488453632
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005546238716148445,
      "loss": 1.4353,
      "theoretical_loss": 3.546930319773169,
      "tokens_seen": 1488519168
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000554603811434303,
      "loss": 1.3991,
      "theoretical_loss": 3.5469166417052698,
      "tokens_seen": 1488584704
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005545837512537612,
      "loss": 1.47,
      "theoretical_loss": 3.5469029644081473,
      "tokens_seen": 1488650240
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005545636910732197,
      "loss": 1.4642,
      "theoretical_loss": 3.5468892878817253,
      "tokens_seen": 1488715776
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005545436308926781,
      "loss": 1.4324,
      "theoretical_loss": 3.5468756121259255,
      "tokens_seen": 1488781312
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005545235707121364,
      "loss": 1.4584,
      "theoretical_loss": 3.5468619371406707,
      "tokens_seen": 1488846848
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005545035105315948,
      "loss": 1.4852,
      "theoretical_loss": 3.5468482629258844,
      "tokens_seen": 1488912384
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005544834503510531,
      "loss": 1.5251,
      "theoretical_loss": 3.546834589481488,
      "tokens_seen": 1488977920
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005544633901705115,
      "loss": 1.4416,
      "theoretical_loss": 3.546820916807405,
      "tokens_seen": 1489043456
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00055444332998997,
      "loss": 1.4484,
      "theoretical_loss": 3.5468072449035577,
      "tokens_seen": 1489108992
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005544232698094282,
      "loss": 1.3994,
      "theoretical_loss": 3.546793573769869,
      "tokens_seen": 1489174528
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005544032096288867,
      "loss": 1.4781,
      "theoretical_loss": 3.5467799034062617,
      "tokens_seen": 1489240064
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000554383149448345,
      "loss": 1.3836,
      "theoretical_loss": 3.546766233812659,
      "tokens_seen": 1489305600
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005543630892678034,
      "loss": 1.4233,
      "theoretical_loss": 3.5467525649889824,
      "tokens_seen": 1489371136
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005543430290872618,
      "loss": 1.4674,
      "theoretical_loss": 3.546738896935156,
      "tokens_seen": 1489436672
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005543229689067202,
      "loss": 1.4098,
      "theoretical_loss": 3.546725229651101,
      "tokens_seen": 1489502208
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005543029087261785,
      "loss": 1.5027,
      "theoretical_loss": 3.5467115631367423,
      "tokens_seen": 1489567744
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000554282848545637,
      "loss": 1.4472,
      "theoretical_loss": 3.5466978973920007,
      "tokens_seen": 1489633280
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005542627883650952,
      "loss": 1.5479,
      "theoretical_loss": 3.5466842324168004,
      "tokens_seen": 1489698816
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005542427281845537,
      "loss": 1.3896,
      "theoretical_loss": 3.5466705682110633,
      "tokens_seen": 1489764352
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000554222668004012,
      "loss": 1.4881,
      "theoretical_loss": 3.5466569047747125,
      "tokens_seen": 1489829888
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005542026078234704,
      "loss": 1.5206,
      "theoretical_loss": 3.5466432421076712,
      "tokens_seen": 1489895424
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005541825476429288,
      "loss": 1.4849,
      "theoretical_loss": 3.5466295802098617,
      "tokens_seen": 1489960960
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005541624874623872,
      "loss": 1.4209,
      "theoretical_loss": 3.546615919081207,
      "tokens_seen": 1490026496
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005541424272818455,
      "loss": 1.4815,
      "theoretical_loss": 3.546602258721631,
      "tokens_seen": 1490092032
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005541223671013039,
      "loss": 1.4339,
      "theoretical_loss": 3.546588599131055,
      "tokens_seen": 1490157568
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005541023069207622,
      "loss": 1.3304,
      "theoretical_loss": 3.546574940309402,
      "tokens_seen": 1490223104
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005540822467402207,
      "loss": 1.3566,
      "theoretical_loss": 3.546561282256596,
      "tokens_seen": 1490288640
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000554062186559679,
      "loss": 1.4001,
      "theoretical_loss": 3.5465476249725594,
      "tokens_seen": 1490354176
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005540421263791374,
      "loss": 1.3628,
      "theoretical_loss": 3.546533968457215,
      "tokens_seen": 1490419712
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005540220661985958,
      "loss": 1.3948,
      "theoretical_loss": 3.546520312710486,
      "tokens_seen": 1490485248
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005540020060180542,
      "loss": 1.4199,
      "theoretical_loss": 3.546506657732295,
      "tokens_seen": 1490550784
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005539819458375126,
      "loss": 1.5258,
      "theoretical_loss": 3.5464930035225652,
      "tokens_seen": 1490616320
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005539618856569709,
      "loss": 1.353,
      "theoretical_loss": 3.5464793500812197,
      "tokens_seen": 1490681856
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005539418254764294,
      "loss": 1.4545,
      "theoretical_loss": 3.5464656974081805,
      "tokens_seen": 1490747392
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005539217652958877,
      "loss": 1.5439,
      "theoretical_loss": 3.546452045503372,
      "tokens_seen": 1490812928
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005539017051153461,
      "loss": 1.4401,
      "theoretical_loss": 3.5464383943667173,
      "tokens_seen": 1490878464
    },
    {
      "epoch": 0.45,
      "objective/train/advantage_avg": 0.4942471385002136,
      "objective/train/docs_used": 544526,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.418752908706665,
      "objective/train/original_loss": 1.418752908706665,
      "objective/train/theoretical_loss": 3.546424743998138,
      "objective/train/tokens_used": 1511404000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24451583623886108,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506678819656372,
      "objective/train/weighted_lm_loss": 1.4905112981796265,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.546424743998138,
      "tokens_seen": 1490944000
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005538816449348044,
      "loss": 1.4554,
      "theoretical_loss": 3.546424743998138,
      "tokens_seen": 1490944000
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005538615847542628,
      "loss": 1.4711,
      "theoretical_loss": 3.5464110943975573,
      "tokens_seen": 1491009536
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005538415245737212,
      "loss": 1.4204,
      "theoretical_loss": 3.5463974455648994,
      "tokens_seen": 1491075072
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005538214643931796,
      "loss": 1.414,
      "theoretical_loss": 3.546383797500087,
      "tokens_seen": 1491140608
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005538014042126379,
      "loss": 1.4791,
      "theoretical_loss": 3.5463701502030425,
      "tokens_seen": 1491206144
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005537813440320964,
      "loss": 1.3931,
      "theoretical_loss": 3.546356503673689,
      "tokens_seen": 1491271680
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005537612838515546,
      "loss": 1.4583,
      "theoretical_loss": 3.5463428579119505,
      "tokens_seen": 1491337216
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005537412236710131,
      "loss": 1.3181,
      "theoretical_loss": 3.5463292129177493,
      "tokens_seen": 1491402752
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005537211634904715,
      "loss": 1.4813,
      "theoretical_loss": 3.546315568691009,
      "tokens_seen": 1491468288
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005537011033099298,
      "loss": 1.4721,
      "theoretical_loss": 3.546301925231652,
      "tokens_seen": 1491533824
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005536810431293882,
      "loss": 1.461,
      "theoretical_loss": 3.546288282539602,
      "tokens_seen": 1491599360
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005536609829488465,
      "loss": 1.3731,
      "theoretical_loss": 3.546274640614782,
      "tokens_seen": 1491664896
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005536409227683049,
      "loss": 1.4163,
      "theoretical_loss": 3.546260999457115,
      "tokens_seen": 1491730432
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005536208625877634,
      "loss": 1.4939,
      "theoretical_loss": 3.546247359066524,
      "tokens_seen": 1491795968
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005536008024072216,
      "loss": 1.5162,
      "theoretical_loss": 3.5462337194429336,
      "tokens_seen": 1491861504
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005535807422266801,
      "loss": 1.418,
      "theoretical_loss": 3.546220080586265,
      "tokens_seen": 1491927040
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005535606820461385,
      "loss": 1.541,
      "theoretical_loss": 3.546206442496443,
      "tokens_seen": 1491992576
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005535406218655968,
      "loss": 1.5423,
      "theoretical_loss": 3.546192805173389,
      "tokens_seen": 1492058112
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005535205616850552,
      "loss": 1.4538,
      "theoretical_loss": 3.546179168617028,
      "tokens_seen": 1492123648
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005535005015045135,
      "loss": 1.4843,
      "theoretical_loss": 3.546165532827282,
      "tokens_seen": 1492189184
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005534804413239719,
      "loss": 1.3915,
      "theoretical_loss": 3.5461518978040747,
      "tokens_seen": 1492254720
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005534603811434304,
      "loss": 1.3909,
      "theoretical_loss": 3.54613826354733,
      "tokens_seen": 1492320256
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005534403209628886,
      "loss": 1.4294,
      "theoretical_loss": 3.54612463005697,
      "tokens_seen": 1492385792
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005534202607823471,
      "loss": 1.6203,
      "theoretical_loss": 3.546110997332919,
      "tokens_seen": 1492451328
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005534002006018054,
      "loss": 1.4643,
      "theoretical_loss": 3.5460973653750996,
      "tokens_seen": 1492516864
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005533801404212638,
      "loss": 1.3648,
      "theoretical_loss": 3.5460837341834353,
      "tokens_seen": 1492582400
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005533600802407222,
      "loss": 1.5716,
      "theoretical_loss": 3.546070103757849,
      "tokens_seen": 1492647936
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005533400200601806,
      "loss": 1.3159,
      "theoretical_loss": 3.5460564740982647,
      "tokens_seen": 1492713472
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005533199598796389,
      "loss": 1.4291,
      "theoretical_loss": 3.5460428452046053,
      "tokens_seen": 1492779008
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005532998996990974,
      "loss": 1.4926,
      "theoretical_loss": 3.546029217076794,
      "tokens_seen": 1492844544
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005532798395185556,
      "loss": 1.3798,
      "theoretical_loss": 3.546015589714755,
      "tokens_seen": 1492910080
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005532597793380141,
      "loss": 1.4778,
      "theoretical_loss": 3.5460019631184108,
      "tokens_seen": 1492975616
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005532397191574724,
      "loss": 1.4476,
      "theoretical_loss": 3.545988337287685,
      "tokens_seen": 1493041152
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005532196589769308,
      "loss": 1.4909,
      "theoretical_loss": 3.5459747122225007,
      "tokens_seen": 1493106688
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005531995987963892,
      "loss": 1.4935,
      "theoretical_loss": 3.545961087922782,
      "tokens_seen": 1493172224
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005531795386158476,
      "loss": 1.319,
      "theoretical_loss": 3.5459474643884517,
      "tokens_seen": 1493237760
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005531594784353059,
      "loss": 1.4752,
      "theoretical_loss": 3.5459338416194335,
      "tokens_seen": 1493303296
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005531394182547643,
      "loss": 1.4666,
      "theoretical_loss": 3.5459202196156507,
      "tokens_seen": 1493368832
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005531193580742226,
      "loss": 1.4923,
      "theoretical_loss": 3.5459065983770266,
      "tokens_seen": 1493434368
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005530992978936811,
      "loss": 1.4333,
      "theoretical_loss": 3.5458929779034856,
      "tokens_seen": 1493499904
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005530792377131394,
      "loss": 1.4184,
      "theoretical_loss": 3.5458793581949495,
      "tokens_seen": 1493565440
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005530591775325978,
      "loss": 1.3185,
      "theoretical_loss": 3.545865739251343,
      "tokens_seen": 1493630976
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005530391173520561,
      "loss": 1.3558,
      "theoretical_loss": 3.5458521210725893,
      "tokens_seen": 1493696512
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005530190571715146,
      "loss": 1.4108,
      "theoretical_loss": 3.5458385036586115,
      "tokens_seen": 1493762048
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005529989969909729,
      "loss": 1.4357,
      "theoretical_loss": 3.545824887009333,
      "tokens_seen": 1493827584
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005529789368104313,
      "loss": 1.3776,
      "theoretical_loss": 3.5458112711246788,
      "tokens_seen": 1493893120
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005529588766298897,
      "loss": 1.4753,
      "theoretical_loss": 3.545797656004571,
      "tokens_seen": 1493958656
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005529388164493481,
      "loss": 1.4778,
      "theoretical_loss": 3.545784041648933,
      "tokens_seen": 1494024192
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005529187562688064,
      "loss": 1.3692,
      "theoretical_loss": 3.545770428057689,
      "tokens_seen": 1494089728
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005528986960882648,
      "loss": 1.4428,
      "theoretical_loss": 3.545756815230763,
      "tokens_seen": 1494155264
    },
    {
      "epoch": 0.45,
      "objective/train/advantage_avg": 0.494049996137619,
      "objective/train/docs_used": 545793,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2766737937927246,
      "objective/train/original_loss": 1.2766737937927246,
      "objective/train/theoretical_loss": 3.545743203168077,
      "objective/train/tokens_used": 1514680800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24451828002929688,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506480932235718,
      "objective/train/weighted_lm_loss": 1.341252088546753,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.545743203168077,
      "tokens_seen": 1494220800
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005528786359077231,
      "loss": 1.334,
      "theoretical_loss": 3.545743203168077,
      "tokens_seen": 1494220800
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005528585757271816,
      "loss": 1.5131,
      "theoretical_loss": 3.545729591869556,
      "tokens_seen": 1494286336
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005528385155466399,
      "loss": 1.4219,
      "theoretical_loss": 3.545715981335123,
      "tokens_seen": 1494351872
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005528184553660983,
      "loss": 1.3961,
      "theoretical_loss": 3.545702371564702,
      "tokens_seen": 1494417408
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005527983951855567,
      "loss": 1.3916,
      "theoretical_loss": 3.545688762558216,
      "tokens_seen": 1494482944
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000552778335005015,
      "loss": 1.4593,
      "theoretical_loss": 3.5456751543155893,
      "tokens_seen": 1494548480
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005527582748244734,
      "loss": 1.4885,
      "theoretical_loss": 3.5456615468367447,
      "tokens_seen": 1494614016
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005527382146439319,
      "loss": 1.4954,
      "theoretical_loss": 3.545647940121607,
      "tokens_seen": 1494679552
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005527181544633901,
      "loss": 1.3752,
      "theoretical_loss": 3.5456343341700984,
      "tokens_seen": 1494745088
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005526980942828486,
      "loss": 1.4352,
      "theoretical_loss": 3.5456207289821435,
      "tokens_seen": 1494810624
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005526780341023068,
      "loss": 1.487,
      "theoretical_loss": 3.5456071245576664,
      "tokens_seen": 1494876160
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005526579739217653,
      "loss": 1.4573,
      "theoretical_loss": 3.5455935208965896,
      "tokens_seen": 1494941696
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005526379137412237,
      "loss": 1.4074,
      "theoretical_loss": 3.545579917998838,
      "tokens_seen": 1495007232
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000552617853560682,
      "loss": 1.3089,
      "theoretical_loss": 3.5455663158643342,
      "tokens_seen": 1495072768
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005525977933801404,
      "loss": 1.4535,
      "theoretical_loss": 3.545552714493003,
      "tokens_seen": 1495138304
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005525777331995989,
      "loss": 1.4735,
      "theoretical_loss": 3.545539113884767,
      "tokens_seen": 1495203840
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005525576730190571,
      "loss": 1.4581,
      "theoretical_loss": 3.5455255140395505,
      "tokens_seen": 1495269376
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005525376128385156,
      "loss": 1.4005,
      "theoretical_loss": 3.5455119149572774,
      "tokens_seen": 1495334912
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005525175526579739,
      "loss": 1.4603,
      "theoretical_loss": 3.5454983166378717,
      "tokens_seen": 1495400448
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005524974924774323,
      "loss": 1.4837,
      "theoretical_loss": 3.5454847190812564,
      "tokens_seen": 1495465984
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005524774322968908,
      "loss": 1.5078,
      "theoretical_loss": 3.5454711222873554,
      "tokens_seen": 1495531520
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000552457372116349,
      "loss": 1.4312,
      "theoretical_loss": 3.5454575262560937,
      "tokens_seen": 1495597056
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005524373119358075,
      "loss": 1.5153,
      "theoretical_loss": 3.5454439309873935,
      "tokens_seen": 1495662592
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005524172517552658,
      "loss": 1.5483,
      "theoretical_loss": 3.5454303364811794,
      "tokens_seen": 1495728128
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005523971915747242,
      "loss": 1.3864,
      "theoretical_loss": 3.545416742737375,
      "tokens_seen": 1495793664
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005523771313941826,
      "loss": 1.4686,
      "theoretical_loss": 3.545403149755905,
      "tokens_seen": 1495859200
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000552357071213641,
      "loss": 1.4309,
      "theoretical_loss": 3.545389557536692,
      "tokens_seen": 1495924736
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005523370110330993,
      "loss": 1.5358,
      "theoretical_loss": 3.5453759660796607,
      "tokens_seen": 1495990272
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005523169508525578,
      "loss": 1.4527,
      "theoretical_loss": 3.5453623753847343,
      "tokens_seen": 1496055808
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000552296890672016,
      "loss": 1.3543,
      "theoretical_loss": 3.5453487854518375,
      "tokens_seen": 1496121344
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005522768304914745,
      "loss": 1.453,
      "theoretical_loss": 3.5453351962808934,
      "tokens_seen": 1496186880
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005522567703109328,
      "loss": 1.4597,
      "theoretical_loss": 3.545321607871826,
      "tokens_seen": 1496252416
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005522367101303912,
      "loss": 1.417,
      "theoretical_loss": 3.54530802022456,
      "tokens_seen": 1496317952
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005522166499498496,
      "loss": 1.5088,
      "theoretical_loss": 3.545294433339019,
      "tokens_seen": 1496383488
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000552196589769308,
      "loss": 1.5645,
      "theoretical_loss": 3.5452808472151265,
      "tokens_seen": 1496449024
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005521765295887663,
      "loss": 1.4153,
      "theoretical_loss": 3.545267261852806,
      "tokens_seen": 1496514560
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005521564694082247,
      "loss": 1.6621,
      "theoretical_loss": 3.545253677251983,
      "tokens_seen": 1496580096
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000552136409227683,
      "loss": 1.5293,
      "theoretical_loss": 3.5452400934125805,
      "tokens_seen": 1496645632
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005521163490471415,
      "loss": 1.3672,
      "theoretical_loss": 3.5452265103345226,
      "tokens_seen": 1496711168
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005520962888665998,
      "loss": 1.4077,
      "theoretical_loss": 3.5452129280177336,
      "tokens_seen": 1496776704
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005520762286860582,
      "loss": 1.4549,
      "theoretical_loss": 3.5451993464621365,
      "tokens_seen": 1496842240
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005520561685055165,
      "loss": 1.5368,
      "theoretical_loss": 3.5451857656676564,
      "tokens_seen": 1496907776
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000552036108324975,
      "loss": 1.4469,
      "theoretical_loss": 3.5451721856342173,
      "tokens_seen": 1496973312
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005520160481444333,
      "loss": 1.4203,
      "theoretical_loss": 3.545158606361743,
      "tokens_seen": 1497038848
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005519959879638917,
      "loss": 1.2797,
      "theoretical_loss": 3.5451450278501566,
      "tokens_seen": 1497104384
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00055197592778335,
      "loss": 1.4777,
      "theoretical_loss": 3.5451314500993836,
      "tokens_seen": 1497169920
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005519558676028085,
      "loss": 1.5292,
      "theoretical_loss": 3.5451178731093473,
      "tokens_seen": 1497235456
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005519358074222668,
      "loss": 1.4606,
      "theoretical_loss": 3.5451042968799724,
      "tokens_seen": 1497300992
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005519157472417252,
      "loss": 1.6132,
      "theoretical_loss": 3.545090721411182,
      "tokens_seen": 1497366528
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005518956870611835,
      "loss": 1.5024,
      "theoretical_loss": 3.545077146702901,
      "tokens_seen": 1497432064
    },
    {
      "epoch": 0.45,
      "objective/train/advantage_avg": 0.491913765668869,
      "objective/train/docs_used": 547043,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2985484600067139,
      "objective/train/original_loss": 1.2985484600067139,
      "objective/train/theoretical_loss": 3.545063572755053,
      "objective/train/tokens_used": 1517957600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2422856092453003,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504230260849,
      "objective/train/weighted_lm_loss": 1.3640440702438354,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.545063572755053,
      "tokens_seen": 1497497600
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000551875626880642,
      "loss": 1.4088,
      "theoretical_loss": 3.545063572755053,
      "tokens_seen": 1497497600
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005518555667001003,
      "loss": 1.4363,
      "theoretical_loss": 3.545049999567563,
      "tokens_seen": 1497563136
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005518355065195587,
      "loss": 1.5174,
      "theoretical_loss": 3.545036427140354,
      "tokens_seen": 1497628672
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005518154463390171,
      "loss": 1.4829,
      "theoretical_loss": 3.545022855473351,
      "tokens_seen": 1497694208
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005517953861584754,
      "loss": 1.5415,
      "theoretical_loss": 3.545009284566478,
      "tokens_seen": 1497759744
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005517753259779338,
      "loss": 1.4873,
      "theoretical_loss": 3.5449957144196587,
      "tokens_seen": 1497825280
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005517552657973923,
      "loss": 1.502,
      "theoretical_loss": 3.544982145032817,
      "tokens_seen": 1497890816
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005517352056168505,
      "loss": 1.3414,
      "theoretical_loss": 3.5449685764058785,
      "tokens_seen": 1497956352
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000551715145436309,
      "loss": 1.4593,
      "theoretical_loss": 3.5449550085387664,
      "tokens_seen": 1498021888
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005516950852557672,
      "loss": 1.4299,
      "theoretical_loss": 3.544941441431405,
      "tokens_seen": 1498087424
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005516750250752257,
      "loss": 1.4772,
      "theoretical_loss": 3.5449278750837188,
      "tokens_seen": 1498152960
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005516549648946841,
      "loss": 1.5411,
      "theoretical_loss": 3.5449143094956317,
      "tokens_seen": 1498218496
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005516349047141424,
      "loss": 1.6133,
      "theoretical_loss": 3.544900744667068,
      "tokens_seen": 1498284032
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005516148445336008,
      "loss": 1.467,
      "theoretical_loss": 3.5448871805979527,
      "tokens_seen": 1498349568
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005515947843530593,
      "loss": 1.4524,
      "theoretical_loss": 3.5448736172882085,
      "tokens_seen": 1498415104
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005515747241725175,
      "loss": 1.4244,
      "theoretical_loss": 3.544860054737761,
      "tokens_seen": 1498480640
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000551554663991976,
      "loss": 1.4537,
      "theoretical_loss": 3.544846492946534,
      "tokens_seen": 1498546176
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005515346038114342,
      "loss": 1.4604,
      "theoretical_loss": 3.5448329319144523,
      "tokens_seen": 1498611712
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005515145436308927,
      "loss": 1.3772,
      "theoretical_loss": 3.5448193716414393,
      "tokens_seen": 1498677248
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005514944834503511,
      "loss": 1.3695,
      "theoretical_loss": 3.5448058121274197,
      "tokens_seen": 1498742784
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005514744232698094,
      "loss": 1.5347,
      "theoretical_loss": 3.544792253372319,
      "tokens_seen": 1498808320
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005514543630892678,
      "loss": 1.5598,
      "theoretical_loss": 3.5447786953760594,
      "tokens_seen": 1498873856
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005514343029087262,
      "loss": 1.4211,
      "theoretical_loss": 3.5447651381385668,
      "tokens_seen": 1498939392
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005514142427281845,
      "loss": 1.4259,
      "theoretical_loss": 3.5447515816597646,
      "tokens_seen": 1499004928
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000551394182547643,
      "loss": 1.4391,
      "theoretical_loss": 3.544738025939578,
      "tokens_seen": 1499070464
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005513741223671012,
      "loss": 1.536,
      "theoretical_loss": 3.5447244709779318,
      "tokens_seen": 1499136000
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005513540621865597,
      "loss": 1.4451,
      "theoretical_loss": 3.544710916774749,
      "tokens_seen": 1499201536
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000551334002006018,
      "loss": 1.5424,
      "theoretical_loss": 3.5446973633299543,
      "tokens_seen": 1499267072
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005513139418254764,
      "loss": 1.4079,
      "theoretical_loss": 3.544683810643473,
      "tokens_seen": 1499332608
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005512938816449348,
      "loss": 1.4708,
      "theoretical_loss": 3.544670258715229,
      "tokens_seen": 1499398144
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005512738214643932,
      "loss": 1.366,
      "theoretical_loss": 3.5446567075451463,
      "tokens_seen": 1499463680
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005512537612838515,
      "loss": 1.4917,
      "theoretical_loss": 3.5446431571331503,
      "tokens_seen": 1499529216
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00055123370110331,
      "loss": 1.4754,
      "theoretical_loss": 3.544629607479165,
      "tokens_seen": 1499594752
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005512136409227683,
      "loss": 1.5019,
      "theoretical_loss": 3.5446160585831143,
      "tokens_seen": 1499660288
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005511935807422267,
      "loss": 1.4249,
      "theoretical_loss": 3.544602510444924,
      "tokens_seen": 1499725824
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000551173520561685,
      "loss": 1.4962,
      "theoretical_loss": 3.544588963064517,
      "tokens_seen": 1499791360
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005511534603811434,
      "loss": 1.4503,
      "theoretical_loss": 3.5445754164418193,
      "tokens_seen": 1499856896
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005511334002006018,
      "loss": 1.3653,
      "theoretical_loss": 3.5445618705767545,
      "tokens_seen": 1499922432
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005511133400200602,
      "loss": 1.4687,
      "theoretical_loss": 3.544548325469247,
      "tokens_seen": 1499987968
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005510932798395185,
      "loss": 1.5985,
      "theoretical_loss": 3.5445347811192223,
      "tokens_seen": 1500053504
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005510732196589769,
      "loss": 1.4535,
      "theoretical_loss": 3.544521237526604,
      "tokens_seen": 1500119040
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005510531594784353,
      "loss": 1.5433,
      "theoretical_loss": 3.544507694691317,
      "tokens_seen": 1500184576
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005510330992978937,
      "loss": 1.4924,
      "theoretical_loss": 3.5444941526132863,
      "tokens_seen": 1500250112
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005510130391173521,
      "loss": 1.5876,
      "theoretical_loss": 3.5444806112924363,
      "tokens_seen": 1500315648
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005509929789368105,
      "loss": 1.4247,
      "theoretical_loss": 3.5444670707286905,
      "tokens_seen": 1500381184
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005509729187562689,
      "loss": 1.4007,
      "theoretical_loss": 3.5444535309219747,
      "tokens_seen": 1500446720
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005509528585757272,
      "loss": 1.5752,
      "theoretical_loss": 3.5444399918722134,
      "tokens_seen": 1500512256
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005509327983951856,
      "loss": 1.4832,
      "theoretical_loss": 3.5444264535793306,
      "tokens_seen": 1500577792
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005509127382146439,
      "loss": 1.4394,
      "theoretical_loss": 3.5444129160432514,
      "tokens_seen": 1500643328
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005508926780341024,
      "loss": 1.3452,
      "theoretical_loss": 3.5443993792639006,
      "tokens_seen": 1500708864
    },
    {
      "epoch": 0.45,
      "objective/train/advantage_avg": 0.49255383014678955,
      "objective/train/docs_used": 548178,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3971058130264282,
      "objective/train/original_loss": 1.3971056938171387,
      "objective/train/theoretical_loss": 3.5443858432412028,
      "objective/train/tokens_used": 1521234400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24296294152736664,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504906177520752,
      "objective/train/weighted_lm_loss": 1.4674906730651855,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.5443858432412028,
      "tokens_seen": 1500774400
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005508726178535607,
      "loss": 1.3684,
      "theoretical_loss": 3.5443858432412028,
      "tokens_seen": 1500774400
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005508525576730191,
      "loss": 1.4479,
      "theoretical_loss": 3.544372307975082,
      "tokens_seen": 1500839936
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005508324974924775,
      "loss": 1.4209,
      "theoretical_loss": 3.544358773465464,
      "tokens_seen": 1500905472
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005508124373119358,
      "loss": 1.367,
      "theoretical_loss": 3.544345239712272,
      "tokens_seen": 1500971008
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005507923771313942,
      "loss": 1.4824,
      "theoretical_loss": 3.5443317067154325,
      "tokens_seen": 1501036544
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005507723169508527,
      "loss": 1.5025,
      "theoretical_loss": 3.544318174474869,
      "tokens_seen": 1501102080
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005507522567703109,
      "loss": 1.5565,
      "theoretical_loss": 3.5443046429905065,
      "tokens_seen": 1501167616
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005507321965897694,
      "loss": 1.418,
      "theoretical_loss": 3.544291112262269,
      "tokens_seen": 1501233152
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005507121364092276,
      "loss": 1.4463,
      "theoretical_loss": 3.544277582290083,
      "tokens_seen": 1501298688
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005506920762286861,
      "loss": 1.463,
      "theoretical_loss": 3.5442640530738716,
      "tokens_seen": 1501364224
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005506720160481445,
      "loss": 1.4513,
      "theoretical_loss": 3.5442505246135605,
      "tokens_seen": 1501429760
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0005506519558676028,
      "loss": 1.3772,
      "theoretical_loss": 3.544236996909074,
      "tokens_seen": 1501495296
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005506318956870612,
      "loss": 1.5978,
      "theoretical_loss": 3.544223469960337,
      "tokens_seen": 1501560832
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005506118355065197,
      "loss": 1.469,
      "theoretical_loss": 3.544209943767275,
      "tokens_seen": 1501626368
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005505917753259779,
      "loss": 1.5731,
      "theoretical_loss": 3.5441964183298116,
      "tokens_seen": 1501691904
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005505717151454364,
      "loss": 1.5245,
      "theoretical_loss": 3.5441828936478723,
      "tokens_seen": 1501757440
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005505516549648946,
      "loss": 1.4009,
      "theoretical_loss": 3.5441693697213816,
      "tokens_seen": 1501822976
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005505315947843531,
      "loss": 1.5632,
      "theoretical_loss": 3.5441558465502645,
      "tokens_seen": 1501888512
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005505115346038115,
      "loss": 1.4323,
      "theoretical_loss": 3.5441423241344463,
      "tokens_seen": 1501954048
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005504914744232698,
      "loss": 1.4214,
      "theoretical_loss": 3.544128802473851,
      "tokens_seen": 1502019584
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005504714142427282,
      "loss": 1.4696,
      "theoretical_loss": 3.5441152815684043,
      "tokens_seen": 1502085120
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005504513540621866,
      "loss": 1.4454,
      "theoretical_loss": 3.5441017614180303,
      "tokens_seen": 1502150656
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005504312938816449,
      "loss": 1.4418,
      "theoretical_loss": 3.5440882420226547,
      "tokens_seen": 1502216192
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005504112337011034,
      "loss": 1.4216,
      "theoretical_loss": 3.5440747233822023,
      "tokens_seen": 1502281728
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005503911735205616,
      "loss": 1.4488,
      "theoretical_loss": 3.5440612054965968,
      "tokens_seen": 1502347264
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005503711133400201,
      "loss": 1.4281,
      "theoretical_loss": 3.544047688365765,
      "tokens_seen": 1502412800
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005503510531594784,
      "loss": 1.4392,
      "theoretical_loss": 3.5440341719896304,
      "tokens_seen": 1502478336
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005503309929789368,
      "loss": 1.6255,
      "theoretical_loss": 3.5440206563681182,
      "tokens_seen": 1502543872
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005503109327983952,
      "loss": 1.4649,
      "theoretical_loss": 3.544007141501154,
      "tokens_seen": 1502609408
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005502908726178536,
      "loss": 1.3713,
      "theoretical_loss": 3.5439936273886623,
      "tokens_seen": 1502674944
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005502708124373119,
      "loss": 1.3936,
      "theoretical_loss": 3.543980114030568,
      "tokens_seen": 1502740480
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005502507522567704,
      "loss": 1.4779,
      "theoretical_loss": 3.5439666014267965,
      "tokens_seen": 1502806016
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005502306920762287,
      "loss": 1.5143,
      "theoretical_loss": 3.543953089577272,
      "tokens_seen": 1502871552
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005502106318956871,
      "loss": 1.4365,
      "theoretical_loss": 3.543939578481921,
      "tokens_seen": 1502937088
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005501905717151454,
      "loss": 1.4959,
      "theoretical_loss": 3.5439260681406664,
      "tokens_seen": 1503002624
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005501705115346038,
      "loss": 1.5091,
      "theoretical_loss": 3.5439125585534352,
      "tokens_seen": 1503068160
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005501504513540622,
      "loss": 1.4725,
      "theoretical_loss": 3.5438990497201512,
      "tokens_seen": 1503133696
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005501303911735206,
      "loss": 1.4565,
      "theoretical_loss": 3.54388554164074,
      "tokens_seen": 1503199232
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005501103309929789,
      "loss": 1.476,
      "theoretical_loss": 3.543872034315127,
      "tokens_seen": 1503264768
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005500902708124373,
      "loss": 1.3416,
      "theoretical_loss": 3.5438585277432364,
      "tokens_seen": 1503330304
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005500702106318957,
      "loss": 1.4431,
      "theoretical_loss": 3.5438450219249935,
      "tokens_seen": 1503395840
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005500501504513541,
      "loss": 1.3754,
      "theoretical_loss": 3.5438315168603243,
      "tokens_seen": 1503461376
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005500300902708124,
      "loss": 1.4882,
      "theoretical_loss": 3.5438180125491527,
      "tokens_seen": 1503526912
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005500100300902709,
      "loss": 1.4491,
      "theoretical_loss": 3.5438045089914043,
      "tokens_seen": 1503592448
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005499899699097291,
      "loss": 1.3401,
      "theoretical_loss": 3.5437910061870044,
      "tokens_seen": 1503657984
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005499699097291876,
      "loss": 1.4935,
      "theoretical_loss": 3.5437775041358783,
      "tokens_seen": 1503723520
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005499498495486459,
      "loss": 1.4652,
      "theoretical_loss": 3.5437640028379507,
      "tokens_seen": 1503789056
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005499297893681043,
      "loss": 1.3454,
      "theoretical_loss": 3.543750502293147,
      "tokens_seen": 1503854592
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005499097291875627,
      "loss": 1.5591,
      "theoretical_loss": 3.543737002501392,
      "tokens_seen": 1503920128
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005498896690070211,
      "loss": 1.5383,
      "theoretical_loss": 3.543723503462611,
      "tokens_seen": 1503985664
    },
    {
      "epoch": 0.46,
      "objective/train/advantage_avg": 0.49412354826927185,
      "objective/train/docs_used": 549351,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.376686692237854,
      "objective/train/original_loss": 1.3766865730285645,
      "objective/train/theoretical_loss": 3.54371000517673,
      "objective/train/tokens_used": 1524511200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2444082796573639,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506548881530762,
      "objective/train/weighted_lm_loss": 1.4463905096054077,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.54371000517673,
      "tokens_seen": 1504051200
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005498696088264794,
      "loss": 1.451,
      "theoretical_loss": 3.54371000517673,
      "tokens_seen": 1504051200
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005498495486459379,
      "loss": 1.3732,
      "theoretical_loss": 3.5436965076436735,
      "tokens_seen": 1504116736
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005498294884653961,
      "loss": 1.4219,
      "theoretical_loss": 3.5436830108633663,
      "tokens_seen": 1504182272
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005498094282848546,
      "loss": 1.3787,
      "theoretical_loss": 3.5436695148357344,
      "tokens_seen": 1504247808
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005497893681043129,
      "loss": 1.4236,
      "theoretical_loss": 3.543656019560703,
      "tokens_seen": 1504313344
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005497693079237713,
      "loss": 1.3481,
      "theoretical_loss": 3.543642525038197,
      "tokens_seen": 1504378880
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005497492477432297,
      "loss": 1.4673,
      "theoretical_loss": 3.5436290312681415,
      "tokens_seen": 1504444416
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000549729187562688,
      "loss": 1.3544,
      "theoretical_loss": 3.543615538250462,
      "tokens_seen": 1504509952
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005497091273821464,
      "loss": 1.4618,
      "theoretical_loss": 3.543602045985084,
      "tokens_seen": 1504575488
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005496890672016049,
      "loss": 1.4159,
      "theoretical_loss": 3.543588554471933,
      "tokens_seen": 1504641024
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005496690070210631,
      "loss": 1.4065,
      "theoretical_loss": 3.543575063710933,
      "tokens_seen": 1504706560
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005496489468405216,
      "loss": 1.3805,
      "theoretical_loss": 3.543561573702011,
      "tokens_seen": 1504772096
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00054962888665998,
      "loss": 1.4569,
      "theoretical_loss": 3.5435480844450913,
      "tokens_seen": 1504837632
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005496088264794383,
      "loss": 1.4568,
      "theoretical_loss": 3.5435345959400992,
      "tokens_seen": 1504903168
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005495887662988967,
      "loss": 1.4886,
      "theoretical_loss": 3.543521108186961,
      "tokens_seen": 1504968704
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000549568706118355,
      "loss": 1.5081,
      "theoretical_loss": 3.5435076211856007,
      "tokens_seen": 1505034240
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005495486459378135,
      "loss": 1.568,
      "theoretical_loss": 3.543494134935944,
      "tokens_seen": 1505099776
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005495285857572719,
      "loss": 1.3991,
      "theoretical_loss": 3.5434806494379174,
      "tokens_seen": 1505165312
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005495085255767302,
      "loss": 1.583,
      "theoretical_loss": 3.543467164691445,
      "tokens_seen": 1505230848
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005494884653961886,
      "loss": 1.3796,
      "theoretical_loss": 3.543453680696453,
      "tokens_seen": 1505296384
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000549468405215647,
      "loss": 1.3091,
      "theoretical_loss": 3.5434401974528664,
      "tokens_seen": 1505361920
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005494483450351053,
      "loss": 1.4705,
      "theoretical_loss": 3.543426714960611,
      "tokens_seen": 1505427456
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005494282848545638,
      "loss": 1.4525,
      "theoretical_loss": 3.5434132332196113,
      "tokens_seen": 1505492992
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000549408224674022,
      "loss": 1.3381,
      "theoretical_loss": 3.543399752229794,
      "tokens_seen": 1505558528
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005493881644934805,
      "loss": 1.478,
      "theoretical_loss": 3.5433862719910834,
      "tokens_seen": 1505624064
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005493681043129388,
      "loss": 1.4936,
      "theoretical_loss": 3.5433727925034058,
      "tokens_seen": 1505689600
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005493480441323972,
      "loss": 1.4595,
      "theoretical_loss": 3.543359313766686,
      "tokens_seen": 1505755136
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005493279839518556,
      "loss": 1.4392,
      "theoretical_loss": 3.5433458357808503,
      "tokens_seen": 1505820672
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000549307923771314,
      "loss": 1.4036,
      "theoretical_loss": 3.543332358545823,
      "tokens_seen": 1505886208
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005492878635907723,
      "loss": 1.4997,
      "theoretical_loss": 3.543318882061531,
      "tokens_seen": 1505951744
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005492678034102308,
      "loss": 1.4619,
      "theoretical_loss": 3.543305406327899,
      "tokens_seen": 1506017280
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000549247743229689,
      "loss": 1.2831,
      "theoretical_loss": 3.5432919313448528,
      "tokens_seen": 1506082816
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005492276830491475,
      "loss": 1.5585,
      "theoretical_loss": 3.5432784571123177,
      "tokens_seen": 1506148352
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005492076228686058,
      "loss": 1.3537,
      "theoretical_loss": 3.5432649836302192,
      "tokens_seen": 1506213888
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005491875626880642,
      "loss": 1.4305,
      "theoretical_loss": 3.5432515108984832,
      "tokens_seen": 1506279424
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005491675025075226,
      "loss": 1.3287,
      "theoretical_loss": 3.543238038917035,
      "tokens_seen": 1506344960
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000549147442326981,
      "loss": 1.497,
      "theoretical_loss": 3.5432245676858,
      "tokens_seen": 1506410496
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005491273821464393,
      "loss": 1.2653,
      "theoretical_loss": 3.543211097204704,
      "tokens_seen": 1506476032
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005491073219658977,
      "loss": 1.4795,
      "theoretical_loss": 3.543197627473673,
      "tokens_seen": 1506541568
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005490872617853561,
      "loss": 1.4988,
      "theoretical_loss": 3.5431841584926325,
      "tokens_seen": 1506607104
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005490672016048145,
      "loss": 1.4692,
      "theoretical_loss": 3.5431706902615074,
      "tokens_seen": 1506672640
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005490471414242728,
      "loss": 1.3753,
      "theoretical_loss": 3.543157222780224,
      "tokens_seen": 1506738176
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005490270812437313,
      "loss": 1.5409,
      "theoretical_loss": 3.543143756048708,
      "tokens_seen": 1506803712
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005490070210631895,
      "loss": 1.4052,
      "theoretical_loss": 3.543130290066884,
      "tokens_seen": 1506869248
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000548986960882648,
      "loss": 1.4911,
      "theoretical_loss": 3.543116824834679,
      "tokens_seen": 1506934784
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005489669007021063,
      "loss": 1.4316,
      "theoretical_loss": 3.543103360352018,
      "tokens_seen": 1507000320
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005489468405215647,
      "loss": 1.4101,
      "theoretical_loss": 3.5430898966188265,
      "tokens_seen": 1507065856
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005489267803410231,
      "loss": 1.4468,
      "theoretical_loss": 3.543076433635031,
      "tokens_seen": 1507131392
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005489067201604815,
      "loss": 1.4877,
      "theoretical_loss": 3.5430629714005564,
      "tokens_seen": 1507196928
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005488866599799398,
      "loss": 1.4747,
      "theoretical_loss": 3.5430495099153285,
      "tokens_seen": 1507262464
    },
    {
      "epoch": 0.46,
      "objective/train/advantage_avg": 0.4903029799461365,
      "objective/train/docs_used": 550589,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4796637296676636,
      "objective/train/original_loss": 1.4796637296676636,
      "objective/train/theoretical_loss": 3.543036049179274,
      "objective/train/tokens_used": 1527788000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24085567891597748,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502548217773438,
      "objective/train/weighted_lm_loss": 1.5539206266403198,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.543036049179274,
      "tokens_seen": 1507328000
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005488665997993983,
      "loss": 1.434,
      "theoretical_loss": 3.543036049179274,
      "tokens_seen": 1507328000
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005488465396188565,
      "loss": 1.4195,
      "theoretical_loss": 3.5430225891923173,
      "tokens_seen": 1507393536
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000548826479438315,
      "loss": 1.5252,
      "theoretical_loss": 3.543009129954385,
      "tokens_seen": 1507459072
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005488064192577733,
      "loss": 1.5043,
      "theoretical_loss": 3.5429956714654023,
      "tokens_seen": 1507524608
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005487863590772317,
      "loss": 1.5889,
      "theoretical_loss": 3.5429822137252955,
      "tokens_seen": 1507590144
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005487662988966901,
      "loss": 1.4682,
      "theoretical_loss": 3.5429687567339903,
      "tokens_seen": 1507655680
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005487462387161484,
      "loss": 1.4913,
      "theoretical_loss": 3.542955300491412,
      "tokens_seen": 1507721216
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005487261785356068,
      "loss": 1.391,
      "theoretical_loss": 3.5429418449974865,
      "tokens_seen": 1507786752
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005487061183550653,
      "loss": 1.4008,
      "theoretical_loss": 3.54292839025214,
      "tokens_seen": 1507852288
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005486860581745235,
      "loss": 1.4113,
      "theoretical_loss": 3.5429149362552987,
      "tokens_seen": 1507917824
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000548665997993982,
      "loss": 1.4401,
      "theoretical_loss": 3.5429014830068875,
      "tokens_seen": 1507983360
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005486459378134404,
      "loss": 1.4223,
      "theoretical_loss": 3.542888030506832,
      "tokens_seen": 1508048896
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005486258776328987,
      "loss": 1.4686,
      "theoretical_loss": 3.5428745787550593,
      "tokens_seen": 1508114432
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005486058174523571,
      "loss": 1.5309,
      "theoretical_loss": 3.542861127751495,
      "tokens_seen": 1508179968
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005485857572718154,
      "loss": 1.46,
      "theoretical_loss": 3.5428476774960638,
      "tokens_seen": 1508245504
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005485656970912738,
      "loss": 1.515,
      "theoretical_loss": 3.542834227988693,
      "tokens_seen": 1508311040
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005485456369107323,
      "loss": 1.3959,
      "theoretical_loss": 3.5428207792293076,
      "tokens_seen": 1508376576
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005485255767301905,
      "loss": 1.4836,
      "theoretical_loss": 3.542807331217834,
      "tokens_seen": 1508442112
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000548505516549649,
      "loss": 1.4171,
      "theoretical_loss": 3.5427938839541975,
      "tokens_seen": 1508507648
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005484854563691072,
      "loss": 1.406,
      "theoretical_loss": 3.542780437438325,
      "tokens_seen": 1508573184
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005484653961885657,
      "loss": 1.4354,
      "theoretical_loss": 3.542766991670142,
      "tokens_seen": 1508638720
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005484453360080241,
      "loss": 1.4614,
      "theoretical_loss": 3.5427535466495734,
      "tokens_seen": 1508704256
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005484252758274824,
      "loss": 1.4972,
      "theoretical_loss": 3.5427401023765466,
      "tokens_seen": 1508769792
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005484052156469408,
      "loss": 1.4444,
      "theoretical_loss": 3.542726658850987,
      "tokens_seen": 1508835328
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005483851554663992,
      "loss": 1.4226,
      "theoretical_loss": 3.542713216072821,
      "tokens_seen": 1508900864
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005483650952858575,
      "loss": 1.666,
      "theoretical_loss": 3.542699774041974,
      "tokens_seen": 1508966400
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000548345035105316,
      "loss": 1.4939,
      "theoretical_loss": 3.5426863327583717,
      "tokens_seen": 1509031936
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005483249749247743,
      "loss": 1.4317,
      "theoretical_loss": 3.5426728922219413,
      "tokens_seen": 1509097472
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005483049147442327,
      "loss": 1.3822,
      "theoretical_loss": 3.542659452432608,
      "tokens_seen": 1509163008
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005482848545636911,
      "loss": 1.4413,
      "theoretical_loss": 3.542646013390298,
      "tokens_seen": 1509228544
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005482647943831495,
      "loss": 1.496,
      "theoretical_loss": 3.5426325750949372,
      "tokens_seen": 1509294080
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005482447342026078,
      "loss": 1.4478,
      "theoretical_loss": 3.5426191375464517,
      "tokens_seen": 1509359616
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005482246740220662,
      "loss": 1.4401,
      "theoretical_loss": 3.542605700744768,
      "tokens_seen": 1509425152
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005482046138415245,
      "loss": 1.4662,
      "theoretical_loss": 3.5425922646898114,
      "tokens_seen": 1509490688
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000548184553660983,
      "loss": 1.4113,
      "theoretical_loss": 3.542578829381509,
      "tokens_seen": 1509556224
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005481644934804413,
      "loss": 1.4389,
      "theoretical_loss": 3.5425653948197855,
      "tokens_seen": 1509621760
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005481444332998997,
      "loss": 1.4204,
      "theoretical_loss": 3.542551961004568,
      "tokens_seen": 1509687296
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000548124373119358,
      "loss": 1.4889,
      "theoretical_loss": 3.5425385279357826,
      "tokens_seen": 1509752832
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005481043129388165,
      "loss": 1.508,
      "theoretical_loss": 3.542525095613356,
      "tokens_seen": 1509818368
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005480842527582748,
      "loss": 1.4832,
      "theoretical_loss": 3.5425116640372125,
      "tokens_seen": 1509883904
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005480641925777332,
      "loss": 1.4257,
      "theoretical_loss": 3.5424982332072794,
      "tokens_seen": 1509949440
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005480441323971917,
      "loss": 1.3955,
      "theoretical_loss": 3.5424848031234832,
      "tokens_seen": 1510014976
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005480240722166499,
      "loss": 1.5901,
      "theoretical_loss": 3.54247137378575,
      "tokens_seen": 1510080512
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005480040120361084,
      "loss": 1.4574,
      "theoretical_loss": 3.542457945194005,
      "tokens_seen": 1510146048
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005479839518555667,
      "loss": 1.5113,
      "theoretical_loss": 3.5424445173481756,
      "tokens_seen": 1510211584
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005479638916750251,
      "loss": 1.4511,
      "theoretical_loss": 3.542431090248187,
      "tokens_seen": 1510277120
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005479438314944835,
      "loss": 1.485,
      "theoretical_loss": 3.5424176638939655,
      "tokens_seen": 1510342656
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005479237713139419,
      "loss": 1.4243,
      "theoretical_loss": 3.542404238285438,
      "tokens_seen": 1510408192
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005479037111334002,
      "loss": 1.503,
      "theoretical_loss": 3.5423908134225304,
      "tokens_seen": 1510473728
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005478836509528587,
      "loss": 1.5393,
      "theoretical_loss": 3.5423773893051695,
      "tokens_seen": 1510539264
    },
    {
      "epoch": 0.46,
      "objective/train/advantage_avg": 0.49604833126068115,
      "objective/train/docs_used": 551665,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3969905376434326,
      "objective/train/original_loss": 1.3969905376434326,
      "objective/train/theoretical_loss": 3.54236396593328,
      "objective/train/tokens_used": 1531064800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2462085634469986,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050856590270996,
      "objective/train/weighted_lm_loss": 1.467955470085144,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0100501775741577,
      "theoretical_loss": 3.54236396593328,
      "tokens_seen": 1510604800
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005478635907723169,
      "loss": 1.4416,
      "theoretical_loss": 3.54236396593328,
      "tokens_seen": 1510604800
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005478435305917754,
      "loss": 1.4443,
      "theoretical_loss": 3.5423505433067897,
      "tokens_seen": 1510670336
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005478234704112337,
      "loss": 1.4306,
      "theoretical_loss": 3.5423371214256245,
      "tokens_seen": 1510735872
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005478034102306921,
      "loss": 1.4424,
      "theoretical_loss": 3.5423237002897103,
      "tokens_seen": 1510801408
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005477833500501505,
      "loss": 1.2937,
      "theoretical_loss": 3.542310279898974,
      "tokens_seen": 1510866944
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005477632898696088,
      "loss": 1.5836,
      "theoretical_loss": 3.542296860253341,
      "tokens_seen": 1510932480
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005477432296890672,
      "loss": 1.5226,
      "theoretical_loss": 3.5422834413527378,
      "tokens_seen": 1510998016
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005477231695085257,
      "loss": 1.5162,
      "theoretical_loss": 3.5422700231970916,
      "tokens_seen": 1511063552
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005477031093279839,
      "loss": 1.4499,
      "theoretical_loss": 3.5422566057863283,
      "tokens_seen": 1511129088
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005476830491474424,
      "loss": 1.5153,
      "theoretical_loss": 3.5422431891203736,
      "tokens_seen": 1511194624
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005476629889669008,
      "loss": 1.4533,
      "theoretical_loss": 3.5422297731991548,
      "tokens_seen": 1511260160
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005476429287863591,
      "loss": 1.4465,
      "theoretical_loss": 3.542216358022598,
      "tokens_seen": 1511325696
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005476228686058175,
      "loss": 1.4495,
      "theoretical_loss": 3.5422029435906293,
      "tokens_seen": 1511391232
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005476028084252758,
      "loss": 1.3619,
      "theoretical_loss": 3.542189529903175,
      "tokens_seen": 1511456768
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005475827482447342,
      "loss": 1.4086,
      "theoretical_loss": 3.542176116960162,
      "tokens_seen": 1511522304
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005475626880641927,
      "loss": 1.4097,
      "theoretical_loss": 3.542162704761516,
      "tokens_seen": 1511587840
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005475426278836509,
      "loss": 1.5129,
      "theoretical_loss": 3.5421492933071645,
      "tokens_seen": 1511653376
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005475225677031094,
      "loss": 1.4597,
      "theoretical_loss": 3.542135882597033,
      "tokens_seen": 1511718912
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005475025075225676,
      "loss": 1.545,
      "theoretical_loss": 3.542122472631048,
      "tokens_seen": 1511784448
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005474824473420261,
      "loss": 1.4326,
      "theoretical_loss": 3.5421090634091366,
      "tokens_seen": 1511849984
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005474623871614845,
      "loss": 1.5445,
      "theoretical_loss": 3.5420956549312246,
      "tokens_seen": 1511915520
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005474423269809428,
      "loss": 1.4479,
      "theoretical_loss": 3.5420822471972384,
      "tokens_seen": 1511981056
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005474222668004012,
      "loss": 1.5197,
      "theoretical_loss": 3.542068840207105,
      "tokens_seen": 1512046592
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005474022066198596,
      "loss": 1.3653,
      "theoretical_loss": 3.5420554339607513,
      "tokens_seen": 1512112128
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005473821464393179,
      "loss": 1.4437,
      "theoretical_loss": 3.542042028458103,
      "tokens_seen": 1512177664
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005473620862587764,
      "loss": 1.4814,
      "theoretical_loss": 3.5420286236990863,
      "tokens_seen": 1512243200
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005473420260782347,
      "loss": 1.4416,
      "theoretical_loss": 3.5420152196836288,
      "tokens_seen": 1512308736
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005473219658976931,
      "loss": 1.4043,
      "theoretical_loss": 3.542001816411656,
      "tokens_seen": 1512374272
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005473019057171515,
      "loss": 1.4871,
      "theoretical_loss": 3.5419884138830953,
      "tokens_seen": 1512439808
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005472818455366099,
      "loss": 1.348,
      "theoretical_loss": 3.541975012097873,
      "tokens_seen": 1512505344
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005472617853560682,
      "loss": 1.4122,
      "theoretical_loss": 3.541961611055915,
      "tokens_seen": 1512570880
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005472417251755266,
      "loss": 1.4929,
      "theoretical_loss": 3.5419482107571483,
      "tokens_seen": 1512636416
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005472216649949849,
      "loss": 1.4008,
      "theoretical_loss": 3.5419348112015,
      "tokens_seen": 1512701952
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005472016048144434,
      "loss": 1.5312,
      "theoretical_loss": 3.5419214123888967,
      "tokens_seen": 1512767488
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005471815446339017,
      "loss": 1.3579,
      "theoretical_loss": 3.5419080143192643,
      "tokens_seen": 1512833024
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005471614844533601,
      "loss": 1.4812,
      "theoretical_loss": 3.541894616992529,
      "tokens_seen": 1512898560
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005471414242728184,
      "loss": 1.4826,
      "theoretical_loss": 3.5418812204086194,
      "tokens_seen": 1512964096
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005471213640922769,
      "loss": 1.4095,
      "theoretical_loss": 3.5418678245674604,
      "tokens_seen": 1513029632
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005471013039117352,
      "loss": 1.4371,
      "theoretical_loss": 3.541854429468979,
      "tokens_seen": 1513095168
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005470812437311936,
      "loss": 1.5386,
      "theoretical_loss": 3.541841035113102,
      "tokens_seen": 1513160704
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005470611835506519,
      "loss": 1.4565,
      "theoretical_loss": 3.541827641499756,
      "tokens_seen": 1513226240
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005470411233701103,
      "loss": 1.4455,
      "theoretical_loss": 3.541814248628868,
      "tokens_seen": 1513291776
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005470210631895687,
      "loss": 1.4182,
      "theoretical_loss": 3.5418008565003647,
      "tokens_seen": 1513357312
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005470010030090271,
      "loss": 1.4487,
      "theoretical_loss": 3.5417874651141723,
      "tokens_seen": 1513422848
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005469809428284854,
      "loss": 1.5151,
      "theoretical_loss": 3.5417740744702177,
      "tokens_seen": 1513488384
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005469608826479439,
      "loss": 1.4674,
      "theoretical_loss": 3.5417606845684277,
      "tokens_seen": 1513553920
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005469408224674022,
      "loss": 1.5056,
      "theoretical_loss": 3.541747295408729,
      "tokens_seen": 1513619456
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005469207622868606,
      "loss": 1.4857,
      "theoretical_loss": 3.5417339069910487,
      "tokens_seen": 1513684992
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000546900702106319,
      "loss": 1.5222,
      "theoretical_loss": 3.541720519315313,
      "tokens_seen": 1513750528
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005468806419257773,
      "loss": 1.3823,
      "theoretical_loss": 3.5417071323814486,
      "tokens_seen": 1513816064
    },
    {
      "epoch": 0.46,
      "objective/train/advantage_avg": 0.4922076165676117,
      "objective/train/docs_used": 552897,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.236923098564148,
      "objective/train/original_loss": 1.236923098564148,
      "objective/train/theoretical_loss": 3.541693746189383,
      "objective/train/tokens_used": 1534341600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2430669218301773,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504564046859741,
      "objective/train/weighted_lm_loss": 1.2992579936981201,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.541693746189383,
      "tokens_seen": 1513881600
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005468605817452357,
      "loss": 1.4001,
      "theoretical_loss": 3.541693746189383,
      "tokens_seen": 1513881600
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005468405215646941,
      "loss": 1.3771,
      "theoretical_loss": 3.5416803607390426,
      "tokens_seen": 1513947136
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005468204613841524,
      "loss": 1.3739,
      "theoretical_loss": 3.541666976030354,
      "tokens_seen": 1514012672
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005468004012036109,
      "loss": 1.5025,
      "theoretical_loss": 3.541653592063244,
      "tokens_seen": 1514078208
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005467803410230691,
      "loss": 1.4162,
      "theoretical_loss": 3.5416402088376397,
      "tokens_seen": 1514143744
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005467602808425276,
      "loss": 1.5391,
      "theoretical_loss": 3.5416268263534683,
      "tokens_seen": 1514209280
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000546740220661986,
      "loss": 1.4214,
      "theoretical_loss": 3.5416134446106553,
      "tokens_seen": 1514274816
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005467201604814443,
      "loss": 1.3879,
      "theoretical_loss": 3.541600063609129,
      "tokens_seen": 1514340352
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005467001003009027,
      "loss": 1.4342,
      "theoretical_loss": 3.5415866833488154,
      "tokens_seen": 1514405888
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005466800401203612,
      "loss": 1.4778,
      "theoretical_loss": 3.541573303829642,
      "tokens_seen": 1514471424
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005466599799398194,
      "loss": 1.4285,
      "theoretical_loss": 3.5415599250515344,
      "tokens_seen": 1514536960
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005466399197592779,
      "loss": 1.4479,
      "theoretical_loss": 3.541546547014421,
      "tokens_seen": 1514602496
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005466198595787361,
      "loss": 1.4921,
      "theoretical_loss": 3.541533169718228,
      "tokens_seen": 1514668032
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005465997993981946,
      "loss": 1.5108,
      "theoretical_loss": 3.5415197931628826,
      "tokens_seen": 1514733568
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000546579739217653,
      "loss": 1.3408,
      "theoretical_loss": 3.5415064173483115,
      "tokens_seen": 1514799104
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005465596790371113,
      "loss": 1.4791,
      "theoretical_loss": 3.5414930422744413,
      "tokens_seen": 1514864640
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005465396188565698,
      "loss": 1.4777,
      "theoretical_loss": 3.541479667941199,
      "tokens_seen": 1514930176
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000546519558676028,
      "loss": 1.465,
      "theoretical_loss": 3.5414662943485125,
      "tokens_seen": 1514995712
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005464994984954865,
      "loss": 1.5371,
      "theoretical_loss": 3.541452921496308,
      "tokens_seen": 1515061248
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005464794383149449,
      "loss": 1.4407,
      "theoretical_loss": 3.5414395493845126,
      "tokens_seen": 1515126784
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005464593781344032,
      "loss": 1.4095,
      "theoretical_loss": 3.5414261780130527,
      "tokens_seen": 1515192320
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005464393179538616,
      "loss": 1.409,
      "theoretical_loss": 3.5414128073818567,
      "tokens_seen": 1515257856
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00054641925777332,
      "loss": 1.5494,
      "theoretical_loss": 3.54139943749085,
      "tokens_seen": 1515323392
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005463991975927783,
      "loss": 1.311,
      "theoretical_loss": 3.5413860683399605,
      "tokens_seen": 1515388928
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005463791374122368,
      "loss": 1.4917,
      "theoretical_loss": 3.5413726999291155,
      "tokens_seen": 1515454464
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005463590772316951,
      "loss": 1.4766,
      "theoretical_loss": 3.5413593322582413,
      "tokens_seen": 1515520000
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005463390170511535,
      "loss": 1.4485,
      "theoretical_loss": 3.5413459653272654,
      "tokens_seen": 1515585536
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005463189568706119,
      "loss": 1.4963,
      "theoretical_loss": 3.5413325991361146,
      "tokens_seen": 1515651072
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005462988966900703,
      "loss": 1.5608,
      "theoretical_loss": 3.5413192336847166,
      "tokens_seen": 1515716608
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005462788365095286,
      "loss": 1.4511,
      "theoretical_loss": 3.5413058689729975,
      "tokens_seen": 1515782144
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000546258776328987,
      "loss": 1.3564,
      "theoretical_loss": 3.541292505000885,
      "tokens_seen": 1515847680
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005462387161484453,
      "loss": 1.3842,
      "theoretical_loss": 3.541279141768306,
      "tokens_seen": 1515913216
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005462186559679038,
      "loss": 1.457,
      "theoretical_loss": 3.5412657792751876,
      "tokens_seen": 1515978752
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005461985957873621,
      "loss": 1.5459,
      "theoretical_loss": 3.5412524175214566,
      "tokens_seen": 1516044288
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005461785356068205,
      "loss": 1.4355,
      "theoretical_loss": 3.541239056507041,
      "tokens_seen": 1516109824
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005461584754262788,
      "loss": 1.3758,
      "theoretical_loss": 3.5412256962318676,
      "tokens_seen": 1516175360
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005461384152457373,
      "loss": 1.4162,
      "theoretical_loss": 3.541212336695863,
      "tokens_seen": 1516240896
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005461183550651956,
      "loss": 1.4223,
      "theoretical_loss": 3.5411989778989548,
      "tokens_seen": 1516306432
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000546098294884654,
      "loss": 1.4719,
      "theoretical_loss": 3.54118561984107,
      "tokens_seen": 1516371968
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005460782347041123,
      "loss": 1.4783,
      "theoretical_loss": 3.541172262522136,
      "tokens_seen": 1516437504
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005460581745235707,
      "loss": 1.4426,
      "theoretical_loss": 3.54115890594208,
      "tokens_seen": 1516503040
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005460381143430291,
      "loss": 1.3767,
      "theoretical_loss": 3.541145550100829,
      "tokens_seen": 1516568576
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005460180541624875,
      "loss": 1.5526,
      "theoretical_loss": 3.5411321949983106,
      "tokens_seen": 1516634112
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005459979939819458,
      "loss": 1.2704,
      "theoretical_loss": 3.5411188406344514,
      "tokens_seen": 1516699648
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005459779338014043,
      "loss": 1.3923,
      "theoretical_loss": 3.5411054870091787,
      "tokens_seen": 1516765184
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005459578736208626,
      "loss": 1.3918,
      "theoretical_loss": 3.5410921341224206,
      "tokens_seen": 1516830720
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000545937813440321,
      "loss": 1.5104,
      "theoretical_loss": 3.5410787819741034,
      "tokens_seen": 1516896256
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005459177532597794,
      "loss": 1.5055,
      "theoretical_loss": 3.5410654305641547,
      "tokens_seen": 1516961792
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005458976930792377,
      "loss": 1.3646,
      "theoretical_loss": 3.541052079892502,
      "tokens_seen": 1517027328
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005458776328986961,
      "loss": 1.4181,
      "theoretical_loss": 3.541038729959072,
      "tokens_seen": 1517092864
    },
    {
      "epoch": 0.46,
      "objective/train/advantage_avg": 0.49417856335639954,
      "objective/train/docs_used": 554075,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4159001111984253,
      "objective/train/original_loss": 1.4158998727798462,
      "objective/train/theoretical_loss": 3.5410253807637924,
      "objective/train/tokens_used": 1537618400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24446405470371246,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506607294082642,
      "objective/train/weighted_lm_loss": 1.48745596408844,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5410253807637924,
      "tokens_seen": 1517158400
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005458575727181545,
      "loss": 1.4776,
      "theoretical_loss": 3.5410253807637924,
      "tokens_seen": 1517158400
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005458375125376128,
      "loss": 1.4746,
      "theoretical_loss": 3.5410120323065906,
      "tokens_seen": 1517223936
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005458174523570713,
      "loss": 1.4857,
      "theoretical_loss": 3.540998684587394,
      "tokens_seen": 1517289472
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005457973921765295,
      "loss": 1.4837,
      "theoretical_loss": 3.540985337606129,
      "tokens_seen": 1517355008
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000545777331995988,
      "loss": 1.4877,
      "theoretical_loss": 3.5409719913627242,
      "tokens_seen": 1517420544
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005457572718154464,
      "loss": 1.4348,
      "theoretical_loss": 3.5409586458571063,
      "tokens_seen": 1517486080
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005457372116349047,
      "loss": 1.4656,
      "theoretical_loss": 3.5409453010892022,
      "tokens_seen": 1517551616
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005457171514543631,
      "loss": 1.5176,
      "theoretical_loss": 3.54093195705894,
      "tokens_seen": 1517617152
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005456970912738216,
      "loss": 1.4412,
      "theoretical_loss": 3.540918613766247,
      "tokens_seen": 1517682688
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005456770310932798,
      "loss": 1.3871,
      "theoretical_loss": 3.540905271211051,
      "tokens_seen": 1517748224
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005456569709127383,
      "loss": 1.3808,
      "theoretical_loss": 3.540891929393278,
      "tokens_seen": 1517813760
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005456369107321965,
      "loss": 1.5387,
      "theoretical_loss": 3.540878588312857,
      "tokens_seen": 1517879296
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000545616850551655,
      "loss": 1.3994,
      "theoretical_loss": 3.5408652479697142,
      "tokens_seen": 1517944832
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005455967903711134,
      "loss": 1.4462,
      "theoretical_loss": 3.540851908363777,
      "tokens_seen": 1518010368
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005455767301905717,
      "loss": 1.4363,
      "theoretical_loss": 3.540838569494974,
      "tokens_seen": 1518075904
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005455566700100301,
      "loss": 1.496,
      "theoretical_loss": 3.540825231363232,
      "tokens_seen": 1518141440
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005455366098294885,
      "loss": 1.3812,
      "theoretical_loss": 3.5408118939684785,
      "tokens_seen": 1518206976
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005455165496489468,
      "loss": 1.3659,
      "theoretical_loss": 3.5407985573106404,
      "tokens_seen": 1518272512
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005454964894684053,
      "loss": 1.3476,
      "theoretical_loss": 3.540785221389646,
      "tokens_seen": 1518338048
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005454764292878635,
      "loss": 1.5566,
      "theoretical_loss": 3.5407718862054223,
      "tokens_seen": 1518403584
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000545456369107322,
      "loss": 1.4573,
      "theoretical_loss": 3.540758551757897,
      "tokens_seen": 1518469120
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005454363089267803,
      "loss": 1.4861,
      "theoretical_loss": 3.5407452180469976,
      "tokens_seen": 1518534656
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005454162487462387,
      "loss": 1.4455,
      "theoretical_loss": 3.5407318850726517,
      "tokens_seen": 1518600192
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005453961885656971,
      "loss": 1.3803,
      "theoretical_loss": 3.5407185528347864,
      "tokens_seen": 1518665728
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005453761283851555,
      "loss": 1.4737,
      "theoretical_loss": 3.54070522133333,
      "tokens_seen": 1518731264
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005453560682046138,
      "loss": 1.5636,
      "theoretical_loss": 3.540691890568209,
      "tokens_seen": 1518796800
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005453360080240723,
      "loss": 1.4815,
      "theoretical_loss": 3.5406785605393525,
      "tokens_seen": 1518862336
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005453159478435305,
      "loss": 1.4939,
      "theoretical_loss": 3.5406652312466864,
      "tokens_seen": 1518927872
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000545295887662989,
      "loss": 1.4301,
      "theoretical_loss": 3.540651902690139,
      "tokens_seen": 1518993408
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005452758274824473,
      "loss": 1.5419,
      "theoretical_loss": 3.540638574869638,
      "tokens_seen": 1519058944
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005452557673019057,
      "loss": 1.4309,
      "theoretical_loss": 3.540625247785111,
      "tokens_seen": 1519124480
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005452357071213641,
      "loss": 1.4061,
      "theoretical_loss": 3.5406119214364855,
      "tokens_seen": 1519190016
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005452156469408225,
      "loss": 1.5019,
      "theoretical_loss": 3.5405985958236896,
      "tokens_seen": 1519255552
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005451955867602808,
      "loss": 1.4787,
      "theoretical_loss": 3.54058527094665,
      "tokens_seen": 1519321088
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005451755265797392,
      "loss": 1.4304,
      "theoretical_loss": 3.5405719468052945,
      "tokens_seen": 1519386624
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005451554663991976,
      "loss": 1.4283,
      "theoretical_loss": 3.540558623399552,
      "tokens_seen": 1519452160
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000545135406218656,
      "loss": 1.3157,
      "theoretical_loss": 3.5405453007293484,
      "tokens_seen": 1519517696
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005451153460381143,
      "loss": 1.4088,
      "theoretical_loss": 3.540531978794612,
      "tokens_seen": 1519583232
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005450952858575727,
      "loss": 1.4883,
      "theoretical_loss": 3.5405186575952716,
      "tokens_seen": 1519648768
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000545075225677031,
      "loss": 1.4587,
      "theoretical_loss": 3.5405053371312536,
      "tokens_seen": 1519714304
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005450551654964895,
      "loss": 1.515,
      "theoretical_loss": 3.540492017402486,
      "tokens_seen": 1519779840
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005450351053159479,
      "loss": 1.3441,
      "theoretical_loss": 3.5404786984088963,
      "tokens_seen": 1519845376
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005450150451354062,
      "loss": 1.4523,
      "theoretical_loss": 3.5404653801504127,
      "tokens_seen": 1519910912
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005449949849548647,
      "loss": 1.5443,
      "theoretical_loss": 3.5404520626269633,
      "tokens_seen": 1519976448
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000544974924774323,
      "loss": 1.4316,
      "theoretical_loss": 3.5404387458384745,
      "tokens_seen": 1520041984
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005449548645937814,
      "loss": 1.4567,
      "theoretical_loss": 3.540425429784875,
      "tokens_seen": 1520107520
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005449348044132398,
      "loss": 1.4666,
      "theoretical_loss": 3.540412114466093,
      "tokens_seen": 1520173056
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005449147442326981,
      "loss": 1.2261,
      "theoretical_loss": 3.5403987998820554,
      "tokens_seen": 1520238592
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005448946840521565,
      "loss": 1.4697,
      "theoretical_loss": 3.54038548603269,
      "tokens_seen": 1520304128
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000544874623871615,
      "loss": 1.3879,
      "theoretical_loss": 3.540372172917925,
      "tokens_seen": 1520369664
    },
    {
      "epoch": 0.46,
      "objective/train/advantage_avg": 0.49111783504486084,
      "objective/train/docs_used": 555119,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4298865795135498,
      "objective/train/original_loss": 1.4298866987228394,
      "objective/train/theoretical_loss": 3.5403588605376886,
      "objective/train/tokens_used": 1540895200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24183444678783417,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503411293029785,
      "objective/train/weighted_lm_loss": 1.5020841360092163,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5403588605376886,
      "tokens_seen": 1520435200
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005448545636910732,
      "loss": 1.5383,
      "theoretical_loss": 3.5403588605376886,
      "tokens_seen": 1520435200
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005448345035105317,
      "loss": 1.4334,
      "theoretical_loss": 3.5403455488919073,
      "tokens_seen": 1520500736
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005448144433299899,
      "loss": 1.4417,
      "theoretical_loss": 3.54033223798051,
      "tokens_seen": 1520566272
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005447943831494484,
      "loss": 1.3862,
      "theoretical_loss": 3.5403189278034244,
      "tokens_seen": 1520631808
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005447743229689068,
      "loss": 1.3734,
      "theoretical_loss": 3.540305618360578,
      "tokens_seen": 1520697344
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005447542627883651,
      "loss": 1.5262,
      "theoretical_loss": 3.540292309651899,
      "tokens_seen": 1520762880
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005447342026078235,
      "loss": 1.5834,
      "theoretical_loss": 3.540279001677315,
      "tokens_seen": 1520828416
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000544714142427282,
      "loss": 1.5032,
      "theoretical_loss": 3.540265694436754,
      "tokens_seen": 1520893952
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005446940822467402,
      "loss": 1.4384,
      "theoretical_loss": 3.540252387930144,
      "tokens_seen": 1520959488
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005446740220661987,
      "loss": 1.4703,
      "theoretical_loss": 3.540239082157413,
      "tokens_seen": 1521025024
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005446539618856569,
      "loss": 1.4062,
      "theoretical_loss": 3.540225777118488,
      "tokens_seen": 1521090560
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005446339017051154,
      "loss": 1.3611,
      "theoretical_loss": 3.540212472813298,
      "tokens_seen": 1521156096
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005446138415245738,
      "loss": 1.4407,
      "theoretical_loss": 3.540199169241771,
      "tokens_seen": 1521221632
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005445937813440321,
      "loss": 1.4904,
      "theoretical_loss": 3.540185866403834,
      "tokens_seen": 1521287168
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005445737211634905,
      "loss": 1.4684,
      "theoretical_loss": 3.5401725642994153,
      "tokens_seen": 1521352704
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005445536609829489,
      "loss": 1.5394,
      "theoretical_loss": 3.540159262928443,
      "tokens_seen": 1521418240
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005445336008024072,
      "loss": 1.3676,
      "theoretical_loss": 3.540145962290845,
      "tokens_seen": 1521483776
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005445135406218657,
      "loss": 1.4075,
      "theoretical_loss": 3.54013266238655,
      "tokens_seen": 1521549312
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005444934804413239,
      "loss": 1.5208,
      "theoretical_loss": 3.5401193632154846,
      "tokens_seen": 1521614848
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005444734202607824,
      "loss": 1.3753,
      "theoretical_loss": 3.5401060647775777,
      "tokens_seen": 1521680384
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005444533600802407,
      "loss": 1.4272,
      "theoretical_loss": 3.5400927670727573,
      "tokens_seen": 1521745920
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005444332998996991,
      "loss": 1.3433,
      "theoretical_loss": 3.540079470100951,
      "tokens_seen": 1521811456
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005444132397191575,
      "loss": 1.4458,
      "theoretical_loss": 3.540066173862087,
      "tokens_seen": 1521876992
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005443931795386159,
      "loss": 1.4502,
      "theoretical_loss": 3.540052878356094,
      "tokens_seen": 1521942528
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005443731193580742,
      "loss": 1.6067,
      "theoretical_loss": 3.5400395835828986,
      "tokens_seen": 1522008064
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005443530591775327,
      "loss": 1.5563,
      "theoretical_loss": 3.5400262895424306,
      "tokens_seen": 1522073600
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005443329989969909,
      "loss": 1.4889,
      "theoretical_loss": 3.5400129962346165,
      "tokens_seen": 1522139136
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005443129388164494,
      "loss": 1.4269,
      "theoretical_loss": 3.5399997036593853,
      "tokens_seen": 1522204672
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005442928786359077,
      "loss": 1.5112,
      "theoretical_loss": 3.539986411816665,
      "tokens_seen": 1522270208
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005442728184553661,
      "loss": 1.5608,
      "theoretical_loss": 3.5399731207063834,
      "tokens_seen": 1522335744
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005442527582748245,
      "loss": 1.5052,
      "theoretical_loss": 3.5399598303284687,
      "tokens_seen": 1522401280
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005442326980942829,
      "loss": 1.4817,
      "theoretical_loss": 3.539946540682849,
      "tokens_seen": 1522466816
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005442126379137412,
      "loss": 1.4314,
      "theoretical_loss": 3.5399332517694533,
      "tokens_seen": 1522532352
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005441925777331996,
      "loss": 1.5594,
      "theoretical_loss": 3.5399199635882086,
      "tokens_seen": 1522597888
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000544172517552658,
      "loss": 1.3201,
      "theoretical_loss": 3.5399066761390428,
      "tokens_seen": 1522663424
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005441524573721164,
      "loss": 1.2537,
      "theoretical_loss": 3.539893389421885,
      "tokens_seen": 1522728960
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005441323971915747,
      "loss": 1.5222,
      "theoretical_loss": 3.539880103436664,
      "tokens_seen": 1522794496
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005441123370110331,
      "loss": 1.4426,
      "theoretical_loss": 3.539866818183306,
      "tokens_seen": 1522860032
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005440922768304914,
      "loss": 1.4077,
      "theoretical_loss": 3.5398535336617405,
      "tokens_seen": 1522925568
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005440722166499499,
      "loss": 1.4652,
      "theoretical_loss": 3.539840249871895,
      "tokens_seen": 1522991104
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005440521564694082,
      "loss": 1.4306,
      "theoretical_loss": 3.5398269668136986,
      "tokens_seen": 1523056640
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005440320962888666,
      "loss": 1.3103,
      "theoretical_loss": 3.539813684487079,
      "tokens_seen": 1523122176
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000544012036108325,
      "loss": 1.4271,
      "theoretical_loss": 3.5398004028919647,
      "tokens_seen": 1523187712
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005439919759277834,
      "loss": 1.4007,
      "theoretical_loss": 3.539787122028283,
      "tokens_seen": 1523253248
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005439719157472417,
      "loss": 1.4436,
      "theoretical_loss": 3.539773841895964,
      "tokens_seen": 1523318784
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005439518555667002,
      "loss": 1.3953,
      "theoretical_loss": 3.5397605624949335,
      "tokens_seen": 1523384320
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005439317953861584,
      "loss": 1.319,
      "theoretical_loss": 3.5397472838251223,
      "tokens_seen": 1523449856
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005439117352056169,
      "loss": 1.5702,
      "theoretical_loss": 3.539734005886457,
      "tokens_seen": 1523515392
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005438916750250752,
      "loss": 1.5546,
      "theoretical_loss": 3.5397207286788666,
      "tokens_seen": 1523580928
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005438716148445336,
      "loss": 1.5009,
      "theoretical_loss": 3.539707452202279,
      "tokens_seen": 1523646464
    },
    {
      "epoch": 0.46,
      "objective/train/advantage_avg": 0.4947527348995209,
      "objective/train/docs_used": 556367,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4194269180297852,
      "objective/train/original_loss": 1.4194267988204956,
      "objective/train/theoretical_loss": 3.5396941764566225,
      "objective/train/tokens_used": 1544172000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24508036673069,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050721287727356,
      "objective/train/weighted_lm_loss": 1.491448163986206,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5396941764566225,
      "tokens_seen": 1523712000
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000543851554663992,
      "loss": 1.5684,
      "theoretical_loss": 3.5396941764566225,
      "tokens_seen": 1523712000
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005438314944834503,
      "loss": 1.5718,
      "theoretical_loss": 3.5396809014418262,
      "tokens_seen": 1523777536
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005438114343029087,
      "loss": 1.4281,
      "theoretical_loss": 3.539667627157818,
      "tokens_seen": 1523843072
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005437913741223672,
      "loss": 1.4187,
      "theoretical_loss": 3.5396543536045253,
      "tokens_seen": 1523908608
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005437713139418254,
      "loss": 1.5173,
      "theoretical_loss": 3.5396410807818777,
      "tokens_seen": 1523974144
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005437512537612839,
      "loss": 1.4471,
      "theoretical_loss": 3.5396278086898034,
      "tokens_seen": 1524039680
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005437311935807421,
      "loss": 1.4209,
      "theoretical_loss": 3.53961453732823,
      "tokens_seen": 1524105216
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005437111334002006,
      "loss": 1.4604,
      "theoretical_loss": 3.539601266697087,
      "tokens_seen": 1524170752
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000543691073219659,
      "loss": 1.5211,
      "theoretical_loss": 3.5395879967963015,
      "tokens_seen": 1524236288
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005436710130391173,
      "loss": 1.3443,
      "theoretical_loss": 3.539574727625803,
      "tokens_seen": 1524301824
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005436509528585757,
      "loss": 1.4246,
      "theoretical_loss": 3.53956145918552,
      "tokens_seen": 1524367360
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005436308926780342,
      "loss": 1.357,
      "theoretical_loss": 3.5395481914753795,
      "tokens_seen": 1524432896
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005436108324974924,
      "loss": 1.4909,
      "theoretical_loss": 3.539534924495311,
      "tokens_seen": 1524498432
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005435907723169509,
      "loss": 1.4751,
      "theoretical_loss": 3.539521658245244,
      "tokens_seen": 1524563968
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005435707121364091,
      "loss": 1.44,
      "theoretical_loss": 3.5395083927251045,
      "tokens_seen": 1524629504
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005435506519558676,
      "loss": 1.4454,
      "theoretical_loss": 3.5394951279348223,
      "tokens_seen": 1524695040
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005435305917753261,
      "loss": 1.5829,
      "theoretical_loss": 3.539481863874326,
      "tokens_seen": 1524760576
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005435105315947843,
      "loss": 1.4813,
      "theoretical_loss": 3.539468600543544,
      "tokens_seen": 1524826112
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005434904714142428,
      "loss": 1.5801,
      "theoretical_loss": 3.539455337942405,
      "tokens_seen": 1524891648
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005434704112337011,
      "loss": 1.4584,
      "theoretical_loss": 3.539442076070837,
      "tokens_seen": 1524957184
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005434503510531595,
      "loss": 1.4227,
      "theoretical_loss": 3.539428814928768,
      "tokens_seen": 1525022720
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005434302908726179,
      "loss": 1.398,
      "theoretical_loss": 3.539415554516128,
      "tokens_seen": 1525088256
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005434102306920763,
      "loss": 1.327,
      "theoretical_loss": 3.5394022948328447,
      "tokens_seen": 1525153792
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005433901705115346,
      "loss": 1.3336,
      "theoretical_loss": 3.5393890358788465,
      "tokens_seen": 1525219328
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005433701103309931,
      "loss": 1.4432,
      "theoretical_loss": 3.5393757776540626,
      "tokens_seen": 1525284864
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005433500501504513,
      "loss": 1.5589,
      "theoretical_loss": 3.5393625201584205,
      "tokens_seen": 1525350400
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005433299899699098,
      "loss": 1.3528,
      "theoretical_loss": 3.5393492633918497,
      "tokens_seen": 1525415936
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005433099297893681,
      "loss": 1.4972,
      "theoretical_loss": 3.5393360073542786,
      "tokens_seen": 1525481472
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005432898696088265,
      "loss": 1.4996,
      "theoretical_loss": 3.539322752045636,
      "tokens_seen": 1525547008
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005432698094282849,
      "loss": 1.4031,
      "theoretical_loss": 3.539309497465849,
      "tokens_seen": 1525612544
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005432497492477433,
      "loss": 1.4079,
      "theoretical_loss": 3.5392962436148485,
      "tokens_seen": 1525678080
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005432296890672016,
      "loss": 1.3637,
      "theoretical_loss": 3.539282990492562,
      "tokens_seen": 1525743616
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00054320962888666,
      "loss": 1.5203,
      "theoretical_loss": 3.539269738098918,
      "tokens_seen": 1525809152
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005431895687061184,
      "loss": 1.4193,
      "theoretical_loss": 3.539256486433845,
      "tokens_seen": 1525874688
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005431695085255768,
      "loss": 1.3336,
      "theoretical_loss": 3.5392432354972723,
      "tokens_seen": 1525940224
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005431494483450351,
      "loss": 1.488,
      "theoretical_loss": 3.5392299852891282,
      "tokens_seen": 1526005760
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005431293881644935,
      "loss": 1.4806,
      "theoretical_loss": 3.539216735809341,
      "tokens_seen": 1526071296
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005431093279839518,
      "loss": 1.5256,
      "theoretical_loss": 3.539203487057841,
      "tokens_seen": 1526136832
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005430892678034103,
      "loss": 1.4549,
      "theoretical_loss": 3.5391902390345544,
      "tokens_seen": 1526202368
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005430692076228686,
      "loss": 1.4984,
      "theoretical_loss": 3.539176991739412,
      "tokens_seen": 1526267904
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000543049147442327,
      "loss": 1.4554,
      "theoretical_loss": 3.5391637451723414,
      "tokens_seen": 1526333440
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005430290872617854,
      "loss": 1.5025,
      "theoretical_loss": 3.539150499333272,
      "tokens_seen": 1526398976
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005430090270812438,
      "loss": 1.3021,
      "theoretical_loss": 3.5391372542221315,
      "tokens_seen": 1526464512
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005429889669007021,
      "loss": 1.423,
      "theoretical_loss": 3.53912400983885,
      "tokens_seen": 1526530048
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005429689067201606,
      "loss": 1.5396,
      "theoretical_loss": 3.5391107661833554,
      "tokens_seen": 1526595584
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005429488465396188,
      "loss": 1.403,
      "theoretical_loss": 3.5390975232555766,
      "tokens_seen": 1526661120
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005429287863590773,
      "loss": 1.3582,
      "theoretical_loss": 3.539084281055443,
      "tokens_seen": 1526726656
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005429087261785356,
      "loss": 1.4929,
      "theoretical_loss": 3.5390710395828817,
      "tokens_seen": 1526792192
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000542888665997994,
      "loss": 1.5032,
      "theoretical_loss": 3.539057798837823,
      "tokens_seen": 1526857728
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005428686058174524,
      "loss": 1.4503,
      "theoretical_loss": 3.539044558820196,
      "tokens_seen": 1526923264
    },
    {
      "epoch": 0.46,
      "objective/train/advantage_avg": 0.4957258403301239,
      "objective/train/docs_used": 557485,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.308136224746704,
      "objective/train/original_loss": 1.308136224746704,
      "objective/train/theoretical_loss": 3.5390313195299283,
      "objective/train/tokens_used": 1547448800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24595850706100464,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508229732513428,
      "objective/train/weighted_lm_loss": 1.3745388984680176,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5390313195299283,
      "tokens_seen": 1526988800
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005428485456369107,
      "loss": 1.4367,
      "theoretical_loss": 3.5390313195299283,
      "tokens_seen": 1526988800
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005428284854563691,
      "loss": 1.4433,
      "theoretical_loss": 3.539018080966949,
      "tokens_seen": 1527054336
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005428084252758276,
      "loss": 1.3974,
      "theoretical_loss": 3.5390048431311882,
      "tokens_seen": 1527119872
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005427883650952858,
      "loss": 1.3162,
      "theoretical_loss": 3.538991606022573,
      "tokens_seen": 1527185408
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005427683049147443,
      "loss": 1.3812,
      "theoretical_loss": 3.538978369641033,
      "tokens_seen": 1527250944
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005427482447342025,
      "loss": 1.4136,
      "theoretical_loss": 3.5389651339864976,
      "tokens_seen": 1527316480
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000542728184553661,
      "loss": 1.4595,
      "theoretical_loss": 3.5389518990588944,
      "tokens_seen": 1527382016
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005427081243731194,
      "loss": 1.4301,
      "theoretical_loss": 3.5389386648581533,
      "tokens_seen": 1527447552
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005426880641925777,
      "loss": 1.4576,
      "theoretical_loss": 3.538925431384203,
      "tokens_seen": 1527513088
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005426680040120361,
      "loss": 1.4234,
      "theoretical_loss": 3.5389121986369725,
      "tokens_seen": 1527578624
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005426479438314946,
      "loss": 1.4917,
      "theoretical_loss": 3.538898966616391,
      "tokens_seen": 1527644160
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005426278836509528,
      "loss": 1.5007,
      "theoretical_loss": 3.538885735322386,
      "tokens_seen": 1527709696
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005426078234704113,
      "loss": 1.3784,
      "theoretical_loss": 3.538872504754888,
      "tokens_seen": 1527775232
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005425877632898695,
      "loss": 1.5697,
      "theoretical_loss": 3.5388592749138255,
      "tokens_seen": 1527840768
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000542567703109328,
      "loss": 1.3651,
      "theoretical_loss": 3.5388460457991266,
      "tokens_seen": 1527906304
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005425476429287864,
      "loss": 1.3988,
      "theoretical_loss": 3.5388328174107215,
      "tokens_seen": 1527971840
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005425275827482447,
      "loss": 1.4316,
      "theoretical_loss": 3.538819589748539,
      "tokens_seen": 1528037376
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005425075225677031,
      "loss": 1.43,
      "theoretical_loss": 3.5388063628125073,
      "tokens_seen": 1528102912
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005424874623871615,
      "loss": 1.366,
      "theoretical_loss": 3.538793136602556,
      "tokens_seen": 1528168448
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005424674022066198,
      "loss": 1.5567,
      "theoretical_loss": 3.5387799111186142,
      "tokens_seen": 1528233984
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005424473420260783,
      "loss": 1.3338,
      "theoretical_loss": 3.5387666863606104,
      "tokens_seen": 1528299520
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005424272818455365,
      "loss": 1.4959,
      "theoretical_loss": 3.538753462328474,
      "tokens_seen": 1528365056
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000542407221664995,
      "loss": 1.5378,
      "theoretical_loss": 3.538740239022134,
      "tokens_seen": 1528430592
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005423871614844534,
      "loss": 1.4644,
      "theoretical_loss": 3.538727016441519,
      "tokens_seen": 1528496128
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005423671013039117,
      "loss": 1.4407,
      "theoretical_loss": 3.5387137945865588,
      "tokens_seen": 1528561664
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005423470411233701,
      "loss": 1.4535,
      "theoretical_loss": 3.5387005734571826,
      "tokens_seen": 1528627200
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005423269809428285,
      "loss": 1.4972,
      "theoretical_loss": 3.538687353053318,
      "tokens_seen": 1528692736
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005423069207622868,
      "loss": 1.4538,
      "theoretical_loss": 3.5386741333748954,
      "tokens_seen": 1528758272
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005422868605817453,
      "loss": 1.387,
      "theoretical_loss": 3.538660914421844,
      "tokens_seen": 1528823808
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005422668004012036,
      "loss": 1.315,
      "theoretical_loss": 3.5386476961940923,
      "tokens_seen": 1528889344
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000542246740220662,
      "loss": 1.483,
      "theoretical_loss": 3.5386344786915696,
      "tokens_seen": 1528954880
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005422266800401203,
      "loss": 1.655,
      "theoretical_loss": 3.538621261914205,
      "tokens_seen": 1529020416
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005422066198595788,
      "loss": 1.4971,
      "theoretical_loss": 3.5386080458619276,
      "tokens_seen": 1529085952
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005421865596790371,
      "loss": 1.4441,
      "theoretical_loss": 3.5385948305346666,
      "tokens_seen": 1529151488
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005421664994984955,
      "loss": 1.393,
      "theoretical_loss": 3.538581615932351,
      "tokens_seen": 1529217024
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005421464393179538,
      "loss": 1.4632,
      "theoretical_loss": 3.5385684020549104,
      "tokens_seen": 1529282560
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005421263791374122,
      "loss": 1.5588,
      "theoretical_loss": 3.538555188902274,
      "tokens_seen": 1529348096
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005421063189568706,
      "loss": 1.4805,
      "theoretical_loss": 3.53854197647437,
      "tokens_seen": 1529413632
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000542086258776329,
      "loss": 1.5141,
      "theoretical_loss": 3.5385287647711285,
      "tokens_seen": 1529479168
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005420661985957873,
      "loss": 1.386,
      "theoretical_loss": 3.538515553792479,
      "tokens_seen": 1529544704
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005420461384152458,
      "loss": 1.4056,
      "theoretical_loss": 3.53850234353835,
      "tokens_seen": 1529610240
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005420260782347042,
      "loss": 1.4017,
      "theoretical_loss": 3.5384891340086706,
      "tokens_seen": 1529675776
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005420060180541625,
      "loss": 1.4343,
      "theoretical_loss": 3.538475925203371,
      "tokens_seen": 1529741312
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000541985957873621,
      "loss": 1.5469,
      "theoretical_loss": 3.5384627171223793,
      "tokens_seen": 1529806848
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005419658976930792,
      "loss": 1.3105,
      "theoretical_loss": 3.5384495097656252,
      "tokens_seen": 1529872384
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005419458375125377,
      "loss": 1.5379,
      "theoretical_loss": 3.5384363031330377,
      "tokens_seen": 1529937920
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000541925777331996,
      "loss": 1.5114,
      "theoretical_loss": 3.5384230972245474,
      "tokens_seen": 1530003456
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005419057171514544,
      "loss": 1.4949,
      "theoretical_loss": 3.538409892040082,
      "tokens_seen": 1530068992
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005418856569709128,
      "loss": 1.4345,
      "theoretical_loss": 3.5383966875795716,
      "tokens_seen": 1530134528
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005418655967903711,
      "loss": 1.4386,
      "theoretical_loss": 3.538383483842945,
      "tokens_seen": 1530200064
    },
    {
      "epoch": 0.46,
      "objective/train/advantage_avg": 0.4933151304721832,
      "objective/train/docs_used": 558616,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2905361652374268,
      "objective/train/original_loss": 1.2905364036560059,
      "objective/train/theoretical_loss": 3.5383702808301325,
      "objective/train/tokens_used": 1550725600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24393880367279053,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505715608596802,
      "objective/train/weighted_lm_loss": 1.3558801412582397,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.8607079386711121,
      "theoretical_loss": 3.5383702808301325,
      "tokens_seen": 1530265600
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005418455366098295,
      "loss": 1.4747,
      "theoretical_loss": 3.5383702808301325,
      "tokens_seen": 1530265600
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000541825476429288,
      "loss": 1.5005,
      "theoretical_loss": 3.538357078541062,
      "tokens_seen": 1530331136
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005418054162487462,
      "loss": 1.5234,
      "theoretical_loss": 3.538343876975664,
      "tokens_seen": 1530396672
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005417853560682047,
      "loss": 1.3748,
      "theoretical_loss": 3.5383306761338673,
      "tokens_seen": 1530462208
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005417652958876629,
      "loss": 1.4999,
      "theoretical_loss": 3.5383174760156013,
      "tokens_seen": 1530527744
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005417452357071214,
      "loss": 1.529,
      "theoretical_loss": 3.5383042766207957,
      "tokens_seen": 1530593280
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005417251755265798,
      "loss": 1.4947,
      "theoretical_loss": 3.5382910779493795,
      "tokens_seen": 1530658816
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005417051153460381,
      "loss": 1.4927,
      "theoretical_loss": 3.5382778800012824,
      "tokens_seen": 1530724352
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005416850551654965,
      "loss": 1.322,
      "theoretical_loss": 3.5382646827764335,
      "tokens_seen": 1530789888
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000541664994984955,
      "loss": 1.4955,
      "theoretical_loss": 3.538251486274762,
      "tokens_seen": 1530855424
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005416449348044132,
      "loss": 1.4838,
      "theoretical_loss": 3.538238290496198,
      "tokens_seen": 1530920960
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005416248746238717,
      "loss": 1.4512,
      "theoretical_loss": 3.5382250954406707,
      "tokens_seen": 1530986496
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005416048144433299,
      "loss": 1.375,
      "theoretical_loss": 3.538211901108109,
      "tokens_seen": 1531052032
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005415847542627884,
      "loss": 1.3998,
      "theoretical_loss": 3.5381987074984425,
      "tokens_seen": 1531117568
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005415646940822468,
      "loss": 1.4534,
      "theoretical_loss": 3.5381855146116017,
      "tokens_seen": 1531183104
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005415446339017051,
      "loss": 1.3732,
      "theoretical_loss": 3.5381723224475143,
      "tokens_seen": 1531248640
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005415245737211635,
      "loss": 1.5325,
      "theoretical_loss": 3.538159131006111,
      "tokens_seen": 1531314176
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005415045135406219,
      "loss": 1.5345,
      "theoretical_loss": 3.538145940287322,
      "tokens_seen": 1531379712
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005414844533600802,
      "loss": 1.4867,
      "theoretical_loss": 3.5381327502910747,
      "tokens_seen": 1531445248
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005414643931795387,
      "loss": 1.3536,
      "theoretical_loss": 3.5381195610173,
      "tokens_seen": 1531510784
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000541444332998997,
      "loss": 1.488,
      "theoretical_loss": 3.538106372465927,
      "tokens_seen": 1531576320
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005414242728184554,
      "loss": 1.615,
      "theoretical_loss": 3.5380931846368853,
      "tokens_seen": 1531641856
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005414042126379138,
      "loss": 1.3997,
      "theoretical_loss": 3.538079997530105,
      "tokens_seen": 1531707392
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005413841524573721,
      "loss": 1.4851,
      "theoretical_loss": 3.5380668111455145,
      "tokens_seen": 1531772928
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005413640922768305,
      "loss": 1.473,
      "theoretical_loss": 3.5380536254830446,
      "tokens_seen": 1531838464
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005413440320962889,
      "loss": 1.4955,
      "theoretical_loss": 3.5380404405426233,
      "tokens_seen": 1531904000
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005413239719157472,
      "loss": 1.4916,
      "theoretical_loss": 3.538027256324182,
      "tokens_seen": 1531969536
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005413039117352057,
      "loss": 1.405,
      "theoretical_loss": 3.5380140728276492,
      "tokens_seen": 1532035072
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000541283851554664,
      "loss": 1.5205,
      "theoretical_loss": 3.5380008900529543,
      "tokens_seen": 1532100608
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005412637913741224,
      "loss": 1.4666,
      "theoretical_loss": 3.5379877080000277,
      "tokens_seen": 1532166144
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005412437311935807,
      "loss": 1.369,
      "theoretical_loss": 3.5379745266687985,
      "tokens_seen": 1532231680
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005412236710130392,
      "loss": 1.3522,
      "theoretical_loss": 3.5379613460591965,
      "tokens_seen": 1532297216
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005412036108324975,
      "loss": 1.3827,
      "theoretical_loss": 3.5379481661711507,
      "tokens_seen": 1532362752
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005411835506519559,
      "loss": 1.4627,
      "theoretical_loss": 3.5379349870045917,
      "tokens_seen": 1532428288
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005411634904714142,
      "loss": 1.3693,
      "theoretical_loss": 3.537921808559449,
      "tokens_seen": 1532493824
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005411434302908726,
      "loss": 1.4995,
      "theoretical_loss": 3.537908630835652,
      "tokens_seen": 1532559360
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000541123370110331,
      "loss": 1.4209,
      "theoretical_loss": 3.53789545383313,
      "tokens_seen": 1532624896
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005411033099297894,
      "loss": 1.4575,
      "theoretical_loss": 3.5378822775518133,
      "tokens_seen": 1532690432
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005410832497492477,
      "loss": 1.4041,
      "theoretical_loss": 3.537869101991631,
      "tokens_seen": 1532755968
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005410631895687062,
      "loss": 1.4776,
      "theoretical_loss": 3.5378559271525134,
      "tokens_seen": 1532821504
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005410431293881645,
      "loss": 1.7002,
      "theoretical_loss": 3.53784275303439,
      "tokens_seen": 1532887040
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005410230692076229,
      "loss": 1.4663,
      "theoretical_loss": 3.5378295796371906,
      "tokens_seen": 1532952576
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005410030090270812,
      "loss": 1.3855,
      "theoretical_loss": 3.537816406960845,
      "tokens_seen": 1533018112
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005409829488465396,
      "loss": 1.4248,
      "theoretical_loss": 3.537803235005282,
      "tokens_seen": 1533083648
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000540962888665998,
      "loss": 1.4297,
      "theoretical_loss": 3.5377900637704327,
      "tokens_seen": 1533149184
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005409428284854564,
      "loss": 1.5077,
      "theoretical_loss": 3.5377768932562264,
      "tokens_seen": 1533214720
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005409227683049147,
      "loss": 1.541,
      "theoretical_loss": 3.537763723462593,
      "tokens_seen": 1533280256
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005409027081243732,
      "loss": 1.4365,
      "theoretical_loss": 3.5377505543894614,
      "tokens_seen": 1533345792
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005408826479438314,
      "loss": 1.445,
      "theoretical_loss": 3.5377373860367616,
      "tokens_seen": 1533411328
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005408625877632899,
      "loss": 1.5843,
      "theoretical_loss": 3.5377242184044246,
      "tokens_seen": 1533476864
    },
    {
      "epoch": 0.46,
      "objective/train/advantage_avg": 0.49114614725112915,
      "objective/train/docs_used": 559697,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4323192834854126,
      "objective/train/original_loss": 1.4323194026947021,
      "objective/train/theoretical_loss": 3.537711051492379,
      "objective/train/tokens_used": 1554002400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24162723124027252,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503427982330322,
      "objective/train/weighted_lm_loss": 1.5043295621871948,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.537711051492379,
      "tokens_seen": 1533542400
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005408425275827483,
      "loss": 1.6177,
      "theoretical_loss": 3.537711051492379,
      "tokens_seen": 1533542400
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005408224674022066,
      "loss": 1.5394,
      "theoretical_loss": 3.5376978853005556,
      "tokens_seen": 1533607936
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000540802407221665,
      "loss": 1.4134,
      "theoretical_loss": 3.537684719828883,
      "tokens_seen": 1533673472
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005407823470411233,
      "loss": 1.4319,
      "theoretical_loss": 3.537671555077292,
      "tokens_seen": 1533739008
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005407622868605817,
      "loss": 1.4606,
      "theoretical_loss": 3.5376583910457127,
      "tokens_seen": 1533804544
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005407422266800402,
      "loss": 1.3582,
      "theoretical_loss": 3.5376452277340737,
      "tokens_seen": 1533870080
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005407221664994984,
      "loss": 1.4655,
      "theoretical_loss": 3.537632065142306,
      "tokens_seen": 1533935616
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005407021063189569,
      "loss": 1.525,
      "theoretical_loss": 3.5376189032703387,
      "tokens_seen": 1534001152
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005406820461384153,
      "loss": 1.5341,
      "theoretical_loss": 3.537605742118102,
      "tokens_seen": 1534066688
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005406619859578736,
      "loss": 1.4858,
      "theoretical_loss": 3.5375925816855265,
      "tokens_seen": 1534132224
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000540641925777332,
      "loss": 1.506,
      "theoretical_loss": 3.5375794219725405,
      "tokens_seen": 1534197760
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005406218655967903,
      "loss": 1.5195,
      "theoretical_loss": 3.537566262979076,
      "tokens_seen": 1534263296
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005406018054162487,
      "loss": 1.4476,
      "theoretical_loss": 3.537553104705061,
      "tokens_seen": 1534328832
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005405817452357072,
      "loss": 1.662,
      "theoretical_loss": 3.5375399471504263,
      "tokens_seen": 1534394368
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0005405616850551654,
      "loss": 1.5083,
      "theoretical_loss": 3.537526790315102,
      "tokens_seen": 1534459904
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005405416248746239,
      "loss": 1.5179,
      "theoretical_loss": 3.5375136341990183,
      "tokens_seen": 1534525440
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005405215646940823,
      "loss": 1.492,
      "theoretical_loss": 3.5375004788021043,
      "tokens_seen": 1534590976
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005405015045135406,
      "loss": 1.4894,
      "theoretical_loss": 3.53748732412429,
      "tokens_seen": 1534656512
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005404814443329991,
      "loss": 1.4954,
      "theoretical_loss": 3.537474170165506,
      "tokens_seen": 1534722048
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005404613841524574,
      "loss": 1.5485,
      "theoretical_loss": 3.5374610169256826,
      "tokens_seen": 1534787584
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005404413239719158,
      "loss": 1.4603,
      "theoretical_loss": 3.537447864404749,
      "tokens_seen": 1534853120
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005404212637913742,
      "loss": 1.5432,
      "theoretical_loss": 3.537434712602635,
      "tokens_seen": 1534918656
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005404012036108325,
      "loss": 1.4169,
      "theoretical_loss": 3.5374215615192717,
      "tokens_seen": 1534984192
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005403811434302909,
      "loss": 1.4981,
      "theoretical_loss": 3.5374084111545887,
      "tokens_seen": 1535049728
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005403610832497493,
      "loss": 1.4496,
      "theoretical_loss": 3.5373952615085154,
      "tokens_seen": 1535115264
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005403410230692076,
      "loss": 1.4673,
      "theoretical_loss": 3.5373821125809823,
      "tokens_seen": 1535180800
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005403209628886661,
      "loss": 1.408,
      "theoretical_loss": 3.53736896437192,
      "tokens_seen": 1535246336
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005403009027081244,
      "loss": 1.4039,
      "theoretical_loss": 3.5373558168812576,
      "tokens_seen": 1535311872
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005402808425275828,
      "loss": 1.4728,
      "theoretical_loss": 3.5373426701089263,
      "tokens_seen": 1535377408
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005402607823470411,
      "loss": 1.5885,
      "theoretical_loss": 3.5373295240548552,
      "tokens_seen": 1535442944
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005402407221664996,
      "loss": 1.4024,
      "theoretical_loss": 3.537316378718975,
      "tokens_seen": 1535508480
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005402206619859579,
      "loss": 1.4896,
      "theoretical_loss": 3.5373032341012154,
      "tokens_seen": 1535574016
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005402006018054163,
      "loss": 1.5291,
      "theoretical_loss": 3.5372900902015063,
      "tokens_seen": 1535639552
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005401805416248746,
      "loss": 1.5874,
      "theoretical_loss": 3.537276947019779,
      "tokens_seen": 1535705088
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000540160481444333,
      "loss": 1.4856,
      "theoretical_loss": 3.5372638045559626,
      "tokens_seen": 1535770624
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005401404212637914,
      "loss": 1.4413,
      "theoretical_loss": 3.5372506628099876,
      "tokens_seen": 1535836160
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005401203610832498,
      "loss": 1.5433,
      "theoretical_loss": 3.537237521781784,
      "tokens_seen": 1535901696
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005401003009027081,
      "loss": 1.4938,
      "theoretical_loss": 3.5372243814712823,
      "tokens_seen": 1535967232
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005400802407221666,
      "loss": 1.4123,
      "theoretical_loss": 3.5372112418784125,
      "tokens_seen": 1536032768
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005400601805416249,
      "loss": 1.5094,
      "theoretical_loss": 3.5371981030031048,
      "tokens_seen": 1536098304
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005400401203610833,
      "loss": 1.5145,
      "theoretical_loss": 3.537184964845289,
      "tokens_seen": 1536163840
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005400200601805416,
      "loss": 1.6271,
      "theoretical_loss": 3.5371718274048956,
      "tokens_seen": 1536229376
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00054,
      "loss": 1.4613,
      "theoretical_loss": 3.5371586906818555,
      "tokens_seen": 1536294912
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005399799398194584,
      "loss": 1.6192,
      "theoretical_loss": 3.5371455546760977,
      "tokens_seen": 1536360448
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005399598796389168,
      "loss": 1.5283,
      "theoretical_loss": 3.5371324193875533,
      "tokens_seen": 1536425984
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005399398194583751,
      "loss": 1.4792,
      "theoretical_loss": 3.5371192848161526,
      "tokens_seen": 1536491520
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005399197592778336,
      "loss": 1.4179,
      "theoretical_loss": 3.537106150961825,
      "tokens_seen": 1536557056
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005398996990972918,
      "loss": 1.5363,
      "theoretical_loss": 3.537093017824502,
      "tokens_seen": 1536622592
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005398796389167503,
      "loss": 1.3736,
      "theoretical_loss": 3.537079885404113,
      "tokens_seen": 1536688128
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005398595787362087,
      "loss": 1.6066,
      "theoretical_loss": 3.5370667537005884,
      "tokens_seen": 1536753664
    },
    {
      "epoch": 0.47,
      "objective/train/advantage_avg": 0.4957691729068756,
      "objective/train/docs_used": 561026,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2708959579467773,
      "objective/train/original_loss": 1.2708958387374878,
      "objective/train/theoretical_loss": 3.5370536227138585,
      "objective/train/tokens_used": 1557279200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24680279195308685,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508315563201904,
      "objective/train/weighted_lm_loss": 1.3354487419128418,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.7046881318092346,
      "theoretical_loss": 3.5370536227138585,
      "tokens_seen": 1536819200
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000539839518555667,
      "loss": 1.4935,
      "theoretical_loss": 3.5370536227138585,
      "tokens_seen": 1536819200
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005398194583751254,
      "loss": 1.4184,
      "theoretical_loss": 3.537040492443854,
      "tokens_seen": 1536884736
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005397993981945837,
      "loss": 1.5192,
      "theoretical_loss": 3.5370273628905045,
      "tokens_seen": 1536950272
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005397793380140421,
      "loss": 1.4064,
      "theoretical_loss": 3.537014234053742,
      "tokens_seen": 1537015808
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005397592778335006,
      "loss": 1.4347,
      "theoretical_loss": 3.537001105933494,
      "tokens_seen": 1537081344
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005397392176529588,
      "loss": 1.5145,
      "theoretical_loss": 3.5369879785296936,
      "tokens_seen": 1537146880
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005397191574724173,
      "loss": 1.4822,
      "theoretical_loss": 3.5369748518422695,
      "tokens_seen": 1537212416
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005396990972918757,
      "loss": 1.5417,
      "theoretical_loss": 3.5369617258711528,
      "tokens_seen": 1537277952
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000539679037111334,
      "loss": 1.4547,
      "theoretical_loss": 3.536948600616274,
      "tokens_seen": 1537343488
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005396589769307924,
      "loss": 1.6276,
      "theoretical_loss": 3.5369354760775624,
      "tokens_seen": 1537409024
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005396389167502507,
      "loss": 1.4541,
      "theoretical_loss": 3.5369223522549493,
      "tokens_seen": 1537474560
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005396188565697091,
      "loss": 1.5859,
      "theoretical_loss": 3.5369092291483657,
      "tokens_seen": 1537540096
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005395987963891676,
      "loss": 1.4167,
      "theoretical_loss": 3.53689610675774,
      "tokens_seen": 1537605632
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005395787362086258,
      "loss": 1.5086,
      "theoretical_loss": 3.536882985083005,
      "tokens_seen": 1537671168
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005395586760280843,
      "loss": 1.3904,
      "theoretical_loss": 3.53686986412409,
      "tokens_seen": 1537736704
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005395386158475426,
      "loss": 1.389,
      "theoretical_loss": 3.536856743880925,
      "tokens_seen": 1537802240
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000539518555667001,
      "loss": 1.4348,
      "theoretical_loss": 3.536843624353441,
      "tokens_seen": 1537867776
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005394984954864594,
      "loss": 1.542,
      "theoretical_loss": 3.5368305055415687,
      "tokens_seen": 1537933312
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005394784353059178,
      "loss": 1.5123,
      "theoretical_loss": 3.5368173874452378,
      "tokens_seen": 1537998848
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005394583751253761,
      "loss": 1.417,
      "theoretical_loss": 3.53680427006438,
      "tokens_seen": 1538064384
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005394383149448346,
      "loss": 1.5178,
      "theoretical_loss": 3.5367911533989242,
      "tokens_seen": 1538129920
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005394182547642928,
      "loss": 1.4164,
      "theoretical_loss": 3.5367780374488023,
      "tokens_seen": 1538195456
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005393981945837513,
      "loss": 1.5572,
      "theoretical_loss": 3.536764922213944,
      "tokens_seen": 1538260992
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005393781344032096,
      "loss": 1.5066,
      "theoretical_loss": 3.53675180769428,
      "tokens_seen": 1538326528
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000539358074222668,
      "loss": 1.3166,
      "theoretical_loss": 3.5367386938897414,
      "tokens_seen": 1538392064
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005393380140421264,
      "loss": 1.5376,
      "theoretical_loss": 3.5367255808002582,
      "tokens_seen": 1538457600
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005393179538615848,
      "loss": 1.5166,
      "theoretical_loss": 3.536712468425761,
      "tokens_seen": 1538523136
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005392978936810431,
      "loss": 1.4858,
      "theoretical_loss": 3.5366993567661797,
      "tokens_seen": 1538588672
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005392778335005015,
      "loss": 1.3773,
      "theoretical_loss": 3.5366862458214463,
      "tokens_seen": 1538654208
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005392577733199598,
      "loss": 1.4724,
      "theoretical_loss": 3.5366731355914904,
      "tokens_seen": 1538719744
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005392377131394183,
      "loss": 1.4893,
      "theoretical_loss": 3.5366600260762433,
      "tokens_seen": 1538785280
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005392176529588766,
      "loss": 1.4304,
      "theoretical_loss": 3.5366469172756343,
      "tokens_seen": 1538850816
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000539197592778335,
      "loss": 1.3261,
      "theoretical_loss": 3.5366338091895955,
      "tokens_seen": 1538916352
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005391775325977933,
      "loss": 1.4442,
      "theoretical_loss": 3.5366207018180567,
      "tokens_seen": 1538981888
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005391574724172518,
      "loss": 1.4918,
      "theoretical_loss": 3.5366075951609486,
      "tokens_seen": 1539047424
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005391374122367101,
      "loss": 1.4963,
      "theoretical_loss": 3.536594489218202,
      "tokens_seen": 1539112960
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005391173520561685,
      "loss": 1.5526,
      "theoretical_loss": 3.5365813839897475,
      "tokens_seen": 1539178496
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005390972918756268,
      "loss": 1.4941,
      "theoretical_loss": 3.5365682794755156,
      "tokens_seen": 1539244032
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005390772316950853,
      "loss": 1.3399,
      "theoretical_loss": 3.5365551756754376,
      "tokens_seen": 1539309568
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005390571715145436,
      "loss": 1.5978,
      "theoretical_loss": 3.5365420725894428,
      "tokens_seen": 1539375104
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000539037111334002,
      "loss": 1.3338,
      "theoretical_loss": 3.5365289702174634,
      "tokens_seen": 1539440640
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005390170511534604,
      "loss": 1.4532,
      "theoretical_loss": 3.5365158685594293,
      "tokens_seen": 1539506176
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005389969909729188,
      "loss": 1.4901,
      "theoretical_loss": 3.5365027676152714,
      "tokens_seen": 1539571712
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005389769307923772,
      "loss": 1.5148,
      "theoretical_loss": 3.53648966738492,
      "tokens_seen": 1539637248
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005389568706118355,
      "loss": 1.5431,
      "theoretical_loss": 3.5364765678683074,
      "tokens_seen": 1539702784
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000538936810431294,
      "loss": 1.4328,
      "theoretical_loss": 3.536463469065362,
      "tokens_seen": 1539768320
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005389167502507522,
      "loss": 1.52,
      "theoretical_loss": 3.5364503709760164,
      "tokens_seen": 1539833856
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005388966900702107,
      "loss": 1.4627,
      "theoretical_loss": 3.5364372736002005,
      "tokens_seen": 1539899392
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000538876629889669,
      "loss": 1.5669,
      "theoretical_loss": 3.536424176937845,
      "tokens_seen": 1539964928
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005388565697091274,
      "loss": 1.5677,
      "theoretical_loss": 3.5364110809888807,
      "tokens_seen": 1540030464
    },
    {
      "epoch": 0.47,
      "objective/train/advantage_avg": 0.49414169788360596,
      "objective/train/docs_used": 562258,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3925516605377197,
      "objective/train/original_loss": 1.3925516605377197,
      "objective/train/theoretical_loss": 3.536397985753239,
      "objective/train/tokens_used": 1560556000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24436485767364502,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506564378738403,
      "objective/train/weighted_lm_loss": 1.4630215167999268,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.536397985753239,
      "tokens_seen": 1540096000
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005388365095285858,
      "loss": 1.456,
      "theoretical_loss": 3.536397985753239,
      "tokens_seen": 1540096000
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005388164493480441,
      "loss": 1.4897,
      "theoretical_loss": 3.5363848912308504,
      "tokens_seen": 1540161536
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005387963891675025,
      "loss": 1.6109,
      "theoretical_loss": 3.536371797421645,
      "tokens_seen": 1540227072
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000538776328986961,
      "loss": 1.7333,
      "theoretical_loss": 3.536358704325555,
      "tokens_seen": 1540292608
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005387562688064192,
      "loss": 1.5009,
      "theoretical_loss": 3.53634561194251,
      "tokens_seen": 1540358144
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005387362086258777,
      "loss": 1.5946,
      "theoretical_loss": 3.536332520272441,
      "tokens_seen": 1540423680
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005387161484453361,
      "loss": 1.4509,
      "theoretical_loss": 3.53631942931528,
      "tokens_seen": 1540489216
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005386960882647944,
      "loss": 1.5083,
      "theoretical_loss": 3.536306339070956,
      "tokens_seen": 1540554752
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005386760280842528,
      "loss": 1.4519,
      "theoretical_loss": 3.5362932495394013,
      "tokens_seen": 1540620288
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005386559679037111,
      "loss": 1.4106,
      "theoretical_loss": 3.536280160720546,
      "tokens_seen": 1540685824
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005386359077231695,
      "loss": 1.5093,
      "theoretical_loss": 3.5362670726143213,
      "tokens_seen": 1540751360
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000538615847542628,
      "loss": 1.4609,
      "theoretical_loss": 3.5362539852206583,
      "tokens_seen": 1540816896
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005385957873620862,
      "loss": 1.3725,
      "theoretical_loss": 3.5362408985394875,
      "tokens_seen": 1540882432
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005385757271815447,
      "loss": 1.3913,
      "theoretical_loss": 3.53622781257074,
      "tokens_seen": 1540947968
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000538555667001003,
      "loss": 1.5527,
      "theoretical_loss": 3.5362147273143467,
      "tokens_seen": 1541013504
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005385356068204614,
      "loss": 1.461,
      "theoretical_loss": 3.5362016427702385,
      "tokens_seen": 1541079040
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005385155466399198,
      "loss": 1.4895,
      "theoretical_loss": 3.5361885589383464,
      "tokens_seen": 1541144576
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005384954864593782,
      "loss": 1.4926,
      "theoretical_loss": 3.5361754758186015,
      "tokens_seen": 1541210112
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005384754262788365,
      "loss": 1.5323,
      "theoretical_loss": 3.536162393410934,
      "tokens_seen": 1541275648
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000538455366098295,
      "loss": 1.4786,
      "theoretical_loss": 3.536149311715276,
      "tokens_seen": 1541341184
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005384353059177532,
      "loss": 1.476,
      "theoretical_loss": 3.5361362307315574,
      "tokens_seen": 1541406720
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005384152457372117,
      "loss": 1.5295,
      "theoretical_loss": 3.5361231504597104,
      "tokens_seen": 1541472256
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00053839518555667,
      "loss": 1.4677,
      "theoretical_loss": 3.5361100708996647,
      "tokens_seen": 1541537792
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005383751253761284,
      "loss": 1.5214,
      "theoretical_loss": 3.536096992051352,
      "tokens_seen": 1541603328
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005383550651955868,
      "loss": 1.5077,
      "theoretical_loss": 3.5360839139147036,
      "tokens_seen": 1541668864
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005383350050150452,
      "loss": 1.4619,
      "theoretical_loss": 3.5360708364896496,
      "tokens_seen": 1541734400
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005383149448345035,
      "loss": 1.5213,
      "theoretical_loss": 3.5360577597761216,
      "tokens_seen": 1541799936
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005382948846539619,
      "loss": 1.3626,
      "theoretical_loss": 3.536044683774051,
      "tokens_seen": 1541865472
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005382748244734202,
      "loss": 1.5663,
      "theoretical_loss": 3.5360316084833685,
      "tokens_seen": 1541931008
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005382547642928787,
      "loss": 1.4759,
      "theoretical_loss": 3.536018533904005,
      "tokens_seen": 1541996544
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000538234704112337,
      "loss": 1.4489,
      "theoretical_loss": 3.5360054600358914,
      "tokens_seen": 1542062080
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005382146439317954,
      "loss": 1.475,
      "theoretical_loss": 3.5359923868789593,
      "tokens_seen": 1542127616
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005381945837512537,
      "loss": 1.4179,
      "theoretical_loss": 3.53597931443314,
      "tokens_seen": 1542193152
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005381745235707122,
      "loss": 1.5228,
      "theoretical_loss": 3.535966242698364,
      "tokens_seen": 1542258688
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005381544633901705,
      "loss": 1.4832,
      "theoretical_loss": 3.5359531716745627,
      "tokens_seen": 1542324224
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005381344032096289,
      "loss": 1.526,
      "theoretical_loss": 3.5359401013616667,
      "tokens_seen": 1542389760
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005381143430290873,
      "loss": 1.4692,
      "theoretical_loss": 3.535927031759608,
      "tokens_seen": 1542455296
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005380942828485457,
      "loss": 1.5347,
      "theoretical_loss": 3.5359139628683165,
      "tokens_seen": 1542520832
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000538074222668004,
      "loss": 1.3727,
      "theoretical_loss": 3.5359008946877255,
      "tokens_seen": 1542586368
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005380541624874624,
      "loss": 1.5715,
      "theoretical_loss": 3.5358878272177634,
      "tokens_seen": 1542651904
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005380341023069207,
      "loss": 1.4938,
      "theoretical_loss": 3.5358747604583636,
      "tokens_seen": 1542717440
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005380140421263792,
      "loss": 1.4543,
      "theoretical_loss": 3.5358616944094567,
      "tokens_seen": 1542782976
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005379939819458375,
      "loss": 1.4022,
      "theoretical_loss": 3.535848629070973,
      "tokens_seen": 1542848512
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005379739217652959,
      "loss": 1.5825,
      "theoretical_loss": 3.535835564442845,
      "tokens_seen": 1542914048
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005379538615847543,
      "loss": 1.4178,
      "theoretical_loss": 3.5358225005250024,
      "tokens_seen": 1542979584
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005379338014042126,
      "loss": 1.5026,
      "theoretical_loss": 3.5358094373173783,
      "tokens_seen": 1543045120
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000537913741223671,
      "loss": 1.5239,
      "theoretical_loss": 3.535796374819902,
      "tokens_seen": 1543110656
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005378936810431295,
      "loss": 1.5168,
      "theoretical_loss": 3.535783313032506,
      "tokens_seen": 1543176192
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005378736208625877,
      "loss": 1.5431,
      "theoretical_loss": 3.535770251955121,
      "tokens_seen": 1543241728
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005378535606820462,
      "loss": 1.4322,
      "theoretical_loss": 3.535757191587679,
      "tokens_seen": 1543307264
    },
    {
      "epoch": 0.47,
      "objective/train/advantage_avg": 0.4918099045753479,
      "objective/train/docs_used": 563510,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.394996166229248,
      "objective/train/original_loss": 1.394996166229248,
      "objective/train/theoretical_loss": 3.5357441319301106,
      "objective/train/tokens_used": 1563832800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24261750280857086,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504143238067627,
      "objective/train/weighted_lm_loss": 1.4652012586593628,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5357441319301106,
      "tokens_seen": 1543372800
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005378335005015044,
      "loss": 1.3622,
      "theoretical_loss": 3.5357441319301106,
      "tokens_seen": 1543372800
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005378134403209629,
      "loss": 1.5421,
      "theoretical_loss": 3.5357310729823466,
      "tokens_seen": 1543438336
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005377933801404213,
      "loss": 1.4588,
      "theoretical_loss": 3.5357180147443197,
      "tokens_seen": 1543503872
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005377733199598796,
      "loss": 1.6027,
      "theoretical_loss": 3.53570495721596,
      "tokens_seen": 1543569408
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000537753259779338,
      "loss": 1.3999,
      "theoretical_loss": 3.5356919003971985,
      "tokens_seen": 1543634944
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005377331995987965,
      "loss": 1.5661,
      "theoretical_loss": 3.5356788442879674,
      "tokens_seen": 1543700480
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005377131394182547,
      "loss": 1.478,
      "theoretical_loss": 3.5356657888881986,
      "tokens_seen": 1543766016
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005376930792377132,
      "loss": 1.4769,
      "theoretical_loss": 3.5356527341978223,
      "tokens_seen": 1543831552
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005376730190571714,
      "loss": 1.4414,
      "theoretical_loss": 3.5356396802167698,
      "tokens_seen": 1543897088
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005376529588766299,
      "loss": 1.5334,
      "theoretical_loss": 3.535626626944973,
      "tokens_seen": 1543962624
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005376328986960883,
      "loss": 1.366,
      "theoretical_loss": 3.5356135743823636,
      "tokens_seen": 1544028160
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005376128385155466,
      "loss": 1.4794,
      "theoretical_loss": 3.5356005225288722,
      "tokens_seen": 1544093696
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000537592778335005,
      "loss": 1.4426,
      "theoretical_loss": 3.5355874713844297,
      "tokens_seen": 1544159232
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005375727181544634,
      "loss": 1.5419,
      "theoretical_loss": 3.5355744209489695,
      "tokens_seen": 1544224768
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005375526579739217,
      "loss": 1.5179,
      "theoretical_loss": 3.5355613712224203,
      "tokens_seen": 1544290304
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005375325977933802,
      "loss": 1.5976,
      "theoretical_loss": 3.535548322204716,
      "tokens_seen": 1544355840
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005375125376128386,
      "loss": 1.5024,
      "theoretical_loss": 3.535535273895787,
      "tokens_seen": 1544421376
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005374924774322969,
      "loss": 1.6337,
      "theoretical_loss": 3.535522226295564,
      "tokens_seen": 1544486912
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005374724172517554,
      "loss": 1.4061,
      "theoretical_loss": 3.53550917940398,
      "tokens_seen": 1544552448
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005374523570712136,
      "loss": 1.4613,
      "theoretical_loss": 3.5354961332209647,
      "tokens_seen": 1544617984
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005374322968906721,
      "loss": 1.3622,
      "theoretical_loss": 3.5354830877464507,
      "tokens_seen": 1544683520
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005374122367101304,
      "loss": 1.3963,
      "theoretical_loss": 3.5354700429803696,
      "tokens_seen": 1544749056
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005373921765295888,
      "loss": 1.5015,
      "theoretical_loss": 3.535456998922652,
      "tokens_seen": 1544814592
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005373721163490472,
      "loss": 1.3601,
      "theoretical_loss": 3.53544395557323,
      "tokens_seen": 1544880128
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005373520561685056,
      "loss": 1.4509,
      "theoretical_loss": 3.5354309129320347,
      "tokens_seen": 1544945664
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005373319959879639,
      "loss": 1.5274,
      "theoretical_loss": 3.5354178709989985,
      "tokens_seen": 1545011200
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005373119358074223,
      "loss": 1.5419,
      "theoretical_loss": 3.535404829774052,
      "tokens_seen": 1545076736
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005372918756268806,
      "loss": 1.4247,
      "theoretical_loss": 3.535391789257127,
      "tokens_seen": 1545142272
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005372718154463391,
      "loss": 1.5539,
      "theoretical_loss": 3.535378749448155,
      "tokens_seen": 1545207808
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005372517552657974,
      "loss": 1.4356,
      "theoretical_loss": 3.535365710347067,
      "tokens_seen": 1545273344
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005372316950852558,
      "loss": 1.3565,
      "theoretical_loss": 3.535352671953796,
      "tokens_seen": 1545338880
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005372116349047141,
      "loss": 1.5993,
      "theoretical_loss": 3.535339634268272,
      "tokens_seen": 1545404416
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005371915747241726,
      "loss": 1.4582,
      "theoretical_loss": 3.5353265972904278,
      "tokens_seen": 1545469952
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005371715145436309,
      "loss": 1.4819,
      "theoretical_loss": 3.5353135610201942,
      "tokens_seen": 1545535488
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005371514543630893,
      "loss": 1.4452,
      "theoretical_loss": 3.5353005254575027,
      "tokens_seen": 1545601024
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005371313941825477,
      "loss": 1.5239,
      "theoretical_loss": 3.535287490602286,
      "tokens_seen": 1545666560
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005371113340020061,
      "loss": 1.5386,
      "theoretical_loss": 3.5352744564544745,
      "tokens_seen": 1545732096
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005370912738214644,
      "loss": 1.5977,
      "theoretical_loss": 3.535261423014,
      "tokens_seen": 1545797632
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005370712136409228,
      "loss": 1.6945,
      "theoretical_loss": 3.5352483902807945,
      "tokens_seen": 1545863168
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005370511534603811,
      "loss": 1.4961,
      "theoretical_loss": 3.53523535825479,
      "tokens_seen": 1545928704
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005370310932798396,
      "loss": 1.4684,
      "theoretical_loss": 3.5352223269359175,
      "tokens_seen": 1545994240
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005370110330992979,
      "loss": 1.4563,
      "theoretical_loss": 3.5352092963241084,
      "tokens_seen": 1546059776
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005369909729187563,
      "loss": 1.4978,
      "theoretical_loss": 3.535196266419295,
      "tokens_seen": 1546125312
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005369709127382147,
      "loss": 1.3891,
      "theoretical_loss": 3.5351832372214087,
      "tokens_seen": 1546190848
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000536950852557673,
      "loss": 1.5238,
      "theoretical_loss": 3.5351702087303813,
      "tokens_seen": 1546256384
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005369307923771314,
      "loss": 1.5892,
      "theoretical_loss": 3.535157180946145,
      "tokens_seen": 1546321920
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005369107321965899,
      "loss": 1.4531,
      "theoretical_loss": 3.5351441538686306,
      "tokens_seen": 1546387456
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005368906720160481,
      "loss": 1.3791,
      "theoretical_loss": 3.5351311274977695,
      "tokens_seen": 1546452992
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005368706118355066,
      "loss": 1.4745,
      "theoretical_loss": 3.535118101833495,
      "tokens_seen": 1546518528
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005368505516549648,
      "loss": 1.3638,
      "theoretical_loss": 3.5351050768757375,
      "tokens_seen": 1546584064
    },
    {
      "epoch": 0.47,
      "objective/train/advantage_avg": 0.49552419781684875,
      "objective/train/docs_used": 564726,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3794103860855103,
      "objective/train/original_loss": 1.3794103860855103,
      "objective/train/theoretical_loss": 3.535092052624429,
      "objective/train/tokens_used": 1567109600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24582239985466003,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050802230834961,
      "objective/train/weighted_lm_loss": 1.449391484260559,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.535092052624429,
      "tokens_seen": 1546649600
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005368304914744233,
      "loss": 1.47,
      "theoretical_loss": 3.535092052624429,
      "tokens_seen": 1546649600
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005368104312938817,
      "loss": 1.5664,
      "theoretical_loss": 3.535079029079502,
      "tokens_seen": 1546715136
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00053679037111334,
      "loss": 1.5453,
      "theoretical_loss": 3.5350660062408874,
      "tokens_seen": 1546780672
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005367703109327984,
      "loss": 1.4774,
      "theoretical_loss": 3.5350529841085168,
      "tokens_seen": 1546846208
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005367502507522569,
      "loss": 1.4975,
      "theoretical_loss": 3.5350399626823226,
      "tokens_seen": 1546911744
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005367301905717151,
      "loss": 1.5659,
      "theoretical_loss": 3.535026941962237,
      "tokens_seen": 1546977280
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005367101303911736,
      "loss": 1.5176,
      "theoretical_loss": 3.535013921948191,
      "tokens_seen": 1547042816
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005366900702106318,
      "loss": 1.5571,
      "theoretical_loss": 3.535000902640116,
      "tokens_seen": 1547108352
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005366700100300903,
      "loss": 1.3718,
      "theoretical_loss": 3.534987884037945,
      "tokens_seen": 1547173888
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005366499498495487,
      "loss": 1.5284,
      "theoretical_loss": 3.534974866141609,
      "tokens_seen": 1547239424
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000536629889669007,
      "loss": 1.4908,
      "theoretical_loss": 3.53496184895104,
      "tokens_seen": 1547304960
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005366098294884654,
      "loss": 1.4361,
      "theoretical_loss": 3.5349488324661706,
      "tokens_seen": 1547370496
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005365897693079238,
      "loss": 1.4842,
      "theoretical_loss": 3.5349358166869314,
      "tokens_seen": 1547436032
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005365697091273821,
      "loss": 1.5812,
      "theoretical_loss": 3.5349228016132552,
      "tokens_seen": 1547501568
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005365496489468406,
      "loss": 1.3905,
      "theoretical_loss": 3.5349097872450734,
      "tokens_seen": 1547567104
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005365295887662988,
      "loss": 1.4534,
      "theoretical_loss": 3.534896773582318,
      "tokens_seen": 1547632640
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005365095285857573,
      "loss": 1.5126,
      "theoretical_loss": 3.534883760624921,
      "tokens_seen": 1547698176
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005364894684052156,
      "loss": 1.6588,
      "theoretical_loss": 3.5348707483728146,
      "tokens_seen": 1547763712
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000536469408224674,
      "loss": 1.4998,
      "theoretical_loss": 3.5348577368259297,
      "tokens_seen": 1547829248
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005364493480441324,
      "loss": 1.587,
      "theoretical_loss": 3.5348447259841995,
      "tokens_seen": 1547894784
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005364292878635908,
      "loss": 1.5712,
      "theoretical_loss": 3.534831715847555,
      "tokens_seen": 1547960320
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005364092276830491,
      "loss": 1.5844,
      "theoretical_loss": 3.534818706415928,
      "tokens_seen": 1548025856
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005363891675025076,
      "loss": 1.5589,
      "theoretical_loss": 3.5348056976892517,
      "tokens_seen": 1548091392
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005363691073219658,
      "loss": 1.4388,
      "theoretical_loss": 3.534792689667457,
      "tokens_seen": 1548156928
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005363490471414243,
      "loss": 1.4756,
      "theoretical_loss": 3.534779682350475,
      "tokens_seen": 1548222464
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005363289869608826,
      "loss": 1.6267,
      "theoretical_loss": 3.5347666757382403,
      "tokens_seen": 1548288000
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000536308926780341,
      "loss": 1.5117,
      "theoretical_loss": 3.5347536698306827,
      "tokens_seen": 1548353536
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005362888665997994,
      "loss": 1.5626,
      "theoretical_loss": 3.534740664627735,
      "tokens_seen": 1548419072
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005362688064192578,
      "loss": 1.4893,
      "theoretical_loss": 3.534727660129329,
      "tokens_seen": 1548484608
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005362487462387161,
      "loss": 1.4767,
      "theoretical_loss": 3.534714656335397,
      "tokens_seen": 1548550144
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005362286860581745,
      "loss": 1.5765,
      "theoretical_loss": 3.534701653245871,
      "tokens_seen": 1548615680
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005362086258776329,
      "loss": 1.3872,
      "theoretical_loss": 3.5346886508606823,
      "tokens_seen": 1548681216
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005361885656970913,
      "loss": 1.4261,
      "theoretical_loss": 3.534675649179764,
      "tokens_seen": 1548746752
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005361685055165496,
      "loss": 1.4517,
      "theoretical_loss": 3.534662648203047,
      "tokens_seen": 1548812288
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000536148445336008,
      "loss": 1.4919,
      "theoretical_loss": 3.5346496479304648,
      "tokens_seen": 1548877824
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005361283851554663,
      "loss": 1.56,
      "theoretical_loss": 3.5346366483619485,
      "tokens_seen": 1548943360
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005361083249749248,
      "loss": 1.584,
      "theoretical_loss": 3.53462364949743,
      "tokens_seen": 1549008896
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005360882647943831,
      "loss": 1.3653,
      "theoretical_loss": 3.534610651336842,
      "tokens_seen": 1549074432
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005360682046138415,
      "loss": 1.3998,
      "theoretical_loss": 3.534597653880117,
      "tokens_seen": 1549139968
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005360481444332999,
      "loss": 1.5339,
      "theoretical_loss": 3.5345846571271857,
      "tokens_seen": 1549205504
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005360280842527583,
      "loss": 1.5514,
      "theoretical_loss": 3.5345716610779814,
      "tokens_seen": 1549271040
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005360080240722167,
      "loss": 1.317,
      "theoretical_loss": 3.534558665732436,
      "tokens_seen": 1549336576
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005359879638916751,
      "loss": 1.566,
      "theoretical_loss": 3.5345456710904806,
      "tokens_seen": 1549402112
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005359679037111334,
      "loss": 1.6158,
      "theoretical_loss": 3.5345326771520487,
      "tokens_seen": 1549467648
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005359478435305918,
      "loss": 1.4572,
      "theoretical_loss": 3.5345196839170723,
      "tokens_seen": 1549533184
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005359277833500503,
      "loss": 1.6629,
      "theoretical_loss": 3.534506691385483,
      "tokens_seen": 1549598720
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005359077231695085,
      "loss": 1.5477,
      "theoretical_loss": 3.5344936995572134,
      "tokens_seen": 1549664256
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000535887662988967,
      "loss": 1.4339,
      "theoretical_loss": 3.5344807084321954,
      "tokens_seen": 1549729792
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005358676028084252,
      "loss": 1.4704,
      "theoretical_loss": 3.534467718010361,
      "tokens_seen": 1549795328
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005358475426278837,
      "loss": 1.5107,
      "theoretical_loss": 3.534454728291643,
      "tokens_seen": 1549860864
    },
    {
      "epoch": 0.47,
      "objective/train/advantage_avg": 0.4930802285671234,
      "objective/train/docs_used": 565981,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.484511137008667,
      "objective/train/original_loss": 1.484511137008667,
      "objective/train/theoretical_loss": 3.534441739275974,
      "objective/train/tokens_used": 1570386400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24380987882614136,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050547480583191,
      "objective/train/weighted_lm_loss": 1.559489369392395,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.534441739275974,
      "tokens_seen": 1549926400
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005358274824473421,
      "loss": 1.582,
      "theoretical_loss": 3.534441739275974,
      "tokens_seen": 1549926400
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005358074222668004,
      "loss": 1.4604,
      "theoretical_loss": 3.5344287509632846,
      "tokens_seen": 1549991936
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005357873620862588,
      "loss": 1.3723,
      "theoretical_loss": 3.5344157633535085,
      "tokens_seen": 1550057472
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005357673019057173,
      "loss": 1.3662,
      "theoretical_loss": 3.534402776446577,
      "tokens_seen": 1550123008
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005357472417251755,
      "loss": 1.5582,
      "theoretical_loss": 3.534389790242423,
      "tokens_seen": 1550188544
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000535727181544634,
      "loss": 1.5138,
      "theoretical_loss": 3.5343768047409787,
      "tokens_seen": 1550254080
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005357071213640922,
      "loss": 1.5469,
      "theoretical_loss": 3.5343638199421763,
      "tokens_seen": 1550319616
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005356870611835507,
      "loss": 1.4674,
      "theoretical_loss": 3.5343508358459474,
      "tokens_seen": 1550385152
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005356670010030091,
      "loss": 1.6057,
      "theoretical_loss": 3.534337852452225,
      "tokens_seen": 1550450688
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005356469408224674,
      "loss": 1.5387,
      "theoretical_loss": 3.534324869760942,
      "tokens_seen": 1550516224
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005356268806419258,
      "loss": 1.4422,
      "theoretical_loss": 3.5343118877720294,
      "tokens_seen": 1550581760
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005356068204613842,
      "loss": 1.5397,
      "theoretical_loss": 3.5342989064854207,
      "tokens_seen": 1550647296
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005355867602808425,
      "loss": 1.4206,
      "theoretical_loss": 3.534285925901047,
      "tokens_seen": 1550712832
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000535566700100301,
      "loss": 1.4954,
      "theoretical_loss": 3.5342729460188416,
      "tokens_seen": 1550778368
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005355466399197592,
      "loss": 1.439,
      "theoretical_loss": 3.534259966838736,
      "tokens_seen": 1550843904
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005355265797392177,
      "loss": 1.4801,
      "theoretical_loss": 3.5342469883606635,
      "tokens_seen": 1550909440
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000535506519558676,
      "loss": 1.4936,
      "theoretical_loss": 3.534234010584556,
      "tokens_seen": 1550974976
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005354864593781344,
      "loss": 1.3866,
      "theoretical_loss": 3.534221033510346,
      "tokens_seen": 1551040512
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005354663991975928,
      "loss": 1.3905,
      "theoretical_loss": 3.534208057137966,
      "tokens_seen": 1551106048
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005354463390170512,
      "loss": 1.455,
      "theoretical_loss": 3.5341950814673475,
      "tokens_seen": 1551171584
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005354262788365095,
      "loss": 1.4972,
      "theoretical_loss": 3.534182106498424,
      "tokens_seen": 1551237120
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000535406218655968,
      "loss": 1.4996,
      "theoretical_loss": 3.5341691322311277,
      "tokens_seen": 1551302656
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005353861584754262,
      "loss": 1.5976,
      "theoretical_loss": 3.5341561586653905,
      "tokens_seen": 1551368192
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005353660982948847,
      "loss": 1.457,
      "theoretical_loss": 3.5341431858011454,
      "tokens_seen": 1551433728
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000535346038114343,
      "loss": 1.4708,
      "theoretical_loss": 3.534130213638324,
      "tokens_seen": 1551499264
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005353259779338014,
      "loss": 1.3981,
      "theoretical_loss": 3.53411724217686,
      "tokens_seen": 1551564800
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005353059177532598,
      "loss": 1.3957,
      "theoretical_loss": 3.5341042714166853,
      "tokens_seen": 1551630336
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005352858575727182,
      "loss": 1.5155,
      "theoretical_loss": 3.5340913013577318,
      "tokens_seen": 1551695872
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005352657973921765,
      "loss": 1.4127,
      "theoretical_loss": 3.5340783319999325,
      "tokens_seen": 1551761408
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005352457372116349,
      "loss": 1.4534,
      "theoretical_loss": 3.5340653633432204,
      "tokens_seen": 1551826944
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005352256770310933,
      "loss": 1.4349,
      "theoretical_loss": 3.5340523953875267,
      "tokens_seen": 1551892480
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005352056168505517,
      "loss": 1.4406,
      "theoretical_loss": 3.534039428132785,
      "tokens_seen": 1551958016
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00053518555667001,
      "loss": 1.5088,
      "theoretical_loss": 3.534026461578928,
      "tokens_seen": 1552023552
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005351654964894685,
      "loss": 1.357,
      "theoretical_loss": 3.5340134957258864,
      "tokens_seen": 1552089088
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005351454363089267,
      "loss": 1.54,
      "theoretical_loss": 3.5340005305735946,
      "tokens_seen": 1552154624
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005351253761283852,
      "loss": 1.4935,
      "theoretical_loss": 3.5339875661219846,
      "tokens_seen": 1552220160
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005351053159478435,
      "loss": 1.3595,
      "theoretical_loss": 3.5339746023709893,
      "tokens_seen": 1552285696
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005350852557673019,
      "loss": 1.4083,
      "theoretical_loss": 3.53396163932054,
      "tokens_seen": 1552351232
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005350651955867603,
      "loss": 1.4105,
      "theoretical_loss": 3.533948676970571,
      "tokens_seen": 1552416768
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005350451354062187,
      "loss": 1.403,
      "theoretical_loss": 3.5339357153210136,
      "tokens_seen": 1552482304
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000535025075225677,
      "loss": 1.447,
      "theoretical_loss": 3.5339227543718006,
      "tokens_seen": 1552547840
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005350050150451355,
      "loss": 1.5252,
      "theoretical_loss": 3.5339097941228648,
      "tokens_seen": 1552613376
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005349849548645937,
      "loss": 1.5713,
      "theoretical_loss": 3.533896834574139,
      "tokens_seen": 1552678912
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005349648946840522,
      "loss": 1.5407,
      "theoretical_loss": 3.533883875725556,
      "tokens_seen": 1552744448
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005349448345035105,
      "loss": 1.5049,
      "theoretical_loss": 3.5338709175770475,
      "tokens_seen": 1552809984
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005349247743229689,
      "loss": 1.4692,
      "theoretical_loss": 3.5338579601285467,
      "tokens_seen": 1552875520
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005349047141424273,
      "loss": 1.4621,
      "theoretical_loss": 3.5338450033799864,
      "tokens_seen": 1552941056
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005348846539618856,
      "loss": 1.4401,
      "theoretical_loss": 3.533832047331299,
      "tokens_seen": 1553006592
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000534864593781344,
      "loss": 1.4337,
      "theoretical_loss": 3.5338190919824175,
      "tokens_seen": 1553072128
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005348445336008025,
      "loss": 1.4292,
      "theoretical_loss": 3.5338061373332748,
      "tokens_seen": 1553137664
    },
    {
      "epoch": 0.47,
      "objective/train/advantage_avg": 0.49172243475914,
      "objective/train/docs_used": 567156,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4032881259918213,
      "objective/train/original_loss": 1.4032881259918213,
      "objective/train/theoretical_loss": 3.533793183383802,
      "objective/train/tokens_used": 1573663200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24228571355342865,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504039525985718,
      "objective/train/weighted_lm_loss": 1.4738452434539795,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.533793183383802,
      "tokens_seen": 1553203200
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005348244734202607,
      "loss": 1.4375,
      "theoretical_loss": 3.533793183383802,
      "tokens_seen": 1553203200
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005348044132397192,
      "loss": 1.4742,
      "theoretical_loss": 3.5337802301339334,
      "tokens_seen": 1553268736
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005347843530591776,
      "loss": 1.346,
      "theoretical_loss": 3.533767277583601,
      "tokens_seen": 1553334272
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005347642928786359,
      "loss": 1.4298,
      "theoretical_loss": 3.5337543257327377,
      "tokens_seen": 1553399808
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005347442326980943,
      "loss": 1.3875,
      "theoretical_loss": 3.5337413745812767,
      "tokens_seen": 1553465344
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005347241725175526,
      "loss": 1.5323,
      "theoretical_loss": 3.5337284241291496,
      "tokens_seen": 1553530880
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000534704112337011,
      "loss": 1.5338,
      "theoretical_loss": 3.53371547437629,
      "tokens_seen": 1553596416
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005346840521564695,
      "loss": 1.5368,
      "theoretical_loss": 3.533702525322631,
      "tokens_seen": 1553661952
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005346639919759277,
      "loss": 1.3147,
      "theoretical_loss": 3.5336895769681043,
      "tokens_seen": 1553727488
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005346439317953862,
      "loss": 1.4348,
      "theoretical_loss": 3.533676629312643,
      "tokens_seen": 1553793024
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005346238716148444,
      "loss": 1.4491,
      "theoretical_loss": 3.5336636823561807,
      "tokens_seen": 1553858560
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005346038114343029,
      "loss": 1.4248,
      "theoretical_loss": 3.5336507360986493,
      "tokens_seen": 1553924096
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005345837512537613,
      "loss": 1.4516,
      "theoretical_loss": 3.533637790539981,
      "tokens_seen": 1553989632
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005345636910732196,
      "loss": 1.4777,
      "theoretical_loss": 3.5336248456801105,
      "tokens_seen": 1554055168
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000534543630892678,
      "loss": 1.3602,
      "theoretical_loss": 3.533611901518969,
      "tokens_seen": 1554120704
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005345235707121364,
      "loss": 1.3131,
      "theoretical_loss": 3.5335989580564897,
      "tokens_seen": 1554186240
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005345035105315948,
      "loss": 1.3906,
      "theoretical_loss": 3.533586015292606,
      "tokens_seen": 1554251776
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005344834503510532,
      "loss": 1.4817,
      "theoretical_loss": 3.53357307322725,
      "tokens_seen": 1554317312
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005344633901705116,
      "loss": 1.4172,
      "theoretical_loss": 3.533560131860355,
      "tokens_seen": 1554382848
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005344433299899699,
      "loss": 1.4216,
      "theoretical_loss": 3.533547191191854,
      "tokens_seen": 1554448384
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005344232698094284,
      "loss": 1.5382,
      "theoretical_loss": 3.5335342512216794,
      "tokens_seen": 1554513920
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005344032096288866,
      "loss": 1.5486,
      "theoretical_loss": 3.533521311949764,
      "tokens_seen": 1554579456
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005343831494483451,
      "loss": 1.5676,
      "theoretical_loss": 3.5335083733760415,
      "tokens_seen": 1554644992
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005343630892678034,
      "loss": 1.4526,
      "theoretical_loss": 3.533495435500444,
      "tokens_seen": 1554710528
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005343430290872618,
      "loss": 1.4729,
      "theoretical_loss": 3.5334824983229045,
      "tokens_seen": 1554776064
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005343229689067202,
      "loss": 1.5672,
      "theoretical_loss": 3.533469561843356,
      "tokens_seen": 1554841600
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005343029087261786,
      "loss": 1.4975,
      "theoretical_loss": 3.533456626061732,
      "tokens_seen": 1554907136
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005342828485456369,
      "loss": 1.3493,
      "theoretical_loss": 3.5334436909779647,
      "tokens_seen": 1554972672
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005342627883650953,
      "loss": 1.5342,
      "theoretical_loss": 3.5334307565919874,
      "tokens_seen": 1555038208
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005342427281845537,
      "loss": 1.4955,
      "theoretical_loss": 3.5334178229037327,
      "tokens_seen": 1555103744
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005342226680040121,
      "loss": 1.3545,
      "theoretical_loss": 3.533404889913134,
      "tokens_seen": 1555169280
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005342026078234704,
      "loss": 1.3407,
      "theoretical_loss": 3.5333919576201236,
      "tokens_seen": 1555234816
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005341825476429289,
      "loss": 1.561,
      "theoretical_loss": 3.533379026024636,
      "tokens_seen": 1555300352
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005341624874623871,
      "loss": 1.4885,
      "theoretical_loss": 3.5333660951266017,
      "tokens_seen": 1555365888
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005341424272818456,
      "loss": 1.4495,
      "theoretical_loss": 3.5333531649259564,
      "tokens_seen": 1555431424
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005341223671013039,
      "loss": 1.4433,
      "theoretical_loss": 3.5333402354226315,
      "tokens_seen": 1555496960
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005341023069207623,
      "loss": 1.5576,
      "theoretical_loss": 3.53332730661656,
      "tokens_seen": 1555562496
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005340822467402207,
      "loss": 1.5615,
      "theoretical_loss": 3.5333143785076757,
      "tokens_seen": 1555628032
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005340621865596791,
      "loss": 1.5052,
      "theoretical_loss": 3.5333014510959107,
      "tokens_seen": 1555693568
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005340421263791374,
      "loss": 1.4676,
      "theoretical_loss": 3.533288524381199,
      "tokens_seen": 1555759104
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005340220661985959,
      "loss": 1.394,
      "theoretical_loss": 3.533275598363473,
      "tokens_seen": 1555824640
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005340020060180541,
      "loss": 1.2333,
      "theoretical_loss": 3.5332626730426666,
      "tokens_seen": 1555890176
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005339819458375126,
      "loss": 1.4229,
      "theoretical_loss": 3.5332497484187115,
      "tokens_seen": 1555955712
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005339618856569709,
      "loss": 1.4035,
      "theoretical_loss": 3.5332368244915417,
      "tokens_seen": 1556021248
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005339418254764293,
      "loss": 1.5164,
      "theoretical_loss": 3.53322390126109,
      "tokens_seen": 1556086784
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005339217652958877,
      "loss": 1.7071,
      "theoretical_loss": 3.5332109787272903,
      "tokens_seen": 1556152320
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000533901705115346,
      "loss": 1.5524,
      "theoretical_loss": 3.5331980568900745,
      "tokens_seen": 1556217856
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005338816449348044,
      "loss": 1.3332,
      "theoretical_loss": 3.533185135749376,
      "tokens_seen": 1556283392
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005338615847542629,
      "loss": 1.4607,
      "theoretical_loss": 3.533172215305129,
      "tokens_seen": 1556348928
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005338415245737211,
      "loss": 1.3625,
      "theoretical_loss": 3.5331592955572653,
      "tokens_seen": 1556414464
    },
    {
      "epoch": 0.47,
      "objective/train/advantage_avg": 0.49402302503585815,
      "objective/train/docs_used": 568457,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2820841073989868,
      "objective/train/original_loss": 1.2820842266082764,
      "objective/train/theoretical_loss": 3.5331463765057185,
      "objective/train/tokens_used": 1576940000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2444804161787033,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506452322006226,
      "objective/train/weighted_lm_loss": 1.3469765186309814,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.5331463765057185,
      "tokens_seen": 1556480000
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005338214643931796,
      "loss": 1.3931,
      "theoretical_loss": 3.5331463765057185,
      "tokens_seen": 1556480000
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000533801404212638,
      "loss": 1.4253,
      "theoretical_loss": 3.533133458150422,
      "tokens_seen": 1556545536
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005337813440320963,
      "loss": 1.4371,
      "theoretical_loss": 3.533120540491309,
      "tokens_seen": 1556611072
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005337612838515547,
      "loss": 1.4457,
      "theoretical_loss": 3.533107623528312,
      "tokens_seen": 1556676608
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000533741223671013,
      "loss": 1.4102,
      "theoretical_loss": 3.5330947072613643,
      "tokens_seen": 1556742144
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005337211634904714,
      "loss": 1.639,
      "theoretical_loss": 3.5330817916904005,
      "tokens_seen": 1556807680
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005337011033099299,
      "loss": 1.5294,
      "theoretical_loss": 3.533068876815352,
      "tokens_seen": 1556873216
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005336810431293881,
      "loss": 1.3594,
      "theoretical_loss": 3.533055962636153,
      "tokens_seen": 1556938752
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005336609829488466,
      "loss": 1.5099,
      "theoretical_loss": 3.5330430491527363,
      "tokens_seen": 1557004288
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005336409227683048,
      "loss": 1.3962,
      "theoretical_loss": 3.5330301363650354,
      "tokens_seen": 1557069824
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005336208625877633,
      "loss": 1.4157,
      "theoretical_loss": 3.5330172242729834,
      "tokens_seen": 1557135360
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005336008024072217,
      "loss": 1.4698,
      "theoretical_loss": 3.5330043128765136,
      "tokens_seen": 1557200896
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00053358074222668,
      "loss": 1.4531,
      "theoretical_loss": 3.5329914021755595,
      "tokens_seen": 1557266432
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005335606820461384,
      "loss": 1.5545,
      "theoretical_loss": 3.5329784921700536,
      "tokens_seen": 1557331968
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005335406218655968,
      "loss": 1.4775,
      "theoretical_loss": 3.53296558285993,
      "tokens_seen": 1557397504
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005335205616850551,
      "loss": 1.3824,
      "theoretical_loss": 3.532952674245122,
      "tokens_seen": 1557463040
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005335005015045136,
      "loss": 1.4746,
      "theoretical_loss": 3.532939766325562,
      "tokens_seen": 1557528576
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005334804413239719,
      "loss": 1.2814,
      "theoretical_loss": 3.5329268591011838,
      "tokens_seen": 1557594112
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005334603811434303,
      "loss": 1.4067,
      "theoretical_loss": 3.532913952571921,
      "tokens_seen": 1557659648
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005334403209628887,
      "loss": 1.4116,
      "theoretical_loss": 3.532901046737707,
      "tokens_seen": 1557725184
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000533420260782347,
      "loss": 1.4273,
      "theoretical_loss": 3.532888141598474,
      "tokens_seen": 1557790720
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005334002006018054,
      "loss": 1.5191,
      "theoretical_loss": 3.532875237154157,
      "tokens_seen": 1557856256
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005333801404212638,
      "loss": 1.423,
      "theoretical_loss": 3.5328623334046885,
      "tokens_seen": 1557921792
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005333600802407221,
      "loss": 1.4507,
      "theoretical_loss": 3.5328494303500015,
      "tokens_seen": 1557987328
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005333400200601806,
      "loss": 1.5237,
      "theoretical_loss": 3.53283652799003,
      "tokens_seen": 1558052864
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005333199598796389,
      "loss": 1.3388,
      "theoretical_loss": 3.5328236263247064,
      "tokens_seen": 1558118400
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005332998996990973,
      "loss": 1.4484,
      "theoretical_loss": 3.532810725353966,
      "tokens_seen": 1558183936
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005332798395185556,
      "loss": 1.3492,
      "theoretical_loss": 3.53279782507774,
      "tokens_seen": 1558249472
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005332597793380141,
      "loss": 1.5659,
      "theoretical_loss": 3.532784925495963,
      "tokens_seen": 1558315008
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005332397191574724,
      "loss": 1.4919,
      "theoretical_loss": 3.5327720266085683,
      "tokens_seen": 1558380544
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005332196589769308,
      "loss": 1.5414,
      "theoretical_loss": 3.5327591284154893,
      "tokens_seen": 1558446080
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005331995987963891,
      "loss": 1.4598,
      "theoretical_loss": 3.5327462309166595,
      "tokens_seen": 1558511616
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005331795386158475,
      "loss": 1.3733,
      "theoretical_loss": 3.532733334112012,
      "tokens_seen": 1558577152
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005331594784353059,
      "loss": 1.4601,
      "theoretical_loss": 3.5327204380014807,
      "tokens_seen": 1558642688
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005331394182547643,
      "loss": 1.4087,
      "theoretical_loss": 3.5327075425849985,
      "tokens_seen": 1558708224
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005331193580742226,
      "loss": 1.315,
      "theoretical_loss": 3.532694647862499,
      "tokens_seen": 1558773760
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005330992978936811,
      "loss": 1.5768,
      "theoretical_loss": 3.5326817538339164,
      "tokens_seen": 1558839296
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005330792377131394,
      "loss": 1.5469,
      "theoretical_loss": 3.532668860499183,
      "tokens_seen": 1558904832
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005330591775325978,
      "loss": 1.3665,
      "theoretical_loss": 3.5326559678582337,
      "tokens_seen": 1558970368
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005330391173520561,
      "loss": 1.4524,
      "theoretical_loss": 3.5326430759110012,
      "tokens_seen": 1559035904
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005330190571715145,
      "loss": 1.3913,
      "theoretical_loss": 3.5326301846574184,
      "tokens_seen": 1559101440
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000532998996990973,
      "loss": 1.5073,
      "theoretical_loss": 3.5326172940974203,
      "tokens_seen": 1559166976
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005329789368104313,
      "loss": 1.4329,
      "theoretical_loss": 3.532604404230939,
      "tokens_seen": 1559232512
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005329588766298897,
      "loss": 1.4195,
      "theoretical_loss": 3.532591515057909,
      "tokens_seen": 1559298048
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005329388164493481,
      "loss": 1.4711,
      "theoretical_loss": 3.532578626578263,
      "tokens_seen": 1559363584
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005329187562688064,
      "loss": 1.4148,
      "theoretical_loss": 3.5325657387919356,
      "tokens_seen": 1559429120
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005328986960882648,
      "loss": 1.394,
      "theoretical_loss": 3.53255285169886,
      "tokens_seen": 1559494656
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005328786359077233,
      "loss": 1.5495,
      "theoretical_loss": 3.532539965298969,
      "tokens_seen": 1559560192
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005328585757271815,
      "loss": 1.5767,
      "theoretical_loss": 3.5325270795921972,
      "tokens_seen": 1559625728
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00053283851554664,
      "loss": 1.3457,
      "theoretical_loss": 3.5325141945784777,
      "tokens_seen": 1559691264
    },
    {
      "epoch": 0.47,
      "objective/train/advantage_avg": 0.49180299043655396,
      "objective/train/docs_used": 569616,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4036917686462402,
      "objective/train/original_loss": 1.4036917686462402,
      "objective/train/theoretical_loss": 3.5325013102577443,
      "objective/train/tokens_used": 1580216800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2426343709230423,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050413727760315,
      "objective/train/weighted_lm_loss": 1.4743260145187378,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5325013102577443,
      "tokens_seen": 1559756800
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005328184553660984,
      "loss": 1.4878,
      "theoretical_loss": 3.5325013102577443,
      "tokens_seen": 1559756800
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005327983951855567,
      "loss": 1.5792,
      "theoretical_loss": 3.5324884266299303,
      "tokens_seen": 1559822336
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005327783350050151,
      "loss": 1.5016,
      "theoretical_loss": 3.5324755436949697,
      "tokens_seen": 1559887872
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005327582748244734,
      "loss": 1.4817,
      "theoretical_loss": 3.5324626614527963,
      "tokens_seen": 1559953408
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005327382146439318,
      "loss": 1.4608,
      "theoretical_loss": 3.532449779903343,
      "tokens_seen": 1560018944
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005327181544633903,
      "loss": 1.406,
      "theoretical_loss": 3.532436899046544,
      "tokens_seen": 1560084480
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005326980942828485,
      "loss": 1.3951,
      "theoretical_loss": 3.532424018882333,
      "tokens_seen": 1560150016
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000532678034102307,
      "loss": 1.5245,
      "theoretical_loss": 3.5324111394106428,
      "tokens_seen": 1560215552
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005326579739217652,
      "loss": 1.4059,
      "theoretical_loss": 3.5323982606314086,
      "tokens_seen": 1560281088
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005326379137412237,
      "loss": 1.5394,
      "theoretical_loss": 3.532385382544563,
      "tokens_seen": 1560346624
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005326178535606821,
      "loss": 1.4097,
      "theoretical_loss": 3.5323725051500396,
      "tokens_seen": 1560412160
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005325977933801404,
      "loss": 1.461,
      "theoretical_loss": 3.5323596284477734,
      "tokens_seen": 1560477696
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005325777331995988,
      "loss": 1.5617,
      "theoretical_loss": 3.5323467524376966,
      "tokens_seen": 1560543232
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005325576730190572,
      "loss": 1.5044,
      "theoretical_loss": 3.5323338771197434,
      "tokens_seen": 1560608768
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005325376128385155,
      "loss": 1.5221,
      "theoretical_loss": 3.532321002493848,
      "tokens_seen": 1560674304
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000532517552657974,
      "loss": 1.3693,
      "theoretical_loss": 3.5323081285599436,
      "tokens_seen": 1560739840
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005324974924774323,
      "loss": 1.5244,
      "theoretical_loss": 3.532295255317964,
      "tokens_seen": 1560805376
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005324774322968907,
      "loss": 1.406,
      "theoretical_loss": 3.532282382767843,
      "tokens_seen": 1560870912
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005324573721163491,
      "loss": 1.417,
      "theoretical_loss": 3.532269510909515,
      "tokens_seen": 1560936448
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005324373119358075,
      "loss": 1.34,
      "theoretical_loss": 3.532256639742913,
      "tokens_seen": 1561001984
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005324172517552658,
      "loss": 1.3984,
      "theoretical_loss": 3.532243769267971,
      "tokens_seen": 1561067520
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005323971915747242,
      "loss": 1.4845,
      "theoretical_loss": 3.532230899484623,
      "tokens_seen": 1561133056
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005323771313941825,
      "loss": 1.4165,
      "theoretical_loss": 3.5322180303928024,
      "tokens_seen": 1561198592
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000532357071213641,
      "loss": 1.4778,
      "theoretical_loss": 3.5322051619924433,
      "tokens_seen": 1561264128
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005323370110330993,
      "loss": 1.4272,
      "theoretical_loss": 3.5321922942834796,
      "tokens_seen": 1561329664
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005323169508525577,
      "loss": 1.532,
      "theoretical_loss": 3.532179427265845,
      "tokens_seen": 1561395200
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000532296890672016,
      "loss": 1.5565,
      "theoretical_loss": 3.5321665609394732,
      "tokens_seen": 1561460736
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005322768304914745,
      "loss": 1.5162,
      "theoretical_loss": 3.5321536953042987,
      "tokens_seen": 1561526272
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005322567703109328,
      "loss": 1.5109,
      "theoretical_loss": 3.5321408303602544,
      "tokens_seen": 1561591808
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005322367101303912,
      "loss": 1.5449,
      "theoretical_loss": 3.5321279661072746,
      "tokens_seen": 1561657344
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005322166499498495,
      "loss": 1.4144,
      "theoretical_loss": 3.5321151025452933,
      "tokens_seen": 1561722880
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005321965897693079,
      "loss": 1.3818,
      "theoretical_loss": 3.532102239674245,
      "tokens_seen": 1561788416
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005321765295887663,
      "loss": 1.4641,
      "theoretical_loss": 3.532089377494062,
      "tokens_seen": 1561853952
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005321564694082247,
      "loss": 1.5522,
      "theoretical_loss": 3.5320765160046794,
      "tokens_seen": 1561919488
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000532136409227683,
      "loss": 1.5627,
      "theoretical_loss": 3.5320636552060307,
      "tokens_seen": 1561985024
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005321163490471415,
      "loss": 1.5673,
      "theoretical_loss": 3.5320507950980504,
      "tokens_seen": 1562050560
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005320962888665998,
      "loss": 1.5303,
      "theoretical_loss": 3.532037935680672,
      "tokens_seen": 1562116096
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005320762286860582,
      "loss": 1.4476,
      "theoretical_loss": 3.532025076953829,
      "tokens_seen": 1562181632
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005320561685055165,
      "loss": 1.4356,
      "theoretical_loss": 3.532012218917456,
      "tokens_seen": 1562247168
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005320361083249749,
      "loss": 1.527,
      "theoretical_loss": 3.5319993615714873,
      "tokens_seen": 1562312704
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005320160481444333,
      "loss": 1.482,
      "theoretical_loss": 3.5319865049158556,
      "tokens_seen": 1562378240
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005319959879638917,
      "loss": 1.4123,
      "theoretical_loss": 3.531973648950496,
      "tokens_seen": 1562443776
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00053197592778335,
      "loss": 1.4244,
      "theoretical_loss": 3.5319607936753417,
      "tokens_seen": 1562509312
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005319558676028085,
      "loss": 1.3823,
      "theoretical_loss": 3.531947939090327,
      "tokens_seen": 1562574848
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005319358074222667,
      "loss": 1.4576,
      "theoretical_loss": 3.5319350851953866,
      "tokens_seen": 1562640384
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005319157472417252,
      "loss": 1.4382,
      "theoretical_loss": 3.5319222319904533,
      "tokens_seen": 1562705920
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005318956870611836,
      "loss": 1.4814,
      "theoretical_loss": 3.5319093794754624,
      "tokens_seen": 1562771456
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005318756268806419,
      "loss": 1.51,
      "theoretical_loss": 3.5318965276503467,
      "tokens_seen": 1562836992
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005318555667001003,
      "loss": 1.3963,
      "theoretical_loss": 3.531883676515041,
      "tokens_seen": 1562902528
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005318355065195588,
      "loss": 1.4884,
      "theoretical_loss": 3.531870826069479,
      "tokens_seen": 1562968064
    },
    {
      "epoch": 0.47,
      "objective/train/advantage_avg": 0.48943787813186646,
      "objective/train/docs_used": 570864,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.19069242477417,
      "objective/train/original_loss": 1.1906923055648804,
      "objective/train/theoretical_loss": 3.5318579763135953,
      "objective/train/tokens_used": 1583493600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2402626872062683,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050165057182312,
      "objective/train/weighted_lm_loss": 1.2502503395080566,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5318579763135953,
      "tokens_seen": 1563033600
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000531815446339017,
      "loss": 1.4874,
      "theoretical_loss": 3.5318579763135953,
      "tokens_seen": 1563033600
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005317953861584755,
      "loss": 1.447,
      "theoretical_loss": 3.5318451272473235,
      "tokens_seen": 1563099136
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005317753259779337,
      "loss": 1.6113,
      "theoretical_loss": 3.5318322788705974,
      "tokens_seen": 1563164672
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005317552657973922,
      "loss": 1.4348,
      "theoretical_loss": 3.531819431183352,
      "tokens_seen": 1563230208
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005317352056168506,
      "loss": 1.3235,
      "theoretical_loss": 3.531806584185521,
      "tokens_seen": 1563295744
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005317151454363089,
      "loss": 1.4126,
      "theoretical_loss": 3.5317937378770377,
      "tokens_seen": 1563361280
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005316950852557673,
      "loss": 1.4488,
      "theoretical_loss": 3.531780892257837,
      "tokens_seen": 1563426816
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005316750250752256,
      "loss": 1.2679,
      "theoretical_loss": 3.531768047327853,
      "tokens_seen": 1563492352
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000531654964894684,
      "loss": 1.53,
      "theoretical_loss": 3.5317552030870205,
      "tokens_seen": 1563557888
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005316349047141425,
      "loss": 1.4434,
      "theoretical_loss": 3.531742359535272,
      "tokens_seen": 1563623424
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005316148445336007,
      "loss": 1.4708,
      "theoretical_loss": 3.5317295166725424,
      "tokens_seen": 1563688960
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005315947843530592,
      "loss": 1.448,
      "theoretical_loss": 3.5317166744987665,
      "tokens_seen": 1563754496
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005315747241725175,
      "loss": 1.4571,
      "theoretical_loss": 3.531703833013878,
      "tokens_seen": 1563820032
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005315546639919759,
      "loss": 1.4219,
      "theoretical_loss": 3.531690992217811,
      "tokens_seen": 1563885568
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005315346038114343,
      "loss": 1.3592,
      "theoretical_loss": 3.5316781521104996,
      "tokens_seen": 1563951104
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005315145436308927,
      "loss": 1.4048,
      "theoretical_loss": 3.531665312691878,
      "tokens_seen": 1564016640
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005314944834503511,
      "loss": 1.4475,
      "theoretical_loss": 3.5316524739618806,
      "tokens_seen": 1564082176
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005314744232698095,
      "loss": 1.4169,
      "theoretical_loss": 3.531639635920442,
      "tokens_seen": 1564147712
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005314543630892679,
      "loss": 1.4192,
      "theoretical_loss": 3.5316267985674954,
      "tokens_seen": 1564213248
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005314343029087262,
      "loss": 1.4958,
      "theoretical_loss": 3.531613961902976,
      "tokens_seen": 1564278784
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005314142427281846,
      "loss": 1.4225,
      "theoretical_loss": 3.5316011259268176,
      "tokens_seen": 1564344320
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005313941825476429,
      "loss": 1.37,
      "theoretical_loss": 3.5315882906389544,
      "tokens_seen": 1564409856
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005313741223671014,
      "loss": 1.4397,
      "theoretical_loss": 3.5315754560393207,
      "tokens_seen": 1564475392
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005313540621865597,
      "loss": 1.3696,
      "theoretical_loss": 3.5315626221278507,
      "tokens_seen": 1564540928
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005313340020060181,
      "loss": 1.4572,
      "theoretical_loss": 3.5315497889044796,
      "tokens_seen": 1564606464
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005313139418254764,
      "loss": 1.5149,
      "theoretical_loss": 3.5315369563691403,
      "tokens_seen": 1564672000
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005312938816449349,
      "loss": 1.3468,
      "theoretical_loss": 3.5315241245217672,
      "tokens_seen": 1564737536
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005312738214643932,
      "loss": 1.4414,
      "theoretical_loss": 3.531511293362296,
      "tokens_seen": 1564803072
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005312537612838516,
      "loss": 1.3984,
      "theoretical_loss": 3.53149846289066,
      "tokens_seen": 1564868608
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005312337011033099,
      "loss": 1.4264,
      "theoretical_loss": 3.531485633106793,
      "tokens_seen": 1564934144
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005312136409227683,
      "loss": 1.5095,
      "theoretical_loss": 3.5314728040106305,
      "tokens_seen": 1564999680
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005311935807422267,
      "loss": 1.5197,
      "theoretical_loss": 3.5314599756021066,
      "tokens_seen": 1565065216
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005311735205616851,
      "loss": 1.5642,
      "theoretical_loss": 3.5314471478811544,
      "tokens_seen": 1565130752
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005311534603811434,
      "loss": 1.4123,
      "theoretical_loss": 3.53143432084771,
      "tokens_seen": 1565196288
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005311334002006019,
      "loss": 1.541,
      "theoretical_loss": 3.531421494501707,
      "tokens_seen": 1565261824
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005311133400200602,
      "loss": 1.5465,
      "theoretical_loss": 3.5314086688430795,
      "tokens_seen": 1565327360
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005310932798395186,
      "loss": 1.4453,
      "theoretical_loss": 3.531395843871762,
      "tokens_seen": 1565392896
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000531073219658977,
      "loss": 1.3988,
      "theoretical_loss": 3.5313830195876896,
      "tokens_seen": 1565458432
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005310531594784353,
      "loss": 1.516,
      "theoretical_loss": 3.5313701959907955,
      "tokens_seen": 1565523968
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005310330992978937,
      "loss": 1.6218,
      "theoretical_loss": 3.5313573730810153,
      "tokens_seen": 1565589504
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005310130391173521,
      "loss": 1.4964,
      "theoretical_loss": 3.5313445508582824,
      "tokens_seen": 1565655040
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005309929789368104,
      "loss": 1.4069,
      "theoretical_loss": 3.5313317293225324,
      "tokens_seen": 1565720576
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005309729187562689,
      "loss": 1.3909,
      "theoretical_loss": 3.5313189084736987,
      "tokens_seen": 1565786112
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005309528585757271,
      "loss": 1.4649,
      "theoretical_loss": 3.531306088311716,
      "tokens_seen": 1565851648
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005309327983951856,
      "loss": 1.4975,
      "theoretical_loss": 3.531293268836519,
      "tokens_seen": 1565917184
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000530912738214644,
      "loss": 1.4015,
      "theoretical_loss": 3.531280450048042,
      "tokens_seen": 1565982720
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005308926780341023,
      "loss": 1.5482,
      "theoretical_loss": 3.5312676319462195,
      "tokens_seen": 1566048256
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005308726178535607,
      "loss": 1.4948,
      "theoretical_loss": 3.5312548145309863,
      "tokens_seen": 1566113792
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005308525576730192,
      "loss": 1.3929,
      "theoretical_loss": 3.531241997802276,
      "tokens_seen": 1566179328
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005308324974924774,
      "loss": 1.361,
      "theoretical_loss": 3.5312291817600245,
      "tokens_seen": 1566244864
    },
    {
      "epoch": 0.47,
      "objective/train/advantage_avg": 0.49122336506843567,
      "objective/train/docs_used": 572070,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3925796747207642,
      "objective/train/original_loss": 1.3925797939300537,
      "objective/train/theoretical_loss": 3.531216366404165,
      "objective/train/tokens_used": 1586770400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24178555607795715,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503513813018799,
      "objective/train/weighted_lm_loss": 1.4628849029541016,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.531216366404165,
      "tokens_seen": 1566310400
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005308124373119359,
      "loss": 1.4053,
      "theoretical_loss": 3.531216366404165,
      "tokens_seen": 1566310400
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005307923771313941,
      "loss": 1.4162,
      "theoretical_loss": 3.5312035517346327,
      "tokens_seen": 1566375936
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005307723169508526,
      "loss": 1.4359,
      "theoretical_loss": 3.531190737751362,
      "tokens_seen": 1566441472
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000530752256770311,
      "loss": 1.4533,
      "theoretical_loss": 3.531177924454287,
      "tokens_seen": 1566507008
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005307321965897693,
      "loss": 1.3303,
      "theoretical_loss": 3.531165111843343,
      "tokens_seen": 1566572544
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005307121364092277,
      "loss": 1.4174,
      "theoretical_loss": 3.5311522999184644,
      "tokens_seen": 1566638080
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.000530692076228686,
      "loss": 1.4724,
      "theoretical_loss": 3.5311394886795853,
      "tokens_seen": 1566703616
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005306720160481444,
      "loss": 1.4711,
      "theoretical_loss": 3.5311266781266406,
      "tokens_seen": 1566769152
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005306519558676029,
      "loss": 1.4978,
      "theoretical_loss": 3.531113868259565,
      "tokens_seen": 1566834688
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005306318956870611,
      "loss": 1.3532,
      "theoretical_loss": 3.531101059078293,
      "tokens_seen": 1566900224
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005306118355065196,
      "loss": 1.4496,
      "theoretical_loss": 3.5310882505827585,
      "tokens_seen": 1566965760
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005305917753259779,
      "loss": 1.4171,
      "theoretical_loss": 3.5310754427728974,
      "tokens_seen": 1567031296
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005305717151454363,
      "loss": 1.4514,
      "theoretical_loss": 3.5310626356486434,
      "tokens_seen": 1567096832
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005305516549648947,
      "loss": 1.4723,
      "theoretical_loss": 3.5310498292099317,
      "tokens_seen": 1567162368
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005305315947843531,
      "loss": 1.3736,
      "theoretical_loss": 3.5310370234566966,
      "tokens_seen": 1567227904
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005305115346038114,
      "loss": 1.4573,
      "theoretical_loss": 3.5310242183888723,
      "tokens_seen": 1567293440
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005304914744232699,
      "loss": 1.4026,
      "theoretical_loss": 3.5310114140063944,
      "tokens_seen": 1567358976
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005304714142427281,
      "loss": 1.4448,
      "theoretical_loss": 3.530998610309197,
      "tokens_seen": 1567424512
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0005304513540621866,
      "loss": 1.4019,
      "theoretical_loss": 3.5309858072972147,
      "tokens_seen": 1567490048
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005304312938816449,
      "loss": 1.568,
      "theoretical_loss": 3.5309730049703827,
      "tokens_seen": 1567555584
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005304112337011033,
      "loss": 1.3927,
      "theoretical_loss": 3.530960203328635,
      "tokens_seen": 1567621120
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005303911735205617,
      "loss": 1.4662,
      "theoretical_loss": 3.5309474023719067,
      "tokens_seen": 1567686656
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005303711133400201,
      "loss": 1.3869,
      "theoretical_loss": 3.5309346021001327,
      "tokens_seen": 1567752192
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005303510531594784,
      "loss": 1.4673,
      "theoretical_loss": 3.530921802513247,
      "tokens_seen": 1567817728
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005303309929789368,
      "loss": 1.492,
      "theoretical_loss": 3.5309090036111854,
      "tokens_seen": 1567883264
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005303109327983951,
      "loss": 1.4127,
      "theoretical_loss": 3.5308962053938817,
      "tokens_seen": 1567948800
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005302908726178536,
      "loss": 1.4265,
      "theoretical_loss": 3.530883407861271,
      "tokens_seen": 1568014336
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005302708124373119,
      "loss": 1.4655,
      "theoretical_loss": 3.530870611013288,
      "tokens_seen": 1568079872
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005302507522567703,
      "loss": 1.5157,
      "theoretical_loss": 3.5308578148498686,
      "tokens_seen": 1568145408
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005302306920762286,
      "loss": 1.5031,
      "theoretical_loss": 3.530845019370945,
      "tokens_seen": 1568210944
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005302106318956871,
      "loss": 1.5587,
      "theoretical_loss": 3.5308322245764545,
      "tokens_seen": 1568276480
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005301905717151454,
      "loss": 1.4021,
      "theoretical_loss": 3.5308194304663303,
      "tokens_seen": 1568342016
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005301705115346038,
      "loss": 1.3837,
      "theoretical_loss": 3.5308066370405076,
      "tokens_seen": 1568407552
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005301504513540622,
      "loss": 1.4159,
      "theoretical_loss": 3.5307938442989215,
      "tokens_seen": 1568473088
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005301303911735206,
      "loss": 1.4583,
      "theoretical_loss": 3.5307810522415073,
      "tokens_seen": 1568538624
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005301103309929789,
      "loss": 1.4709,
      "theoretical_loss": 3.5307682608681983,
      "tokens_seen": 1568604160
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005300902708124374,
      "loss": 1.3997,
      "theoretical_loss": 3.530755470178931,
      "tokens_seen": 1568669696
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005300702106318956,
      "loss": 1.4375,
      "theoretical_loss": 3.5307426801736392,
      "tokens_seen": 1568735232
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005300501504513541,
      "loss": 1.4643,
      "theoretical_loss": 3.5307298908522577,
      "tokens_seen": 1568800768
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005300300902708124,
      "loss": 1.4564,
      "theoretical_loss": 3.530717102214722,
      "tokens_seen": 1568866304
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005300100300902708,
      "loss": 1.3933,
      "theoretical_loss": 3.5307043142609666,
      "tokens_seen": 1568931840
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005299899699097293,
      "loss": 1.4358,
      "theoretical_loss": 3.530691526990927,
      "tokens_seen": 1568997376
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005299699097291875,
      "loss": 1.3582,
      "theoretical_loss": 3.5306787404045368,
      "tokens_seen": 1569062912
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000529949849548646,
      "loss": 1.5337,
      "theoretical_loss": 3.530665954501732,
      "tokens_seen": 1569128448
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005299297893681044,
      "loss": 1.4239,
      "theoretical_loss": 3.530653169282447,
      "tokens_seen": 1569193984
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005299097291875627,
      "loss": 1.3523,
      "theoretical_loss": 3.5306403847466172,
      "tokens_seen": 1569259520
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005298896690070211,
      "loss": 1.4975,
      "theoretical_loss": 3.530627600894177,
      "tokens_seen": 1569325056
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005298696088264796,
      "loss": 1.3806,
      "theoretical_loss": 3.5306148177250614,
      "tokens_seen": 1569390592
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005298495486459378,
      "loss": 1.4628,
      "theoretical_loss": 3.5306020352392053,
      "tokens_seen": 1569456128
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005298294884653963,
      "loss": 1.4123,
      "theoretical_loss": 3.5305892534365446,
      "tokens_seen": 1569521664
    },
    {
      "epoch": 0.48,
      "objective/train/advantage_avg": 0.49602967500686646,
      "objective/train/docs_used": 573246,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3148502111434937,
      "objective/train/original_loss": 1.314850091934204,
      "objective/train/theoretical_loss": 3.5305764723170125,
      "objective/train/tokens_used": 1590047200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24636462330818176,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508555173873901,
      "objective/train/weighted_lm_loss": 1.3817310333251953,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5305764723170125,
      "tokens_seen": 1569587200
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005298094282848545,
      "loss": 1.4751,
      "theoretical_loss": 3.5305764723170125,
      "tokens_seen": 1569587200
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000529789368104313,
      "loss": 1.4441,
      "theoretical_loss": 3.5305636918805456,
      "tokens_seen": 1569652736
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005297693079237714,
      "loss": 1.4612,
      "theoretical_loss": 3.5305509121270777,
      "tokens_seen": 1569718272
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005297492477432297,
      "loss": 1.3112,
      "theoretical_loss": 3.5305381330565444,
      "tokens_seen": 1569783808
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005297291875626881,
      "loss": 1.4424,
      "theoretical_loss": 3.5305253546688813,
      "tokens_seen": 1569849344
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005297091273821465,
      "loss": 1.3921,
      "theoretical_loss": 3.530512576964022,
      "tokens_seen": 1569914880
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005296890672016048,
      "loss": 1.3251,
      "theoretical_loss": 3.5304997999419028,
      "tokens_seen": 1569980416
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005296690070210633,
      "loss": 1.3606,
      "theoretical_loss": 3.5304870236024577,
      "tokens_seen": 1570045952
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005296489468405215,
      "loss": 1.5448,
      "theoretical_loss": 3.530474247945622,
      "tokens_seen": 1570111488
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00052962888665998,
      "loss": 1.3993,
      "theoretical_loss": 3.5304614729713313,
      "tokens_seen": 1570177024
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005296088264794383,
      "loss": 1.4183,
      "theoretical_loss": 3.5304486986795203,
      "tokens_seen": 1570242560
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005295887662988967,
      "loss": 1.5435,
      "theoretical_loss": 3.5304359250701234,
      "tokens_seen": 1570308096
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005295687061183551,
      "loss": 1.3975,
      "theoretical_loss": 3.530423152143077,
      "tokens_seen": 1570373632
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005295486459378135,
      "loss": 1.3726,
      "theoretical_loss": 3.5304103798983153,
      "tokens_seen": 1570439168
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005295285857572718,
      "loss": 1.4719,
      "theoretical_loss": 3.5303976083357735,
      "tokens_seen": 1570504704
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005295085255767303,
      "loss": 1.3163,
      "theoretical_loss": 3.5303848374553866,
      "tokens_seen": 1570570240
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005294884653961885,
      "loss": 1.3392,
      "theoretical_loss": 3.53037206725709,
      "tokens_seen": 1570635776
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000529468405215647,
      "loss": 1.3768,
      "theoretical_loss": 3.5303592977408185,
      "tokens_seen": 1570701312
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005294483450351053,
      "loss": 1.3107,
      "theoretical_loss": 3.5303465289065077,
      "tokens_seen": 1570766848
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005294282848545637,
      "loss": 1.396,
      "theoretical_loss": 3.5303337607540923,
      "tokens_seen": 1570832384
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005294082246740221,
      "loss": 1.3656,
      "theoretical_loss": 3.5303209932835076,
      "tokens_seen": 1570897920
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005293881644934805,
      "loss": 1.3874,
      "theoretical_loss": 3.530308226494688,
      "tokens_seen": 1570963456
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005293681043129388,
      "loss": 1.4177,
      "theoretical_loss": 3.5302954603875696,
      "tokens_seen": 1571028992
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005293480441323972,
      "loss": 1.4716,
      "theoretical_loss": 3.530282694962088,
      "tokens_seen": 1571094528
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005293279839518555,
      "loss": 1.5034,
      "theoretical_loss": 3.5302699302181773,
      "tokens_seen": 1571160064
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000529307923771314,
      "loss": 1.3505,
      "theoretical_loss": 3.530257166155773,
      "tokens_seen": 1571225600
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005292878635907723,
      "loss": 1.4136,
      "theoretical_loss": 3.5302444027748106,
      "tokens_seen": 1571291136
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005292678034102307,
      "loss": 1.4768,
      "theoretical_loss": 3.5302316400752245,
      "tokens_seen": 1571356672
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000529247743229689,
      "loss": 1.4344,
      "theoretical_loss": 3.5302188780569512,
      "tokens_seen": 1571422208
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005292276830491475,
      "loss": 1.4761,
      "theoretical_loss": 3.5302061167199246,
      "tokens_seen": 1571487744
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005292076228686058,
      "loss": 1.4188,
      "theoretical_loss": 3.530193356064081,
      "tokens_seen": 1571553280
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005291875626880642,
      "loss": 1.3352,
      "theoretical_loss": 3.5301805960893544,
      "tokens_seen": 1571618816
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005291675025075226,
      "loss": 1.3721,
      "theoretical_loss": 3.5301678367956812,
      "tokens_seen": 1571684352
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000529147442326981,
      "loss": 1.4487,
      "theoretical_loss": 3.5301550781829967,
      "tokens_seen": 1571749888
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005291273821464393,
      "loss": 1.5303,
      "theoretical_loss": 3.530142320251235,
      "tokens_seen": 1571815424
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005291073219658978,
      "loss": 1.492,
      "theoretical_loss": 3.5301295630003326,
      "tokens_seen": 1571880960
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000529087261785356,
      "loss": 1.4356,
      "theoretical_loss": 3.530116806430224,
      "tokens_seen": 1571946496
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005290672016048145,
      "loss": 1.4067,
      "theoretical_loss": 3.530104050540845,
      "tokens_seen": 1572012032
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005290471414242728,
      "loss": 1.3555,
      "theoretical_loss": 3.5300912953321304,
      "tokens_seen": 1572077568
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005290270812437312,
      "loss": 1.416,
      "theoretical_loss": 3.5300785408040154,
      "tokens_seen": 1572143104
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005290070210631896,
      "loss": 1.3115,
      "theoretical_loss": 3.5300657869564365,
      "tokens_seen": 1572208640
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005289869608826479,
      "loss": 1.5115,
      "theoretical_loss": 3.5300530337893274,
      "tokens_seen": 1572274176
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005289669007021063,
      "loss": 1.4274,
      "theoretical_loss": 3.530040281302625,
      "tokens_seen": 1572339712
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005289468405215648,
      "loss": 1.5862,
      "theoretical_loss": 3.5300275294962633,
      "tokens_seen": 1572405248
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000528926780341023,
      "loss": 1.4788,
      "theoretical_loss": 3.5300147783701785,
      "tokens_seen": 1572470784
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005289067201604815,
      "loss": 1.5714,
      "theoretical_loss": 3.530002027924305,
      "tokens_seen": 1572536320
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005288866599799397,
      "loss": 1.4558,
      "theoretical_loss": 3.5299892781585793,
      "tokens_seen": 1572601856
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005288665997993982,
      "loss": 1.4844,
      "theoretical_loss": 3.529976529072937,
      "tokens_seen": 1572667392
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005288465396188566,
      "loss": 1.4857,
      "theoretical_loss": 3.5299637806673116,
      "tokens_seen": 1572732928
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005288264794383149,
      "loss": 1.4449,
      "theoretical_loss": 3.5299510329416406,
      "tokens_seen": 1572798464
    },
    {
      "epoch": 0.48,
      "objective/train/advantage_avg": 0.4913088381290436,
      "objective/train/docs_used": 574423,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4549490213394165,
      "objective/train/original_loss": 1.454948902130127,
      "objective/train/theoretical_loss": 3.529938285895858,
      "objective/train/tokens_used": 1593324000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24197573959827423,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503610372543335,
      "objective/train/weighted_lm_loss": 1.5284897089004517,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.529938285895858,
      "tokens_seen": 1572864000
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005288064192577733,
      "loss": 1.5476,
      "theoretical_loss": 3.529938285895858,
      "tokens_seen": 1572864000
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005287863590772318,
      "loss": 1.3528,
      "theoretical_loss": 3.5299255395299,
      "tokens_seen": 1572929536
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00052876629889669,
      "loss": 1.351,
      "theoretical_loss": 3.5299127938437014,
      "tokens_seen": 1572995072
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005287462387161485,
      "loss": 1.469,
      "theoretical_loss": 3.5299000488371983,
      "tokens_seen": 1573060608
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005287261785356067,
      "loss": 1.3842,
      "theoretical_loss": 3.529887304510326,
      "tokens_seen": 1573126144
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005287061183550652,
      "loss": 1.399,
      "theoretical_loss": 3.5298745608630195,
      "tokens_seen": 1573191680
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005286860581745236,
      "loss": 1.3046,
      "theoretical_loss": 3.529861817895214,
      "tokens_seen": 1573257216
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005286659979939819,
      "loss": 1.3693,
      "theoretical_loss": 3.529849075606846,
      "tokens_seen": 1573322752
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005286459378134403,
      "loss": 1.4427,
      "theoretical_loss": 3.5298363339978502,
      "tokens_seen": 1573388288
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005286258776328987,
      "loss": 1.4932,
      "theoretical_loss": 3.529823593068163,
      "tokens_seen": 1573453824
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000528605817452357,
      "loss": 1.4335,
      "theoretical_loss": 3.529810852817718,
      "tokens_seen": 1573519360
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005285857572718155,
      "loss": 1.2839,
      "theoretical_loss": 3.529798113246453,
      "tokens_seen": 1573584896
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005285656970912737,
      "loss": 1.355,
      "theoretical_loss": 3.529785374354302,
      "tokens_seen": 1573650432
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005285456369107322,
      "loss": 1.391,
      "theoretical_loss": 3.529772636141201,
      "tokens_seen": 1573715968
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005285255767301907,
      "loss": 1.6312,
      "theoretical_loss": 3.529759898607086,
      "tokens_seen": 1573781504
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005285055165496489,
      "loss": 1.3402,
      "theoretical_loss": 3.5297471617518914,
      "tokens_seen": 1573847040
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005284854563691074,
      "loss": 1.3266,
      "theoretical_loss": 3.529734425575553,
      "tokens_seen": 1573912576
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005284653961885657,
      "loss": 1.4385,
      "theoretical_loss": 3.5297216900780075,
      "tokens_seen": 1573978112
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005284453360080241,
      "loss": 1.5045,
      "theoretical_loss": 3.5297089552591894,
      "tokens_seen": 1574043648
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005284252758274825,
      "loss": 1.4131,
      "theoretical_loss": 3.529696221119034,
      "tokens_seen": 1574109184
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005284052156469409,
      "loss": 1.3194,
      "theoretical_loss": 3.5296834876574783,
      "tokens_seen": 1574174720
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005283851554663992,
      "loss": 1.4485,
      "theoretical_loss": 3.5296707548744566,
      "tokens_seen": 1574240256
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005283650952858576,
      "loss": 1.4431,
      "theoretical_loss": 3.5296580227699046,
      "tokens_seen": 1574305792
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000528345035105316,
      "loss": 1.4013,
      "theoretical_loss": 3.5296452913437584,
      "tokens_seen": 1574371328
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005283249749247744,
      "loss": 1.4846,
      "theoretical_loss": 3.529632560595954,
      "tokens_seen": 1574436864
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005283049147442327,
      "loss": 1.3408,
      "theoretical_loss": 3.5296198305264257,
      "tokens_seen": 1574502400
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005282848545636911,
      "loss": 1.4688,
      "theoretical_loss": 3.5296071011351104,
      "tokens_seen": 1574567936
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005282647943831494,
      "loss": 1.2875,
      "theoretical_loss": 3.529594372421943,
      "tokens_seen": 1574633472
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005282447342026079,
      "loss": 1.4956,
      "theoretical_loss": 3.5295816443868593,
      "tokens_seen": 1574699008
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005282246740220662,
      "loss": 1.2044,
      "theoretical_loss": 3.529568917029795,
      "tokens_seen": 1574764544
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005282046138415246,
      "loss": 1.2398,
      "theoretical_loss": 3.529556190350686,
      "tokens_seen": 1574830080
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000528184553660983,
      "loss": 1.4164,
      "theoretical_loss": 3.5295434643494676,
      "tokens_seen": 1574895616
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005281644934804414,
      "loss": 1.4644,
      "theoretical_loss": 3.529530739026076,
      "tokens_seen": 1574961152
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005281444332998997,
      "loss": 1.4691,
      "theoretical_loss": 3.5295180143804465,
      "tokens_seen": 1575026688
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005281243731193582,
      "loss": 1.3798,
      "theoretical_loss": 3.5295052904125144,
      "tokens_seen": 1575092224
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005281043129388164,
      "loss": 1.3162,
      "theoretical_loss": 3.529492567122216,
      "tokens_seen": 1575157760
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005280842527582749,
      "loss": 1.4287,
      "theoretical_loss": 3.5294798445094875,
      "tokens_seen": 1575223296
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005280641925777332,
      "loss": 1.4852,
      "theoretical_loss": 3.529467122574263,
      "tokens_seen": 1575288832
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005280441323971916,
      "loss": 1.546,
      "theoretical_loss": 3.52945440131648,
      "tokens_seen": 1575354368
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00052802407221665,
      "loss": 1.3467,
      "theoretical_loss": 3.5294416807360736,
      "tokens_seen": 1575419904
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005280040120361083,
      "loss": 1.4404,
      "theoretical_loss": 3.529428960832979,
      "tokens_seen": 1575485440
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005279839518555667,
      "loss": 1.3251,
      "theoretical_loss": 3.529416241607133,
      "tokens_seen": 1575550976
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005279638916750252,
      "loss": 1.4172,
      "theoretical_loss": 3.52940352305847,
      "tokens_seen": 1575616512
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005279438314944834,
      "loss": 1.4041,
      "theoretical_loss": 3.529390805186927,
      "tokens_seen": 1575682048
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005279237713139419,
      "loss": 1.4614,
      "theoretical_loss": 3.5293780879924395,
      "tokens_seen": 1575747584
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005279037111334001,
      "loss": 1.2463,
      "theoretical_loss": 3.529365371474943,
      "tokens_seen": 1575813120
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005278836509528586,
      "loss": 1.4067,
      "theoretical_loss": 3.5293526556343737,
      "tokens_seen": 1575878656
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000527863590772317,
      "loss": 1.438,
      "theoretical_loss": 3.529339940470667,
      "tokens_seen": 1575944192
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005278435305917753,
      "loss": 1.4093,
      "theoretical_loss": 3.529327225983759,
      "tokens_seen": 1576009728
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005278234704112337,
      "loss": 1.468,
      "theoretical_loss": 3.529314512173585,
      "tokens_seen": 1576075264
    },
    {
      "epoch": 0.48,
      "objective/train/advantage_avg": 0.49186986684799194,
      "objective/train/docs_used": 575737,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.209204077720642,
      "objective/train/original_loss": 1.209204077720642,
      "objective/train/theoretical_loss": 3.529301799040082,
      "objective/train/tokens_used": 1596600800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24241523444652557,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504193305969238,
      "objective/train/weighted_lm_loss": 1.270138144493103,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.529301799040082,
      "tokens_seen": 1576140800
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005278034102306922,
      "loss": 1.3366,
      "theoretical_loss": 3.529301799040082,
      "tokens_seen": 1576140800
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005277833500501504,
      "loss": 1.5105,
      "theoretical_loss": 3.5292890865831845,
      "tokens_seen": 1576206336
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005277632898696089,
      "loss": 1.3317,
      "theoretical_loss": 3.5292763748028295,
      "tokens_seen": 1576271872
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005277432296890671,
      "loss": 1.3583,
      "theoretical_loss": 3.5292636636989525,
      "tokens_seen": 1576337408
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005277231695085256,
      "loss": 1.457,
      "theoretical_loss": 3.5292509532714886,
      "tokens_seen": 1576402944
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000527703109327984,
      "loss": 1.4393,
      "theoretical_loss": 3.529238243520375,
      "tokens_seen": 1576468480
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005276830491474423,
      "loss": 1.3963,
      "theoretical_loss": 3.5292255344455468,
      "tokens_seen": 1576534016
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005276629889669007,
      "loss": 1.3952,
      "theoretical_loss": 3.5292128260469395,
      "tokens_seen": 1576599552
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005276429287863591,
      "loss": 1.4037,
      "theoretical_loss": 3.52920011832449,
      "tokens_seen": 1576665088
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005276228686058174,
      "loss": 1.3746,
      "theoretical_loss": 3.5291874112781336,
      "tokens_seen": 1576730624
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005276028084252759,
      "loss": 1.4205,
      "theoretical_loss": 3.529174704907807,
      "tokens_seen": 1576796160
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005275827482447341,
      "loss": 1.4829,
      "theoretical_loss": 3.529161999213445,
      "tokens_seen": 1576861696
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005275626880641926,
      "loss": 1.3451,
      "theoretical_loss": 3.5291492941949842,
      "tokens_seen": 1576927232
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000527542627883651,
      "loss": 1.4454,
      "theoretical_loss": 3.52913658985236,
      "tokens_seen": 1576992768
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005275225677031093,
      "loss": 1.4242,
      "theoretical_loss": 3.529123886185509,
      "tokens_seen": 1577058304
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005275025075225677,
      "loss": 1.3811,
      "theoretical_loss": 3.5291111831943676,
      "tokens_seen": 1577123840
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005274824473420261,
      "loss": 1.3822,
      "theoretical_loss": 3.5290984808788712,
      "tokens_seen": 1577189376
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005274623871614844,
      "loss": 1.4063,
      "theoretical_loss": 3.5290857792389554,
      "tokens_seen": 1577254912
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005274423269809429,
      "loss": 1.3636,
      "theoretical_loss": 3.529073078274556,
      "tokens_seen": 1577320448
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005274222668004012,
      "loss": 1.3769,
      "theoretical_loss": 3.5290603779856102,
      "tokens_seen": 1577385984
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005274022066198596,
      "loss": 1.3798,
      "theoretical_loss": 3.5290476783720535,
      "tokens_seen": 1577451520
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005273821464393179,
      "loss": 1.3902,
      "theoretical_loss": 3.5290349794338214,
      "tokens_seen": 1577517056
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005273620862587764,
      "loss": 1.4772,
      "theoretical_loss": 3.5290222811708505,
      "tokens_seen": 1577582592
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005273420260782347,
      "loss": 1.4401,
      "theoretical_loss": 3.5290095835830773,
      "tokens_seen": 1577648128
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005273219658976931,
      "loss": 1.505,
      "theoretical_loss": 3.5289968866704364,
      "tokens_seen": 1577713664
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005273019057171514,
      "loss": 1.459,
      "theoretical_loss": 3.528984190432865,
      "tokens_seen": 1577779200
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005272818455366098,
      "loss": 1.3991,
      "theoretical_loss": 3.528971494870299,
      "tokens_seen": 1577844736
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005272617853560682,
      "loss": 1.4208,
      "theoretical_loss": 3.528958799982674,
      "tokens_seen": 1577910272
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005272417251755266,
      "loss": 1.2946,
      "theoretical_loss": 3.5289461057699265,
      "tokens_seen": 1577975808
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005272216649949849,
      "loss": 1.4294,
      "theoretical_loss": 3.528933412231993,
      "tokens_seen": 1578041344
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005272016048144434,
      "loss": 1.3477,
      "theoretical_loss": 3.5289207193688092,
      "tokens_seen": 1578106880
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005271815446339017,
      "loss": 1.4033,
      "theoretical_loss": 3.528908027180311,
      "tokens_seen": 1578172416
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005271614844533601,
      "loss": 1.4142,
      "theoretical_loss": 3.528895335666434,
      "tokens_seen": 1578237952
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005271414242728184,
      "loss": 1.3354,
      "theoretical_loss": 3.5288826448271156,
      "tokens_seen": 1578303488
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005271213640922768,
      "loss": 1.3974,
      "theoretical_loss": 3.5288699546622913,
      "tokens_seen": 1578369024
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005271013039117352,
      "loss": 1.4652,
      "theoretical_loss": 3.5288572651718972,
      "tokens_seen": 1578434560
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005270812437311936,
      "loss": 1.4639,
      "theoretical_loss": 3.52884457635587,
      "tokens_seen": 1578500096
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005270611835506519,
      "loss": 1.3809,
      "theoretical_loss": 3.528831888214145,
      "tokens_seen": 1578565632
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005270411233701104,
      "loss": 1.486,
      "theoretical_loss": 3.528819200746659,
      "tokens_seen": 1578631168
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005270210631895687,
      "loss": 1.4333,
      "theoretical_loss": 3.5288065139533478,
      "tokens_seen": 1578696704
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005270010030090271,
      "loss": 1.478,
      "theoretical_loss": 3.5287938278341473,
      "tokens_seen": 1578762240
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005269809428284856,
      "loss": 1.4403,
      "theoretical_loss": 3.528781142388995,
      "tokens_seen": 1578827776
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005269608826479438,
      "loss": 1.5068,
      "theoretical_loss": 3.5287684576178258,
      "tokens_seen": 1578893312
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005269408224674023,
      "loss": 1.4248,
      "theoretical_loss": 3.528755773520577,
      "tokens_seen": 1578958848
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005269207622868605,
      "loss": 1.4688,
      "theoretical_loss": 3.5287430900971835,
      "tokens_seen": 1579024384
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000526900702106319,
      "loss": 1.5235,
      "theoretical_loss": 3.5287304073475827,
      "tokens_seen": 1579089920
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005268806419257774,
      "loss": 1.445,
      "theoretical_loss": 3.5287177252717097,
      "tokens_seen": 1579155456
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005268605817452357,
      "loss": 1.4139,
      "theoretical_loss": 3.5287050438695022,
      "tokens_seen": 1579220992
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005268405215646941,
      "loss": 1.4623,
      "theoretical_loss": 3.5286923631408955,
      "tokens_seen": 1579286528
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005268204613841526,
      "loss": 1.4547,
      "theoretical_loss": 3.5286796830858265,
      "tokens_seen": 1579352064
    },
    {
      "epoch": 0.48,
      "objective/train/advantage_avg": 0.49563488364219666,
      "objective/train/docs_used": 576827,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3918602466583252,
      "objective/train/original_loss": 1.3918602466583252,
      "objective/train/theoretical_loss": 3.5286670037042303,
      "objective/train/tokens_used": 1599877600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24598407745361328,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050814151763916,
      "objective/train/weighted_lm_loss": 1.4624632596969604,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5286670037042303,
      "tokens_seen": 1579417600
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005268004012036108,
      "loss": 1.4762,
      "theoretical_loss": 3.5286670037042303,
      "tokens_seen": 1579417600
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005267803410230693,
      "loss": 1.5081,
      "theoretical_loss": 3.5286543249960447,
      "tokens_seen": 1579483136
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005267602808425275,
      "loss": 1.3477,
      "theoretical_loss": 3.5286416469612045,
      "tokens_seen": 1579548672
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000526740220661986,
      "loss": 1.3564,
      "theoretical_loss": 3.528628969599647,
      "tokens_seen": 1579614208
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005267201604814444,
      "loss": 1.4335,
      "theoretical_loss": 3.528616292911309,
      "tokens_seen": 1579679744
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005267001003009027,
      "loss": 1.4321,
      "theoretical_loss": 3.5286036168961252,
      "tokens_seen": 1579745280
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005266800401203611,
      "loss": 1.3012,
      "theoretical_loss": 3.5285909415540333,
      "tokens_seen": 1579810816
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005266599799398195,
      "loss": 1.5698,
      "theoretical_loss": 3.528578266884969,
      "tokens_seen": 1579876352
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005266399197592778,
      "loss": 1.4873,
      "theoretical_loss": 3.5285655928888686,
      "tokens_seen": 1579941888
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005266198595787363,
      "loss": 1.4049,
      "theoretical_loss": 3.528552919565669,
      "tokens_seen": 1580007424
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005265997993981945,
      "loss": 1.4337,
      "theoretical_loss": 3.528540246915306,
      "tokens_seen": 1580072960
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000526579739217653,
      "loss": 1.4731,
      "theoretical_loss": 3.528527574937716,
      "tokens_seen": 1580138496
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005265596790371114,
      "loss": 1.3883,
      "theoretical_loss": 3.5285149036328356,
      "tokens_seen": 1580204032
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005265396188565697,
      "loss": 1.4244,
      "theoretical_loss": 3.5285022330006015,
      "tokens_seen": 1580269568
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005265195586760281,
      "loss": 1.473,
      "theoretical_loss": 3.5284895630409503,
      "tokens_seen": 1580335104
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005264994984954865,
      "loss": 1.3669,
      "theoretical_loss": 3.528476893753817,
      "tokens_seen": 1580400640
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005264794383149448,
      "loss": 1.4924,
      "theoretical_loss": 3.528464225139139,
      "tokens_seen": 1580466176
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005264593781344033,
      "loss": 1.3962,
      "theoretical_loss": 3.528451557196853,
      "tokens_seen": 1580531712
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005264393179538616,
      "loss": 1.6332,
      "theoretical_loss": 3.528438889926895,
      "tokens_seen": 1580597248
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00052641925777332,
      "loss": 1.5788,
      "theoretical_loss": 3.528426223329201,
      "tokens_seen": 1580662784
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005263991975927783,
      "loss": 1.4479,
      "theoretical_loss": 3.5284135574037085,
      "tokens_seen": 1580728320
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005263791374122368,
      "loss": 1.4274,
      "theoretical_loss": 3.528400892150353,
      "tokens_seen": 1580793856
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005263590772316951,
      "loss": 1.4659,
      "theoretical_loss": 3.528388227569072,
      "tokens_seen": 1580859392
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005263390170511535,
      "loss": 1.4956,
      "theoretical_loss": 3.528375563659801,
      "tokens_seen": 1580924928
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005263189568706118,
      "loss": 1.3945,
      "theoretical_loss": 3.5283629004224766,
      "tokens_seen": 1580990464
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005262988966900702,
      "loss": 1.4743,
      "theoretical_loss": 3.528350237857036,
      "tokens_seen": 1581056000
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005262788365095286,
      "loss": 1.3968,
      "theoretical_loss": 3.5283375759634152,
      "tokens_seen": 1581121536
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000526258776328987,
      "loss": 1.5738,
      "theoretical_loss": 3.5283249147415505,
      "tokens_seen": 1581187072
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005262387161484453,
      "loss": 1.4258,
      "theoretical_loss": 3.5283122541913787,
      "tokens_seen": 1581252608
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005262186559679038,
      "loss": 1.3888,
      "theoretical_loss": 3.5282995943128364,
      "tokens_seen": 1581318144
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005261985957873621,
      "loss": 1.4169,
      "theoretical_loss": 3.52828693510586,
      "tokens_seen": 1581383680
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005261785356068205,
      "loss": 1.3957,
      "theoretical_loss": 3.528274276570386,
      "tokens_seen": 1581449216
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005261584754262788,
      "loss": 1.4764,
      "theoretical_loss": 3.5282616187063516,
      "tokens_seen": 1581514752
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005261384152457372,
      "loss": 1.3607,
      "theoretical_loss": 3.5282489615136923,
      "tokens_seen": 1581580288
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005261183550651956,
      "loss": 1.4915,
      "theoretical_loss": 3.528236304992345,
      "tokens_seen": 1581645824
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000526098294884654,
      "loss": 1.4494,
      "theoretical_loss": 3.5282236491422467,
      "tokens_seen": 1581711360
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005260782347041123,
      "loss": 1.3871,
      "theoretical_loss": 3.528210993963334,
      "tokens_seen": 1581776896
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005260581745235708,
      "loss": 1.4669,
      "theoretical_loss": 3.528198339455543,
      "tokens_seen": 1581842432
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000526038114343029,
      "loss": 1.4488,
      "theoretical_loss": 3.5281856856188103,
      "tokens_seen": 1581907968
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005260180541624875,
      "loss": 1.5659,
      "theoretical_loss": 3.5281730324530733,
      "tokens_seen": 1581973504
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005259979939819458,
      "loss": 1.4232,
      "theoretical_loss": 3.528160379958268,
      "tokens_seen": 1582039040
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005259779338014042,
      "loss": 1.488,
      "theoretical_loss": 3.528147728134331,
      "tokens_seen": 1582104576
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005259578736208626,
      "loss": 1.4,
      "theoretical_loss": 3.528135076981199,
      "tokens_seen": 1582170112
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005259378134403209,
      "loss": 1.4114,
      "theoretical_loss": 3.528122426498809,
      "tokens_seen": 1582235648
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005259177532597793,
      "loss": 1.5465,
      "theoretical_loss": 3.528109776687097,
      "tokens_seen": 1582301184
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005258976930792378,
      "loss": 1.5293,
      "theoretical_loss": 3.5280971275460002,
      "tokens_seen": 1582366720
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000525877632898696,
      "loss": 1.4126,
      "theoretical_loss": 3.5280844790754555,
      "tokens_seen": 1582432256
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005258575727181545,
      "loss": 1.4559,
      "theoretical_loss": 3.5280718312753985,
      "tokens_seen": 1582497792
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005258375125376129,
      "loss": 1.5653,
      "theoretical_loss": 3.528059184145767,
      "tokens_seen": 1582563328
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005258174523570712,
      "loss": 1.5058,
      "theoretical_loss": 3.528046537686498,
      "tokens_seen": 1582628864
    },
    {
      "epoch": 0.48,
      "objective/train/advantage_avg": 0.49373170733451843,
      "objective/train/docs_used": 578056,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.204235315322876,
      "objective/train/original_loss": 1.204235315322876,
      "objective/train/theoretical_loss": 3.5280338918975263,
      "objective/train/tokens_used": 1603154400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2440899759531021,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506141185760498,
      "objective/train/weighted_lm_loss": 1.2652686834335327,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5280338918975263,
      "tokens_seen": 1582694400
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005257973921765296,
      "loss": 1.3284,
      "theoretical_loss": 3.5280338918975263,
      "tokens_seen": 1582694400
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005257773319959879,
      "loss": 1.5169,
      "theoretical_loss": 3.52802124677879,
      "tokens_seen": 1582759936
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005257572718154463,
      "loss": 1.4684,
      "theoretical_loss": 3.5280086023302264,
      "tokens_seen": 1582825472
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005257372116349048,
      "loss": 1.4102,
      "theoretical_loss": 3.527995958551771,
      "tokens_seen": 1582891008
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000525717151454363,
      "loss": 1.4719,
      "theoretical_loss": 3.527983315443361,
      "tokens_seen": 1582956544
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005256970912738215,
      "loss": 1.3999,
      "theoretical_loss": 3.5279706730049334,
      "tokens_seen": 1583022080
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005256770310932798,
      "loss": 1.3954,
      "theoretical_loss": 3.527958031236425,
      "tokens_seen": 1583087616
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005256569709127382,
      "loss": 1.3913,
      "theoretical_loss": 3.5279453901377718,
      "tokens_seen": 1583153152
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005256369107321966,
      "loss": 1.3681,
      "theoretical_loss": 3.5279327497089117,
      "tokens_seen": 1583218688
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000525616850551655,
      "loss": 1.4794,
      "theoretical_loss": 3.527920109949781,
      "tokens_seen": 1583284224
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005255967903711133,
      "loss": 1.5032,
      "theoretical_loss": 3.527907470860315,
      "tokens_seen": 1583349760
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005255767301905718,
      "loss": 1.4594,
      "theoretical_loss": 3.5278948324404533,
      "tokens_seen": 1583415296
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00052555667001003,
      "loss": 1.5006,
      "theoretical_loss": 3.5278821946901306,
      "tokens_seen": 1583480832
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005255366098294885,
      "loss": 1.4909,
      "theoretical_loss": 3.527869557609285,
      "tokens_seen": 1583546368
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005255165496489469,
      "loss": 1.4408,
      "theoretical_loss": 3.527856921197852,
      "tokens_seen": 1583611904
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005254964894684052,
      "loss": 1.4497,
      "theoretical_loss": 3.52784428545577,
      "tokens_seen": 1583677440
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005254764292878637,
      "loss": 1.4338,
      "theoretical_loss": 3.527831650382975,
      "tokens_seen": 1583742976
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000525456369107322,
      "loss": 1.4973,
      "theoretical_loss": 3.527819015979403,
      "tokens_seen": 1583808512
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005254363089267804,
      "loss": 1.4711,
      "theoretical_loss": 3.5278063822449925,
      "tokens_seen": 1583874048
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005254162487462387,
      "loss": 1.4658,
      "theoretical_loss": 3.5277937491796796,
      "tokens_seen": 1583939584
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005253961885656972,
      "loss": 1.4786,
      "theoretical_loss": 3.5277811167834012,
      "tokens_seen": 1584005120
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005253761283851555,
      "loss": 1.4051,
      "theoretical_loss": 3.527768485056094,
      "tokens_seen": 1584070656
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005253560682046139,
      "loss": 1.4509,
      "theoretical_loss": 3.5277558539976956,
      "tokens_seen": 1584136192
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005253360080240722,
      "loss": 1.5335,
      "theoretical_loss": 3.527743223608142,
      "tokens_seen": 1584201728
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005253159478435306,
      "loss": 1.3127,
      "theoretical_loss": 3.5277305938873704,
      "tokens_seen": 1584267264
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000525295887662989,
      "loss": 1.4223,
      "theoretical_loss": 3.5277179648353183,
      "tokens_seen": 1584332800
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005252758274824474,
      "loss": 1.5416,
      "theoretical_loss": 3.5277053364519215,
      "tokens_seen": 1584398336
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005252557673019057,
      "loss": 1.4275,
      "theoretical_loss": 3.5276927087371184,
      "tokens_seen": 1584463872
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005252357071213642,
      "loss": 1.4322,
      "theoretical_loss": 3.527680081690845,
      "tokens_seen": 1584529408
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005252156469408225,
      "loss": 1.4707,
      "theoretical_loss": 3.527667455313038,
      "tokens_seen": 1584594944
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005251955867602809,
      "loss": 1.532,
      "theoretical_loss": 3.5276548296036356,
      "tokens_seen": 1584660480
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005251755265797392,
      "loss": 1.4126,
      "theoretical_loss": 3.5276422045625733,
      "tokens_seen": 1584726016
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005251554663991976,
      "loss": 1.4097,
      "theoretical_loss": 3.5276295801897892,
      "tokens_seen": 1584791552
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000525135406218656,
      "loss": 1.4708,
      "theoretical_loss": 3.5276169564852196,
      "tokens_seen": 1584857088
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005251153460381144,
      "loss": 1.3719,
      "theoretical_loss": 3.527604333448802,
      "tokens_seen": 1584922624
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005250952858575727,
      "loss": 1.4515,
      "theoretical_loss": 3.527591711080473,
      "tokens_seen": 1584988160
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005250752256770312,
      "loss": 1.4363,
      "theoretical_loss": 3.5275790893801697,
      "tokens_seen": 1585053696
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005250551654964894,
      "loss": 1.4025,
      "theoretical_loss": 3.527566468347829,
      "tokens_seen": 1585119232
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005250351053159479,
      "loss": 1.3907,
      "theoretical_loss": 3.527553847983389,
      "tokens_seen": 1585184768
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005250150451354063,
      "loss": 1.3557,
      "theoretical_loss": 3.5275412282867853,
      "tokens_seen": 1585250304
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005249949849548646,
      "loss": 1.5644,
      "theoretical_loss": 3.5275286092579554,
      "tokens_seen": 1585315840
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000524974924774323,
      "loss": 1.5808,
      "theoretical_loss": 3.527515990896837,
      "tokens_seen": 1585381376
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005249548645937813,
      "loss": 1.3072,
      "theoretical_loss": 3.5275033732033667,
      "tokens_seen": 1585446912
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005249348044132397,
      "loss": 1.3407,
      "theoretical_loss": 3.5274907561774813,
      "tokens_seen": 1585512448
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005249147442326982,
      "loss": 1.36,
      "theoretical_loss": 3.5274781398191184,
      "tokens_seen": 1585577984
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005248946840521564,
      "loss": 1.4515,
      "theoretical_loss": 3.527465524128215,
      "tokens_seen": 1585643520
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005248746238716149,
      "loss": 1.4183,
      "theoretical_loss": 3.5274529091047073,
      "tokens_seen": 1585709056
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005248545636910733,
      "loss": 1.4444,
      "theoretical_loss": 3.5274402947485335,
      "tokens_seen": 1585774592
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005248345035105316,
      "loss": 1.4857,
      "theoretical_loss": 3.5274276810596303,
      "tokens_seen": 1585840128
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00052481444332999,
      "loss": 1.3334,
      "theoretical_loss": 3.5274150680379357,
      "tokens_seen": 1585905664
    },
    {
      "epoch": 0.48,
      "objective/train/advantage_avg": 0.4924222230911255,
      "objective/train/docs_used": 579278,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2216386795043945,
      "objective/train/original_loss": 1.2216386795043945,
      "objective/train/theoretical_loss": 3.527402455683385,
      "objective/train/tokens_used": 1606431200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2428509145975113,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504767894744873,
      "objective/train/weighted_lm_loss": 1.2835732698440552,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.527402455683385,
      "tokens_seen": 1585971200
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005247943831494483,
      "loss": 1.3517,
      "theoretical_loss": 3.527402455683385,
      "tokens_seen": 1585971200
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005247743229689067,
      "loss": 1.443,
      "theoretical_loss": 3.527389843995917,
      "tokens_seen": 1586036736
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005247542627883652,
      "loss": 1.4817,
      "theoretical_loss": 3.527377232975468,
      "tokens_seen": 1586102272
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005247342026078234,
      "loss": 1.5019,
      "theoretical_loss": 3.5273646226219757,
      "tokens_seen": 1586167808
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005247141424272819,
      "loss": 1.3304,
      "theoretical_loss": 3.527352012935377,
      "tokens_seen": 1586233344
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005246940822467402,
      "loss": 1.5629,
      "theoretical_loss": 3.527339403915609,
      "tokens_seen": 1586298880
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005246740220661986,
      "loss": 1.3846,
      "theoretical_loss": 3.5273267955626086,
      "tokens_seen": 1586364416
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000524653961885657,
      "loss": 1.4189,
      "theoretical_loss": 3.5273141878763137,
      "tokens_seen": 1586429952
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005246339017051153,
      "loss": 1.3943,
      "theoretical_loss": 3.527301580856661,
      "tokens_seen": 1586495488
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005246138415245737,
      "loss": 1.5483,
      "theoretical_loss": 3.527288974503588,
      "tokens_seen": 1586561024
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005245937813440322,
      "loss": 1.3518,
      "theoretical_loss": 3.5272763688170317,
      "tokens_seen": 1586626560
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005245737211634904,
      "loss": 1.3916,
      "theoretical_loss": 3.5272637637969297,
      "tokens_seen": 1586692096
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005245536609829489,
      "loss": 1.5098,
      "theoretical_loss": 3.527251159443219,
      "tokens_seen": 1586757632
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005245336008024072,
      "loss": 1.4011,
      "theoretical_loss": 3.5272385557558366,
      "tokens_seen": 1586823168
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005245135406218656,
      "loss": 1.535,
      "theoretical_loss": 3.52722595273472,
      "tokens_seen": 1586888704
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000524493480441324,
      "loss": 1.4962,
      "theoretical_loss": 3.5272133503798067,
      "tokens_seen": 1586954240
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005244734202607824,
      "loss": 1.531,
      "theoretical_loss": 3.5272007486910333,
      "tokens_seen": 1587019776
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005244533600802407,
      "loss": 1.5407,
      "theoretical_loss": 3.5271881476683378,
      "tokens_seen": 1587085312
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005244332998996991,
      "loss": 1.3993,
      "theoretical_loss": 3.527175547311657,
      "tokens_seen": 1587150848
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005244132397191574,
      "loss": 1.4315,
      "theoretical_loss": 3.5271629476209285,
      "tokens_seen": 1587216384
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005243931795386159,
      "loss": 1.4251,
      "theoretical_loss": 3.5271503485960896,
      "tokens_seen": 1587281920
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005243731193580742,
      "loss": 1.4574,
      "theoretical_loss": 3.5271377502370775,
      "tokens_seen": 1587347456
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005243530591775326,
      "loss": 1.4282,
      "theoretical_loss": 3.527125152543829,
      "tokens_seen": 1587412992
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005243329989969909,
      "loss": 1.4313,
      "theoretical_loss": 3.5271125555162826,
      "tokens_seen": 1587478528
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005243129388164494,
      "loss": 1.4375,
      "theoretical_loss": 3.5270999591543752,
      "tokens_seen": 1587544064
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005242928786359077,
      "loss": 1.4089,
      "theoretical_loss": 3.5270873634580435,
      "tokens_seen": 1587609600
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005242728184553661,
      "loss": 1.4792,
      "theoretical_loss": 3.5270747684272257,
      "tokens_seen": 1587675136
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005242527582748244,
      "loss": 1.3545,
      "theoretical_loss": 3.527062174061858,
      "tokens_seen": 1587740672
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005242326980942829,
      "loss": 1.4261,
      "theoretical_loss": 3.5270495803618793,
      "tokens_seen": 1587806208
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005242126379137412,
      "loss": 1.356,
      "theoretical_loss": 3.5270369873272265,
      "tokens_seen": 1587871744
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005241925777331996,
      "loss": 1.3338,
      "theoretical_loss": 3.527024394957836,
      "tokens_seen": 1587937280
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005241725175526579,
      "loss": 1.3515,
      "theoretical_loss": 3.5270118032536466,
      "tokens_seen": 1588002816
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005241524573721164,
      "loss": 1.3812,
      "theoretical_loss": 3.5269992122145952,
      "tokens_seen": 1588068352
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005241323971915747,
      "loss": 1.4938,
      "theoretical_loss": 3.5269866218406185,
      "tokens_seen": 1588133888
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005241123370110331,
      "loss": 1.4366,
      "theoretical_loss": 3.5269740321316547,
      "tokens_seen": 1588199424
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005240922768304915,
      "loss": 1.3248,
      "theoretical_loss": 3.5269614430876413,
      "tokens_seen": 1588264960
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005240722166499498,
      "loss": 1.3622,
      "theoretical_loss": 3.526948854708515,
      "tokens_seen": 1588330496
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005240521564694082,
      "loss": 1.4253,
      "theoretical_loss": 3.5269362669942144,
      "tokens_seen": 1588396032
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005240320962888667,
      "loss": 1.4063,
      "theoretical_loss": 3.5269236799446757,
      "tokens_seen": 1588461568
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000524012036108325,
      "loss": 1.3929,
      "theoretical_loss": 3.5269110935598373,
      "tokens_seen": 1588527104
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005239919759277834,
      "loss": 1.4611,
      "theoretical_loss": 3.526898507839636,
      "tokens_seen": 1588592640
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005239719157472417,
      "loss": 1.3477,
      "theoretical_loss": 3.52688592278401,
      "tokens_seen": 1588658176
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005239518555667001,
      "loss": 1.4003,
      "theoretical_loss": 3.526873338392896,
      "tokens_seen": 1588723712
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005239317953861586,
      "loss": 1.3992,
      "theoretical_loss": 3.5268607546662327,
      "tokens_seen": 1588789248
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005239117352056168,
      "loss": 1.4186,
      "theoretical_loss": 3.526848171603956,
      "tokens_seen": 1588854784
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005238916750250753,
      "loss": 1.4544,
      "theoretical_loss": 3.526835589206005,
      "tokens_seen": 1588920320
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005238716148445337,
      "loss": 1.4194,
      "theoretical_loss": 3.5268230074723164,
      "tokens_seen": 1588985856
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000523851554663992,
      "loss": 1.4479,
      "theoretical_loss": 3.5268104264028275,
      "tokens_seen": 1589051392
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005238314944834504,
      "loss": 1.512,
      "theoretical_loss": 3.526797845997476,
      "tokens_seen": 1589116928
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005238114343029087,
      "loss": 1.4631,
      "theoretical_loss": 3.5267852662562,
      "tokens_seen": 1589182464
    },
    {
      "epoch": 0.48,
      "objective/train/advantage_avg": 0.49218177795410156,
      "objective/train/docs_used": 580493,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3156033754348755,
      "objective/train/original_loss": 1.3156033754348755,
      "objective/train/theoretical_loss": 3.5267726871789367,
      "objective/train/tokens_used": 1609708000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24304485321044922,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050453782081604,
      "objective/train/weighted_lm_loss": 1.3820679187774658,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5267726871789367,
      "tokens_seen": 1589248000
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005237913741223671,
      "loss": 1.355,
      "theoretical_loss": 3.5267726871789367,
      "tokens_seen": 1589248000
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005237713139418256,
      "loss": 1.3529,
      "theoretical_loss": 3.5267601087656235,
      "tokens_seen": 1589313536
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005237512537612838,
      "loss": 1.3714,
      "theoretical_loss": 3.5267475310161984,
      "tokens_seen": 1589379072
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005237311935807423,
      "loss": 1.4619,
      "theoretical_loss": 3.5267349539305988,
      "tokens_seen": 1589444608
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005237111334002006,
      "loss": 1.377,
      "theoretical_loss": 3.5267223775087615,
      "tokens_seen": 1589510144
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000523691073219659,
      "loss": 1.5891,
      "theoretical_loss": 3.5267098017506253,
      "tokens_seen": 1589575680
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005236710130391174,
      "loss": 1.4602,
      "theoretical_loss": 3.5266972266561276,
      "tokens_seen": 1589641216
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005236509528585757,
      "loss": 1.4423,
      "theoretical_loss": 3.5266846522252058,
      "tokens_seen": 1589706752
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005236308926780341,
      "loss": 1.3507,
      "theoretical_loss": 3.526672078457797,
      "tokens_seen": 1589772288
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005236108324974926,
      "loss": 1.4197,
      "theoretical_loss": 3.5266595053538397,
      "tokens_seen": 1589837824
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005235907723169508,
      "loss": 1.3578,
      "theoretical_loss": 3.526646932913271,
      "tokens_seen": 1589903360
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005235707121364093,
      "loss": 1.329,
      "theoretical_loss": 3.526634361136029,
      "tokens_seen": 1589968896
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005235506519558676,
      "loss": 1.4151,
      "theoretical_loss": 3.526621790022051,
      "tokens_seen": 1590034432
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000523530591775326,
      "loss": 1.3674,
      "theoretical_loss": 3.5266092195712746,
      "tokens_seen": 1590099968
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005235105315947844,
      "loss": 1.4174,
      "theoretical_loss": 3.5265966497836376,
      "tokens_seen": 1590165504
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005234904714142428,
      "loss": 1.4983,
      "theoretical_loss": 3.526584080659078,
      "tokens_seen": 1590231040
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005234704112337011,
      "loss": 1.4856,
      "theoretical_loss": 3.526571512197533,
      "tokens_seen": 1590296576
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005234503510531595,
      "loss": 1.4579,
      "theoretical_loss": 3.526558944398941,
      "tokens_seen": 1590362112
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005234302908726178,
      "loss": 1.5735,
      "theoretical_loss": 3.5265463772632386,
      "tokens_seen": 1590427648
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005234102306920763,
      "loss": 1.3877,
      "theoretical_loss": 3.5265338107903643,
      "tokens_seen": 1590493184
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005233901705115346,
      "loss": 1.4195,
      "theoretical_loss": 3.5265212449802563,
      "tokens_seen": 1590558720
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000523370110330993,
      "loss": 1.4663,
      "theoretical_loss": 3.526508679832851,
      "tokens_seen": 1590624256
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005233500501504513,
      "loss": 1.3634,
      "theoretical_loss": 3.5264961153480874,
      "tokens_seen": 1590689792
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005233299899699098,
      "loss": 1.4097,
      "theoretical_loss": 3.526483551525903,
      "tokens_seen": 1590755328
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005233099297893681,
      "loss": 1.4533,
      "theoretical_loss": 3.526470988366235,
      "tokens_seen": 1590820864
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005232898696088265,
      "loss": 1.5101,
      "theoretical_loss": 3.5264584258690213,
      "tokens_seen": 1590886400
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005232698094282848,
      "loss": 1.4684,
      "theoretical_loss": 3.5264458640342,
      "tokens_seen": 1590951936
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005232497492477433,
      "loss": 1.3989,
      "theoretical_loss": 3.526433302861709,
      "tokens_seen": 1591017472
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005232296890672016,
      "loss": 1.5437,
      "theoretical_loss": 3.526420742351485,
      "tokens_seen": 1591083008
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00052320962888666,
      "loss": 1.3679,
      "theoretical_loss": 3.5264081825034674,
      "tokens_seen": 1591148544
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005231895687061183,
      "loss": 1.4688,
      "theoretical_loss": 3.5263956233175935,
      "tokens_seen": 1591214080
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005231695085255768,
      "loss": 1.4092,
      "theoretical_loss": 3.5263830647938006,
      "tokens_seen": 1591279616
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005231494483450351,
      "loss": 1.3763,
      "theoretical_loss": 3.526370506932027,
      "tokens_seen": 1591345152
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005231293881644935,
      "loss": 1.3098,
      "theoretical_loss": 3.52635794973221,
      "tokens_seen": 1591410688
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005231093279839519,
      "loss": 1.436,
      "theoretical_loss": 3.5263453931942883,
      "tokens_seen": 1591476224
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005230892678034102,
      "loss": 1.2847,
      "theoretical_loss": 3.526332837318199,
      "tokens_seen": 1591541760
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005230692076228686,
      "loss": 1.5116,
      "theoretical_loss": 3.5263202821038804,
      "tokens_seen": 1591607296
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000523049147442327,
      "loss": 1.3129,
      "theoretical_loss": 3.52630772755127,
      "tokens_seen": 1591672832
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005230290872617853,
      "loss": 1.4481,
      "theoretical_loss": 3.5262951736603063,
      "tokens_seen": 1591738368
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005230090270812438,
      "loss": 1.4164,
      "theoretical_loss": 3.5262826204309263,
      "tokens_seen": 1591803904
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000522988966900702,
      "loss": 1.3281,
      "theoretical_loss": 3.5262700678630687,
      "tokens_seen": 1591869440
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005229689067201605,
      "loss": 1.5257,
      "theoretical_loss": 3.5262575159566714,
      "tokens_seen": 1591934976
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005229488465396189,
      "loss": 1.4303,
      "theoretical_loss": 3.5262449647116716,
      "tokens_seen": 1592000512
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005229287863590772,
      "loss": 1.3065,
      "theoretical_loss": 3.5262324141280077,
      "tokens_seen": 1592066048
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005229087261785356,
      "loss": 1.4045,
      "theoretical_loss": 3.5262198642056175,
      "tokens_seen": 1592131584
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005228886659979941,
      "loss": 1.4212,
      "theoretical_loss": 3.526207314944439,
      "tokens_seen": 1592197120
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005228686058174523,
      "loss": 1.3585,
      "theoretical_loss": 3.5261947663444104,
      "tokens_seen": 1592262656
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005228485456369108,
      "loss": 1.3742,
      "theoretical_loss": 3.526182218405469,
      "tokens_seen": 1592328192
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000522828485456369,
      "loss": 1.3845,
      "theoretical_loss": 3.526169671127554,
      "tokens_seen": 1592393728
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005228084252758275,
      "loss": 1.3604,
      "theoretical_loss": 3.526157124510602,
      "tokens_seen": 1592459264
    },
    {
      "epoch": 0.48,
      "objective/train/advantage_avg": 0.49332740902900696,
      "objective/train/docs_used": 581545,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4762202501296997,
      "objective/train/original_loss": 1.4762201309204102,
      "objective/train/theoretical_loss": 3.5261445785545513,
      "objective/train/tokens_used": 1612984800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24372516572475433,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505717992782593,
      "objective/train/weighted_lm_loss": 1.5509356260299683,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5261445785545513,
      "tokens_seen": 1592524800
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005227883650952859,
      "loss": 1.4919,
      "theoretical_loss": 3.5261445785545513,
      "tokens_seen": 1592524800
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005227683049147442,
      "loss": 1.401,
      "theoretical_loss": 3.5261320332593407,
      "tokens_seen": 1592590336
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005227482447342026,
      "loss": 1.3224,
      "theoretical_loss": 3.526119488624907,
      "tokens_seen": 1592655872
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000522728184553661,
      "loss": 1.4466,
      "theoretical_loss": 3.5261069446511897,
      "tokens_seen": 1592721408
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005227081243731193,
      "loss": 1.5393,
      "theoretical_loss": 3.526094401338125,
      "tokens_seen": 1592786944
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005226880641925778,
      "loss": 1.4447,
      "theoretical_loss": 3.526081858685653,
      "tokens_seen": 1592852480
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000522668004012036,
      "loss": 1.4709,
      "theoretical_loss": 3.52606931669371,
      "tokens_seen": 1592918016
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005226479438314945,
      "loss": 1.4827,
      "theoretical_loss": 3.5260567753622345,
      "tokens_seen": 1592983552
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005226278836509528,
      "loss": 1.3595,
      "theoretical_loss": 3.526044234691165,
      "tokens_seen": 1593049088
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005226078234704112,
      "loss": 1.4834,
      "theoretical_loss": 3.526031694680439,
      "tokens_seen": 1593114624
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005225877632898696,
      "loss": 1.3724,
      "theoretical_loss": 3.526019155329995,
      "tokens_seen": 1593180160
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000522567703109328,
      "loss": 1.3538,
      "theoretical_loss": 3.526006616639771,
      "tokens_seen": 1593245696
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005225476429287863,
      "loss": 1.4189,
      "theoretical_loss": 3.525994078609705,
      "tokens_seen": 1593311232
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005225275827482448,
      "loss": 1.4216,
      "theoretical_loss": 3.5259815412397346,
      "tokens_seen": 1593376768
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005225075225677032,
      "loss": 1.4441,
      "theoretical_loss": 3.525969004529799,
      "tokens_seen": 1593442304
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005224874623871615,
      "loss": 1.3875,
      "theoretical_loss": 3.5259564684798352,
      "tokens_seen": 1593507840
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005224674022066199,
      "loss": 1.5061,
      "theoretical_loss": 3.525943933089782,
      "tokens_seen": 1593573376
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005224473420260782,
      "loss": 1.5082,
      "theoretical_loss": 3.5259313983595777,
      "tokens_seen": 1593638912
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005224272818455367,
      "loss": 1.4807,
      "theoretical_loss": 3.52591886428916,
      "tokens_seen": 1593704448
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000522407221664995,
      "loss": 1.4738,
      "theoretical_loss": 3.525906330878467,
      "tokens_seen": 1593769984
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005223871614844534,
      "loss": 1.474,
      "theoretical_loss": 3.5258937981274365,
      "tokens_seen": 1593835520
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005223671013039117,
      "loss": 1.4511,
      "theoretical_loss": 3.5258812660360075,
      "tokens_seen": 1593901056
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005223470411233702,
      "loss": 1.3322,
      "theoretical_loss": 3.525868734604118,
      "tokens_seen": 1593966592
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005223269809428285,
      "loss": 1.4404,
      "theoretical_loss": 3.5258562038317054,
      "tokens_seen": 1594032128
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005223069207622869,
      "loss": 1.4769,
      "theoretical_loss": 3.5258436737187084,
      "tokens_seen": 1594097664
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005222868605817452,
      "loss": 1.469,
      "theoretical_loss": 3.525831144265066,
      "tokens_seen": 1594163200
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005222668004012037,
      "loss": 1.3429,
      "theoretical_loss": 3.5258186154707154,
      "tokens_seen": 1594228736
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000522246740220662,
      "loss": 1.4304,
      "theoretical_loss": 3.5258060873355945,
      "tokens_seen": 1594294272
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005222266800401204,
      "loss": 1.3075,
      "theoretical_loss": 3.5257935598596424,
      "tokens_seen": 1594359808
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005222066198595787,
      "loss": 1.3962,
      "theoretical_loss": 3.525781033042797,
      "tokens_seen": 1594425344
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005221865596790372,
      "loss": 1.3921,
      "theoretical_loss": 3.5257685068849964,
      "tokens_seen": 1594490880
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005221664994984955,
      "loss": 1.4178,
      "theoretical_loss": 3.5257559813861796,
      "tokens_seen": 1594556416
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005221464393179539,
      "loss": 1.3855,
      "theoretical_loss": 3.5257434565462833,
      "tokens_seen": 1594621952
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005221263791374123,
      "loss": 1.3785,
      "theoretical_loss": 3.525730932365247,
      "tokens_seen": 1594687488
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005221063189568706,
      "loss": 1.4882,
      "theoretical_loss": 3.5257184088430087,
      "tokens_seen": 1594753024
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000522086258776329,
      "loss": 1.4579,
      "theoretical_loss": 3.525705885979507,
      "tokens_seen": 1594818560
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005220661985957875,
      "loss": 1.4465,
      "theoretical_loss": 3.5256933637746792,
      "tokens_seen": 1594884096
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005220461384152457,
      "loss": 1.4046,
      "theoretical_loss": 3.5256808422284642,
      "tokens_seen": 1594949632
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005220260782347042,
      "loss": 1.3431,
      "theoretical_loss": 3.5256683213408007,
      "tokens_seen": 1595015168
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005220060180541624,
      "loss": 1.2065,
      "theoretical_loss": 3.5256558011116264,
      "tokens_seen": 1595080704
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005219859578736209,
      "loss": 1.421,
      "theoretical_loss": 3.5256432815408796,
      "tokens_seen": 1595146240
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005219658976930793,
      "loss": 1.3896,
      "theoretical_loss": 3.525630762628499,
      "tokens_seen": 1595211776
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005219458375125376,
      "loss": 1.3093,
      "theoretical_loss": 3.5256182443744226,
      "tokens_seen": 1595277312
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000521925777331996,
      "loss": 1.3301,
      "theoretical_loss": 3.525605726778589,
      "tokens_seen": 1595342848
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005219057171514545,
      "loss": 1.4331,
      "theoretical_loss": 3.5255932098409364,
      "tokens_seen": 1595408384
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005218856569709127,
      "loss": 1.4744,
      "theoretical_loss": 3.5255806935614036,
      "tokens_seen": 1595473920
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005218655967903712,
      "loss": 1.4659,
      "theoretical_loss": 3.5255681779399284,
      "tokens_seen": 1595539456
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005218455366098294,
      "loss": 1.3844,
      "theoretical_loss": 3.525555662976449,
      "tokens_seen": 1595604992
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005218254764292879,
      "loss": 1.3053,
      "theoretical_loss": 3.5255431486709043,
      "tokens_seen": 1595670528
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005218054162487463,
      "loss": 1.3202,
      "theoretical_loss": 3.525530635023233,
      "tokens_seen": 1595736064
    },
    {
      "epoch": 0.48,
      "objective/train/advantage_avg": 0.4902195632457733,
      "objective/train/docs_used": 582715,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3351233005523682,
      "objective/train/original_loss": 1.3351234197616577,
      "objective/train/theoretical_loss": 3.5255181220333722,
      "objective/train/tokens_used": 1616261600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2414616048336029,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050249457359314,
      "objective/train/weighted_lm_loss": 1.402158260345459,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5255181220333722,
      "tokens_seen": 1595801600
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005217853560682046,
      "loss": 1.3337,
      "theoretical_loss": 3.5255181220333722,
      "tokens_seen": 1595801600
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000521765295887663,
      "loss": 1.3897,
      "theoretical_loss": 3.525505609701262,
      "tokens_seen": 1595867136
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005217452357071214,
      "loss": 1.3296,
      "theoretical_loss": 3.525493098026839,
      "tokens_seen": 1595932672
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005217251755265797,
      "loss": 1.3738,
      "theoretical_loss": 3.525480587010043,
      "tokens_seen": 1595998208
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005217051153460382,
      "loss": 1.3888,
      "theoretical_loss": 3.5254680766508124,
      "tokens_seen": 1596063744
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005216850551654964,
      "loss": 1.5225,
      "theoretical_loss": 3.525455566949085,
      "tokens_seen": 1596129280
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005216649949849549,
      "loss": 1.4447,
      "theoretical_loss": 3.5254430579047993,
      "tokens_seen": 1596194816
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005216449348044132,
      "loss": 1.3851,
      "theoretical_loss": 3.5254305495178944,
      "tokens_seen": 1596260352
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005216248746238716,
      "loss": 1.4152,
      "theoretical_loss": 3.525418041788308,
      "tokens_seen": 1596325888
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00052160481444333,
      "loss": 1.3252,
      "theoretical_loss": 3.525405534715979,
      "tokens_seen": 1596391424
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005215847542627884,
      "loss": 1.3155,
      "theoretical_loss": 3.5253930283008454,
      "tokens_seen": 1596456960
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005215646940822467,
      "loss": 1.4092,
      "theoretical_loss": 3.5253805225428465,
      "tokens_seen": 1596522496
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005215446339017052,
      "loss": 1.4881,
      "theoretical_loss": 3.52536801744192,
      "tokens_seen": 1596588032
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005215245737211634,
      "loss": 1.3747,
      "theoretical_loss": 3.525355512998005,
      "tokens_seen": 1596653568
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005215045135406219,
      "loss": 1.4322,
      "theoretical_loss": 3.5253430092110403,
      "tokens_seen": 1596719104
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005214844533600802,
      "loss": 1.3691,
      "theoretical_loss": 3.525330506080963,
      "tokens_seen": 1596784640
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005214643931795386,
      "loss": 1.4526,
      "theoretical_loss": 3.5253180036077136,
      "tokens_seen": 1596850176
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000521444332998997,
      "loss": 1.3541,
      "theoretical_loss": 3.5253055017912285,
      "tokens_seen": 1596915712
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005214242728184554,
      "loss": 1.3632,
      "theoretical_loss": 3.5252930006314482,
      "tokens_seen": 1596981248
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005214042126379137,
      "loss": 1.4537,
      "theoretical_loss": 3.52528050012831,
      "tokens_seen": 1597046784
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005213841524573721,
      "loss": 1.4105,
      "theoretical_loss": 3.525268000281753,
      "tokens_seen": 1597112320
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005213640922768305,
      "loss": 1.4,
      "theoretical_loss": 3.5252555010917153,
      "tokens_seen": 1597177856
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005213440320962889,
      "loss": 1.3559,
      "theoretical_loss": 3.5252430025581356,
      "tokens_seen": 1597243392
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005213239719157472,
      "loss": 1.3359,
      "theoretical_loss": 3.5252305046809536,
      "tokens_seen": 1597308928
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005213039117352056,
      "loss": 1.3315,
      "theoretical_loss": 3.525218007460106,
      "tokens_seen": 1597374464
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005212838515546639,
      "loss": 1.4641,
      "theoretical_loss": 3.525205510895533,
      "tokens_seen": 1597440000
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005212637913741224,
      "loss": 1.4945,
      "theoretical_loss": 3.5251930149871726,
      "tokens_seen": 1597505536
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005212437311935807,
      "loss": 1.3704,
      "theoretical_loss": 3.5251805197349633,
      "tokens_seen": 1597571072
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005212236710130391,
      "loss": 1.3652,
      "theoretical_loss": 3.525168025138844,
      "tokens_seen": 1597636608
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005212036108324975,
      "loss": 1.3396,
      "theoretical_loss": 3.5251555311987532,
      "tokens_seen": 1597702144
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005211835506519559,
      "loss": 1.497,
      "theoretical_loss": 3.5251430379146296,
      "tokens_seen": 1597767680
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005211634904714142,
      "loss": 1.3877,
      "theoretical_loss": 3.525130545286412,
      "tokens_seen": 1597833216
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005211434302908727,
      "loss": 1.3913,
      "theoretical_loss": 3.5251180533140385,
      "tokens_seen": 1597898752
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005211233701103309,
      "loss": 1.4174,
      "theoretical_loss": 3.5251055619974485,
      "tokens_seen": 1597964288
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005211033099297894,
      "loss": 1.3834,
      "theoretical_loss": 3.5250930713365802,
      "tokens_seen": 1598029824
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005210832497492477,
      "loss": 1.3811,
      "theoretical_loss": 3.5250805813313724,
      "tokens_seen": 1598095360
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005210631895687061,
      "loss": 1.3376,
      "theoretical_loss": 3.5250680919817636,
      "tokens_seen": 1598160896
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005210431293881645,
      "loss": 1.3614,
      "theoretical_loss": 3.5250556032876927,
      "tokens_seen": 1598226432
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005210230692076228,
      "loss": 1.3378,
      "theoretical_loss": 3.525043115249099,
      "tokens_seen": 1598291968
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005210030090270813,
      "loss": 1.4622,
      "theoretical_loss": 3.5250306278659203,
      "tokens_seen": 1598357504
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005209829488465397,
      "loss": 1.4905,
      "theoretical_loss": 3.5250181411380956,
      "tokens_seen": 1598423040
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000520962888665998,
      "loss": 1.3968,
      "theoretical_loss": 3.5250056550655637,
      "tokens_seen": 1598488576
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005209428284854564,
      "loss": 1.4316,
      "theoretical_loss": 3.5249931696482637,
      "tokens_seen": 1598554112
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005209227683049149,
      "loss": 1.4562,
      "theoretical_loss": 3.524980684886134,
      "tokens_seen": 1598619648
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005209027081243731,
      "loss": 1.4234,
      "theoretical_loss": 3.524968200779113,
      "tokens_seen": 1598685184
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005208826479438316,
      "loss": 1.471,
      "theoretical_loss": 3.52495571732714,
      "tokens_seen": 1598750720
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005208625877632898,
      "loss": 1.4172,
      "theoretical_loss": 3.524943234530154,
      "tokens_seen": 1598816256
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005208425275827483,
      "loss": 1.5247,
      "theoretical_loss": 3.5249307523880935,
      "tokens_seen": 1598881792
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005208224674022067,
      "loss": 1.3178,
      "theoretical_loss": 3.524918270900897,
      "tokens_seen": 1598947328
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000520802407221665,
      "loss": 1.4248,
      "theoretical_loss": 3.5249057900685035,
      "tokens_seen": 1599012864
    },
    {
      "epoch": 0.48,
      "objective/train/advantage_avg": 0.49344751238822937,
      "objective/train/docs_used": 583998,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4556032419204712,
      "objective/train/original_loss": 1.4556032419204712,
      "objective/train/theoretical_loss": 3.5248933098908517,
      "objective/train/tokens_used": 1619538400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24378977715969086,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050584077835083,
      "objective/train/weighted_lm_loss": 1.5293455123901367,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5248933098908517,
      "tokens_seen": 1599078400
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005207823470411234,
      "loss": 1.4285,
      "theoretical_loss": 3.5248933098908517,
      "tokens_seen": 1599078400
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005207622868605818,
      "loss": 1.3538,
      "theoretical_loss": 3.5248808303678807,
      "tokens_seen": 1599143936
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005207422266800401,
      "loss": 1.3689,
      "theoretical_loss": 3.5248683514995296,
      "tokens_seen": 1599209472
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005207221664994986,
      "loss": 1.537,
      "theoretical_loss": 3.5248558732857362,
      "tokens_seen": 1599275008
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005207021063189568,
      "loss": 1.3758,
      "theoretical_loss": 3.5248433957264402,
      "tokens_seen": 1599340544
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005206820461384153,
      "loss": 1.4661,
      "theoretical_loss": 3.5248309188215803,
      "tokens_seen": 1599406080
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005206619859578736,
      "loss": 1.4703,
      "theoretical_loss": 3.524818442571095,
      "tokens_seen": 1599471616
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000520641925777332,
      "loss": 1.4113,
      "theoretical_loss": 3.524805966974924,
      "tokens_seen": 1599537152
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005206218655967904,
      "loss": 1.4728,
      "theoretical_loss": 3.5247934920330053,
      "tokens_seen": 1599602688
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005206018054162488,
      "loss": 1.4267,
      "theoretical_loss": 3.5247810177452785,
      "tokens_seen": 1599668224
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005205817452357071,
      "loss": 1.4469,
      "theoretical_loss": 3.524768544111682,
      "tokens_seen": 1599733760
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005205616850551656,
      "loss": 1.4255,
      "theoretical_loss": 3.5247560711321544,
      "tokens_seen": 1599799296
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005205416248746238,
      "loss": 1.5092,
      "theoretical_loss": 3.5247435988066353,
      "tokens_seen": 1599864832
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005205215646940823,
      "loss": 1.5292,
      "theoretical_loss": 3.5247311271350634,
      "tokens_seen": 1599930368
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005205015045135406,
      "loss": 1.4129,
      "theoretical_loss": 3.5247186561173782,
      "tokens_seen": 1599995904
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000520481444332999,
      "loss": 1.3974,
      "theoretical_loss": 3.524706185753517,
      "tokens_seen": 1600061440
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005204613841524574,
      "loss": 1.4105,
      "theoretical_loss": 3.52469371604342,
      "tokens_seen": 1600126976
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005204413239719158,
      "loss": 1.508,
      "theoretical_loss": 3.5246812469870266,
      "tokens_seen": 1600192512
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005204212637913741,
      "loss": 1.475,
      "theoretical_loss": 3.5246687785842745,
      "tokens_seen": 1600258048
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005204012036108325,
      "loss": 1.4034,
      "theoretical_loss": 3.5246563108351037,
      "tokens_seen": 1600323584
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005203811434302909,
      "loss": 1.3485,
      "theoretical_loss": 3.524643843739452,
      "tokens_seen": 1600389120
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0005203610832497493,
      "loss": 1.549,
      "theoretical_loss": 3.5246313772972595,
      "tokens_seen": 1600454656
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005203410230692076,
      "loss": 1.3749,
      "theoretical_loss": 3.524618911508465,
      "tokens_seen": 1600520192
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000520320962888666,
      "loss": 1.5466,
      "theoretical_loss": 3.524606446373007,
      "tokens_seen": 1600585728
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005203009027081243,
      "loss": 1.5416,
      "theoretical_loss": 3.524593981890825,
      "tokens_seen": 1600651264
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005202808425275828,
      "loss": 1.3547,
      "theoretical_loss": 3.524581518061858,
      "tokens_seen": 1600716800
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005202607823470411,
      "loss": 1.3806,
      "theoretical_loss": 3.5245690548860447,
      "tokens_seen": 1600782336
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005202407221664995,
      "loss": 1.4574,
      "theoretical_loss": 3.524556592363324,
      "tokens_seen": 1600847872
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005202206619859579,
      "loss": 1.5123,
      "theoretical_loss": 3.524544130493635,
      "tokens_seen": 1600913408
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005202006018054163,
      "loss": 1.4163,
      "theoretical_loss": 3.524531669276918,
      "tokens_seen": 1600978944
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005201805416248746,
      "loss": 1.4388,
      "theoretical_loss": 3.5245192087131105,
      "tokens_seen": 1601044480
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005201604814443331,
      "loss": 1.4572,
      "theoretical_loss": 3.524506748802152,
      "tokens_seen": 1601110016
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005201404212637913,
      "loss": 1.3739,
      "theoretical_loss": 3.5244942895439815,
      "tokens_seen": 1601175552
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005201203610832498,
      "loss": 1.4003,
      "theoretical_loss": 3.524481830938539,
      "tokens_seen": 1601241088
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005201003009027081,
      "loss": 1.5313,
      "theoretical_loss": 3.524469372985762,
      "tokens_seen": 1601306624
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005200802407221665,
      "loss": 1.2918,
      "theoretical_loss": 3.524456915685591,
      "tokens_seen": 1601372160
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005200601805416249,
      "loss": 1.4845,
      "theoretical_loss": 3.524444459037965,
      "tokens_seen": 1601437696
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005200401203610832,
      "loss": 1.2939,
      "theoretical_loss": 3.524432003042822,
      "tokens_seen": 1601503232
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005200200601805416,
      "loss": 1.3571,
      "theoretical_loss": 3.524419547700102,
      "tokens_seen": 1601568768
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005200000000000001,
      "loss": 1.5091,
      "theoretical_loss": 3.524407093009744,
      "tokens_seen": 1601634304
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005199799398194583,
      "loss": 1.5134,
      "theoretical_loss": 3.5243946389716867,
      "tokens_seen": 1601699840
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005199598796389168,
      "loss": 1.4344,
      "theoretical_loss": 3.52438218558587,
      "tokens_seen": 1601765376
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005199398194583751,
      "loss": 1.4169,
      "theoretical_loss": 3.5243697328522323,
      "tokens_seen": 1601830912
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005199197592778335,
      "loss": 1.4684,
      "theoretical_loss": 3.5243572807707135,
      "tokens_seen": 1601896448
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005198996990972919,
      "loss": 1.5119,
      "theoretical_loss": 3.5243448293412527,
      "tokens_seen": 1601961984
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005198796389167502,
      "loss": 1.4664,
      "theoretical_loss": 3.5243323785637886,
      "tokens_seen": 1602027520
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005198595787362086,
      "loss": 1.375,
      "theoretical_loss": 3.5243199284382603,
      "tokens_seen": 1602093056
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005198395185556671,
      "loss": 1.5863,
      "theoretical_loss": 3.524307478964608,
      "tokens_seen": 1602158592
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005198194583751253,
      "loss": 1.3999,
      "theoretical_loss": 3.5242950301427696,
      "tokens_seen": 1602224128
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005197993981945838,
      "loss": 1.3664,
      "theoretical_loss": 3.5242825819726846,
      "tokens_seen": 1602289664
    },
    {
      "epoch": 0.49,
      "objective/train/advantage_avg": 0.4948594570159912,
      "objective/train/docs_used": 585126,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3824971914291382,
      "objective/train/original_loss": 1.3824970722198486,
      "objective/train/theoretical_loss": 3.524270134454293,
      "objective/train/tokens_used": 1622815200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24526368081569672,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507328510284424,
      "objective/train/weighted_lm_loss": 1.4526634216308594,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.524270134454293,
      "tokens_seen": 1602355200
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000519779338014042,
      "loss": 1.4365,
      "theoretical_loss": 3.524270134454293,
      "tokens_seen": 1602355200
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005197592778335005,
      "loss": 1.4346,
      "theoretical_loss": 3.5242576875875335,
      "tokens_seen": 1602420736
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005197392176529589,
      "loss": 1.3437,
      "theoretical_loss": 3.5242452413723457,
      "tokens_seen": 1602486272
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005197191574724172,
      "loss": 1.428,
      "theoretical_loss": 3.5242327958086683,
      "tokens_seen": 1602551808
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005196990972918756,
      "loss": 1.4268,
      "theoretical_loss": 3.5242203508964405,
      "tokens_seen": 1602617344
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000519679037111334,
      "loss": 1.397,
      "theoretical_loss": 3.5242079066356022,
      "tokens_seen": 1602682880
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005196589769307923,
      "loss": 1.4505,
      "theoretical_loss": 3.5241954630260923,
      "tokens_seen": 1602748416
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005196389167502508,
      "loss": 1.4111,
      "theoretical_loss": 3.5241830200678503,
      "tokens_seen": 1602813952
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000519618856569709,
      "loss": 1.3584,
      "theoretical_loss": 3.524170577760815,
      "tokens_seen": 1602879488
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005195987963891675,
      "loss": 1.1936,
      "theoretical_loss": 3.5241581361049263,
      "tokens_seen": 1602945024
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005195787362086259,
      "loss": 1.4621,
      "theoretical_loss": 3.524145695100123,
      "tokens_seen": 1603010560
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005195586760280842,
      "loss": 1.5324,
      "theoretical_loss": 3.524133254746345,
      "tokens_seen": 1603076096
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005195386158475426,
      "loss": 1.4817,
      "theoretical_loss": 3.5241208150435304,
      "tokens_seen": 1603141632
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000519518555667001,
      "loss": 1.3428,
      "theoretical_loss": 3.52410837599162,
      "tokens_seen": 1603207168
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005194984954864594,
      "loss": 1.313,
      "theoretical_loss": 3.5240959375905527,
      "tokens_seen": 1603272704
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005194784353059178,
      "loss": 1.3796,
      "theoretical_loss": 3.524083499840267,
      "tokens_seen": 1603338240
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005194583751253762,
      "loss": 1.472,
      "theoretical_loss": 3.5240710627407035,
      "tokens_seen": 1603403776
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005194383149448345,
      "loss": 1.4156,
      "theoretical_loss": 3.5240586262918003,
      "tokens_seen": 1603469312
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005194182547642929,
      "loss": 1.4508,
      "theoretical_loss": 3.5240461904934977,
      "tokens_seen": 1603534848
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005193981945837513,
      "loss": 1.4161,
      "theoretical_loss": 3.524033755345735,
      "tokens_seen": 1603600384
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005193781344032097,
      "loss": 1.4063,
      "theoretical_loss": 3.524021320848451,
      "tokens_seen": 1603665920
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000519358074222668,
      "loss": 1.5093,
      "theoretical_loss": 3.524008887001586,
      "tokens_seen": 1603731456
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005193380140421265,
      "loss": 1.443,
      "theoretical_loss": 3.5239964538050788,
      "tokens_seen": 1603796992
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005193179538615847,
      "loss": 1.3216,
      "theoretical_loss": 3.523984021258869,
      "tokens_seen": 1603862528
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005192978936810432,
      "loss": 1.5155,
      "theoretical_loss": 3.5239715893628953,
      "tokens_seen": 1603928064
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005192778335005015,
      "loss": 1.4309,
      "theoretical_loss": 3.5239591581170986,
      "tokens_seen": 1603993600
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005192577733199599,
      "loss": 1.46,
      "theoretical_loss": 3.5239467275214165,
      "tokens_seen": 1604059136
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005192377131394183,
      "loss": 1.4173,
      "theoretical_loss": 3.52393429757579,
      "tokens_seen": 1604124672
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005192176529588767,
      "loss": 1.569,
      "theoretical_loss": 3.5239218682801576,
      "tokens_seen": 1604190208
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000519197592778335,
      "loss": 1.4608,
      "theoretical_loss": 3.5239094396344592,
      "tokens_seen": 1604255744
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005191775325977935,
      "loss": 1.4118,
      "theoretical_loss": 3.523897011638635,
      "tokens_seen": 1604321280
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005191574724172517,
      "loss": 1.3512,
      "theoretical_loss": 3.523884584292623,
      "tokens_seen": 1604386816
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005191374122367102,
      "loss": 1.4971,
      "theoretical_loss": 3.523872157596363,
      "tokens_seen": 1604452352
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005191173520561685,
      "loss": 1.3411,
      "theoretical_loss": 3.5238597315497953,
      "tokens_seen": 1604517888
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005190972918756269,
      "loss": 1.5166,
      "theoretical_loss": 3.5238473061528586,
      "tokens_seen": 1604583424
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005190772316950853,
      "loss": 1.334,
      "theoretical_loss": 3.523834881405493,
      "tokens_seen": 1604648960
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005190571715145436,
      "loss": 1.3722,
      "theoretical_loss": 3.5238224573076375,
      "tokens_seen": 1604714496
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000519037111334002,
      "loss": 1.4217,
      "theoretical_loss": 3.5238100338592324,
      "tokens_seen": 1604780032
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005190170511534605,
      "loss": 1.3222,
      "theoretical_loss": 3.5237976110602163,
      "tokens_seen": 1604845568
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005189969909729187,
      "loss": 1.4648,
      "theoretical_loss": 3.523785188910529,
      "tokens_seen": 1604911104
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005189769307923772,
      "loss": 1.5501,
      "theoretical_loss": 3.5237727674101107,
      "tokens_seen": 1604976640
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005189568706118355,
      "loss": 1.3906,
      "theoretical_loss": 3.5237603465589,
      "tokens_seen": 1605042176
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005189368104312939,
      "loss": 1.5029,
      "theoretical_loss": 3.523747926356837,
      "tokens_seen": 1605107712
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005189167502507523,
      "loss": 1.5126,
      "theoretical_loss": 3.5237355068038614,
      "tokens_seen": 1605173248
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005188966900702106,
      "loss": 1.4181,
      "theoretical_loss": 3.5237230878999126,
      "tokens_seen": 1605238784
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000518876629889669,
      "loss": 1.4016,
      "theoretical_loss": 3.5237106696449296,
      "tokens_seen": 1605304320
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005188565697091275,
      "loss": 1.3253,
      "theoretical_loss": 3.523698252038853,
      "tokens_seen": 1605369856
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005188365095285857,
      "loss": 1.406,
      "theoretical_loss": 3.5236858350816216,
      "tokens_seen": 1605435392
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005188164493480442,
      "loss": 1.4442,
      "theoretical_loss": 3.5236734187731757,
      "tokens_seen": 1605500928
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005187963891675024,
      "loss": 1.3119,
      "theoretical_loss": 3.5236610031134545,
      "tokens_seen": 1605566464
    },
    {
      "epoch": 0.49,
      "objective/train/advantage_avg": 0.49325236678123474,
      "objective/train/docs_used": 586391,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3480722904205322,
      "objective/train/original_loss": 1.3480721712112427,
      "objective/train/theoretical_loss": 3.5236485881023976,
      "objective/train/tokens_used": 1626092000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24369022250175476,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505640506744385,
      "objective/train/weighted_lm_loss": 1.4163076877593994,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5236485881023976,
      "tokens_seen": 1605632000
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005187763289869609,
      "loss": 1.4174,
      "theoretical_loss": 3.5236485881023976,
      "tokens_seen": 1605632000
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005187562688064193,
      "loss": 1.4072,
      "theoretical_loss": 3.523636173739945,
      "tokens_seen": 1605697536
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005187362086258776,
      "loss": 1.4034,
      "theoretical_loss": 3.5236237600260356,
      "tokens_seen": 1605763072
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000518716148445336,
      "loss": 1.3079,
      "theoretical_loss": 3.5236113469606103,
      "tokens_seen": 1605828608
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005186960882647944,
      "loss": 1.4396,
      "theoretical_loss": 3.5235989345436076,
      "tokens_seen": 1605894144
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005186760280842527,
      "loss": 1.4519,
      "theoretical_loss": 3.523586522774967,
      "tokens_seen": 1605959680
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005186559679037112,
      "loss": 1.4781,
      "theoretical_loss": 3.5235741116546295,
      "tokens_seen": 1606025216
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005186359077231695,
      "loss": 1.4759,
      "theoretical_loss": 3.5235617011825338,
      "tokens_seen": 1606090752
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005186158475426279,
      "loss": 1.3396,
      "theoretical_loss": 3.52354929135862,
      "tokens_seen": 1606156288
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005185957873620863,
      "loss": 1.3767,
      "theoretical_loss": 3.5235368821828272,
      "tokens_seen": 1606221824
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005185757271815446,
      "loss": 1.4403,
      "theoretical_loss": 3.5235244736550957,
      "tokens_seen": 1606287360
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000518555667001003,
      "loss": 1.3995,
      "theoretical_loss": 3.523512065775365,
      "tokens_seen": 1606352896
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005185356068204614,
      "loss": 1.2998,
      "theoretical_loss": 3.5234996585435754,
      "tokens_seen": 1606418432
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005185155466399197,
      "loss": 1.405,
      "theoretical_loss": 3.5234872519596654,
      "tokens_seen": 1606483968
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005184954864593782,
      "loss": 1.4222,
      "theoretical_loss": 3.523474846023576,
      "tokens_seen": 1606549504
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005184754262788365,
      "loss": 1.4371,
      "theoretical_loss": 3.523462440735246,
      "tokens_seen": 1606615040
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005184553660982949,
      "loss": 1.4073,
      "theoretical_loss": 3.5234500360946157,
      "tokens_seen": 1606680576
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005184353059177532,
      "loss": 1.447,
      "theoretical_loss": 3.5234376321016248,
      "tokens_seen": 1606746112
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005184152457372117,
      "loss": 1.3654,
      "theoretical_loss": 3.5234252287562127,
      "tokens_seen": 1606811648
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00051839518555667,
      "loss": 1.2035,
      "theoretical_loss": 3.52341282605832,
      "tokens_seen": 1606877184
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005183751253761284,
      "loss": 1.4657,
      "theoretical_loss": 3.5234004240078853,
      "tokens_seen": 1606942720
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005183550651955867,
      "loss": 1.3608,
      "theoretical_loss": 3.5233880226048493,
      "tokens_seen": 1607008256
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005183350050150451,
      "loss": 1.5268,
      "theoretical_loss": 3.523375621849152,
      "tokens_seen": 1607073792
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005183149448345035,
      "loss": 1.4538,
      "theoretical_loss": 3.5233632217407322,
      "tokens_seen": 1607139328
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005182948846539619,
      "loss": 1.3694,
      "theoretical_loss": 3.5233508222795304,
      "tokens_seen": 1607204864
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005182748244734202,
      "loss": 1.4059,
      "theoretical_loss": 3.5233384234654865,
      "tokens_seen": 1607270400
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005182547642928787,
      "loss": 1.3889,
      "theoretical_loss": 3.52332602529854,
      "tokens_seen": 1607335936
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000518234704112337,
      "loss": 1.4281,
      "theoretical_loss": 3.523313627778631,
      "tokens_seen": 1607401472
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005182146439317954,
      "loss": 1.3417,
      "theoretical_loss": 3.5233012309057,
      "tokens_seen": 1607467008
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005181945837512537,
      "loss": 1.5965,
      "theoretical_loss": 3.5232888346796853,
      "tokens_seen": 1607532544
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005181745235707121,
      "loss": 1.3222,
      "theoretical_loss": 3.5232764391005276,
      "tokens_seen": 1607598080
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005181544633901705,
      "loss": 1.3772,
      "theoretical_loss": 3.523264044168167,
      "tokens_seen": 1607663616
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005181344032096289,
      "loss": 1.373,
      "theoretical_loss": 3.5232516498825426,
      "tokens_seen": 1607729152
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005181143430290872,
      "loss": 1.5103,
      "theoretical_loss": 3.5232392562435955,
      "tokens_seen": 1607794688
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005180942828485457,
      "loss": 1.4829,
      "theoretical_loss": 3.523226863251265,
      "tokens_seen": 1607860224
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005180742226680039,
      "loss": 1.4717,
      "theoretical_loss": 3.523214470905491,
      "tokens_seen": 1607925760
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005180541624874624,
      "loss": 1.3678,
      "theoretical_loss": 3.5232020792062126,
      "tokens_seen": 1607991296
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005180341023069208,
      "loss": 1.3547,
      "theoretical_loss": 3.523189688153371,
      "tokens_seen": 1608056832
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005180140421263791,
      "loss": 1.3525,
      "theoretical_loss": 3.523177297746906,
      "tokens_seen": 1608122368
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005179939819458376,
      "loss": 1.4962,
      "theoretical_loss": 3.523164907986757,
      "tokens_seen": 1608187904
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000517973921765296,
      "loss": 1.3432,
      "theoretical_loss": 3.523152518872864,
      "tokens_seen": 1608253440
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005179538615847543,
      "loss": 1.4525,
      "theoretical_loss": 3.5231401304051673,
      "tokens_seen": 1608318976
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005179338014042127,
      "loss": 1.4362,
      "theoretical_loss": 3.523127742583606,
      "tokens_seen": 1608384512
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000517913741223671,
      "loss": 1.3548,
      "theoretical_loss": 3.5231153554081214,
      "tokens_seen": 1608450048
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005178936810431294,
      "loss": 1.5066,
      "theoretical_loss": 3.5231029688786526,
      "tokens_seen": 1608515584
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005178736208625879,
      "loss": 1.4821,
      "theoretical_loss": 3.52309058299514,
      "tokens_seen": 1608581120
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005178535606820461,
      "loss": 1.5107,
      "theoretical_loss": 3.523078197757523,
      "tokens_seen": 1608646656
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005178335005015046,
      "loss": 1.4261,
      "theoretical_loss": 3.5230658131657426,
      "tokens_seen": 1608712192
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005178134403209628,
      "loss": 1.319,
      "theoretical_loss": 3.523053429219738,
      "tokens_seen": 1608777728
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005177933801404213,
      "loss": 1.5244,
      "theoretical_loss": 3.523041045919449,
      "tokens_seen": 1608843264
    },
    {
      "epoch": 0.49,
      "objective/train/advantage_avg": 0.49116790294647217,
      "objective/train/docs_used": 587617,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.590345859527588,
      "objective/train/original_loss": 1.5903457403182983,
      "objective/train/theoretical_loss": 3.5230286632648165,
      "objective/train/tokens_used": 1629368800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24188819527626038,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503463745117188,
      "objective/train/weighted_lm_loss": 1.6705447435379028,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.5230286632648165,
      "tokens_seen": 1608908800
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005177733199598797,
      "loss": 1.5786,
      "theoretical_loss": 3.5230286632648165,
      "tokens_seen": 1608908800
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000517753259779338,
      "loss": 1.4782,
      "theoretical_loss": 3.52301628125578,
      "tokens_seen": 1608974336
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005177331995987964,
      "loss": 1.5726,
      "theoretical_loss": 3.52300389989228,
      "tokens_seen": 1609039872
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005177131394182548,
      "loss": 1.4847,
      "theoretical_loss": 3.522991519174256,
      "tokens_seen": 1609105408
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005176930792377131,
      "loss": 1.4056,
      "theoretical_loss": 3.522979139101648,
      "tokens_seen": 1609170944
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005176730190571716,
      "loss": 1.5179,
      "theoretical_loss": 3.5229667596743965,
      "tokens_seen": 1609236480
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005176529588766299,
      "loss": 1.5322,
      "theoretical_loss": 3.5229543808924415,
      "tokens_seen": 1609302016
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005176328986960883,
      "loss": 1.4625,
      "theoretical_loss": 3.5229420027557232,
      "tokens_seen": 1609367552
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005176128385155467,
      "loss": 1.4896,
      "theoretical_loss": 3.5229296252641813,
      "tokens_seen": 1609433088
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000517592778335005,
      "loss": 1.3572,
      "theoretical_loss": 3.522917248417756,
      "tokens_seen": 1609498624
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005175727181544634,
      "loss": 1.3631,
      "theoretical_loss": 3.522904872216388,
      "tokens_seen": 1609564160
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005175526579739218,
      "loss": 1.4076,
      "theoretical_loss": 3.522892496660017,
      "tokens_seen": 1609629696
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005175325977933801,
      "loss": 1.4207,
      "theoretical_loss": 3.522880121748582,
      "tokens_seen": 1609695232
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005175125376128386,
      "loss": 1.4912,
      "theoretical_loss": 3.5228677474820254,
      "tokens_seen": 1609760768
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005174924774322969,
      "loss": 1.3872,
      "theoretical_loss": 3.5228553738602857,
      "tokens_seen": 1609826304
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005174724172517553,
      "loss": 1.449,
      "theoretical_loss": 3.522843000883303,
      "tokens_seen": 1609891840
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005174523570712136,
      "loss": 1.5491,
      "theoretical_loss": 3.5228306285510183,
      "tokens_seen": 1609957376
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005174322968906721,
      "loss": 1.4715,
      "theoretical_loss": 3.5228182568633715,
      "tokens_seen": 1610022912
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005174122367101304,
      "loss": 1.3693,
      "theoretical_loss": 3.5228058858203033,
      "tokens_seen": 1610088448
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005173921765295888,
      "loss": 1.4827,
      "theoretical_loss": 3.5227935154217525,
      "tokens_seen": 1610153984
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005173721163490471,
      "loss": 1.4103,
      "theoretical_loss": 3.5227811456676603,
      "tokens_seen": 1610219520
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005173520561685055,
      "loss": 1.4662,
      "theoretical_loss": 3.5227687765579665,
      "tokens_seen": 1610285056
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005173319959879639,
      "loss": 1.4536,
      "theoretical_loss": 3.5227564080926115,
      "tokens_seen": 1610350592
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005173119358074223,
      "loss": 1.4574,
      "theoretical_loss": 3.5227440402715358,
      "tokens_seen": 1610416128
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005172918756268806,
      "loss": 1.4333,
      "theoretical_loss": 3.522731673094679,
      "tokens_seen": 1610481664
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005172718154463391,
      "loss": 1.4534,
      "theoretical_loss": 3.522719306561982,
      "tokens_seen": 1610547200
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005172517552657974,
      "loss": 1.4175,
      "theoretical_loss": 3.5227069406733844,
      "tokens_seen": 1610612736
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005172316950852558,
      "loss": 1.4112,
      "theoretical_loss": 3.522694575428827,
      "tokens_seen": 1610678272
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005172116349047141,
      "loss": 1.4571,
      "theoretical_loss": 3.5226822108282496,
      "tokens_seen": 1610743808
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005171915747241725,
      "loss": 1.5447,
      "theoretical_loss": 3.522669846871592,
      "tokens_seen": 1610809344
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005171715145436309,
      "loss": 1.4587,
      "theoretical_loss": 3.5226574835587963,
      "tokens_seen": 1610874880
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005171514543630893,
      "loss": 1.4918,
      "theoretical_loss": 3.522645120889801,
      "tokens_seen": 1610940416
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005171313941825476,
      "loss": 1.3602,
      "theoretical_loss": 3.522632758864547,
      "tokens_seen": 1611005952
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005171113340020061,
      "loss": 1.3182,
      "theoretical_loss": 3.5226203974829744,
      "tokens_seen": 1611071488
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005170912738214643,
      "loss": 1.424,
      "theoretical_loss": 3.522608036745024,
      "tokens_seen": 1611137024
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005170712136409228,
      "loss": 1.3344,
      "theoretical_loss": 3.5225956766506354,
      "tokens_seen": 1611202560
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005170511534603812,
      "loss": 1.5451,
      "theoretical_loss": 3.522583317199749,
      "tokens_seen": 1611268096
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005170310932798395,
      "loss": 1.5105,
      "theoretical_loss": 3.5225709583923064,
      "tokens_seen": 1611333632
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005170110330992979,
      "loss": 1.3726,
      "theoretical_loss": 3.5225586002282467,
      "tokens_seen": 1611399168
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005169909729187564,
      "loss": 1.3986,
      "theoretical_loss": 3.5225462427075103,
      "tokens_seen": 1611464704
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005169709127382146,
      "loss": 1.3628,
      "theoretical_loss": 3.5225338858300375,
      "tokens_seen": 1611530240
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005169508525576731,
      "loss": 1.5651,
      "theoretical_loss": 3.5225215295957693,
      "tokens_seen": 1611595776
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005169307923771313,
      "loss": 1.4771,
      "theoretical_loss": 3.5225091740046457,
      "tokens_seen": 1611661312
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005169107321965898,
      "loss": 1.4385,
      "theoretical_loss": 3.522496819056607,
      "tokens_seen": 1611726848
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005168906720160482,
      "loss": 1.5324,
      "theoretical_loss": 3.522484464751593,
      "tokens_seen": 1611792384
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005168706118355065,
      "loss": 1.5987,
      "theoretical_loss": 3.522472111089545,
      "tokens_seen": 1611857920
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005168505516549649,
      "loss": 1.3893,
      "theoretical_loss": 3.5224597580704033,
      "tokens_seen": 1611923456
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005168304914744232,
      "loss": 1.4985,
      "theoretical_loss": 3.5224474056941086,
      "tokens_seen": 1611988992
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005168104312938816,
      "loss": 1.3759,
      "theoretical_loss": 3.5224350539606,
      "tokens_seen": 1612054528
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005167903711133401,
      "loss": 1.4532,
      "theoretical_loss": 3.52242270286982,
      "tokens_seen": 1612120064
    },
    {
      "epoch": 0.49,
      "objective/train/advantage_avg": 0.48874321579933167,
      "objective/train/docs_used": 588885,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.420301914215088,
      "objective/train/original_loss": 1.420301914215088,
      "objective/train/theoretical_loss": 3.5224103524217067,
      "objective/train/tokens_used": 1632645600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.240156888961792,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500949621200562,
      "objective/train/weighted_lm_loss": 1.491533637046814,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5224103524217067,
      "tokens_seen": 1612185600
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005167703109327983,
      "loss": 1.4164,
      "theoretical_loss": 3.5224103524217067,
      "tokens_seen": 1612185600
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005167502507522568,
      "loss": 1.438,
      "theoretical_loss": 3.5223980026162014,
      "tokens_seen": 1612251136
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005167301905717151,
      "loss": 1.3935,
      "theoretical_loss": 3.5223856534532456,
      "tokens_seen": 1612316672
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005167101303911735,
      "loss": 1.5561,
      "theoretical_loss": 3.5223733049327786,
      "tokens_seen": 1612382208
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005166900702106319,
      "loss": 1.4709,
      "theoretical_loss": 3.522360957054741,
      "tokens_seen": 1612447744
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005166700100300903,
      "loss": 1.2713,
      "theoretical_loss": 3.5223486098190735,
      "tokens_seen": 1612513280
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005166499498495486,
      "loss": 1.4313,
      "theoretical_loss": 3.522336263225717,
      "tokens_seen": 1612578816
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005166298896690071,
      "loss": 1.3135,
      "theoretical_loss": 3.522323917274611,
      "tokens_seen": 1612644352
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005166098294884653,
      "loss": 1.5242,
      "theoretical_loss": 3.5223115719656963,
      "tokens_seen": 1612709888
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005165897693079238,
      "loss": 1.4804,
      "theoretical_loss": 3.522299227298914,
      "tokens_seen": 1612775424
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005165697091273821,
      "loss": 1.4444,
      "theoretical_loss": 3.5222868832742042,
      "tokens_seen": 1612840960
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005165496489468405,
      "loss": 1.4271,
      "theoretical_loss": 3.522274539891508,
      "tokens_seen": 1612906496
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005165295887662989,
      "loss": 1.4783,
      "theoretical_loss": 3.5222621971507646,
      "tokens_seen": 1612972032
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005165095285857573,
      "loss": 1.4494,
      "theoretical_loss": 3.5222498550519155,
      "tokens_seen": 1613037568
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005164894684052157,
      "loss": 1.4425,
      "theoretical_loss": 3.522237513594901,
      "tokens_seen": 1613103104
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000516469408224674,
      "loss": 1.3666,
      "theoretical_loss": 3.5222251727796614,
      "tokens_seen": 1613168640
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005164493480441325,
      "loss": 1.4963,
      "theoretical_loss": 3.522212832606138,
      "tokens_seen": 1613234176
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005164292878635908,
      "loss": 1.4433,
      "theoretical_loss": 3.5222004930742705,
      "tokens_seen": 1613299712
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005164092276830492,
      "loss": 1.4579,
      "theoretical_loss": 3.5221881541840006,
      "tokens_seen": 1613365248
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005163891675025075,
      "loss": 1.5037,
      "theoretical_loss": 3.522175815935267,
      "tokens_seen": 1613430784
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005163691073219659,
      "loss": 1.5121,
      "theoretical_loss": 3.5221634783280122,
      "tokens_seen": 1613496320
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005163490471414243,
      "loss": 1.4352,
      "theoretical_loss": 3.522151141362176,
      "tokens_seen": 1613561856
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005163289869608827,
      "loss": 1.3942,
      "theoretical_loss": 3.5221388050376987,
      "tokens_seen": 1613627392
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000516308926780341,
      "loss": 1.4357,
      "theoretical_loss": 3.5221264693545216,
      "tokens_seen": 1613692928
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005162888665997995,
      "loss": 1.4357,
      "theoretical_loss": 3.5221141343125852,
      "tokens_seen": 1613758464
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005162688064192578,
      "loss": 1.4783,
      "theoretical_loss": 3.5221017999118294,
      "tokens_seen": 1613824000
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005162487462387162,
      "loss": 1.4676,
      "theoretical_loss": 3.5220894661521953,
      "tokens_seen": 1613889536
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005162286860581745,
      "loss": 1.5222,
      "theoretical_loss": 3.522077133033624,
      "tokens_seen": 1613955072
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005162086258776329,
      "loss": 1.4356,
      "theoretical_loss": 3.5220648005560555,
      "tokens_seen": 1614020608
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005161885656970913,
      "loss": 1.5077,
      "theoretical_loss": 3.522052468719431,
      "tokens_seen": 1614086144
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005161685055165497,
      "loss": 1.5676,
      "theoretical_loss": 3.52204013752369,
      "tokens_seen": 1614151680
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000516148445336008,
      "loss": 1.5272,
      "theoretical_loss": 3.522027806968775,
      "tokens_seen": 1614217216
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005161283851554665,
      "loss": 1.5867,
      "theoretical_loss": 3.522015477054625,
      "tokens_seen": 1614282752
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005161083249749247,
      "loss": 1.4945,
      "theoretical_loss": 3.5220031477811817,
      "tokens_seen": 1614348288
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005160882647943832,
      "loss": 1.4974,
      "theoretical_loss": 3.5219908191483853,
      "tokens_seen": 1614413824
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005160682046138416,
      "loss": 1.4962,
      "theoretical_loss": 3.521978491156177,
      "tokens_seen": 1614479360
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005160481444332999,
      "loss": 1.5429,
      "theoretical_loss": 3.521966163804497,
      "tokens_seen": 1614544896
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005160280842527583,
      "loss": 1.3517,
      "theoretical_loss": 3.521953837093286,
      "tokens_seen": 1614610432
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005160080240722168,
      "loss": 1.5294,
      "theoretical_loss": 3.5219415110224856,
      "tokens_seen": 1614675968
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000515987963891675,
      "loss": 1.476,
      "theoretical_loss": 3.521929185592036,
      "tokens_seen": 1614741504
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005159679037111335,
      "loss": 1.4899,
      "theoretical_loss": 3.521916860801877,
      "tokens_seen": 1614807040
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005159478435305917,
      "loss": 1.4146,
      "theoretical_loss": 3.521904536651951,
      "tokens_seen": 1614872576
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005159277833500502,
      "loss": 1.4364,
      "theoretical_loss": 3.5218922131421975,
      "tokens_seen": 1614938112
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005159077231695086,
      "loss": 1.5803,
      "theoretical_loss": 3.521879890272558,
      "tokens_seen": 1615003648
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005158876629889669,
      "loss": 1.4089,
      "theoretical_loss": 3.521867568042973,
      "tokens_seen": 1615069184
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005158676028084253,
      "loss": 1.4539,
      "theoretical_loss": 3.521855246453383,
      "tokens_seen": 1615134720
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005158475426278836,
      "loss": 1.3594,
      "theoretical_loss": 3.521842925503729,
      "tokens_seen": 1615200256
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000515827482447342,
      "loss": 1.412,
      "theoretical_loss": 3.5218306051939523,
      "tokens_seen": 1615265792
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005158074222668005,
      "loss": 1.565,
      "theoretical_loss": 3.5218182855239935,
      "tokens_seen": 1615331328
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005157873620862587,
      "loss": 1.4057,
      "theoretical_loss": 3.521805966493793,
      "tokens_seen": 1615396864
    },
    {
      "epoch": 0.49,
      "objective/train/advantage_avg": 0.4895901679992676,
      "objective/train/docs_used": 590099,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3162305355072021,
      "objective/train/original_loss": 1.3162305355072021,
      "objective/train/theoretical_loss": 3.5217936481032917,
      "objective/train/tokens_used": 1635922400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24111151695251465,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501846075057983,
      "objective/train/weighted_lm_loss": 1.3820867538452148,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5217936481032917,
      "tokens_seen": 1615462400
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005157673019057172,
      "loss": 1.4165,
      "theoretical_loss": 3.5217936481032917,
      "tokens_seen": 1615462400
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005157472417251755,
      "loss": 1.3606,
      "theoretical_loss": 3.5217813303524306,
      "tokens_seen": 1615527936
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005157271815446339,
      "loss": 1.4567,
      "theoretical_loss": 3.5217690132411508,
      "tokens_seen": 1615593472
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005157071213640923,
      "loss": 1.5102,
      "theoretical_loss": 3.5217566967693927,
      "tokens_seen": 1615659008
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005156870611835507,
      "loss": 1.5716,
      "theoretical_loss": 3.521744380937097,
      "tokens_seen": 1615724544
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000515667001003009,
      "loss": 1.4422,
      "theoretical_loss": 3.5217320657442057,
      "tokens_seen": 1615790080
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005156469408224675,
      "loss": 1.3392,
      "theoretical_loss": 3.521719751190658,
      "tokens_seen": 1615855616
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005156268806419257,
      "loss": 1.482,
      "theoretical_loss": 3.5217074372763957,
      "tokens_seen": 1615921152
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005156068204613842,
      "loss": 1.401,
      "theoretical_loss": 3.5216951240013605,
      "tokens_seen": 1615986688
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005155867602808425,
      "loss": 1.3458,
      "theoretical_loss": 3.5216828113654914,
      "tokens_seen": 1616052224
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005155667001003009,
      "loss": 1.5264,
      "theoretical_loss": 3.5216704993687307,
      "tokens_seen": 1616117760
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005155466399197593,
      "loss": 1.3797,
      "theoretical_loss": 3.5216581880110196,
      "tokens_seen": 1616183296
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005155265797392177,
      "loss": 1.4177,
      "theoretical_loss": 3.5216458772922974,
      "tokens_seen": 1616248832
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000515506519558676,
      "loss": 1.431,
      "theoretical_loss": 3.5216335672125063,
      "tokens_seen": 1616314368
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005154864593781344,
      "loss": 1.5064,
      "theoretical_loss": 3.5216212577715873,
      "tokens_seen": 1616379904
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005154663991975927,
      "loss": 1.4769,
      "theoretical_loss": 3.5216089489694804,
      "tokens_seen": 1616445440
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005154463390170512,
      "loss": 1.4447,
      "theoretical_loss": 3.5215966408061274,
      "tokens_seen": 1616510976
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005154262788365095,
      "loss": 1.5609,
      "theoretical_loss": 3.521584333281469,
      "tokens_seen": 1616576512
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005154062186559679,
      "loss": 1.4059,
      "theoretical_loss": 3.5215720263954458,
      "tokens_seen": 1616642048
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005153861584754262,
      "loss": 1.4894,
      "theoretical_loss": 3.521559720148,
      "tokens_seen": 1616707584
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005153660982948847,
      "loss": 1.425,
      "theoretical_loss": 3.5215474145390706,
      "tokens_seen": 1616773120
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000515346038114343,
      "loss": 1.4034,
      "theoretical_loss": 3.5215351095686005,
      "tokens_seen": 1616838656
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005153259779338014,
      "loss": 1.4001,
      "theoretical_loss": 3.521522805236529,
      "tokens_seen": 1616904192
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005153059177532598,
      "loss": 1.4516,
      "theoretical_loss": 3.521510501542799,
      "tokens_seen": 1616969728
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005152858575727182,
      "loss": 1.5179,
      "theoretical_loss": 3.52149819848735,
      "tokens_seen": 1617035264
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005152657973921765,
      "loss": 1.5579,
      "theoretical_loss": 3.521485896070123,
      "tokens_seen": 1617100800
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000515245737211635,
      "loss": 1.4762,
      "theoretical_loss": 3.52147359429106,
      "tokens_seen": 1617166336
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005152256770310932,
      "loss": 1.5291,
      "theoretical_loss": 3.521461293150102,
      "tokens_seen": 1617231872
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005152056168505517,
      "loss": 1.4104,
      "theoretical_loss": 3.5214489926471897,
      "tokens_seen": 1617297408
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00051518555667001,
      "loss": 1.4347,
      "theoretical_loss": 3.5214366927822636,
      "tokens_seen": 1617362944
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005151654964894684,
      "loss": 1.5109,
      "theoretical_loss": 3.5214243935552654,
      "tokens_seen": 1617428480
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005151454363089268,
      "loss": 1.4338,
      "theoretical_loss": 3.521412094966136,
      "tokens_seen": 1617494016
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005151253761283851,
      "loss": 1.3663,
      "theoretical_loss": 3.5213997970148165,
      "tokens_seen": 1617559552
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005151053159478435,
      "loss": 1.3604,
      "theoretical_loss": 3.5213874997012478,
      "tokens_seen": 1617625088
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000515085255767302,
      "loss": 1.3341,
      "theoretical_loss": 3.5213752030253715,
      "tokens_seen": 1617690624
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005150651955867602,
      "loss": 1.4467,
      "theoretical_loss": 3.521362906987128,
      "tokens_seen": 1617756160
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005150451354062187,
      "loss": 1.4072,
      "theoretical_loss": 3.521350611586459,
      "tokens_seen": 1617821696
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005150250752256769,
      "loss": 1.4363,
      "theoretical_loss": 3.5213383168233054,
      "tokens_seen": 1617887232
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005150050150451354,
      "loss": 1.4862,
      "theoretical_loss": 3.5213260226976084,
      "tokens_seen": 1617952768
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005149849548645939,
      "loss": 1.4432,
      "theoretical_loss": 3.5213137292093086,
      "tokens_seen": 1618018304
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005149648946840521,
      "loss": 1.5084,
      "theoretical_loss": 3.521301436358348,
      "tokens_seen": 1618083840
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005149448345035106,
      "loss": 1.4535,
      "theoretical_loss": 3.521289144144667,
      "tokens_seen": 1618149376
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000514924774322969,
      "loss": 1.413,
      "theoretical_loss": 3.5212768525682074,
      "tokens_seen": 1618214912
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005149047141424273,
      "loss": 1.546,
      "theoretical_loss": 3.52126456162891,
      "tokens_seen": 1618280448
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005148846539618857,
      "loss": 1.4526,
      "theoretical_loss": 3.521252271326716,
      "tokens_seen": 1618345984
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000514864593781344,
      "loss": 1.4403,
      "theoretical_loss": 3.5212399816615663,
      "tokens_seen": 1618411520
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005148445336008024,
      "loss": 1.4721,
      "theoretical_loss": 3.5212276926334027,
      "tokens_seen": 1618477056
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005148244734202609,
      "loss": 1.3813,
      "theoretical_loss": 3.5212154042421657,
      "tokens_seen": 1618542592
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005148044132397191,
      "loss": 1.4304,
      "theoretical_loss": 3.521203116487797,
      "tokens_seen": 1618608128
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005147843530591776,
      "loss": 1.5078,
      "theoretical_loss": 3.521190829370238,
      "tokens_seen": 1618673664
    },
    {
      "epoch": 0.49,
      "objective/train/advantage_avg": 0.4944377541542053,
      "objective/train/docs_used": 591377,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2216219902038574,
      "objective/train/original_loss": 1.2216219902038574,
      "objective/train/theoretical_loss": 3.5211785428894293,
      "objective/train/tokens_used": 1639199200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24467982351779938,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506876707077026,
      "objective/train/weighted_lm_loss": 1.2835488319396973,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5211785428894293,
      "tokens_seen": 1618739200
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005147642928786359,
      "loss": 1.2481,
      "theoretical_loss": 3.5211785428894293,
      "tokens_seen": 1618739200
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005147442326980943,
      "loss": 1.4707,
      "theoretical_loss": 3.521166257045313,
      "tokens_seen": 1618804736
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005147241725175527,
      "loss": 1.4722,
      "theoretical_loss": 3.521153971837829,
      "tokens_seen": 1618870272
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005147041123370111,
      "loss": 1.5074,
      "theoretical_loss": 3.52114168726692,
      "tokens_seen": 1618935808
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005146840521564694,
      "loss": 1.5013,
      "theoretical_loss": 3.5211294033325258,
      "tokens_seen": 1619001344
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005146639919759279,
      "loss": 1.444,
      "theoretical_loss": 3.5211171200345888,
      "tokens_seen": 1619066880
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005146439317953861,
      "loss": 1.503,
      "theoretical_loss": 3.5211048373730494,
      "tokens_seen": 1619132416
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005146238716148446,
      "loss": 1.3102,
      "theoretical_loss": 3.52109255534785,
      "tokens_seen": 1619197952
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005146038114343029,
      "loss": 1.4229,
      "theoretical_loss": 3.5210802739589306,
      "tokens_seen": 1619263488
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005145837512537613,
      "loss": 1.4694,
      "theoretical_loss": 3.521067993206233,
      "tokens_seen": 1619329024
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005145636910732197,
      "loss": 1.3453,
      "theoretical_loss": 3.5210557130897,
      "tokens_seen": 1619394560
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005145436308926781,
      "loss": 1.3898,
      "theoretical_loss": 3.52104343360927,
      "tokens_seen": 1619460096
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005145235707121364,
      "loss": 1.5648,
      "theoretical_loss": 3.521031154764886,
      "tokens_seen": 1619525632
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005145035105315948,
      "loss": 1.4598,
      "theoretical_loss": 3.52101887655649,
      "tokens_seen": 1619591168
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005144834503510531,
      "loss": 1.3985,
      "theoretical_loss": 3.5210065989840214,
      "tokens_seen": 1619656704
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005144633901705116,
      "loss": 1.5359,
      "theoretical_loss": 3.5209943220474234,
      "tokens_seen": 1619722240
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005144433299899699,
      "loss": 1.4747,
      "theoretical_loss": 3.520982045746636,
      "tokens_seen": 1619787776
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005144232698094283,
      "loss": 1.4301,
      "theoretical_loss": 3.520969770081601,
      "tokens_seen": 1619853312
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005144032096288866,
      "loss": 1.4708,
      "theoretical_loss": 3.5209574950522606,
      "tokens_seen": 1619918848
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005143831494483451,
      "loss": 1.4637,
      "theoretical_loss": 3.5209452206585548,
      "tokens_seen": 1619984384
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005143630892678034,
      "loss": 1.4209,
      "theoretical_loss": 3.5209329469004254,
      "tokens_seen": 1620049920
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005143430290872618,
      "loss": 1.4456,
      "theoretical_loss": 3.5209206737778143,
      "tokens_seen": 1620115456
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005143229689067202,
      "loss": 1.4833,
      "theoretical_loss": 3.5209084012906624,
      "tokens_seen": 1620180992
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005143029087261786,
      "loss": 1.5879,
      "theoretical_loss": 3.520896129438911,
      "tokens_seen": 1620246528
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005142828485456369,
      "loss": 1.3958,
      "theoretical_loss": 3.5208838582225024,
      "tokens_seen": 1620312064
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005142627883650953,
      "loss": 1.4608,
      "theoretical_loss": 3.520871587641377,
      "tokens_seen": 1620377600
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005142427281845536,
      "loss": 1.3417,
      "theoretical_loss": 3.5208593176954763,
      "tokens_seen": 1620443136
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005142226680040121,
      "loss": 1.4474,
      "theoretical_loss": 3.5208470483847423,
      "tokens_seen": 1620508672
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005142026078234704,
      "loss": 1.4144,
      "theoretical_loss": 3.5208347797091157,
      "tokens_seen": 1620574208
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005141825476429288,
      "loss": 1.5275,
      "theoretical_loss": 3.5208225116685385,
      "tokens_seen": 1620639744
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005141624874623872,
      "loss": 1.3324,
      "theoretical_loss": 3.520810244262953,
      "tokens_seen": 1620705280
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005141424272818455,
      "loss": 1.3822,
      "theoretical_loss": 3.5207979774922986,
      "tokens_seen": 1620770816
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005141223671013039,
      "loss": 1.3982,
      "theoretical_loss": 3.5207857113565177,
      "tokens_seen": 1620836352
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005141023069207624,
      "loss": 1.388,
      "theoretical_loss": 3.5207734458555526,
      "tokens_seen": 1620901888
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005140822467402206,
      "loss": 1.459,
      "theoretical_loss": 3.5207611809893438,
      "tokens_seen": 1620967424
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005140621865596791,
      "loss": 1.4195,
      "theoretical_loss": 3.5207489167578334,
      "tokens_seen": 1621032960
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005140421263791373,
      "loss": 1.464,
      "theoretical_loss": 3.520736653160962,
      "tokens_seen": 1621098496
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005140220661985958,
      "loss": 1.4381,
      "theoretical_loss": 3.520724390198672,
      "tokens_seen": 1621164032
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005140020060180542,
      "loss": 1.4458,
      "theoretical_loss": 3.520712127870905,
      "tokens_seen": 1621229568
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005139819458375125,
      "loss": 1.3847,
      "theoretical_loss": 3.5206998661776017,
      "tokens_seen": 1621295104
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005139618856569709,
      "loss": 1.4679,
      "theoretical_loss": 3.520687605118704,
      "tokens_seen": 1621360640
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005139418254764294,
      "loss": 1.4813,
      "theoretical_loss": 3.520675344694154,
      "tokens_seen": 1621426176
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005139217652958876,
      "loss": 1.5564,
      "theoretical_loss": 3.520663084903892,
      "tokens_seen": 1621491712
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005139017051153461,
      "loss": 1.4116,
      "theoretical_loss": 3.5206508257478606,
      "tokens_seen": 1621557248
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005138816449348043,
      "loss": 1.5347,
      "theoretical_loss": 3.520638567226001,
      "tokens_seen": 1621622784
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005138615847542628,
      "loss": 1.5123,
      "theoretical_loss": 3.5206263093382546,
      "tokens_seen": 1621688320
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005138415245737212,
      "loss": 1.3882,
      "theoretical_loss": 3.5206140520845635,
      "tokens_seen": 1621753856
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005138214643931795,
      "loss": 1.4831,
      "theoretical_loss": 3.520601795464869,
      "tokens_seen": 1621819392
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005138014042126379,
      "loss": 1.3776,
      "theoretical_loss": 3.5205895394791127,
      "tokens_seen": 1621884928
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005137813440320963,
      "loss": 1.46,
      "theoretical_loss": 3.520577284127236,
      "tokens_seen": 1621950464
    },
    {
      "epoch": 0.49,
      "objective/train/advantage_avg": 0.49273592233657837,
      "objective/train/docs_used": 592532,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2885770797729492,
      "objective/train/original_loss": 1.2885769605636597,
      "objective/train/theoretical_loss": 3.5205650294091804,
      "objective/train/tokens_used": 1642476000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24373948574066162,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505127906799316,
      "objective/train/weighted_lm_loss": 1.3535220623016357,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5205650294091804,
      "tokens_seen": 1622016000
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005137612838515546,
      "loss": 1.3918,
      "theoretical_loss": 3.5205650294091804,
      "tokens_seen": 1622016000
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005137412236710131,
      "loss": 1.4156,
      "theoretical_loss": 3.520552775324888,
      "tokens_seen": 1622081536
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005137211634904713,
      "loss": 1.4092,
      "theoretical_loss": 3.5205405218743007,
      "tokens_seen": 1622147072
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005137011033099298,
      "loss": 1.4367,
      "theoretical_loss": 3.520528269057359,
      "tokens_seen": 1622212608
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005136810431293881,
      "loss": 1.3933,
      "theoretical_loss": 3.520516016874005,
      "tokens_seen": 1622278144
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005136609829488465,
      "loss": 1.4716,
      "theoretical_loss": 3.520503765324181,
      "tokens_seen": 1622343680
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005136409227683049,
      "loss": 1.5016,
      "theoretical_loss": 3.520491514407828,
      "tokens_seen": 1622409216
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005136208625877633,
      "loss": 1.3791,
      "theoretical_loss": 3.5204792641248877,
      "tokens_seen": 1622474752
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005136008024072216,
      "loss": 1.4061,
      "theoretical_loss": 3.520467014475302,
      "tokens_seen": 1622540288
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005135807422266801,
      "loss": 1.5141,
      "theoretical_loss": 3.5204547654590126,
      "tokens_seen": 1622605824
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005135606820461384,
      "loss": 1.4947,
      "theoretical_loss": 3.520442517075961,
      "tokens_seen": 1622671360
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005135406218655968,
      "loss": 1.4929,
      "theoretical_loss": 3.5204302693260883,
      "tokens_seen": 1622736896
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005135205616850551,
      "loss": 1.4366,
      "theoretical_loss": 3.5204180222093373,
      "tokens_seen": 1622802432
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005135005015045135,
      "loss": 1.4245,
      "theoretical_loss": 3.5204057757256493,
      "tokens_seen": 1622867968
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000513480441323972,
      "loss": 1.2851,
      "theoretical_loss": 3.5203935298749656,
      "tokens_seen": 1622933504
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005134603811434303,
      "loss": 1.4682,
      "theoretical_loss": 3.520381284657229,
      "tokens_seen": 1622999040
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005134403209628887,
      "loss": 1.3677,
      "theoretical_loss": 3.5203690400723797,
      "tokens_seen": 1623064576
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000513420260782347,
      "loss": 1.4476,
      "theoretical_loss": 3.5203567961203603,
      "tokens_seen": 1623130112
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005134002006018055,
      "loss": 1.4217,
      "theoretical_loss": 3.520344552801113,
      "tokens_seen": 1623195648
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005133801404212638,
      "loss": 1.4457,
      "theoretical_loss": 3.5203323101145787,
      "tokens_seen": 1623261184
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005133600802407222,
      "loss": 1.4314,
      "theoretical_loss": 3.5203200680606996,
      "tokens_seen": 1623326720
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005133400200601806,
      "loss": 1.441,
      "theoretical_loss": 3.520307826639417,
      "tokens_seen": 1623392256
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000513319959879639,
      "loss": 1.5204,
      "theoretical_loss": 3.5202955858506737,
      "tokens_seen": 1623457792
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005132998996990973,
      "loss": 1.5003,
      "theoretical_loss": 3.5202833456944105,
      "tokens_seen": 1623523328
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005132798395185558,
      "loss": 1.4916,
      "theoretical_loss": 3.5202711061705694,
      "tokens_seen": 1623588864
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000513259779338014,
      "loss": 1.5026,
      "theoretical_loss": 3.520258867279092,
      "tokens_seen": 1623654400
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005132397191574725,
      "loss": 1.5152,
      "theoretical_loss": 3.520246629019921,
      "tokens_seen": 1623719936
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005132196589769308,
      "loss": 1.4659,
      "theoretical_loss": 3.5202343913929974,
      "tokens_seen": 1623785472
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005131995987963892,
      "loss": 1.3743,
      "theoretical_loss": 3.520222154398263,
      "tokens_seen": 1623851008
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005131795386158476,
      "loss": 1.5066,
      "theoretical_loss": 3.5202099180356603,
      "tokens_seen": 1623916544
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005131594784353059,
      "loss": 1.3954,
      "theoretical_loss": 3.5201976823051306,
      "tokens_seen": 1623982080
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005131394182547643,
      "loss": 1.3975,
      "theoretical_loss": 3.5201854472066154,
      "tokens_seen": 1624047616
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005131193580742228,
      "loss": 1.5217,
      "theoretical_loss": 3.5201732127400573,
      "tokens_seen": 1624113152
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000513099297893681,
      "loss": 1.4817,
      "theoretical_loss": 3.5201609789053983,
      "tokens_seen": 1624178688
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005130792377131395,
      "loss": 1.4752,
      "theoretical_loss": 3.5201487457025795,
      "tokens_seen": 1624244224
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005130591775325977,
      "loss": 1.4078,
      "theoretical_loss": 3.520136513131543,
      "tokens_seen": 1624309760
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005130391173520562,
      "loss": 1.5055,
      "theoretical_loss": 3.520124281192231,
      "tokens_seen": 1624375296
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005130190571715146,
      "loss": 1.5188,
      "theoretical_loss": 3.520112049884585,
      "tokens_seen": 1624440832
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005129989969909729,
      "loss": 1.3225,
      "theoretical_loss": 3.5200998192085473,
      "tokens_seen": 1624506368
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005129789368104313,
      "loss": 1.4679,
      "theoretical_loss": 3.520087589164059,
      "tokens_seen": 1624571904
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005129588766298898,
      "loss": 1.3912,
      "theoretical_loss": 3.520075359751063,
      "tokens_seen": 1624637440
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000512938816449348,
      "loss": 1.4284,
      "theoretical_loss": 3.5200631309695014,
      "tokens_seen": 1624702976
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005129187562688065,
      "loss": 1.4134,
      "theoretical_loss": 3.5200509028193148,
      "tokens_seen": 1624768512
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005128986960882647,
      "loss": 1.4578,
      "theoretical_loss": 3.5200386753004462,
      "tokens_seen": 1624834048
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005128786359077232,
      "loss": 1.5794,
      "theoretical_loss": 3.520026448412837,
      "tokens_seen": 1624899584
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005128585757271816,
      "loss": 1.4093,
      "theoretical_loss": 3.5200142221564295,
      "tokens_seen": 1624965120
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005128385155466399,
      "loss": 1.522,
      "theoretical_loss": 3.5200019965311657,
      "tokens_seen": 1625030656
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005128184553660983,
      "loss": 1.4246,
      "theoretical_loss": 3.519989771536987,
      "tokens_seen": 1625096192
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005127983951855567,
      "loss": 1.5152,
      "theoretical_loss": 3.5199775471738364,
      "tokens_seen": 1625161728
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000512778335005015,
      "loss": 1.4151,
      "theoretical_loss": 3.519965323441655,
      "tokens_seen": 1625227264
    },
    {
      "epoch": 0.49,
      "objective/train/advantage_avg": 0.4928131699562073,
      "objective/train/docs_used": 593743,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.588471531867981,
      "objective/train/original_loss": 1.5884714126586914,
      "objective/train/theoretical_loss": 3.519953100340385,
      "objective/train/tokens_used": 1645752800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24342887103557587,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505188703536987,
      "objective/train/weighted_lm_loss": 1.6688486337661743,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.519953100340385,
      "tokens_seen": 1625292800
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005127582748244735,
      "loss": 1.5407,
      "theoretical_loss": 3.519953100340385,
      "tokens_seen": 1625292800
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005127382146439317,
      "loss": 1.4568,
      "theoretical_loss": 3.5199408778699683,
      "tokens_seen": 1625358336
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005127181544633902,
      "loss": 1.5122,
      "theoretical_loss": 3.5199286560303475,
      "tokens_seen": 1625423872
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005126980942828485,
      "loss": 1.6539,
      "theoretical_loss": 3.5199164348214635,
      "tokens_seen": 1625489408
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005126780341023069,
      "loss": 1.4194,
      "theoretical_loss": 3.51990421424326,
      "tokens_seen": 1625554944
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005126579739217653,
      "loss": 1.4811,
      "theoretical_loss": 3.5198919942956772,
      "tokens_seen": 1625620480
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005126379137412237,
      "loss": 1.4598,
      "theoretical_loss": 3.5198797749786586,
      "tokens_seen": 1625686016
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000512617853560682,
      "loss": 1.4493,
      "theoretical_loss": 3.519867556292145,
      "tokens_seen": 1625751552
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005125977933801405,
      "loss": 1.431,
      "theoretical_loss": 3.519855338236079,
      "tokens_seen": 1625817088
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005125777331995988,
      "loss": 1.3155,
      "theoretical_loss": 3.5198431208104033,
      "tokens_seen": 1625882624
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005125576730190572,
      "loss": 1.4685,
      "theoretical_loss": 3.519830904015059,
      "tokens_seen": 1625948160
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005125376128385155,
      "loss": 1.4149,
      "theoretical_loss": 3.519818687849989,
      "tokens_seen": 1626013696
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000512517552657974,
      "loss": 1.4636,
      "theoretical_loss": 3.5198064723151345,
      "tokens_seen": 1626079232
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005124974924774323,
      "loss": 1.389,
      "theoretical_loss": 3.519794257410438,
      "tokens_seen": 1626144768
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005124774322968907,
      "loss": 1.5225,
      "theoretical_loss": 3.5197820431358418,
      "tokens_seen": 1626210304
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000512457372116349,
      "loss": 1.536,
      "theoretical_loss": 3.5197698294912882,
      "tokens_seen": 1626275840
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005124373119358074,
      "loss": 1.4133,
      "theoretical_loss": 3.519757616476719,
      "tokens_seen": 1626341376
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005124172517552658,
      "loss": 1.4445,
      "theoretical_loss": 3.5197454040920757,
      "tokens_seen": 1626406912
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005123971915747242,
      "loss": 1.4078,
      "theoretical_loss": 3.519733192337301,
      "tokens_seen": 1626472448
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005123771313941825,
      "loss": 1.3915,
      "theoretical_loss": 3.519720981212337,
      "tokens_seen": 1626537984
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000512357071213641,
      "loss": 1.4938,
      "theoretical_loss": 3.519708770717126,
      "tokens_seen": 1626603520
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005123370110330993,
      "loss": 1.5993,
      "theoretical_loss": 3.5196965608516106,
      "tokens_seen": 1626669056
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005123169508525577,
      "loss": 1.3959,
      "theoretical_loss": 3.519684351615732,
      "tokens_seen": 1626734592
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000512296890672016,
      "loss": 1.4458,
      "theoretical_loss": 3.5196721430094327,
      "tokens_seen": 1626800128
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005122768304914744,
      "loss": 1.5007,
      "theoretical_loss": 3.519659935032655,
      "tokens_seen": 1626865664
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005122567703109328,
      "loss": 1.3519,
      "theoretical_loss": 3.5196477276853413,
      "tokens_seen": 1626931200
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005122367101303912,
      "loss": 1.5504,
      "theoretical_loss": 3.519635520967433,
      "tokens_seen": 1626996736
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005122166499498495,
      "loss": 1.4964,
      "theoretical_loss": 3.5196233148788734,
      "tokens_seen": 1627062272
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000512196589769308,
      "loss": 1.4648,
      "theoretical_loss": 3.5196111094196034,
      "tokens_seen": 1627127808
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005121765295887662,
      "loss": 1.4161,
      "theoretical_loss": 3.519598904589566,
      "tokens_seen": 1627193344
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005121564694082247,
      "loss": 1.3668,
      "theoretical_loss": 3.519586700388704,
      "tokens_seen": 1627258880
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000512136409227683,
      "loss": 1.393,
      "theoretical_loss": 3.5195744968169587,
      "tokens_seen": 1627324416
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005121163490471414,
      "loss": 1.48,
      "theoretical_loss": 3.5195622938742726,
      "tokens_seen": 1627389952
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005120962888665998,
      "loss": 1.4286,
      "theoretical_loss": 3.5195500915605877,
      "tokens_seen": 1627455488
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005120762286860581,
      "loss": 1.4175,
      "theoretical_loss": 3.5195378898758465,
      "tokens_seen": 1627521024
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005120561685055165,
      "loss": 1.5432,
      "theoretical_loss": 3.519525688819992,
      "tokens_seen": 1627586560
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000512036108324975,
      "loss": 1.4954,
      "theoretical_loss": 3.519513488392965,
      "tokens_seen": 1627652096
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005120160481444332,
      "loss": 1.5546,
      "theoretical_loss": 3.519501288594708,
      "tokens_seen": 1627717632
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005119959879638917,
      "loss": 1.3708,
      "theoretical_loss": 3.519489089425165,
      "tokens_seen": 1627783168
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005119759277833502,
      "loss": 1.505,
      "theoretical_loss": 3.5194768908842766,
      "tokens_seen": 1627848704
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005119558676028084,
      "loss": 1.4073,
      "theoretical_loss": 3.5194646929719853,
      "tokens_seen": 1627914240
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005119358074222669,
      "loss": 1.4478,
      "theoretical_loss": 3.519452495688234,
      "tokens_seen": 1627979776
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005119157472417251,
      "loss": 1.4677,
      "theoretical_loss": 3.519440299032964,
      "tokens_seen": 1628045312
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005118956870611836,
      "loss": 1.5363,
      "theoretical_loss": 3.519428103006119,
      "tokens_seen": 1628110848
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000511875626880642,
      "loss": 1.3957,
      "theoretical_loss": 3.519415907607641,
      "tokens_seen": 1628176384
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005118555667001003,
      "loss": 1.3616,
      "theoretical_loss": 3.5194037128374713,
      "tokens_seen": 1628241920
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005118355065195587,
      "loss": 1.4641,
      "theoretical_loss": 3.519391518695553,
      "tokens_seen": 1628307456
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005118154463390171,
      "loss": 1.3699,
      "theoretical_loss": 3.5193793251818284,
      "tokens_seen": 1628372992
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005117953861584754,
      "loss": 1.4798,
      "theoretical_loss": 3.5193671322962397,
      "tokens_seen": 1628438528
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005117753259779339,
      "loss": 1.4665,
      "theoretical_loss": 3.5193549400387294,
      "tokens_seen": 1628504064
    },
    {
      "epoch": 0.49,
      "objective/train/advantage_avg": 0.49220892786979675,
      "objective/train/docs_used": 595003,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3413543701171875,
      "objective/train/original_loss": 1.3413543701171875,
      "objective/train/theoretical_loss": 3.5193427484092403,
      "objective/train/tokens_used": 1649029600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24361661076545715,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504592657089233,
      "objective/train/weighted_lm_loss": 1.4089620113372803,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5193427484092403,
      "tokens_seen": 1628569600
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005117552657973921,
      "loss": 1.3646,
      "theoretical_loss": 3.5193427484092403,
      "tokens_seen": 1628569600
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005117352056168506,
      "loss": 1.4568,
      "theoretical_loss": 3.519330557407714,
      "tokens_seen": 1628635136
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005117151454363089,
      "loss": 1.543,
      "theoretical_loss": 3.519318367034093,
      "tokens_seen": 1628700672
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005116950852557673,
      "loss": 1.4145,
      "theoretical_loss": 3.5193061772883203,
      "tokens_seen": 1628766208
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005116750250752257,
      "loss": 1.3922,
      "theoretical_loss": 3.519293988170338,
      "tokens_seen": 1628831744
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005116549648946841,
      "loss": 1.5015,
      "theoretical_loss": 3.519281799680088,
      "tokens_seen": 1628897280
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005116349047141424,
      "loss": 1.4153,
      "theoretical_loss": 3.519269611817513,
      "tokens_seen": 1628962816
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005116148445336009,
      "loss": 1.4745,
      "theoretical_loss": 3.5192574245825563,
      "tokens_seen": 1629028352
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005115947843530592,
      "loss": 1.456,
      "theoretical_loss": 3.5192452379751593,
      "tokens_seen": 1629093888
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005115747241725176,
      "loss": 1.4306,
      "theoretical_loss": 3.5192330519952644,
      "tokens_seen": 1629159424
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005115546639919759,
      "loss": 1.5392,
      "theoretical_loss": 3.5192208666428146,
      "tokens_seen": 1629224960
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005115346038114343,
      "loss": 1.5362,
      "theoretical_loss": 3.5192086819177524,
      "tokens_seen": 1629290496
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005115145436308927,
      "loss": 1.3792,
      "theoretical_loss": 3.51919649782002,
      "tokens_seen": 1629356032
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005114944834503511,
      "loss": 1.4496,
      "theoretical_loss": 3.5191843143495594,
      "tokens_seen": 1629421568
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005114744232698094,
      "loss": 1.4414,
      "theoretical_loss": 3.5191721315063136,
      "tokens_seen": 1629487104
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005114543630892678,
      "loss": 1.4484,
      "theoretical_loss": 3.5191599492902252,
      "tokens_seen": 1629552640
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005114343029087262,
      "loss": 1.3812,
      "theoretical_loss": 3.519147767701237,
      "tokens_seen": 1629618176
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005114142427281846,
      "loss": 1.36,
      "theoretical_loss": 3.519135586739291,
      "tokens_seen": 1629683712
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005113941825476429,
      "loss": 1.4703,
      "theoretical_loss": 3.5191234064043293,
      "tokens_seen": 1629749248
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005113741223671014,
      "loss": 1.3989,
      "theoretical_loss": 3.519111226696295,
      "tokens_seen": 1629814784
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005113540621865597,
      "loss": 1.3904,
      "theoretical_loss": 3.5190990476151307,
      "tokens_seen": 1629880320
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005113340020060181,
      "loss": 1.4403,
      "theoretical_loss": 3.5190868691607786,
      "tokens_seen": 1629945856
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005113139418254764,
      "loss": 1.393,
      "theoretical_loss": 3.5190746913331816,
      "tokens_seen": 1630011392
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005112938816449348,
      "loss": 1.4698,
      "theoretical_loss": 3.5190625141322815,
      "tokens_seen": 1630076928
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005112738214643932,
      "loss": 1.501,
      "theoretical_loss": 3.5190503375580215,
      "tokens_seen": 1630142464
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005112537612838516,
      "loss": 1.3957,
      "theoretical_loss": 3.5190381616103448,
      "tokens_seen": 1630208000
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005112337011033099,
      "loss": 1.4805,
      "theoretical_loss": 3.5190259862891926,
      "tokens_seen": 1630273536
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005112136409227684,
      "loss": 1.4449,
      "theoretical_loss": 3.5190138115945078,
      "tokens_seen": 1630339072
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005111935807422266,
      "loss": 1.3245,
      "theoretical_loss": 3.5190016375262334,
      "tokens_seen": 1630404608
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005111735205616851,
      "loss": 1.4834,
      "theoretical_loss": 3.5189894640843122,
      "tokens_seen": 1630470144
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005111534603811434,
      "loss": 1.4029,
      "theoretical_loss": 3.518977291268686,
      "tokens_seen": 1630535680
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005111334002006018,
      "loss": 1.438,
      "theoretical_loss": 3.5189651190792985,
      "tokens_seen": 1630601216
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005111133400200602,
      "loss": 1.4284,
      "theoretical_loss": 3.518952947516091,
      "tokens_seen": 1630666752
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005110932798395185,
      "loss": 1.5068,
      "theoretical_loss": 3.5189407765790075,
      "tokens_seen": 1630732288
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005110732196589769,
      "loss": 1.4893,
      "theoretical_loss": 3.5189286062679894,
      "tokens_seen": 1630797824
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005110531594784354,
      "loss": 1.473,
      "theoretical_loss": 3.5189164365829804,
      "tokens_seen": 1630863360
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005110330992978936,
      "loss": 1.4284,
      "theoretical_loss": 3.518904267523922,
      "tokens_seen": 1630928896
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005110130391173521,
      "loss": 1.3846,
      "theoretical_loss": 3.518892099090758,
      "tokens_seen": 1630994432
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005109929789368105,
      "loss": 1.4881,
      "theoretical_loss": 3.51887993128343,
      "tokens_seen": 1631059968
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005109729187562688,
      "loss": 1.3844,
      "theoretical_loss": 3.518867764101882,
      "tokens_seen": 1631125504
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005109528585757272,
      "loss": 1.4262,
      "theoretical_loss": 3.5188555975460556,
      "tokens_seen": 1631191040
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005109327983951855,
      "loss": 1.2931,
      "theoretical_loss": 3.518843431615893,
      "tokens_seen": 1631256576
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005109127382146439,
      "loss": 1.373,
      "theoretical_loss": 3.518831266311339,
      "tokens_seen": 1631322112
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005108926780341024,
      "loss": 1.5087,
      "theoretical_loss": 3.5188191016323334,
      "tokens_seen": 1631387648
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005108726178535606,
      "loss": 1.3485,
      "theoretical_loss": 3.5188069375788213,
      "tokens_seen": 1631453184
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005108525576730191,
      "loss": 1.5223,
      "theoretical_loss": 3.518794774150745,
      "tokens_seen": 1631518720
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005108324974924774,
      "loss": 1.4668,
      "theoretical_loss": 3.518782611348046,
      "tokens_seen": 1631584256
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005108124373119358,
      "loss": 1.5214,
      "theoretical_loss": 3.5187704491706686,
      "tokens_seen": 1631649792
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005107923771313942,
      "loss": 1.4162,
      "theoretical_loss": 3.518758287618554,
      "tokens_seen": 1631715328
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005107723169508525,
      "loss": 1.4174,
      "theoretical_loss": 3.5187461266916458,
      "tokens_seen": 1631780864
    },
    {
      "epoch": 0.49,
      "objective/train/advantage_avg": 0.4951614439487457,
      "objective/train/docs_used": 596193,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4828615188598633,
      "objective/train/original_loss": 1.4828615188598633,
      "objective/train/theoretical_loss": 3.5187339663898873,
      "objective/train/tokens_used": 1652306400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24534478783607483,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050763487815857,
      "objective/train/weighted_lm_loss": 1.5581560134887695,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5187339663898873,
      "tokens_seen": 1631846400
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005107522567703109,
      "loss": 1.4907,
      "theoretical_loss": 3.5187339663898873,
      "tokens_seen": 1631846400
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005107321965897693,
      "loss": 1.4851,
      "theoretical_loss": 3.51872180671322,
      "tokens_seen": 1631911936
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005107121364092276,
      "loss": 1.4882,
      "theoretical_loss": 3.518709647661588,
      "tokens_seen": 1631977472
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005106920762286861,
      "loss": 1.4275,
      "theoretical_loss": 3.518697489234932,
      "tokens_seen": 1632043008
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005106720160481444,
      "loss": 1.471,
      "theoretical_loss": 3.5186853314331974,
      "tokens_seen": 1632108544
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005106519558676028,
      "loss": 1.4868,
      "theoretical_loss": 3.518673174256325,
      "tokens_seen": 1632174080
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005106318956870612,
      "loss": 1.4042,
      "theoretical_loss": 3.5186610177042588,
      "tokens_seen": 1632239616
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005106118355065196,
      "loss": 1.466,
      "theoretical_loss": 3.518648861776941,
      "tokens_seen": 1632305152
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005105917753259779,
      "loss": 1.4563,
      "theoretical_loss": 3.5186367064743145,
      "tokens_seen": 1632370688
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005105717151454363,
      "loss": 1.5281,
      "theoretical_loss": 3.518624551796322,
      "tokens_seen": 1632436224
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005105516549648946,
      "loss": 1.5054,
      "theoretical_loss": 3.5186123977429067,
      "tokens_seen": 1632501760
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005105315947843531,
      "loss": 1.5189,
      "theoretical_loss": 3.518600244314011,
      "tokens_seen": 1632567296
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005105115346038114,
      "loss": 1.4845,
      "theoretical_loss": 3.518588091509578,
      "tokens_seen": 1632632832
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005104914744232698,
      "loss": 1.4663,
      "theoretical_loss": 3.518575939329551,
      "tokens_seen": 1632698368
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005104714142427282,
      "loss": 1.4789,
      "theoretical_loss": 3.5185637877738714,
      "tokens_seen": 1632763904
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005104513540621866,
      "loss": 1.4584,
      "theoretical_loss": 3.518551636842484,
      "tokens_seen": 1632829440
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000510431293881645,
      "loss": 1.3415,
      "theoretical_loss": 3.51853948653533,
      "tokens_seen": 1632894976
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005104112337011033,
      "loss": 1.3569,
      "theoretical_loss": 3.5185273368523537,
      "tokens_seen": 1632960512
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005103911735205618,
      "loss": 1.483,
      "theoretical_loss": 3.518515187793497,
      "tokens_seen": 1633026048
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005103711133400201,
      "loss": 1.4525,
      "theoretical_loss": 3.518503039358703,
      "tokens_seen": 1633091584
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005103510531594785,
      "loss": 1.4045,
      "theoretical_loss": 3.5184908915479145,
      "tokens_seen": 1633157120
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005103309929789368,
      "loss": 1.363,
      "theoretical_loss": 3.518478744361074,
      "tokens_seen": 1633222656
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005103109327983952,
      "loss": 1.4174,
      "theoretical_loss": 3.5184665977981258,
      "tokens_seen": 1633288192
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005102908726178536,
      "loss": 1.433,
      "theoretical_loss": 3.518454451859012,
      "tokens_seen": 1633353728
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000510270812437312,
      "loss": 1.5431,
      "theoretical_loss": 3.5184423065436756,
      "tokens_seen": 1633419264
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0005102507522567703,
      "loss": 1.4173,
      "theoretical_loss": 3.5184301618520593,
      "tokens_seen": 1633484800
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005102306920762288,
      "loss": 1.4513,
      "theoretical_loss": 3.5184180177841062,
      "tokens_seen": 1633550336
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000510210631895687,
      "loss": 1.5099,
      "theoretical_loss": 3.5184058743397593,
      "tokens_seen": 1633615872
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005101905717151455,
      "loss": 1.3596,
      "theoretical_loss": 3.5183937315189615,
      "tokens_seen": 1633681408
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005101705115346038,
      "loss": 1.4337,
      "theoretical_loss": 3.518381589321656,
      "tokens_seen": 1633746944
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005101504513540622,
      "loss": 1.4424,
      "theoretical_loss": 3.5183694477477854,
      "tokens_seen": 1633812480
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005101303911735206,
      "loss": 1.3498,
      "theoretical_loss": 3.5183573067972933,
      "tokens_seen": 1633878016
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005101103309929789,
      "loss": 1.4065,
      "theoretical_loss": 3.5183451664701217,
      "tokens_seen": 1633943552
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005100902708124373,
      "loss": 1.4141,
      "theoretical_loss": 3.5183330267662143,
      "tokens_seen": 1634009088
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005100702106318958,
      "loss": 1.4796,
      "theoretical_loss": 3.5183208876855145,
      "tokens_seen": 1634074624
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000510050150451354,
      "loss": 1.4219,
      "theoretical_loss": 3.5183087492279643,
      "tokens_seen": 1634140160
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005100300902708125,
      "loss": 1.4264,
      "theoretical_loss": 3.5182966113935072,
      "tokens_seen": 1634205696
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005100100300902709,
      "loss": 1.459,
      "theoretical_loss": 3.518284474182086,
      "tokens_seen": 1634271232
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005099899699097292,
      "loss": 1.3812,
      "theoretical_loss": 3.5182723375936447,
      "tokens_seen": 1634336768
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005099699097291876,
      "loss": 1.4315,
      "theoretical_loss": 3.5182602016281255,
      "tokens_seen": 1634402304
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005099498495486459,
      "loss": 1.4926,
      "theoretical_loss": 3.518248066285471,
      "tokens_seen": 1634467840
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005099297893681043,
      "loss": 1.4144,
      "theoretical_loss": 3.5182359315656253,
      "tokens_seen": 1634533376
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005099097291875628,
      "loss": 1.416,
      "theoretical_loss": 3.518223797468531,
      "tokens_seen": 1634598912
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000509889669007021,
      "loss": 1.3594,
      "theoretical_loss": 3.518211663994131,
      "tokens_seen": 1634664448
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005098696088264795,
      "loss": 1.4154,
      "theoretical_loss": 3.518199531142369,
      "tokens_seen": 1634729984
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005098495486459378,
      "loss": 1.4882,
      "theoretical_loss": 3.518187398913187,
      "tokens_seen": 1634795520
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005098294884653962,
      "loss": 1.4328,
      "theoretical_loss": 3.518175267306529,
      "tokens_seen": 1634861056
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005098094282848546,
      "loss": 1.4493,
      "theoretical_loss": 3.5181631363223382,
      "tokens_seen": 1634926592
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000509789368104313,
      "loss": 1.4685,
      "theoretical_loss": 3.518151005960557,
      "tokens_seen": 1634992128
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005097693079237713,
      "loss": 1.4587,
      "theoretical_loss": 3.518138876221129,
      "tokens_seen": 1635057664
    },
    {
      "epoch": 0.5,
      "objective/train/advantage_avg": 0.4936535060405731,
      "objective/train/docs_used": 597373,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4009824991226196,
      "objective/train/original_loss": 1.40098237991333,
      "objective/train/theoretical_loss": 3.518126747103997,
      "objective/train/tokens_used": 1655583200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24421419203281403,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506069660186768,
      "objective/train/weighted_lm_loss": 1.4719345569610596,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.518126747103997,
      "tokens_seen": 1635123200
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005097492477432297,
      "loss": 1.4195,
      "theoretical_loss": 3.518126747103997,
      "tokens_seen": 1635123200
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000509729187562688,
      "loss": 1.44,
      "theoretical_loss": 3.5181146186091046,
      "tokens_seen": 1635188736
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005097091273821465,
      "loss": 1.4648,
      "theoretical_loss": 3.5181024907363945,
      "tokens_seen": 1635254272
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005096890672016048,
      "loss": 1.4999,
      "theoretical_loss": 3.5180903634858103,
      "tokens_seen": 1635319808
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005096690070210632,
      "loss": 1.3041,
      "theoretical_loss": 3.5180782368572947,
      "tokens_seen": 1635385344
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005096489468405216,
      "loss": 1.5518,
      "theoretical_loss": 3.5180661108507914,
      "tokens_seen": 1635450880
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00050962888665998,
      "loss": 1.3715,
      "theoretical_loss": 3.518053985466243,
      "tokens_seen": 1635516416
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005096088264794383,
      "loss": 1.3727,
      "theoretical_loss": 3.5180418607035926,
      "tokens_seen": 1635581952
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005095887662988967,
      "loss": 1.349,
      "theoretical_loss": 3.518029736562784,
      "tokens_seen": 1635647488
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000509568706118355,
      "loss": 1.3814,
      "theoretical_loss": 3.51801761304376,
      "tokens_seen": 1635713024
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005095486459378135,
      "loss": 1.3724,
      "theoretical_loss": 3.518005490146464,
      "tokens_seen": 1635778560
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005095285857572718,
      "loss": 1.4146,
      "theoretical_loss": 3.517993367870839,
      "tokens_seen": 1635844096
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005095085255767302,
      "loss": 1.4044,
      "theoretical_loss": 3.517981246216828,
      "tokens_seen": 1635909632
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005094884653961885,
      "loss": 1.2723,
      "theoretical_loss": 3.517969125184375,
      "tokens_seen": 1635975168
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000509468405215647,
      "loss": 1.359,
      "theoretical_loss": 3.5179570047734225,
      "tokens_seen": 1636040704
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005094483450351053,
      "loss": 1.4737,
      "theoretical_loss": 3.5179448849839146,
      "tokens_seen": 1636106240
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005094282848545637,
      "loss": 1.5032,
      "theoretical_loss": 3.5179327658157935,
      "tokens_seen": 1636171776
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000509408224674022,
      "loss": 1.5555,
      "theoretical_loss": 3.517920647269003,
      "tokens_seen": 1636237312
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005093881644934805,
      "loss": 1.5415,
      "theoretical_loss": 3.5179085293434857,
      "tokens_seen": 1636302848
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005093681043129388,
      "loss": 1.232,
      "theoretical_loss": 3.5178964120391862,
      "tokens_seen": 1636368384
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005093480441323972,
      "loss": 1.3615,
      "theoretical_loss": 3.5178842953560467,
      "tokens_seen": 1636433920
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005093279839518555,
      "loss": 1.431,
      "theoretical_loss": 3.5178721792940113,
      "tokens_seen": 1636499456
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000509307923771314,
      "loss": 1.3797,
      "theoretical_loss": 3.517860063853022,
      "tokens_seen": 1636564992
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005092878635907723,
      "loss": 1.3994,
      "theoretical_loss": 3.5178479490330234,
      "tokens_seen": 1636630528
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005092678034102307,
      "loss": 1.3639,
      "theoretical_loss": 3.5178358348339582,
      "tokens_seen": 1636696064
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005092477432296891,
      "loss": 1.4543,
      "theoretical_loss": 3.5178237212557697,
      "tokens_seen": 1636761600
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005092276830491474,
      "loss": 1.5536,
      "theoretical_loss": 3.517811608298401,
      "tokens_seen": 1636827136
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005092076228686058,
      "loss": 1.3269,
      "theoretical_loss": 3.517799495961796,
      "tokens_seen": 1636892672
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005091875626880642,
      "loss": 1.53,
      "theoretical_loss": 3.5177873842458984,
      "tokens_seen": 1636958208
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005091675025075225,
      "loss": 1.5147,
      "theoretical_loss": 3.5177752731506504,
      "tokens_seen": 1637023744
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000509147442326981,
      "loss": 1.4758,
      "theoretical_loss": 3.5177631626759958,
      "tokens_seen": 1637089280
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005091273821464392,
      "loss": 1.3881,
      "theoretical_loss": 3.5177510528218785,
      "tokens_seen": 1637154816
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005091073219658977,
      "loss": 1.3757,
      "theoretical_loss": 3.517738943588241,
      "tokens_seen": 1637220352
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005090872617853561,
      "loss": 1.4075,
      "theoretical_loss": 3.5177268349750275,
      "tokens_seen": 1637285888
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005090672016048144,
      "loss": 1.3407,
      "theoretical_loss": 3.5177147269821805,
      "tokens_seen": 1637351424
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005090471414242728,
      "loss": 1.4138,
      "theoretical_loss": 3.5177026196096444,
      "tokens_seen": 1637416960
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005090270812437313,
      "loss": 1.4246,
      "theoretical_loss": 3.517690512857361,
      "tokens_seen": 1637482496
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005090070210631896,
      "loss": 1.3554,
      "theoretical_loss": 3.5176784067252758,
      "tokens_seen": 1637548032
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000508986960882648,
      "loss": 1.403,
      "theoretical_loss": 3.517666301213331,
      "tokens_seen": 1637613568
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005089669007021063,
      "loss": 1.4179,
      "theoretical_loss": 3.5176541963214696,
      "tokens_seen": 1637679104
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005089468405215647,
      "loss": 1.2757,
      "theoretical_loss": 3.517642092049636,
      "tokens_seen": 1637744640
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005089267803410232,
      "loss": 1.431,
      "theoretical_loss": 3.5176299883977733,
      "tokens_seen": 1637810176
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005089067201604814,
      "loss": 1.5824,
      "theoretical_loss": 3.5176178853658246,
      "tokens_seen": 1637875712
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005088866599799399,
      "loss": 1.5174,
      "theoretical_loss": 3.517605782953734,
      "tokens_seen": 1637941248
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005088665997993982,
      "loss": 1.4067,
      "theoretical_loss": 3.5175936811614443,
      "tokens_seen": 1638006784
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005088465396188566,
      "loss": 1.4094,
      "theoretical_loss": 3.5175815799888994,
      "tokens_seen": 1638072320
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000508826479438315,
      "loss": 1.5601,
      "theoretical_loss": 3.5175694794360424,
      "tokens_seen": 1638137856
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005088064192577733,
      "loss": 1.5274,
      "theoretical_loss": 3.517557379502817,
      "tokens_seen": 1638203392
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005087863590772317,
      "loss": 1.4456,
      "theoretical_loss": 3.517545280189167,
      "tokens_seen": 1638268928
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005087662988966901,
      "loss": 1.4837,
      "theoretical_loss": 3.5175331814950352,
      "tokens_seen": 1638334464
    },
    {
      "debugging/Compilability": 1.0,
      "debugging/distinct-1-grams": 0.7474626180525646,
      "debugging/entropy-1-grams": 5.599013278347353,
      "debugging/length": 499.42857142857144,
      "debugging/num_segments": 14,
      "debugging/raw_token_scores_avg": 0.007199673913419247,
      "debugging/raw_token_scores_std": 0.028890423476696014,
      "debugging/score": 0.0110880360129877,
      "debugging/score_std": 0.01573414919444883,
      "epoch": 0.5,
      "objective/train/advantage_avg": 0.4927961230278015,
      "objective/train/docs_used": 598618,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3778165578842163,
      "objective/train/original_loss": 1.3778164386749268,
      "objective/train/theoretical_loss": 3.5175210834203656,
      "objective/train/tokens_used": 1658860000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24368339776992798,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505183935165405,
      "objective/train/weighted_lm_loss": 1.4475257396697998,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5175210834203656,
      "tokens_seen": 1638400000
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005087462387161484,
      "loss": 1.3705,
      "theoretical_loss": 3.5175210834203656,
      "tokens_seen": 1638400000
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005087261785356069,
      "loss": 1.4866,
      "theoretical_loss": 3.517508985965102,
      "tokens_seen": 1638465536
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005087061183550652,
      "loss": 1.5003,
      "theoretical_loss": 3.5174968891291867,
      "tokens_seen": 1638531072
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005086860581745236,
      "loss": 1.5442,
      "theoretical_loss": 3.5174847929125646,
      "tokens_seen": 1638596608
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000508665997993982,
      "loss": 1.5592,
      "theoretical_loss": 3.5174726973151778,
      "tokens_seen": 1638662144
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005086459378134404,
      "loss": 1.5096,
      "theoretical_loss": 3.517460602336971,
      "tokens_seen": 1638727680
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005086258776328987,
      "loss": 1.4653,
      "theoretical_loss": 3.517448507977888,
      "tokens_seen": 1638793216
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005086058174523571,
      "loss": 1.2198,
      "theoretical_loss": 3.5174364142378716,
      "tokens_seen": 1638858752
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005085857572718154,
      "loss": 1.4434,
      "theoretical_loss": 3.517424321116865,
      "tokens_seen": 1638924288
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005085656970912739,
      "loss": 1.4503,
      "theoretical_loss": 3.5174122286148126,
      "tokens_seen": 1638989824
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005085456369107322,
      "loss": 1.4634,
      "theoretical_loss": 3.5174001367316583,
      "tokens_seen": 1639055360
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005085255767301906,
      "loss": 1.4677,
      "theoretical_loss": 3.517388045467344,
      "tokens_seen": 1639120896
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005085055165496489,
      "loss": 1.4961,
      "theoretical_loss": 3.517375954821815,
      "tokens_seen": 1639186432
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005084854563691074,
      "loss": 1.4132,
      "theoretical_loss": 3.517363864795014,
      "tokens_seen": 1639251968
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005084653961885657,
      "loss": 1.3001,
      "theoretical_loss": 3.5173517753868846,
      "tokens_seen": 1639317504
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005084453360080241,
      "loss": 1.5192,
      "theoretical_loss": 3.5173396865973707,
      "tokens_seen": 1639383040
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005084252758274824,
      "loss": 1.4767,
      "theoretical_loss": 3.517327598426416,
      "tokens_seen": 1639448576
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005084052156469409,
      "loss": 1.3634,
      "theoretical_loss": 3.5173155108739644,
      "tokens_seen": 1639514112
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005083851554663992,
      "loss": 1.414,
      "theoretical_loss": 3.5173034239399583,
      "tokens_seen": 1639579648
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005083650952858576,
      "loss": 1.3725,
      "theoretical_loss": 3.517291337624343,
      "tokens_seen": 1639645184
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005083450351053159,
      "loss": 1.5357,
      "theoretical_loss": 3.5172792519270604,
      "tokens_seen": 1639710720
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005083249749247744,
      "loss": 1.3945,
      "theoretical_loss": 3.5172671668480557,
      "tokens_seen": 1639776256
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005083049147442327,
      "loss": 1.4755,
      "theoretical_loss": 3.5172550823872717,
      "tokens_seen": 1639841792
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005082848545636911,
      "loss": 1.4419,
      "theoretical_loss": 3.517242998544652,
      "tokens_seen": 1639907328
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005082647943831495,
      "loss": 1.6178,
      "theoretical_loss": 3.517230915320141,
      "tokens_seen": 1639972864
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005082447342026078,
      "loss": 1.3847,
      "theoretical_loss": 3.5172188327136817,
      "tokens_seen": 1640038400
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005082246740220662,
      "loss": 1.5196,
      "theoretical_loss": 3.517206750725218,
      "tokens_seen": 1640103936
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005082046138415246,
      "loss": 1.4081,
      "theoretical_loss": 3.5171946693546934,
      "tokens_seen": 1640169472
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005081845536609829,
      "loss": 1.4695,
      "theoretical_loss": 3.5171825886020525,
      "tokens_seen": 1640235008
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005081644934804414,
      "loss": 1.4182,
      "theoretical_loss": 3.517170508467238,
      "tokens_seen": 1640300544
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005081444332998996,
      "loss": 1.3081,
      "theoretical_loss": 3.517158428950194,
      "tokens_seen": 1640366080
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005081243731193581,
      "loss": 1.4812,
      "theoretical_loss": 3.517146350050864,
      "tokens_seen": 1640431616
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005081043129388165,
      "loss": 1.3737,
      "theoretical_loss": 3.5171342717691925,
      "tokens_seen": 1640497152
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005080842527582748,
      "loss": 1.4476,
      "theoretical_loss": 3.517122194105122,
      "tokens_seen": 1640562688
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005080641925777332,
      "loss": 1.237,
      "theoretical_loss": 3.517110117058597,
      "tokens_seen": 1640628224
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005080441323971917,
      "loss": 1.3808,
      "theoretical_loss": 3.5170980406295618,
      "tokens_seen": 1640693760
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005080240722166499,
      "loss": 1.5292,
      "theoretical_loss": 3.517085964817959,
      "tokens_seen": 1640759296
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005080040120361084,
      "loss": 1.3489,
      "theoretical_loss": 3.5170738896237332,
      "tokens_seen": 1640824832
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005079839518555666,
      "loss": 1.4086,
      "theoretical_loss": 3.517061815046828,
      "tokens_seen": 1640890368
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005079638916750251,
      "loss": 1.4773,
      "theoretical_loss": 3.517049741087187,
      "tokens_seen": 1640955904
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005079438314944835,
      "loss": 1.5173,
      "theoretical_loss": 3.517037667744754,
      "tokens_seen": 1641021440
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005079237713139418,
      "loss": 1.3938,
      "theoretical_loss": 3.5170255950194727,
      "tokens_seen": 1641086976
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005079037111334002,
      "loss": 1.4475,
      "theoretical_loss": 3.517013522911287,
      "tokens_seen": 1641152512
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005078836509528586,
      "loss": 1.4238,
      "theoretical_loss": 3.5170014514201413,
      "tokens_seen": 1641218048
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005078635907723169,
      "loss": 1.4309,
      "theoretical_loss": 3.516989380545979,
      "tokens_seen": 1641283584
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005078435305917754,
      "loss": 1.4409,
      "theoretical_loss": 3.516977310288744,
      "tokens_seen": 1641349120
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005078234704112336,
      "loss": 1.3334,
      "theoretical_loss": 3.5169652406483793,
      "tokens_seen": 1641414656
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005078034102306921,
      "loss": 1.6331,
      "theoretical_loss": 3.51695317162483,
      "tokens_seen": 1641480192
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005077833500501504,
      "loss": 1.4492,
      "theoretical_loss": 3.5169411032180387,
      "tokens_seen": 1641545728
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005077632898696088,
      "loss": 1.5165,
      "theoretical_loss": 3.5169290354279505,
      "tokens_seen": 1641611264
    },
    {
      "epoch": 0.5,
      "objective/train/advantage_avg": 0.4894565939903259,
      "objective/train/docs_used": 599774,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3587998151779175,
      "objective/train/original_loss": 1.358799934387207,
      "objective/train/theoretical_loss": 3.5169169682545087,
      "objective/train/tokens_used": 1662136800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24226564168930054,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501770973205566,
      "objective/train/weighted_lm_loss": 1.4273089170455933,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5169169682545087,
      "tokens_seen": 1641676800
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005077432296890672,
      "loss": 1.4146,
      "theoretical_loss": 3.5169169682545087,
      "tokens_seen": 1641676800
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005077231695085256,
      "loss": 1.4709,
      "theoretical_loss": 3.5169049016976572,
      "tokens_seen": 1641742336
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005077031093279839,
      "loss": 1.4966,
      "theoretical_loss": 3.51689283575734,
      "tokens_seen": 1641807872
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005076830491474424,
      "loss": 1.489,
      "theoretical_loss": 3.516880770433501,
      "tokens_seen": 1641873408
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005076629889669006,
      "loss": 1.4853,
      "theoretical_loss": 3.5168687057260835,
      "tokens_seen": 1641938944
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005076429287863591,
      "loss": 1.5398,
      "theoretical_loss": 3.5168566416350324,
      "tokens_seen": 1642004480
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005076228686058174,
      "loss": 1.4937,
      "theoretical_loss": 3.516844578160291,
      "tokens_seen": 1642070016
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005076028084252758,
      "loss": 1.2679,
      "theoretical_loss": 3.516832515301803,
      "tokens_seen": 1642135552
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005075827482447342,
      "loss": 1.4239,
      "theoretical_loss": 3.516820453059513,
      "tokens_seen": 1642201088
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005075626880641926,
      "loss": 1.4314,
      "theoretical_loss": 3.516808391433364,
      "tokens_seen": 1642266624
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005075426278836509,
      "loss": 1.5129,
      "theoretical_loss": 3.5167963304233014,
      "tokens_seen": 1642332160
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005075225677031093,
      "loss": 1.3986,
      "theoretical_loss": 3.5167842700292677,
      "tokens_seen": 1642397696
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005075025075225678,
      "loss": 1.4748,
      "theoretical_loss": 3.516772210251208,
      "tokens_seen": 1642463232
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005074824473420261,
      "loss": 1.4624,
      "theoretical_loss": 3.5167601510890654,
      "tokens_seen": 1642528768
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005074623871614845,
      "loss": 1.4628,
      "theoretical_loss": 3.516748092542784,
      "tokens_seen": 1642594304
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005074423269809428,
      "loss": 1.4463,
      "theoretical_loss": 3.5167360346123075,
      "tokens_seen": 1642659840
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005074222668004013,
      "loss": 1.5469,
      "theoretical_loss": 3.516723977297581,
      "tokens_seen": 1642725376
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005074022066198596,
      "loss": 1.4407,
      "theoretical_loss": 3.5167119205985475,
      "tokens_seen": 1642790912
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000507382146439318,
      "loss": 1.519,
      "theoretical_loss": 3.5166998645151515,
      "tokens_seen": 1642856448
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005073620862587763,
      "loss": 1.4402,
      "theoretical_loss": 3.516687809047337,
      "tokens_seen": 1642921984
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005073420260782348,
      "loss": 1.4919,
      "theoretical_loss": 3.516675754195047,
      "tokens_seen": 1642987520
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005073219658976931,
      "loss": 1.5124,
      "theoretical_loss": 3.5166636999582277,
      "tokens_seen": 1643053056
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005073019057171515,
      "loss": 1.5488,
      "theoretical_loss": 3.516651646336821,
      "tokens_seen": 1643118592
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005072818455366099,
      "loss": 1.3845,
      "theoretical_loss": 3.5166395933307717,
      "tokens_seen": 1643184128
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005072617853560682,
      "loss": 1.4153,
      "theoretical_loss": 3.516627540940024,
      "tokens_seen": 1643249664
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005072417251755266,
      "loss": 1.4885,
      "theoretical_loss": 3.516615489164522,
      "tokens_seen": 1643315200
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000507221664994985,
      "loss": 1.5267,
      "theoretical_loss": 3.5166034380042093,
      "tokens_seen": 1643380736
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005072016048144433,
      "loss": 1.5139,
      "theoretical_loss": 3.51659138745903,
      "tokens_seen": 1643446272
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005071815446339018,
      "loss": 1.3488,
      "theoretical_loss": 3.5165793375289294,
      "tokens_seen": 1643511808
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00050716148445336,
      "loss": 1.4497,
      "theoretical_loss": 3.51656728821385,
      "tokens_seen": 1643577344
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005071414242728185,
      "loss": 1.4936,
      "theoretical_loss": 3.5165552395137363,
      "tokens_seen": 1643642880
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005071213640922769,
      "loss": 1.4529,
      "theoretical_loss": 3.516543191428533,
      "tokens_seen": 1643708416
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005071013039117352,
      "loss": 1.3481,
      "theoretical_loss": 3.5165311439581837,
      "tokens_seen": 1643773952
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005070812437311936,
      "loss": 1.4512,
      "theoretical_loss": 3.5165190971026323,
      "tokens_seen": 1643839488
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005070611835506521,
      "loss": 1.3745,
      "theoretical_loss": 3.516507050861823,
      "tokens_seen": 1643905024
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005070411233701103,
      "loss": 1.4574,
      "theoretical_loss": 3.5164950052357007,
      "tokens_seen": 1643970560
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005070210631895688,
      "loss": 1.3901,
      "theoretical_loss": 3.516482960224209,
      "tokens_seen": 1644036096
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000507001003009027,
      "loss": 1.383,
      "theoretical_loss": 3.516470915827292,
      "tokens_seen": 1644101632
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005069809428284855,
      "loss": 1.3218,
      "theoretical_loss": 3.5164588720448937,
      "tokens_seen": 1644167168
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005069608826479439,
      "loss": 1.429,
      "theoretical_loss": 3.5164468288769584,
      "tokens_seen": 1644232704
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005069408224674022,
      "loss": 1.4358,
      "theoretical_loss": 3.5164347863234307,
      "tokens_seen": 1644298240
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005069207622868606,
      "loss": 1.4776,
      "theoretical_loss": 3.5164227443842537,
      "tokens_seen": 1644363776
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000506900702106319,
      "loss": 1.4487,
      "theoretical_loss": 3.5164107030593725,
      "tokens_seen": 1644429312
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005068806419257773,
      "loss": 1.4712,
      "theoretical_loss": 3.516398662348731,
      "tokens_seen": 1644494848
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005068605817452358,
      "loss": 1.4636,
      "theoretical_loss": 3.5163866222522735,
      "tokens_seen": 1644560384
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000506840521564694,
      "loss": 1.5116,
      "theoretical_loss": 3.516374582769944,
      "tokens_seen": 1644625920
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005068204613841525,
      "loss": 1.393,
      "theoretical_loss": 3.516362543901687,
      "tokens_seen": 1644691456
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005068004012036108,
      "loss": 1.5865,
      "theoretical_loss": 3.516350505647446,
      "tokens_seen": 1644756992
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005067803410230692,
      "loss": 1.4908,
      "theoretical_loss": 3.516338468007166,
      "tokens_seen": 1644822528
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005067602808425276,
      "loss": 1.4963,
      "theoretical_loss": 3.516326430980791,
      "tokens_seen": 1644888064
    },
    {
      "epoch": 0.5,
      "objective/train/advantage_avg": 0.4947347342967987,
      "objective/train/docs_used": 601010,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.395460844039917,
      "objective/train/original_loss": 1.395460844039917,
      "objective/train/theoretical_loss": 3.5163143945682656,
      "objective/train/tokens_used": 1665413600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24497202038764954,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050718903541565,
      "objective/train/weighted_lm_loss": 1.46614670753479,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.5163143945682656,
      "tokens_seen": 1644953600
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000506740220661986,
      "loss": 1.4059,
      "theoretical_loss": 3.5163143945682656,
      "tokens_seen": 1644953600
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005067201604814443,
      "loss": 1.4333,
      "theoretical_loss": 3.516302358769533,
      "tokens_seen": 1645019136
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005067001003009028,
      "loss": 1.3845,
      "theoretical_loss": 3.5162903235845384,
      "tokens_seen": 1645084672
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000506680040120361,
      "loss": 1.516,
      "theoretical_loss": 3.5162782890132256,
      "tokens_seen": 1645150208
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005066599799398195,
      "loss": 1.5837,
      "theoretical_loss": 3.5162662550555392,
      "tokens_seen": 1645215744
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005066399197592778,
      "loss": 1.5264,
      "theoretical_loss": 3.516254221711423,
      "tokens_seen": 1645281280
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005066198595787362,
      "loss": 1.4363,
      "theoretical_loss": 3.5162421889808217,
      "tokens_seen": 1645346816
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005065997993981946,
      "loss": 1.5321,
      "theoretical_loss": 3.5162301568636796,
      "tokens_seen": 1645412352
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000506579739217653,
      "loss": 1.4766,
      "theoretical_loss": 3.5162181253599405,
      "tokens_seen": 1645477888
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005065596790371113,
      "loss": 1.5172,
      "theoretical_loss": 3.516206094469549,
      "tokens_seen": 1645543424
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005065396188565697,
      "loss": 1.5095,
      "theoretical_loss": 3.5161940641924496,
      "tokens_seen": 1645608960
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000506519558676028,
      "loss": 1.3756,
      "theoretical_loss": 3.516182034528587,
      "tokens_seen": 1645674496
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005064994984954865,
      "loss": 1.4123,
      "theoretical_loss": 3.516170005477904,
      "tokens_seen": 1645740032
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005064794383149448,
      "loss": 1.4862,
      "theoretical_loss": 3.516157977040346,
      "tokens_seen": 1645805568
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005064593781344032,
      "loss": 1.4768,
      "theoretical_loss": 3.5161459492158578,
      "tokens_seen": 1645871104
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005064393179538615,
      "loss": 1.4426,
      "theoretical_loss": 3.516133922004383,
      "tokens_seen": 1645936640
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00050641925777332,
      "loss": 1.3646,
      "theoretical_loss": 3.516121895405866,
      "tokens_seen": 1646002176
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005063991975927783,
      "loss": 1.5113,
      "theoretical_loss": 3.5161098694202515,
      "tokens_seen": 1646067712
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005063791374122367,
      "loss": 1.566,
      "theoretical_loss": 3.5160978440474837,
      "tokens_seen": 1646133248
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005063590772316951,
      "loss": 1.4509,
      "theoretical_loss": 3.5160858192875066,
      "tokens_seen": 1646198784
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005063390170511535,
      "loss": 1.5365,
      "theoretical_loss": 3.516073795140265,
      "tokens_seen": 1646264320
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005063189568706118,
      "loss": 1.3503,
      "theoretical_loss": 3.5160617716057034,
      "tokens_seen": 1646329856
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005062988966900703,
      "loss": 1.3904,
      "theoretical_loss": 3.5160497486837654,
      "tokens_seen": 1646395392
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005062788365095285,
      "loss": 1.4928,
      "theoretical_loss": 3.516037726374397,
      "tokens_seen": 1646460928
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000506258776328987,
      "loss": 1.4877,
      "theoretical_loss": 3.5160257046775407,
      "tokens_seen": 1646526464
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005062387161484453,
      "loss": 1.4487,
      "theoretical_loss": 3.516013683593142,
      "tokens_seen": 1646592000
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005062186559679037,
      "loss": 1.3923,
      "theoretical_loss": 3.5160016631211457,
      "tokens_seen": 1646657536
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005061985957873621,
      "loss": 1.578,
      "theoretical_loss": 3.515989643261495,
      "tokens_seen": 1646723072
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005061785356068204,
      "loss": 1.3324,
      "theoretical_loss": 3.515977624014135,
      "tokens_seen": 1646788608
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005061584754262788,
      "loss": 1.4485,
      "theoretical_loss": 3.5159656053790105,
      "tokens_seen": 1646854144
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005061384152457373,
      "loss": 1.3654,
      "theoretical_loss": 3.5159535873560657,
      "tokens_seen": 1646919680
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005061183550651955,
      "loss": 1.4285,
      "theoretical_loss": 3.5159415699452445,
      "tokens_seen": 1646985216
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000506098294884654,
      "loss": 1.4024,
      "theoretical_loss": 3.515929553146492,
      "tokens_seen": 1647050752
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005060782347041123,
      "loss": 1.4398,
      "theoretical_loss": 3.515917536959753,
      "tokens_seen": 1647116288
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005060581745235707,
      "loss": 1.3289,
      "theoretical_loss": 3.515905521384971,
      "tokens_seen": 1647181824
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005060381143430291,
      "loss": 1.4292,
      "theoretical_loss": 3.5158935064220906,
      "tokens_seen": 1647247360
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005060180541624874,
      "loss": 1.427,
      "theoretical_loss": 3.515881492071057,
      "tokens_seen": 1647312896
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005059979939819459,
      "loss": 1.4038,
      "theoretical_loss": 3.5158694783318145,
      "tokens_seen": 1647378432
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005059779338014043,
      "loss": 1.4656,
      "theoretical_loss": 3.5158574652043075,
      "tokens_seen": 1647443968
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005059578736208626,
      "loss": 1.4949,
      "theoretical_loss": 3.5158454526884806,
      "tokens_seen": 1647509504
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000505937813440321,
      "loss": 1.455,
      "theoretical_loss": 3.515833440784278,
      "tokens_seen": 1647575040
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005059177532597794,
      "loss": 1.4904,
      "theoretical_loss": 3.5158214294916443,
      "tokens_seen": 1647640576
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005058976930792377,
      "loss": 1.3306,
      "theoretical_loss": 3.5158094188105244,
      "tokens_seen": 1647706112
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005058776328986962,
      "loss": 1.5325,
      "theoretical_loss": 3.5157974087408626,
      "tokens_seen": 1647771648
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005058575727181544,
      "loss": 1.3405,
      "theoretical_loss": 3.5157853992826036,
      "tokens_seen": 1647837184
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005058375125376129,
      "loss": 1.4069,
      "theoretical_loss": 3.515773390435692,
      "tokens_seen": 1647902720
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005058174523570712,
      "loss": 1.4516,
      "theoretical_loss": 3.515761382200072,
      "tokens_seen": 1647968256
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005057973921765296,
      "loss": 1.4335,
      "theoretical_loss": 3.5157493745756883,
      "tokens_seen": 1648033792
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000505777331995988,
      "loss": 1.4697,
      "theoretical_loss": 3.5157373675624854,
      "tokens_seen": 1648099328
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005057572718154464,
      "loss": 1.443,
      "theoretical_loss": 3.5157253611604085,
      "tokens_seen": 1648164864
    },
    {
      "epoch": 0.5,
      "objective/train/advantage_avg": 0.49225878715515137,
      "objective/train/docs_used": 602144,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4322959184646606,
      "objective/train/original_loss": 1.4322959184646606,
      "objective/train/theoretical_loss": 3.5157133553694013,
      "objective/train/tokens_used": 1668690400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24289138615131378,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050460696220398,
      "objective/train/weighted_lm_loss": 1.504431128501892,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.5157133553694013,
      "tokens_seen": 1648230400
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005057372116349047,
      "loss": 1.4687,
      "theoretical_loss": 3.5157133553694013,
      "tokens_seen": 1648230400
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005057171514543632,
      "loss": 1.5849,
      "theoretical_loss": 3.5157013501894094,
      "tokens_seen": 1648295936
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005056970912738214,
      "loss": 1.4101,
      "theoretical_loss": 3.515689345620377,
      "tokens_seen": 1648361472
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005056770310932799,
      "loss": 1.4924,
      "theoretical_loss": 3.515677341662248,
      "tokens_seen": 1648427008
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005056569709127382,
      "loss": 1.4881,
      "theoretical_loss": 3.515665338314968,
      "tokens_seen": 1648492544
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005056369107321966,
      "loss": 1.4096,
      "theoretical_loss": 3.515653335578481,
      "tokens_seen": 1648558080
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000505616850551655,
      "loss": 1.5207,
      "theoretical_loss": 3.5156413334527317,
      "tokens_seen": 1648623616
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005055967903711134,
      "loss": 1.5418,
      "theoretical_loss": 3.5156293319376655,
      "tokens_seen": 1648689152
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005055767301905717,
      "loss": 1.4581,
      "theoretical_loss": 3.5156173310332264,
      "tokens_seen": 1648754688
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005055566700100301,
      "loss": 1.3629,
      "theoretical_loss": 3.5156053307393584,
      "tokens_seen": 1648820224
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005055366098294885,
      "loss": 1.4301,
      "theoretical_loss": 3.515593331056008,
      "tokens_seen": 1648885760
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005055165496489469,
      "loss": 1.3292,
      "theoretical_loss": 3.5155813319831184,
      "tokens_seen": 1648951296
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005054964894684052,
      "loss": 1.4383,
      "theoretical_loss": 3.5155693335206344,
      "tokens_seen": 1649016832
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005054764292878636,
      "loss": 1.3423,
      "theoretical_loss": 3.5155573356685013,
      "tokens_seen": 1649082368
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005054563691073219,
      "loss": 1.5312,
      "theoretical_loss": 3.5155453384266635,
      "tokens_seen": 1649147904
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005054363089267804,
      "loss": 1.3971,
      "theoretical_loss": 3.5155333417950656,
      "tokens_seen": 1649213440
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005054162487462387,
      "loss": 1.4002,
      "theoretical_loss": 3.5155213457736525,
      "tokens_seen": 1649278976
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005053961885656971,
      "loss": 1.4882,
      "theoretical_loss": 3.5155093503623687,
      "tokens_seen": 1649344512
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005053761283851555,
      "loss": 1.3585,
      "theoretical_loss": 3.515497355561159,
      "tokens_seen": 1649410048
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005053560682046139,
      "loss": 1.5447,
      "theoretical_loss": 3.5154853613699686,
      "tokens_seen": 1649475584
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005053360080240722,
      "loss": 1.3593,
      "theoretical_loss": 3.5154733677887413,
      "tokens_seen": 1649541120
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005053159478435307,
      "loss": 1.5116,
      "theoretical_loss": 3.5154613748174226,
      "tokens_seen": 1649606656
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005052958876629889,
      "loss": 1.4974,
      "theoretical_loss": 3.515449382455957,
      "tokens_seen": 1649672192
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005052758274824474,
      "loss": 1.5102,
      "theoretical_loss": 3.515437390704289,
      "tokens_seen": 1649737728
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005052557673019057,
      "loss": 1.4495,
      "theoretical_loss": 3.515425399562364,
      "tokens_seen": 1649803264
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005052357071213641,
      "loss": 1.3905,
      "theoretical_loss": 3.5154134090301263,
      "tokens_seen": 1649868800
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005052156469408225,
      "loss": 1.4479,
      "theoretical_loss": 3.515401419107521,
      "tokens_seen": 1649934336
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005051955867602808,
      "loss": 1.4083,
      "theoretical_loss": 3.5153894297944923,
      "tokens_seen": 1649999872
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005051755265797392,
      "loss": 1.482,
      "theoretical_loss": 3.515377441090986,
      "tokens_seen": 1650065408
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005051554663991977,
      "loss": 1.4483,
      "theoretical_loss": 3.515365452996946,
      "tokens_seen": 1650130944
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005051354062186559,
      "loss": 1.3054,
      "theoretical_loss": 3.515353465512317,
      "tokens_seen": 1650196480
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005051153460381144,
      "loss": 1.4016,
      "theoretical_loss": 3.515341478637045,
      "tokens_seen": 1650262016
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005050952858575727,
      "loss": 1.5323,
      "theoretical_loss": 3.5153294923710736,
      "tokens_seen": 1650327552
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005050752256770311,
      "loss": 1.5635,
      "theoretical_loss": 3.5153175067143483,
      "tokens_seen": 1650393088
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005050551654964895,
      "loss": 1.466,
      "theoretical_loss": 3.5153055216668134,
      "tokens_seen": 1650458624
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005050351053159478,
      "loss": 1.4159,
      "theoretical_loss": 3.5152935372284144,
      "tokens_seen": 1650524160
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005050150451354062,
      "loss": 1.4742,
      "theoretical_loss": 3.515281553399096,
      "tokens_seen": 1650589696
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005049949849548647,
      "loss": 1.4995,
      "theoretical_loss": 3.515269570178803,
      "tokens_seen": 1650655232
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005049749247743229,
      "loss": 1.4388,
      "theoretical_loss": 3.51525758756748,
      "tokens_seen": 1650720768
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005049548645937814,
      "loss": 1.4392,
      "theoretical_loss": 3.515245605565072,
      "tokens_seen": 1650786304
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005049348044132396,
      "loss": 1.4701,
      "theoretical_loss": 3.515233624171524,
      "tokens_seen": 1650851840
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005049147442326981,
      "loss": 1.418,
      "theoretical_loss": 3.5152216433867807,
      "tokens_seen": 1650917376
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005048946840521565,
      "loss": 1.3811,
      "theoretical_loss": 3.5152096632107876,
      "tokens_seen": 1650982912
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005048746238716148,
      "loss": 1.515,
      "theoretical_loss": 3.515197683643489,
      "tokens_seen": 1651048448
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005048545636910732,
      "loss": 1.3538,
      "theoretical_loss": 3.5151857046848294,
      "tokens_seen": 1651113984
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005048345035105316,
      "loss": 1.469,
      "theoretical_loss": 3.515173726334755,
      "tokens_seen": 1651179520
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005048144433299899,
      "loss": 1.4868,
      "theoretical_loss": 3.5151617485932096,
      "tokens_seen": 1651245056
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005047943831494484,
      "loss": 1.3476,
      "theoretical_loss": 3.5151497714601385,
      "tokens_seen": 1651310592
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005047743229689067,
      "loss": 1.3144,
      "theoretical_loss": 3.515137794935487,
      "tokens_seen": 1651376128
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005047542627883651,
      "loss": 1.429,
      "theoretical_loss": 3.5151258190191994,
      "tokens_seen": 1651441664
    },
    {
      "epoch": 0.5,
      "objective/train/advantage_avg": 0.491573303937912,
      "objective/train/docs_used": 603335,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2921056747436523,
      "objective/train/original_loss": 1.2921056747436523,
      "objective/train/theoretical_loss": 3.5151138437112213,
      "objective/train/tokens_used": 1671967200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2424304038286209,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503896474838257,
      "objective/train/weighted_lm_loss": 1.3570951223373413,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5151138437112213,
      "tokens_seen": 1651507200
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005047342026078235,
      "loss": 1.3931,
      "theoretical_loss": 3.5151138437112213,
      "tokens_seen": 1651507200
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005047141424272818,
      "loss": 1.4898,
      "theoretical_loss": 3.5151018690114975,
      "tokens_seen": 1651572736
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005046940822467402,
      "loss": 1.5805,
      "theoretical_loss": 3.5150898949199725,
      "tokens_seen": 1651638272
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005046740220661986,
      "loss": 1.4267,
      "theoretical_loss": 3.5150779214365917,
      "tokens_seen": 1651703808
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005046539618856569,
      "loss": 1.4237,
      "theoretical_loss": 3.5150659485613005,
      "tokens_seen": 1651769344
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005046339017051154,
      "loss": 1.5511,
      "theoretical_loss": 3.515053976294043,
      "tokens_seen": 1651834880
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005046138415245737,
      "loss": 1.4276,
      "theoretical_loss": 3.515042004634764,
      "tokens_seen": 1651900416
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005045937813440321,
      "loss": 1.3835,
      "theoretical_loss": 3.5150300335834097,
      "tokens_seen": 1651965952
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005045737211634904,
      "loss": 1.375,
      "theoretical_loss": 3.5150180631399244,
      "tokens_seen": 1652031488
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005045536609829489,
      "loss": 1.4865,
      "theoretical_loss": 3.5150060933042533,
      "tokens_seen": 1652097024
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005045336008024072,
      "loss": 1.43,
      "theoretical_loss": 3.514994124076341,
      "tokens_seen": 1652162560
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005045135406218656,
      "loss": 1.4221,
      "theoretical_loss": 3.5149821554561336,
      "tokens_seen": 1652228096
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000504493480441324,
      "loss": 1.3474,
      "theoretical_loss": 3.5149701874435753,
      "tokens_seen": 1652293632
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005044734202607823,
      "loss": 1.4868,
      "theoretical_loss": 3.5149582200386105,
      "tokens_seen": 1652359168
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005044533600802408,
      "loss": 1.6483,
      "theoretical_loss": 3.514946253241186,
      "tokens_seen": 1652424704
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005044332998996991,
      "loss": 1.4543,
      "theoretical_loss": 3.514934287051245,
      "tokens_seen": 1652490240
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005044132397191575,
      "loss": 1.5436,
      "theoretical_loss": 3.514922321468734,
      "tokens_seen": 1652555776
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005043931795386159,
      "loss": 1.597,
      "theoretical_loss": 3.5149103564935977,
      "tokens_seen": 1652621312
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005043731193580743,
      "loss": 1.4153,
      "theoretical_loss": 3.5148983921257813,
      "tokens_seen": 1652686848
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005043530591775326,
      "loss": 1.4132,
      "theoretical_loss": 3.5148864283652292,
      "tokens_seen": 1652752384
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005043329989969911,
      "loss": 1.4547,
      "theoretical_loss": 3.5148744652118866,
      "tokens_seen": 1652817920
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005043129388164493,
      "loss": 1.4853,
      "theoretical_loss": 3.514862502665699,
      "tokens_seen": 1652883456
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005042928786359078,
      "loss": 1.4641,
      "theoretical_loss": 3.514850540726612,
      "tokens_seen": 1652948992
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005042728184553661,
      "loss": 1.4343,
      "theoretical_loss": 3.51483857939457,
      "tokens_seen": 1653014528
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005042527582748245,
      "loss": 1.4897,
      "theoretical_loss": 3.5148266186695185,
      "tokens_seen": 1653080064
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005042326980942829,
      "loss": 1.4958,
      "theoretical_loss": 3.514814658551402,
      "tokens_seen": 1653145600
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005042126379137412,
      "loss": 1.387,
      "theoretical_loss": 3.5148026990401666,
      "tokens_seen": 1653211136
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005041925777331996,
      "loss": 1.452,
      "theoretical_loss": 3.5147907401357563,
      "tokens_seen": 1653276672
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005041725175526581,
      "loss": 1.3677,
      "theoretical_loss": 3.5147787818381175,
      "tokens_seen": 1653342208
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005041524573721163,
      "loss": 1.3673,
      "theoretical_loss": 3.514766824147194,
      "tokens_seen": 1653407744
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005041323971915748,
      "loss": 1.4303,
      "theoretical_loss": 3.5147548670629325,
      "tokens_seen": 1653473280
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005041123370110331,
      "loss": 1.4382,
      "theoretical_loss": 3.514742910585277,
      "tokens_seen": 1653538816
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005040922768304915,
      "loss": 1.5256,
      "theoretical_loss": 3.514730954714173,
      "tokens_seen": 1653604352
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005040722166499499,
      "loss": 1.3388,
      "theoretical_loss": 3.514718999449566,
      "tokens_seen": 1653669888
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005040521564694082,
      "loss": 1.3608,
      "theoretical_loss": 3.5147070447914013,
      "tokens_seen": 1653735424
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005040320962888666,
      "loss": 1.5678,
      "theoretical_loss": 3.5146950907396235,
      "tokens_seen": 1653800960
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005040120361083251,
      "loss": 1.4754,
      "theoretical_loss": 3.5146831372941776,
      "tokens_seen": 1653866496
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005039919759277833,
      "loss": 1.4406,
      "theoretical_loss": 3.5146711844550103,
      "tokens_seen": 1653932032
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005039719157472418,
      "loss": 1.552,
      "theoretical_loss": 3.514659232222065,
      "tokens_seen": 1653997568
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005039518555667,
      "loss": 1.4527,
      "theoretical_loss": 3.5146472805952884,
      "tokens_seen": 1654063104
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005039317953861585,
      "loss": 1.3891,
      "theoretical_loss": 3.5146353295746247,
      "tokens_seen": 1654128640
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005039117352056169,
      "loss": 1.2288,
      "theoretical_loss": 3.5146233791600197,
      "tokens_seen": 1654194176
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005038916750250752,
      "loss": 1.4814,
      "theoretical_loss": 3.514611429351419,
      "tokens_seen": 1654259712
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005038716148445336,
      "loss": 1.4723,
      "theoretical_loss": 3.5145994801487666,
      "tokens_seen": 1654325248
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000503851554663992,
      "loss": 1.4814,
      "theoretical_loss": 3.514587531552009,
      "tokens_seen": 1654390784
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005038314944834503,
      "loss": 1.3754,
      "theoretical_loss": 3.514575583561091,
      "tokens_seen": 1654456320
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005038114343029088,
      "loss": 1.3402,
      "theoretical_loss": 3.5145636361759576,
      "tokens_seen": 1654521856
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000503791374122367,
      "loss": 1.4543,
      "theoretical_loss": 3.514551689396555,
      "tokens_seen": 1654587392
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005037713139418255,
      "loss": 1.4786,
      "theoretical_loss": 3.5145397432228274,
      "tokens_seen": 1654652928
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005037512537612839,
      "loss": 1.4809,
      "theoretical_loss": 3.514527797654721,
      "tokens_seen": 1654718464
    },
    {
      "epoch": 0.5,
      "objective/train/advantage_avg": 0.49156254529953003,
      "objective/train/docs_used": 604403,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.463065266609192,
      "objective/train/original_loss": 1.4630653858184814,
      "objective/train/theoretical_loss": 3.5145158526921807,
      "objective/train/tokens_used": 1675244000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24229401350021362,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050387978553772,
      "objective/train/weighted_lm_loss": 1.5367331504821777,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.5145158526921807,
      "tokens_seen": 1654784000
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005037311935807422,
      "loss": 1.5399,
      "theoretical_loss": 3.5145158526921807,
      "tokens_seen": 1654784000
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005037111334002006,
      "loss": 1.5462,
      "theoretical_loss": 3.514503908335152,
      "tokens_seen": 1654849536
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000503691073219659,
      "loss": 1.4094,
      "theoretical_loss": 3.5144919645835797,
      "tokens_seen": 1654915072
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005036710130391173,
      "loss": 1.4362,
      "theoretical_loss": 3.51448002143741,
      "tokens_seen": 1654980608
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005036509528585758,
      "loss": 1.3503,
      "theoretical_loss": 3.5144680788965874,
      "tokens_seen": 1655046144
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005036308926780341,
      "loss": 1.3955,
      "theoretical_loss": 3.5144561369610576,
      "tokens_seen": 1655111680
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005036108324974925,
      "loss": 1.4519,
      "theoretical_loss": 3.514444195630766,
      "tokens_seen": 1655177216
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005035907723169508,
      "loss": 1.4131,
      "theoretical_loss": 3.5144322549056586,
      "tokens_seen": 1655242752
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005035707121364093,
      "loss": 1.4594,
      "theoretical_loss": 3.514420314785679,
      "tokens_seen": 1655308288
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005035506519558676,
      "loss": 1.3942,
      "theoretical_loss": 3.5144083752707744,
      "tokens_seen": 1655373824
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000503530591775326,
      "loss": 1.3739,
      "theoretical_loss": 3.5143964363608893,
      "tokens_seen": 1655439360
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005035105315947843,
      "loss": 1.4581,
      "theoretical_loss": 3.5143844980559695,
      "tokens_seen": 1655504896
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005034904714142427,
      "loss": 1.5134,
      "theoretical_loss": 3.5143725603559597,
      "tokens_seen": 1655570432
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005034704112337011,
      "loss": 1.3822,
      "theoretical_loss": 3.514360623260806,
      "tokens_seen": 1655635968
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005034503510531595,
      "loss": 1.5775,
      "theoretical_loss": 3.514348686770454,
      "tokens_seen": 1655701504
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005034302908726178,
      "loss": 1.4112,
      "theoretical_loss": 3.5143367508848478,
      "tokens_seen": 1655767040
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005034102306920763,
      "loss": 1.4015,
      "theoretical_loss": 3.5143248156039344,
      "tokens_seen": 1655832576
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005033901705115346,
      "loss": 1.4027,
      "theoretical_loss": 3.5143128809276583,
      "tokens_seen": 1655898112
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000503370110330993,
      "loss": 1.4367,
      "theoretical_loss": 3.5143009468559656,
      "tokens_seen": 1655963648
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005033500501504513,
      "loss": 1.4102,
      "theoretical_loss": 3.514289013388801,
      "tokens_seen": 1656029184
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005033299899699097,
      "loss": 1.5018,
      "theoretical_loss": 3.51427708052611,
      "tokens_seen": 1656094720
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005033099297893681,
      "loss": 1.3944,
      "theoretical_loss": 3.514265148267839,
      "tokens_seen": 1656160256
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005032898696088265,
      "loss": 1.3419,
      "theoretical_loss": 3.514253216613932,
      "tokens_seen": 1656225792
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005032698094282848,
      "loss": 1.5361,
      "theoretical_loss": 3.5142412855643363,
      "tokens_seen": 1656291328
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005032497492477433,
      "loss": 1.4733,
      "theoretical_loss": 3.514229355118996,
      "tokens_seen": 1656356864
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005032296890672015,
      "loss": 1.4757,
      "theoretical_loss": 3.514217425277857,
      "tokens_seen": 1656422400
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00050320962888666,
      "loss": 1.4135,
      "theoretical_loss": 3.514205496040865,
      "tokens_seen": 1656487936
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005031895687061184,
      "loss": 1.4743,
      "theoretical_loss": 3.5141935674079647,
      "tokens_seen": 1656553472
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005031695085255767,
      "loss": 1.4091,
      "theoretical_loss": 3.5141816393791023,
      "tokens_seen": 1656619008
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005031494483450351,
      "loss": 1.4193,
      "theoretical_loss": 3.5141697119542235,
      "tokens_seen": 1656684544
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005031293881644935,
      "loss": 1.4212,
      "theoretical_loss": 3.5141577851332735,
      "tokens_seen": 1656750080
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005031093279839518,
      "loss": 1.4106,
      "theoretical_loss": 3.514145858916198,
      "tokens_seen": 1656815616
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005030892678034103,
      "loss": 1.4352,
      "theoretical_loss": 3.514133933302942,
      "tokens_seen": 1656881152
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005030692076228685,
      "loss": 1.4589,
      "theoretical_loss": 3.5141220082934517,
      "tokens_seen": 1656946688
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000503049147442327,
      "loss": 1.3494,
      "theoretical_loss": 3.5141100838876724,
      "tokens_seen": 1657012224
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005030290872617854,
      "loss": 1.5096,
      "theoretical_loss": 3.51409816008555,
      "tokens_seen": 1657077760
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005030090270812437,
      "loss": 1.3452,
      "theoretical_loss": 3.51408623688703,
      "tokens_seen": 1657143296
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005029889669007022,
      "loss": 1.3142,
      "theoretical_loss": 3.5140743142920567,
      "tokens_seen": 1657208832
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005029689067201604,
      "loss": 1.4053,
      "theoretical_loss": 3.5140623923005774,
      "tokens_seen": 1657274368
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005029488465396189,
      "loss": 1.4891,
      "theoretical_loss": 3.5140504709125366,
      "tokens_seen": 1657339904
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005029287863590773,
      "loss": 1.4115,
      "theoretical_loss": 3.514038550127881,
      "tokens_seen": 1657405440
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005029087261785356,
      "loss": 1.3088,
      "theoretical_loss": 3.514026629946555,
      "tokens_seen": 1657470976
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000502888665997994,
      "loss": 1.3865,
      "theoretical_loss": 3.514014710368505,
      "tokens_seen": 1657536512
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005028686058174524,
      "loss": 1.3938,
      "theoretical_loss": 3.5140027913936756,
      "tokens_seen": 1657602048
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005028485456369107,
      "loss": 1.4595,
      "theoretical_loss": 3.5139908730220135,
      "tokens_seen": 1657667584
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005028284854563692,
      "loss": 1.3878,
      "theoretical_loss": 3.5139789552534646,
      "tokens_seen": 1657733120
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005028084252758275,
      "loss": 1.3413,
      "theoretical_loss": 3.513967038087973,
      "tokens_seen": 1657798656
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005027883650952859,
      "loss": 1.5371,
      "theoretical_loss": 3.513955121525486,
      "tokens_seen": 1657864192
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005027683049147443,
      "loss": 1.3817,
      "theoretical_loss": 3.513943205565948,
      "tokens_seen": 1657929728
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005027482447342026,
      "loss": 1.4257,
      "theoretical_loss": 3.513931290209306,
      "tokens_seen": 1657995264
    },
    {
      "epoch": 0.5,
      "objective/train/advantage_avg": 0.49387332797050476,
      "objective/train/docs_used": 605625,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.344884991645813,
      "objective/train/original_loss": 1.344884991645813,
      "objective/train/theoretical_loss": 3.5139193754555036,
      "objective/train/tokens_used": 1678520800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2443666309118271,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050629734992981,
      "objective/train/weighted_lm_loss": 1.4130091667175293,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5139193754555036,
      "tokens_seen": 1658060800
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000502728184553661,
      "loss": 1.3714,
      "theoretical_loss": 3.5139193754555036,
      "tokens_seen": 1658060800
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005027081243731194,
      "loss": 1.3583,
      "theoretical_loss": 3.5139074613044885,
      "tokens_seen": 1658126336
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005026880641925777,
      "loss": 1.535,
      "theoretical_loss": 3.5138955477562055,
      "tokens_seen": 1658191872
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005026680040120362,
      "loss": 1.3965,
      "theoretical_loss": 3.5138836348106004,
      "tokens_seen": 1658257408
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005026479438314945,
      "loss": 1.2966,
      "theoretical_loss": 3.513871722467619,
      "tokens_seen": 1658322944
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005026278836509529,
      "loss": 1.4524,
      "theoretical_loss": 3.5138598107272068,
      "tokens_seen": 1658388480
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005026078234704112,
      "loss": 1.4743,
      "theoretical_loss": 3.51384789958931,
      "tokens_seen": 1658454016
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005025877632898697,
      "loss": 1.4552,
      "theoretical_loss": 3.513835989053873,
      "tokens_seen": 1658519552
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000502567703109328,
      "loss": 1.4067,
      "theoretical_loss": 3.513824079120843,
      "tokens_seen": 1658585088
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005025476429287864,
      "loss": 1.5167,
      "theoretical_loss": 3.513812169790165,
      "tokens_seen": 1658650624
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005025275827482447,
      "loss": 1.5237,
      "theoretical_loss": 3.5138002610617853,
      "tokens_seen": 1658716160
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005025075225677031,
      "loss": 1.3055,
      "theoretical_loss": 3.5137883529356486,
      "tokens_seen": 1658781696
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005024874623871615,
      "loss": 1.4172,
      "theoretical_loss": 3.513776445411702,
      "tokens_seen": 1658847232
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005024674022066199,
      "loss": 1.3841,
      "theoretical_loss": 3.51376453848989,
      "tokens_seen": 1658912768
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005024473420260782,
      "loss": 1.3179,
      "theoretical_loss": 3.513752632170159,
      "tokens_seen": 1658978304
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005024272818455367,
      "loss": 1.429,
      "theoretical_loss": 3.5137407264524554,
      "tokens_seen": 1659043840
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000502407221664995,
      "loss": 1.4036,
      "theoretical_loss": 3.5137288213367235,
      "tokens_seen": 1659109376
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005023871614844534,
      "loss": 1.4532,
      "theoretical_loss": 3.5137169168229105,
      "tokens_seen": 1659174912
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005023671013039117,
      "loss": 1.4413,
      "theoretical_loss": 3.513705012910961,
      "tokens_seen": 1659240448
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005023470411233701,
      "loss": 1.4288,
      "theoretical_loss": 3.5136931096008217,
      "tokens_seen": 1659305984
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005023269809428285,
      "loss": 1.4342,
      "theoretical_loss": 3.5136812068924375,
      "tokens_seen": 1659371520
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005023069207622869,
      "loss": 1.3264,
      "theoretical_loss": 3.5136693047857555,
      "tokens_seen": 1659437056
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005022868605817452,
      "loss": 1.4189,
      "theoretical_loss": 3.5136574032807206,
      "tokens_seen": 1659502592
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005022668004012037,
      "loss": 1.4023,
      "theoretical_loss": 3.513645502377279,
      "tokens_seen": 1659568128
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005022467402206619,
      "loss": 1.548,
      "theoretical_loss": 3.513633602075376,
      "tokens_seen": 1659633664
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005022266800401204,
      "loss": 1.3916,
      "theoretical_loss": 3.513621702374958,
      "tokens_seen": 1659699200
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005022066198595788,
      "loss": 1.4407,
      "theoretical_loss": 3.5136098032759704,
      "tokens_seen": 1659764736
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005021865596790371,
      "loss": 1.4865,
      "theoretical_loss": 3.5135979047783596,
      "tokens_seen": 1659830272
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005021664994984955,
      "loss": 1.3529,
      "theoretical_loss": 3.513586006882071,
      "tokens_seen": 1659895808
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000502146439317954,
      "loss": 1.4451,
      "theoretical_loss": 3.5135741095870507,
      "tokens_seen": 1659961344
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005021263791374122,
      "loss": 1.4622,
      "theoretical_loss": 3.5135622128932447,
      "tokens_seen": 1660026880
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005021063189568707,
      "loss": 1.4717,
      "theoretical_loss": 3.5135503168005986,
      "tokens_seen": 1660092416
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005020862587763289,
      "loss": 1.3378,
      "theoretical_loss": 3.513538421309059,
      "tokens_seen": 1660157952
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005020661985957874,
      "loss": 1.3964,
      "theoretical_loss": 3.5135265264185707,
      "tokens_seen": 1660223488
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005020461384152458,
      "loss": 1.3907,
      "theoretical_loss": 3.5135146321290796,
      "tokens_seen": 1660289024
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005020260782347041,
      "loss": 1.4657,
      "theoretical_loss": 3.5135027384405326,
      "tokens_seen": 1660354560
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005020060180541625,
      "loss": 1.5194,
      "theoretical_loss": 3.5134908453528757,
      "tokens_seen": 1660420096
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005019859578736208,
      "loss": 1.3952,
      "theoretical_loss": 3.5134789528660537,
      "tokens_seen": 1660485632
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005019658976930792,
      "loss": 1.3909,
      "theoretical_loss": 3.513467060980013,
      "tokens_seen": 1660551168
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005019458375125377,
      "loss": 1.5274,
      "theoretical_loss": 3.5134551696946996,
      "tokens_seen": 1660616704
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005019257773319959,
      "loss": 1.3525,
      "theoretical_loss": 3.51344327901006,
      "tokens_seen": 1660682240
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005019057171514544,
      "loss": 1.5408,
      "theoretical_loss": 3.5134313889260396,
      "tokens_seen": 1660747776
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005018856569709127,
      "loss": 1.2938,
      "theoretical_loss": 3.513419499442584,
      "tokens_seen": 1660813312
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005018655967903711,
      "loss": 1.5202,
      "theoretical_loss": 3.5134076105596397,
      "tokens_seen": 1660878848
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005018455366098295,
      "loss": 1.4137,
      "theoretical_loss": 3.513395722277153,
      "tokens_seen": 1660944384
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005018254764292879,
      "loss": 1.4452,
      "theoretical_loss": 3.5133838345950688,
      "tokens_seen": 1661009920
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005018054162487462,
      "loss": 1.4967,
      "theoretical_loss": 3.513371947513334,
      "tokens_seen": 1661075456
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005017853560682047,
      "loss": 1.5161,
      "theoretical_loss": 3.5133600610318942,
      "tokens_seen": 1661140992
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005017652958876629,
      "loss": 1.3968,
      "theoretical_loss": 3.513348175150696,
      "tokens_seen": 1661206528
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005017452357071214,
      "loss": 1.384,
      "theoretical_loss": 3.5133362898696845,
      "tokens_seen": 1661272064
    },
    {
      "epoch": 0.5,
      "objective/train/advantage_avg": 0.4922904074192047,
      "objective/train/docs_used": 606795,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.407870888710022,
      "objective/train/original_loss": 1.407870888710022,
      "objective/train/theoretical_loss": 3.513324405188806,
      "objective/train/tokens_used": 1681797600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2435891330242157,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504673719406128,
      "objective/train/weighted_lm_loss": 1.4789938926696777,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.513324405188806,
      "tokens_seen": 1661337600
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005017251755265797,
      "loss": 1.4212,
      "theoretical_loss": 3.513324405188806,
      "tokens_seen": 1661337600
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005017051153460381,
      "loss": 1.3659,
      "theoretical_loss": 3.5133125211080074,
      "tokens_seen": 1661403136
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005016850551654965,
      "loss": 1.2744,
      "theoretical_loss": 3.5133006376272338,
      "tokens_seen": 1661468672
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005016649949849549,
      "loss": 1.4329,
      "theoretical_loss": 3.513288754746431,
      "tokens_seen": 1661534208
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005016449348044132,
      "loss": 1.4612,
      "theoretical_loss": 3.513276872465546,
      "tokens_seen": 1661599744
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005016248746238716,
      "loss": 1.3999,
      "theoretical_loss": 3.513264990784524,
      "tokens_seen": 1661665280
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00050160481444333,
      "loss": 1.3787,
      "theoretical_loss": 3.5132531097033115,
      "tokens_seen": 1661730816
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005015847542627884,
      "loss": 1.2799,
      "theoretical_loss": 3.5132412292218547,
      "tokens_seen": 1661796352
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005015646940822467,
      "loss": 1.3891,
      "theoretical_loss": 3.513229349340099,
      "tokens_seen": 1661861888
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005015446339017051,
      "loss": 1.3207,
      "theoretical_loss": 3.5132174700579917,
      "tokens_seen": 1661927424
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005015245737211634,
      "loss": 1.4281,
      "theoretical_loss": 3.5132055913754776,
      "tokens_seen": 1661992960
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005015045135406219,
      "loss": 1.4372,
      "theoretical_loss": 3.5131937132925035,
      "tokens_seen": 1662058496
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005014844533600803,
      "loss": 1.421,
      "theoretical_loss": 3.5131818358090148,
      "tokens_seen": 1662124032
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005014643931795386,
      "loss": 1.4352,
      "theoretical_loss": 3.5131699589249585,
      "tokens_seen": 1662189568
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005014443329989971,
      "loss": 1.4644,
      "theoretical_loss": 3.5131580826402806,
      "tokens_seen": 1662255104
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005014242728184554,
      "loss": 1.4344,
      "theoretical_loss": 3.513146206954927,
      "tokens_seen": 1662320640
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005014042126379138,
      "loss": 1.4657,
      "theoretical_loss": 3.5131343318688435,
      "tokens_seen": 1662386176
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005013841524573721,
      "loss": 1.5087,
      "theoretical_loss": 3.513122457381977,
      "tokens_seen": 1662451712
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005013640922768305,
      "loss": 1.4193,
      "theoretical_loss": 3.5131105834942726,
      "tokens_seen": 1662517248
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005013440320962889,
      "loss": 1.469,
      "theoretical_loss": 3.5130987102056768,
      "tokens_seen": 1662582784
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005013239719157473,
      "loss": 1.476,
      "theoretical_loss": 3.513086837516137,
      "tokens_seen": 1662648320
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005013039117352056,
      "loss": 1.3895,
      "theoretical_loss": 3.5130749654255977,
      "tokens_seen": 1662713856
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005012838515546641,
      "loss": 1.4447,
      "theoretical_loss": 3.5130630939340053,
      "tokens_seen": 1662779392
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005012637913741223,
      "loss": 1.6183,
      "theoretical_loss": 3.513051223041307,
      "tokens_seen": 1662844928
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005012437311935808,
      "loss": 1.3635,
      "theoretical_loss": 3.513039352747448,
      "tokens_seen": 1662910464
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005012236710130392,
      "loss": 1.3207,
      "theoretical_loss": 3.513027483052376,
      "tokens_seen": 1662976000
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005012036108324975,
      "loss": 1.3989,
      "theoretical_loss": 3.5130156139560347,
      "tokens_seen": 1663041536
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005011835506519559,
      "loss": 1.4173,
      "theoretical_loss": 3.5130037454583727,
      "tokens_seen": 1663107072
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005011634904714143,
      "loss": 1.4569,
      "theoretical_loss": 3.5129918775593345,
      "tokens_seen": 1663172608
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005011434302908726,
      "loss": 1.4064,
      "theoretical_loss": 3.5129800102588673,
      "tokens_seen": 1663238144
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005011233701103311,
      "loss": 1.5413,
      "theoretical_loss": 3.512968143556917,
      "tokens_seen": 1663303680
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005011033099297893,
      "loss": 1.4454,
      "theoretical_loss": 3.51295627745343,
      "tokens_seen": 1663369216
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005010832497492478,
      "loss": 1.4805,
      "theoretical_loss": 3.512944411948352,
      "tokens_seen": 1663434752
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005010631895687062,
      "loss": 1.4416,
      "theoretical_loss": 3.51293254704163,
      "tokens_seen": 1663500288
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005010431293881645,
      "loss": 1.3551,
      "theoretical_loss": 3.51292068273321,
      "tokens_seen": 1663565824
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005010230692076229,
      "loss": 1.4354,
      "theoretical_loss": 3.512908819023038,
      "tokens_seen": 1663631360
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005010030090270812,
      "loss": 1.397,
      "theoretical_loss": 3.5128969559110605,
      "tokens_seen": 1663696896
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005009829488465396,
      "loss": 1.3188,
      "theoretical_loss": 3.512885093397224,
      "tokens_seen": 1663762432
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005009628886659981,
      "loss": 1.4755,
      "theoretical_loss": 3.512873231481474,
      "tokens_seen": 1663827968
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005009428284854563,
      "loss": 1.472,
      "theoretical_loss": 3.5128613701637574,
      "tokens_seen": 1663893504
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005009227683049148,
      "loss": 1.3939,
      "theoretical_loss": 3.512849509444021,
      "tokens_seen": 1663959040
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005009027081243731,
      "loss": 1.3981,
      "theoretical_loss": 3.5128376493222095,
      "tokens_seen": 1664024576
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005008826479438315,
      "loss": 1.3848,
      "theoretical_loss": 3.512825789798271,
      "tokens_seen": 1664090112
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005008625877632899,
      "loss": 1.4467,
      "theoretical_loss": 3.512813930872151,
      "tokens_seen": 1664155648
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005008425275827483,
      "loss": 1.2157,
      "theoretical_loss": 3.5128020725437956,
      "tokens_seen": 1664221184
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005008224674022066,
      "loss": 1.4707,
      "theoretical_loss": 3.512790214813151,
      "tokens_seen": 1664286720
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005008024072216651,
      "loss": 1.528,
      "theoretical_loss": 3.5127783576801646,
      "tokens_seen": 1664352256
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005007823470411233,
      "loss": 1.4762,
      "theoretical_loss": 3.5127665011447817,
      "tokens_seen": 1664417792
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005007622868605818,
      "loss": 1.4667,
      "theoretical_loss": 3.5127546452069494,
      "tokens_seen": 1664483328
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005007422266800401,
      "loss": 1.5301,
      "theoretical_loss": 3.5127427898666133,
      "tokens_seen": 1664548864
    },
    {
      "epoch": 0.5,
      "objective/train/advantage_avg": 0.49012744426727295,
      "objective/train/docs_used": 608106,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2502481937408447,
      "objective/train/original_loss": 1.2502479553222656,
      "objective/train/theoretical_loss": 3.51273093512372,
      "objective/train/tokens_used": 1685074400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24083980917930603,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502370595932007,
      "objective/train/weighted_lm_loss": 1.3129973411560059,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.51273093512372,
      "tokens_seen": 1664614400
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005007221664994985,
      "loss": 1.3971,
      "theoretical_loss": 3.51273093512372,
      "tokens_seen": 1664614400
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005007021063189569,
      "loss": 1.4822,
      "theoretical_loss": 3.5127190809782167,
      "tokens_seen": 1664679936
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005006820461384153,
      "loss": 1.4601,
      "theoretical_loss": 3.512707227430049,
      "tokens_seen": 1664745472
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005006619859578736,
      "loss": 1.3867,
      "theoretical_loss": 3.5126953744791627,
      "tokens_seen": 1664811008
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000500641925777332,
      "loss": 1.3288,
      "theoretical_loss": 3.512683522125505,
      "tokens_seen": 1664876544
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005006218655967903,
      "loss": 1.4418,
      "theoretical_loss": 3.5126716703690226,
      "tokens_seen": 1664942080
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005006018054162488,
      "loss": 1.5012,
      "theoretical_loss": 3.512659819209661,
      "tokens_seen": 1665007616
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005005817452357071,
      "loss": 1.4038,
      "theoretical_loss": 3.5126479686473675,
      "tokens_seen": 1665073152
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005005616850551655,
      "loss": 1.3025,
      "theoretical_loss": 3.512636118682088,
      "tokens_seen": 1665138688
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005005416248746238,
      "loss": 1.5316,
      "theoretical_loss": 3.5126242693137693,
      "tokens_seen": 1665204224
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005005215646940823,
      "loss": 1.4033,
      "theoretical_loss": 3.5126124205423572,
      "tokens_seen": 1665269760
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005005015045135406,
      "loss": 1.418,
      "theoretical_loss": 3.5126005723677984,
      "tokens_seen": 1665335296
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000500481444332999,
      "loss": 1.3401,
      "theoretical_loss": 3.5125887247900396,
      "tokens_seen": 1665400832
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005004613841524574,
      "loss": 1.5513,
      "theoretical_loss": 3.5125768778090274,
      "tokens_seen": 1665466368
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005004413239719158,
      "loss": 1.4146,
      "theoretical_loss": 3.5125650314247077,
      "tokens_seen": 1665531904
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005004212637913741,
      "loss": 1.4341,
      "theoretical_loss": 3.5125531856370276,
      "tokens_seen": 1665597440
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005004012036108325,
      "loss": 1.4887,
      "theoretical_loss": 3.512541340445933,
      "tokens_seen": 1665662976
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005003811434302908,
      "loss": 1.51,
      "theoretical_loss": 3.5125294958513704,
      "tokens_seen": 1665728512
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005003610832497493,
      "loss": 1.4905,
      "theoretical_loss": 3.512517651853287,
      "tokens_seen": 1665794048
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005003410230692076,
      "loss": 1.3675,
      "theoretical_loss": 3.5125058084516283,
      "tokens_seen": 1665859584
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000500320962888666,
      "loss": 1.3942,
      "theoretical_loss": 3.5124939656463416,
      "tokens_seen": 1665925120
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005003009027081244,
      "loss": 1.4266,
      "theoretical_loss": 3.512482123437373,
      "tokens_seen": 1665990656
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005002808425275827,
      "loss": 1.3509,
      "theoretical_loss": 3.5124702818246694,
      "tokens_seen": 1666056192
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005002607823470411,
      "loss": 1.3745,
      "theoretical_loss": 3.5124584408081767,
      "tokens_seen": 1666121728
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005002407221664996,
      "loss": 1.4645,
      "theoretical_loss": 3.5124466003878423,
      "tokens_seen": 1666187264
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005002206619859578,
      "loss": 1.4317,
      "theoretical_loss": 3.5124347605636115,
      "tokens_seen": 1666252800
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005002006018054163,
      "loss": 1.4693,
      "theoretical_loss": 3.512422921335432,
      "tokens_seen": 1666318336
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0005001805416248745,
      "loss": 1.4647,
      "theoretical_loss": 3.5124110827032498,
      "tokens_seen": 1666383872
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000500160481444333,
      "loss": 1.3741,
      "theoretical_loss": 3.512399244667012,
      "tokens_seen": 1666449408
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0005001404212637914,
      "loss": 1.4634,
      "theoretical_loss": 3.5123874072266643,
      "tokens_seen": 1666514944
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0005001203610832497,
      "loss": 1.4925,
      "theoretical_loss": 3.5123755703821535,
      "tokens_seen": 1666580480
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0005001003009027081,
      "loss": 1.5384,
      "theoretical_loss": 3.512363734133427,
      "tokens_seen": 1666646016
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0005000802407221666,
      "loss": 1.3981,
      "theoretical_loss": 3.5123518984804303,
      "tokens_seen": 1666711552
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0005000601805416248,
      "loss": 1.5116,
      "theoretical_loss": 3.5123400634231112,
      "tokens_seen": 1666777088
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0005000401203610833,
      "loss": 1.5179,
      "theoretical_loss": 3.5123282289614153,
      "tokens_seen": 1666842624
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0005000200601805415,
      "loss": 1.5754,
      "theoretical_loss": 3.512316395095289,
      "tokens_seen": 1666908160
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0005,
      "loss": 1.4304,
      "theoretical_loss": 3.5123045618246795,
      "tokens_seen": 1666973696
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004999799398194584,
      "loss": 1.4321,
      "theoretical_loss": 3.5122927291495336,
      "tokens_seen": 1667039232
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004999598796389167,
      "loss": 1.3087,
      "theoretical_loss": 3.5122808970697976,
      "tokens_seen": 1667104768
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004999398194583751,
      "loss": 1.5075,
      "theoretical_loss": 3.5122690655854187,
      "tokens_seen": 1667170304
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004999197592778335,
      "loss": 1.4358,
      "theoretical_loss": 3.5122572346963423,
      "tokens_seen": 1667235840
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004998996990972919,
      "loss": 1.4595,
      "theoretical_loss": 3.512245404402516,
      "tokens_seen": 1667301376
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004998796389167503,
      "loss": 1.3346,
      "theoretical_loss": 3.5122335747038864,
      "tokens_seen": 1667366912
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004998595787362087,
      "loss": 1.3109,
      "theoretical_loss": 3.5122217456004,
      "tokens_seen": 1667432448
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000499839518555667,
      "loss": 1.5521,
      "theoretical_loss": 3.512209917092003,
      "tokens_seen": 1667497984
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004998194583751254,
      "loss": 1.3336,
      "theoretical_loss": 3.512198089178643,
      "tokens_seen": 1667563520
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004997993981945837,
      "loss": 1.439,
      "theoretical_loss": 3.512186261860266,
      "tokens_seen": 1667629056
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004997793380140421,
      "loss": 1.5244,
      "theoretical_loss": 3.5121744351368194,
      "tokens_seen": 1667694592
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004997592778335005,
      "loss": 1.4795,
      "theoretical_loss": 3.5121626090082487,
      "tokens_seen": 1667760128
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004997392176529588,
      "loss": 1.458,
      "theoretical_loss": 3.512150783474502,
      "tokens_seen": 1667825664
    },
    {
      "epoch": 0.51,
      "objective/train/advantage_avg": 0.49619147181510925,
      "objective/train/docs_used": 609331,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4826010465621948,
      "objective/train/original_loss": 1.4826009273529053,
      "objective/train/theoretical_loss": 3.512138958535525,
      "objective/train/tokens_used": 1688351200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2465878427028656,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508729219436646,
      "objective/train/weighted_lm_loss": 1.5580132007598877,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.512138958535525,
      "tokens_seen": 1667891200
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004997191574724173,
      "loss": 1.5018,
      "theoretical_loss": 3.512138958535525,
      "tokens_seen": 1667891200
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004996990972918757,
      "loss": 1.3841,
      "theoretical_loss": 3.5121271341912648,
      "tokens_seen": 1667956736
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000499679037111334,
      "loss": 1.412,
      "theoretical_loss": 3.5121153104416676,
      "tokens_seen": 1668022272
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004996589769307924,
      "loss": 1.3532,
      "theoretical_loss": 3.5121034872866814,
      "tokens_seen": 1668087808
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004996389167502507,
      "loss": 1.4173,
      "theoretical_loss": 3.5120916647262517,
      "tokens_seen": 1668153344
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004996188565697091,
      "loss": 1.2459,
      "theoretical_loss": 3.512079842760326,
      "tokens_seen": 1668218880
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004995987963891675,
      "loss": 1.4072,
      "theoretical_loss": 3.5120680213888504,
      "tokens_seen": 1668284416
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004995787362086258,
      "loss": 1.3685,
      "theoretical_loss": 3.512056200611772,
      "tokens_seen": 1668349952
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004995586760280842,
      "loss": 1.3964,
      "theoretical_loss": 3.512044380429038,
      "tokens_seen": 1668415488
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004995386158475427,
      "loss": 1.4253,
      "theoretical_loss": 3.512032560840594,
      "tokens_seen": 1668481024
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000499518555667001,
      "loss": 1.4958,
      "theoretical_loss": 3.512020741846388,
      "tokens_seen": 1668546560
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004994984954864594,
      "loss": 1.3265,
      "theoretical_loss": 3.5120089234463663,
      "tokens_seen": 1668612096
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004994784353059178,
      "loss": 1.4646,
      "theoretical_loss": 3.511997105640476,
      "tokens_seen": 1668677632
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004994583751253761,
      "loss": 1.4517,
      "theoretical_loss": 3.511985288428663,
      "tokens_seen": 1668743168
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004994383149448345,
      "loss": 1.333,
      "theoretical_loss": 3.511973471810875,
      "tokens_seen": 1668808704
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004994182547642928,
      "loss": 1.4085,
      "theoretical_loss": 3.5119616557870583,
      "tokens_seen": 1668874240
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004993981945837512,
      "loss": 1.2898,
      "theoretical_loss": 3.5119498403571603,
      "tokens_seen": 1668939776
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004993781344032096,
      "loss": 1.4782,
      "theoretical_loss": 3.511938025521127,
      "tokens_seen": 1669005312
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000499358074222668,
      "loss": 1.5369,
      "theoretical_loss": 3.5119262112789063,
      "tokens_seen": 1669070848
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004993380140421264,
      "loss": 1.4209,
      "theoretical_loss": 3.5119143976304437,
      "tokens_seen": 1669136384
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004993179538615848,
      "loss": 1.3862,
      "theoretical_loss": 3.511902584575687,
      "tokens_seen": 1669201920
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004992978936810431,
      "loss": 1.4786,
      "theoretical_loss": 3.511890772114583,
      "tokens_seen": 1669267456
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004992778335005015,
      "loss": 1.3488,
      "theoretical_loss": 3.5118789602470786,
      "tokens_seen": 1669332992
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004992577733199598,
      "loss": 1.5981,
      "theoretical_loss": 3.51186714897312,
      "tokens_seen": 1669398528
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004992377131394183,
      "loss": 1.5406,
      "theoretical_loss": 3.511855338292655,
      "tokens_seen": 1669464064
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004992176529588767,
      "loss": 1.4995,
      "theoretical_loss": 3.5118435282056297,
      "tokens_seen": 1669529600
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000499197592778335,
      "loss": 1.4408,
      "theoretical_loss": 3.5118317187119916,
      "tokens_seen": 1669595136
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004991775325977934,
      "loss": 1.4028,
      "theoretical_loss": 3.5118199098116873,
      "tokens_seen": 1669660672
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004991574724172518,
      "loss": 1.3894,
      "theoretical_loss": 3.5118081015046636,
      "tokens_seen": 1669726208
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004991374122367101,
      "loss": 1.5057,
      "theoretical_loss": 3.5117962937908676,
      "tokens_seen": 1669791744
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004991173520561685,
      "loss": 1.497,
      "theoretical_loss": 3.511784486670246,
      "tokens_seen": 1669857280
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000499097291875627,
      "loss": 1.4635,
      "theoretical_loss": 3.5117726801427454,
      "tokens_seen": 1669922816
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004990772316950853,
      "loss": 1.2625,
      "theoretical_loss": 3.5117608742083135,
      "tokens_seen": 1669988352
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004990571715145437,
      "loss": 1.372,
      "theoretical_loss": 3.5117490688668975,
      "tokens_seen": 1670053888
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000499037111334002,
      "loss": 1.4809,
      "theoretical_loss": 3.5117372641184432,
      "tokens_seen": 1670119424
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004990170511534604,
      "loss": 1.5266,
      "theoretical_loss": 3.511725459962898,
      "tokens_seen": 1670184960
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004989969909729188,
      "loss": 1.4475,
      "theoretical_loss": 3.5117136564002096,
      "tokens_seen": 1670250496
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004989769307923771,
      "loss": 1.3869,
      "theoretical_loss": 3.511701853430324,
      "tokens_seen": 1670316032
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004989568706118355,
      "loss": 1.5093,
      "theoretical_loss": 3.5116900510531885,
      "tokens_seen": 1670381568
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004989368104312939,
      "loss": 1.4206,
      "theoretical_loss": 3.5116782492687504,
      "tokens_seen": 1670447104
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004989167502507523,
      "loss": 1.4863,
      "theoretical_loss": 3.511666448076956,
      "tokens_seen": 1670512640
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004988966900702107,
      "loss": 1.3839,
      "theoretical_loss": 3.5116546474777524,
      "tokens_seen": 1670578176
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004988766298896691,
      "loss": 1.4283,
      "theoretical_loss": 3.5116428474710872,
      "tokens_seen": 1670643712
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004988565697091274,
      "loss": 1.4822,
      "theoretical_loss": 3.5116310480569073,
      "tokens_seen": 1670709248
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004988365095285858,
      "loss": 1.378,
      "theoretical_loss": 3.5116192492351592,
      "tokens_seen": 1670774784
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004988164493480441,
      "loss": 1.4256,
      "theoretical_loss": 3.51160745100579,
      "tokens_seen": 1670840320
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004987963891675025,
      "loss": 1.5586,
      "theoretical_loss": 3.5115956533687473,
      "tokens_seen": 1670905856
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004987763289869609,
      "loss": 1.4658,
      "theoretical_loss": 3.511583856323978,
      "tokens_seen": 1670971392
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004987562688064192,
      "loss": 1.59,
      "theoretical_loss": 3.5115720598714284,
      "tokens_seen": 1671036928
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004987362086258777,
      "loss": 1.6269,
      "theoretical_loss": 3.511560264011046,
      "tokens_seen": 1671102464
    },
    {
      "epoch": 0.51,
      "objective/train/advantage_avg": 0.4945238530635834,
      "objective/train/docs_used": 610761,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3090906143188477,
      "objective/train/original_loss": 1.3090906143188477,
      "objective/train/theoretical_loss": 3.5115484687427783,
      "objective/train/tokens_used": 1691628000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24479834735393524,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506969690322876,
      "objective/train/weighted_lm_loss": 1.3752391338348389,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.5115484687427783,
      "tokens_seen": 1671168000
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004987161484453361,
      "loss": 1.3914,
      "theoretical_loss": 3.5115484687427783,
      "tokens_seen": 1671168000
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004986960882647944,
      "loss": 1.4475,
      "theoretical_loss": 3.5115366740665714,
      "tokens_seen": 1671233536
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004986760280842528,
      "loss": 1.3459,
      "theoretical_loss": 3.5115248799823733,
      "tokens_seen": 1671299072
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004986559679037111,
      "loss": 1.3637,
      "theoretical_loss": 3.5115130864901305,
      "tokens_seen": 1671364608
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004986359077231695,
      "loss": 1.2299,
      "theoretical_loss": 3.5115012935897907,
      "tokens_seen": 1671430144
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004986158475426279,
      "loss": 1.5468,
      "theoretical_loss": 3.5114895012813,
      "tokens_seen": 1671495680
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004985957873620862,
      "loss": 1.3325,
      "theoretical_loss": 3.5114777095646064,
      "tokens_seen": 1671561216
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004985757271815446,
      "loss": 1.2738,
      "theoretical_loss": 3.5114659184396566,
      "tokens_seen": 1671626752
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004985556670010031,
      "loss": 1.482,
      "theoretical_loss": 3.5114541279063975,
      "tokens_seen": 1671692288
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004985356068204614,
      "loss": 1.4577,
      "theoretical_loss": 3.511442337964777,
      "tokens_seen": 1671757824
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004985155466399198,
      "loss": 1.3866,
      "theoretical_loss": 3.511430548614741,
      "tokens_seen": 1671823360
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004984954864593782,
      "loss": 1.4124,
      "theoretical_loss": 3.5114187598562374,
      "tokens_seen": 1671888896
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004984754262788365,
      "loss": 1.4753,
      "theoretical_loss": 3.511406971689214,
      "tokens_seen": 1671954432
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004984553660982949,
      "loss": 1.4448,
      "theoretical_loss": 3.5113951841136166,
      "tokens_seen": 1672019968
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004984353059177532,
      "loss": 1.4049,
      "theoretical_loss": 3.511383397129393,
      "tokens_seen": 1672085504
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004984152457372116,
      "loss": 1.3644,
      "theoretical_loss": 3.5113716107364903,
      "tokens_seen": 1672151040
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00049839518555667,
      "loss": 1.3356,
      "theoretical_loss": 3.511359824934856,
      "tokens_seen": 1672216576
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004983751253761284,
      "loss": 1.3364,
      "theoretical_loss": 3.5113480397244365,
      "tokens_seen": 1672282112
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004983550651955868,
      "loss": 1.4638,
      "theoretical_loss": 3.5113362551051797,
      "tokens_seen": 1672347648
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004983350050150452,
      "loss": 1.3873,
      "theoretical_loss": 3.5113244710770326,
      "tokens_seen": 1672413184
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004983149448345035,
      "loss": 1.3348,
      "theoretical_loss": 3.5113126876399416,
      "tokens_seen": 1672478720
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004982948846539619,
      "loss": 1.4591,
      "theoretical_loss": 3.5113009047938553,
      "tokens_seen": 1672544256
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004982748244734202,
      "loss": 1.337,
      "theoretical_loss": 3.5112891225387193,
      "tokens_seen": 1672609792
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004982547642928786,
      "loss": 1.4341,
      "theoretical_loss": 3.5112773408744826,
      "tokens_seen": 1672675328
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000498234704112337,
      "loss": 1.376,
      "theoretical_loss": 3.511265559801091,
      "tokens_seen": 1672740864
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004982146439317953,
      "loss": 1.4533,
      "theoretical_loss": 3.5112537793184924,
      "tokens_seen": 1672806400
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004981945837512538,
      "loss": 1.4101,
      "theoretical_loss": 3.511241999426634,
      "tokens_seen": 1672871936
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004981745235707122,
      "loss": 1.4811,
      "theoretical_loss": 3.511230220125462,
      "tokens_seen": 1672937472
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004981544633901705,
      "loss": 1.3561,
      "theoretical_loss": 3.5112184414149255,
      "tokens_seen": 1673003008
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004981344032096289,
      "loss": 1.4315,
      "theoretical_loss": 3.5112066632949706,
      "tokens_seen": 1673068544
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004981143430290873,
      "loss": 1.2327,
      "theoretical_loss": 3.5111948857655446,
      "tokens_seen": 1673134080
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004980942828485456,
      "loss": 1.479,
      "theoretical_loss": 3.511183108826595,
      "tokens_seen": 1673199616
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000498074222668004,
      "loss": 1.4998,
      "theoretical_loss": 3.5111713324780687,
      "tokens_seen": 1673265152
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004980541624874623,
      "loss": 1.3851,
      "theoretical_loss": 3.5111595567199134,
      "tokens_seen": 1673330688
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004980341023069207,
      "loss": 1.4558,
      "theoretical_loss": 3.511147781552076,
      "tokens_seen": 1673396224
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004980140421263792,
      "loss": 1.5328,
      "theoretical_loss": 3.5111360069745046,
      "tokens_seen": 1673461760
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004979939819458375,
      "loss": 1.4395,
      "theoretical_loss": 3.5111242329871457,
      "tokens_seen": 1673527296
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004979739217652959,
      "loss": 1.4558,
      "theoretical_loss": 3.5111124595899463,
      "tokens_seen": 1673592832
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004979538615847543,
      "loss": 1.4296,
      "theoretical_loss": 3.5111006867828545,
      "tokens_seen": 1673658368
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004979338014042126,
      "loss": 1.407,
      "theoretical_loss": 3.5110889145658177,
      "tokens_seen": 1673723904
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000497913741223671,
      "loss": 1.4167,
      "theoretical_loss": 3.5110771429387824,
      "tokens_seen": 1673789440
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004978936810431293,
      "loss": 1.4816,
      "theoretical_loss": 3.5110653719016964,
      "tokens_seen": 1673854976
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004978736208625877,
      "loss": 1.3181,
      "theoretical_loss": 3.511053601454507,
      "tokens_seen": 1673920512
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004978535606820461,
      "loss": 1.4154,
      "theoretical_loss": 3.5110418315971623,
      "tokens_seen": 1673986048
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004978335005015045,
      "loss": 1.4678,
      "theoretical_loss": 3.511030062329608,
      "tokens_seen": 1674051584
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004978134403209629,
      "loss": 1.4949,
      "theoretical_loss": 3.511018293651793,
      "tokens_seen": 1674117120
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004977933801404213,
      "loss": 1.3863,
      "theoretical_loss": 3.5110065255636638,
      "tokens_seen": 1674182656
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004977733199598796,
      "loss": 1.4116,
      "theoretical_loss": 3.510994758065168,
      "tokens_seen": 1674248192
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000497753259779338,
      "loss": 1.4398,
      "theoretical_loss": 3.5109829911562533,
      "tokens_seen": 1674313728
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004977331995987965,
      "loss": 1.4239,
      "theoretical_loss": 3.5109712248368665,
      "tokens_seen": 1674379264
    },
    {
      "epoch": 0.51,
      "objective/train/advantage_avg": 0.4949960708618164,
      "objective/train/docs_used": 611357,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.33742356300354,
      "objective/train/original_loss": 1.33742356300354,
      "objective/train/theoretical_loss": 3.5109594591069553,
      "objective/train/tokens_used": 1694904800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24525581300258636,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507464408874512,
      "objective/train/weighted_lm_loss": 1.4052034616470337,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5109594591069553,
      "tokens_seen": 1674444800
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004977131394182548,
      "loss": 1.3469,
      "theoretical_loss": 3.5109594591069553,
      "tokens_seen": 1674444800
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004976930792377132,
      "loss": 1.4624,
      "theoretical_loss": 3.510947693966467,
      "tokens_seen": 1674510336
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004976730190571715,
      "loss": 1.2508,
      "theoretical_loss": 3.5109359294153495,
      "tokens_seen": 1674575872
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004976529588766299,
      "loss": 1.4457,
      "theoretical_loss": 3.5109241654535497,
      "tokens_seen": 1674641408
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004976328986960883,
      "loss": 1.4016,
      "theoretical_loss": 3.510912402081015,
      "tokens_seen": 1674706944
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004976128385155466,
      "loss": 1.3718,
      "theoretical_loss": 3.5109006392976925,
      "tokens_seen": 1674772480
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000497592778335005,
      "loss": 1.3966,
      "theoretical_loss": 3.5108888771035307,
      "tokens_seen": 1674838016
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004975727181544635,
      "loss": 1.4712,
      "theoretical_loss": 3.510877115498476,
      "tokens_seen": 1674903552
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004975526579739218,
      "loss": 1.3562,
      "theoretical_loss": 3.510865354482476,
      "tokens_seen": 1674969088
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004975325977933802,
      "loss": 1.3775,
      "theoretical_loss": 3.510853594055479,
      "tokens_seen": 1675034624
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004975125376128386,
      "loss": 1.4421,
      "theoretical_loss": 3.5108418342174317,
      "tokens_seen": 1675100160
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004974924774322969,
      "loss": 1.3397,
      "theoretical_loss": 3.5108300749682817,
      "tokens_seen": 1675165696
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004974724172517553,
      "loss": 1.4672,
      "theoretical_loss": 3.5108183163079767,
      "tokens_seen": 1675231232
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004974523570712136,
      "loss": 1.4136,
      "theoretical_loss": 3.510806558236464,
      "tokens_seen": 1675296768
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000497432296890672,
      "loss": 1.3971,
      "theoretical_loss": 3.5107948007536907,
      "tokens_seen": 1675362304
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004974122367101304,
      "loss": 1.3793,
      "theoretical_loss": 3.510783043859605,
      "tokens_seen": 1675427840
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004973921765295888,
      "loss": 1.4197,
      "theoretical_loss": 3.5107712875541535,
      "tokens_seen": 1675493376
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004973721163490472,
      "loss": 1.4327,
      "theoretical_loss": 3.5107595318372846,
      "tokens_seen": 1675558912
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004973520561685056,
      "loss": 1.3538,
      "theoretical_loss": 3.5107477767089454,
      "tokens_seen": 1675624448
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004973319959879639,
      "loss": 1.4676,
      "theoretical_loss": 3.510736022169084,
      "tokens_seen": 1675689984
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004973119358074223,
      "loss": 1.285,
      "theoretical_loss": 3.510724268217647,
      "tokens_seen": 1675755520
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004972918756268806,
      "loss": 1.3668,
      "theoretical_loss": 3.5107125148545824,
      "tokens_seen": 1675821056
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000497271815446339,
      "loss": 1.3644,
      "theoretical_loss": 3.5107007620798374,
      "tokens_seen": 1675886592
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004972517552657974,
      "loss": 1.3004,
      "theoretical_loss": 3.51068900989336,
      "tokens_seen": 1675952128
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004972316950852557,
      "loss": 1.4525,
      "theoretical_loss": 3.510677258295098,
      "tokens_seen": 1676017664
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004972116349047142,
      "loss": 1.4792,
      "theoretical_loss": 3.5106655072849984,
      "tokens_seen": 1676083200
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004971915747241726,
      "loss": 1.4177,
      "theoretical_loss": 3.510653756863009,
      "tokens_seen": 1676148736
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004971715145436309,
      "loss": 1.4501,
      "theoretical_loss": 3.5106420070290767,
      "tokens_seen": 1676214272
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004971514543630893,
      "loss": 1.3917,
      "theoretical_loss": 3.5106302577831503,
      "tokens_seen": 1676279808
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004971313941825477,
      "loss": 1.4267,
      "theoretical_loss": 3.5106185091251767,
      "tokens_seen": 1676345344
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000497111334002006,
      "loss": 1.315,
      "theoretical_loss": 3.5106067610551035,
      "tokens_seen": 1676410880
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004970912738214644,
      "loss": 1.384,
      "theoretical_loss": 3.5105950135728783,
      "tokens_seen": 1676476416
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004970712136409227,
      "loss": 1.484,
      "theoretical_loss": 3.5105832666784487,
      "tokens_seen": 1676541952
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004970511534603811,
      "loss": 1.5299,
      "theoretical_loss": 3.5105715203717627,
      "tokens_seen": 1676607488
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004970310932798396,
      "loss": 1.4902,
      "theoretical_loss": 3.5105597746527675,
      "tokens_seen": 1676673024
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004970110330992979,
      "loss": 1.377,
      "theoretical_loss": 3.5105480295214107,
      "tokens_seen": 1676738560
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004969909729187563,
      "loss": 1.2686,
      "theoretical_loss": 3.51053628497764,
      "tokens_seen": 1676804096
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004969709127382147,
      "loss": 1.4454,
      "theoretical_loss": 3.510524541021403,
      "tokens_seen": 1676869632
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000496950852557673,
      "loss": 1.3643,
      "theoretical_loss": 3.5105127976526482,
      "tokens_seen": 1676935168
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004969307923771314,
      "loss": 1.419,
      "theoretical_loss": 3.5105010548713214,
      "tokens_seen": 1677000704
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004969107321965897,
      "loss": 1.3642,
      "theoretical_loss": 3.5104893126773726,
      "tokens_seen": 1677066240
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004968906720160481,
      "loss": 1.3915,
      "theoretical_loss": 3.5104775710707474,
      "tokens_seen": 1677131776
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004968706118355065,
      "loss": 1.452,
      "theoretical_loss": 3.5104658300513942,
      "tokens_seen": 1677197312
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004968505516549649,
      "loss": 1.4052,
      "theoretical_loss": 3.510454089619261,
      "tokens_seen": 1677262848
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004968304914744233,
      "loss": 1.3004,
      "theoretical_loss": 3.5104423497742956,
      "tokens_seen": 1677328384
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004968104312938817,
      "loss": 1.3949,
      "theoretical_loss": 3.510430610516445,
      "tokens_seen": 1677393920
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00049679037111334,
      "loss": 1.3367,
      "theoretical_loss": 3.5104188718456575,
      "tokens_seen": 1677459456
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004967703109327984,
      "loss": 1.4382,
      "theoretical_loss": 3.5104071337618805,
      "tokens_seen": 1677524992
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004967502507522568,
      "loss": 1.4364,
      "theoretical_loss": 3.510395396265062,
      "tokens_seen": 1677590528
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004967301905717151,
      "loss": 1.3419,
      "theoretical_loss": 3.510383659355149,
      "tokens_seen": 1677656064
    },
    {
      "epoch": 0.51,
      "objective/train/advantage_avg": 0.49177229404449463,
      "objective/train/docs_used": 612648,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.271415114402771,
      "objective/train/original_loss": 1.2714149951934814,
      "objective/train/theoretical_loss": 3.5103719230320904,
      "objective/train/tokens_used": 1698181600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24274088442325592,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504112243652344,
      "objective/train/weighted_lm_loss": 1.3353408575057983,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5103719230320904,
      "tokens_seen": 1677721600
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004967101303911735,
      "loss": 1.3795,
      "theoretical_loss": 3.5103719230320904,
      "tokens_seen": 1677721600
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000496690070210632,
      "loss": 1.3874,
      "theoretical_loss": 3.5103601872958325,
      "tokens_seen": 1677787136
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004966700100300903,
      "loss": 1.4156,
      "theoretical_loss": 3.5103484521463244,
      "tokens_seen": 1677852672
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004966499498495487,
      "loss": 1.4841,
      "theoretical_loss": 3.510336717583513,
      "tokens_seen": 1677918208
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000496629889669007,
      "loss": 1.4204,
      "theoretical_loss": 3.510324983607347,
      "tokens_seen": 1677983744
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004966098294884654,
      "loss": 1.4507,
      "theoretical_loss": 3.5103132502177727,
      "tokens_seen": 1678049280
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004965897693079238,
      "loss": 1.446,
      "theoretical_loss": 3.510301517414739,
      "tokens_seen": 1678114816
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004965697091273821,
      "loss": 1.492,
      "theoretical_loss": 3.5102897851981933,
      "tokens_seen": 1678180352
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004965496489468405,
      "loss": 1.2587,
      "theoretical_loss": 3.5102780535680838,
      "tokens_seen": 1678245888
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004965295887662988,
      "loss": 1.3896,
      "theoretical_loss": 3.510266322524357,
      "tokens_seen": 1678311424
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004965095285857573,
      "loss": 1.4822,
      "theoretical_loss": 3.5102545920669623,
      "tokens_seen": 1678376960
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004964894684052157,
      "loss": 1.5296,
      "theoretical_loss": 3.510242862195846,
      "tokens_seen": 1678442496
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000496469408224674,
      "loss": 1.375,
      "theoretical_loss": 3.510231132910958,
      "tokens_seen": 1678508032
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004964493480441324,
      "loss": 1.4067,
      "theoretical_loss": 3.510219404212244,
      "tokens_seen": 1678573568
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004964292878635908,
      "loss": 1.3806,
      "theoretical_loss": 3.5102076760996526,
      "tokens_seen": 1678639104
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004964092276830491,
      "loss": 1.3515,
      "theoretical_loss": 3.5101959485731316,
      "tokens_seen": 1678704640
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004963891675025075,
      "loss": 1.4651,
      "theoretical_loss": 3.5101842216326293,
      "tokens_seen": 1678770176
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004963691073219659,
      "loss": 1.4077,
      "theoretical_loss": 3.510172495278093,
      "tokens_seen": 1678835712
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004963490471414242,
      "loss": 1.4825,
      "theoretical_loss": 3.5101607695094708,
      "tokens_seen": 1678901248
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004963289869608827,
      "loss": 1.3084,
      "theoretical_loss": 3.51014904432671,
      "tokens_seen": 1678966784
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000496308926780341,
      "loss": 1.388,
      "theoretical_loss": 3.5101373197297594,
      "tokens_seen": 1679032320
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004962888665997994,
      "loss": 1.4251,
      "theoretical_loss": 3.5101255957185664,
      "tokens_seen": 1679097856
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004962688064192578,
      "loss": 1.2816,
      "theoretical_loss": 3.5101138722930783,
      "tokens_seen": 1679163392
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004962487462387161,
      "loss": 1.4684,
      "theoretical_loss": 3.5101021494532443,
      "tokens_seen": 1679228928
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004962286860581746,
      "loss": 1.3773,
      "theoretical_loss": 3.510090427199011,
      "tokens_seen": 1679294464
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000496208625877633,
      "loss": 1.4501,
      "theoretical_loss": 3.5100787055303266,
      "tokens_seen": 1679360000
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004961885656970913,
      "loss": 1.4616,
      "theoretical_loss": 3.51006698444714,
      "tokens_seen": 1679425536
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004961685055165497,
      "loss": 1.4324,
      "theoretical_loss": 3.510055263949398,
      "tokens_seen": 1679491072
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004961484453360081,
      "loss": 1.3736,
      "theoretical_loss": 3.5100435440370483,
      "tokens_seen": 1679556608
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004961283851554664,
      "loss": 1.4642,
      "theoretical_loss": 3.51003182471004,
      "tokens_seen": 1679622144
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004961083249749248,
      "loss": 1.4127,
      "theoretical_loss": 3.5100201059683203,
      "tokens_seen": 1679687680
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004960882647943831,
      "loss": 1.4665,
      "theoretical_loss": 3.510008387811837,
      "tokens_seen": 1679753216
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004960682046138415,
      "loss": 1.3176,
      "theoretical_loss": 3.5099966702405383,
      "tokens_seen": 1679818752
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004960481444333,
      "loss": 1.2865,
      "theoretical_loss": 3.5099849532543725,
      "tokens_seen": 1679884288
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004960280842527583,
      "loss": 1.5855,
      "theoretical_loss": 3.5099732368532868,
      "tokens_seen": 1679949824
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004960080240722167,
      "loss": 1.5396,
      "theoretical_loss": 3.5099615210372295,
      "tokens_seen": 1680015360
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004959879638916751,
      "loss": 1.4615,
      "theoretical_loss": 3.5099498058061487,
      "tokens_seen": 1680080896
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004959679037111334,
      "loss": 1.4376,
      "theoretical_loss": 3.5099380911599924,
      "tokens_seen": 1680146432
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004959478435305918,
      "loss": 1.354,
      "theoretical_loss": 3.5099263770987084,
      "tokens_seen": 1680211968
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004959277833500501,
      "loss": 1.394,
      "theoretical_loss": 3.5099146636222445,
      "tokens_seen": 1680277504
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004959077231695085,
      "loss": 1.447,
      "theoretical_loss": 3.5099029507305497,
      "tokens_seen": 1680343040
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004958876629889669,
      "loss": 1.4736,
      "theoretical_loss": 3.5098912384235703,
      "tokens_seen": 1680408576
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004958676028084253,
      "loss": 1.3696,
      "theoretical_loss": 3.5098795267012557,
      "tokens_seen": 1680474112
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004958475426278837,
      "loss": 1.4823,
      "theoretical_loss": 3.5098678155635534,
      "tokens_seen": 1680539648
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004958274824473421,
      "loss": 1.4818,
      "theoretical_loss": 3.509856105010412,
      "tokens_seen": 1680605184
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004958074222668004,
      "loss": 1.4665,
      "theoretical_loss": 3.509844395041778,
      "tokens_seen": 1680670720
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004957873620862588,
      "loss": 1.4583,
      "theoretical_loss": 3.509832685657601,
      "tokens_seen": 1680736256
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004957673019057172,
      "loss": 1.5739,
      "theoretical_loss": 3.509820976857829,
      "tokens_seen": 1680801792
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004957472417251755,
      "loss": 1.5177,
      "theoretical_loss": 3.509809268642409,
      "tokens_seen": 1680867328
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004957271815446339,
      "loss": 1.4442,
      "theoretical_loss": 3.5097975610112897,
      "tokens_seen": 1680932864
    },
    {
      "epoch": 0.51,
      "objective/train/advantage_avg": 0.48790156841278076,
      "objective/train/docs_used": 613922,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3153777122497559,
      "objective/train/original_loss": 1.3153777122497559,
      "objective/train/theoretical_loss": 3.509785853964419,
      "objective/train/tokens_used": 1701458400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2402409166097641,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050011396408081,
      "objective/train/weighted_lm_loss": 1.3815168142318726,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.509785853964419,
      "tokens_seen": 1680998400
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004957071213640923,
      "loss": 1.4292,
      "theoretical_loss": 3.509785853964419,
      "tokens_seen": 1680998400
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004956870611835507,
      "loss": 1.4536,
      "theoretical_loss": 3.509774147501745,
      "tokens_seen": 1681063936
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004956670010030091,
      "loss": 1.5557,
      "theoretical_loss": 3.5097624416232156,
      "tokens_seen": 1681129472
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004956469408224674,
      "loss": 1.4925,
      "theoretical_loss": 3.5097507363287797,
      "tokens_seen": 1681195008
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004956268806419258,
      "loss": 1.3666,
      "theoretical_loss": 3.5097390316183845,
      "tokens_seen": 1681260544
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004956068204613842,
      "loss": 1.5124,
      "theoretical_loss": 3.5097273274919782,
      "tokens_seen": 1681326080
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004955867602808425,
      "loss": 1.4547,
      "theoretical_loss": 3.509715623949509,
      "tokens_seen": 1681391616
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004955667001003009,
      "loss": 1.4439,
      "theoretical_loss": 3.5097039209909253,
      "tokens_seen": 1681457152
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004955466399197592,
      "loss": 1.5022,
      "theoretical_loss": 3.509692218616175,
      "tokens_seen": 1681522688
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004955265797392177,
      "loss": 1.4807,
      "theoretical_loss": 3.509680516825206,
      "tokens_seen": 1681588224
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004955065195586761,
      "loss": 1.4785,
      "theoretical_loss": 3.509668815617967,
      "tokens_seen": 1681653760
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004954864593781344,
      "loss": 1.4326,
      "theoretical_loss": 3.509657114994406,
      "tokens_seen": 1681719296
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004954663991975928,
      "loss": 1.455,
      "theoretical_loss": 3.50964541495447,
      "tokens_seen": 1681784832
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004954463390170512,
      "loss": 1.4082,
      "theoretical_loss": 3.5096337154981088,
      "tokens_seen": 1681850368
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004954262788365095,
      "loss": 1.4372,
      "theoretical_loss": 3.5096220166252694,
      "tokens_seen": 1681915904
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004954062186559679,
      "loss": 1.4351,
      "theoretical_loss": 3.5096103183359,
      "tokens_seen": 1681981440
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004953861584754263,
      "loss": 1.3774,
      "theoretical_loss": 3.5095986206299505,
      "tokens_seen": 1682046976
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004953660982948846,
      "loss": 1.4163,
      "theoretical_loss": 3.5095869235073667,
      "tokens_seen": 1682112512
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004953460381143431,
      "loss": 1.4221,
      "theoretical_loss": 3.509575226968098,
      "tokens_seen": 1682178048
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004953259779338014,
      "loss": 1.5509,
      "theoretical_loss": 3.509563531012092,
      "tokens_seen": 1682243584
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004953059177532598,
      "loss": 1.3529,
      "theoretical_loss": 3.5095518356392983,
      "tokens_seen": 1682309120
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004952858575727182,
      "loss": 1.4064,
      "theoretical_loss": 3.509540140849663,
      "tokens_seen": 1682374656
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004952657973921765,
      "loss": 1.402,
      "theoretical_loss": 3.5095284466431362,
      "tokens_seen": 1682440192
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004952457372116349,
      "loss": 1.3709,
      "theoretical_loss": 3.509516753019665,
      "tokens_seen": 1682505728
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004952256770310933,
      "loss": 1.4498,
      "theoretical_loss": 3.5095050599791975,
      "tokens_seen": 1682571264
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004952056168505516,
      "loss": 1.3848,
      "theoretical_loss": 3.5094933675216833,
      "tokens_seen": 1682636800
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00049518555667001,
      "loss": 1.4336,
      "theoretical_loss": 3.509481675647069,
      "tokens_seen": 1682702336
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004951654964894685,
      "loss": 1.4054,
      "theoretical_loss": 3.509469984355303,
      "tokens_seen": 1682767872
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004951454363089268,
      "loss": 1.4167,
      "theoretical_loss": 3.509458293646335,
      "tokens_seen": 1682833408
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004951253761283852,
      "loss": 1.3459,
      "theoretical_loss": 3.5094466035201126,
      "tokens_seen": 1682898944
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004951053159478435,
      "loss": 1.5032,
      "theoretical_loss": 3.509434913976583,
      "tokens_seen": 1682964480
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004950852557673019,
      "loss": 1.3173,
      "theoretical_loss": 3.5094232250156954,
      "tokens_seen": 1683030016
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004950651955867603,
      "loss": 1.4952,
      "theoretical_loss": 3.509411536637398,
      "tokens_seen": 1683095552
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004950451354062186,
      "loss": 1.4548,
      "theoretical_loss": 3.509399848841639,
      "tokens_seen": 1683161088
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000495025075225677,
      "loss": 1.4914,
      "theoretical_loss": 3.509388161628367,
      "tokens_seen": 1683226624
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004950050150451354,
      "loss": 1.4475,
      "theoretical_loss": 3.50937647499753,
      "tokens_seen": 1683292160
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004949849548645938,
      "loss": 1.426,
      "theoretical_loss": 3.5093647889490756,
      "tokens_seen": 1683357696
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004949648946840522,
      "loss": 1.4441,
      "theoretical_loss": 3.509353103482953,
      "tokens_seen": 1683423232
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004949448345035105,
      "loss": 1.4669,
      "theoretical_loss": 3.509341418599111,
      "tokens_seen": 1683488768
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004949247743229689,
      "loss": 1.3711,
      "theoretical_loss": 3.5093297342974967,
      "tokens_seen": 1683554304
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004949047141424273,
      "loss": 1.5031,
      "theoretical_loss": 3.5093180505780586,
      "tokens_seen": 1683619840
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004948846539618856,
      "loss": 1.415,
      "theoretical_loss": 3.509306367440746,
      "tokens_seen": 1683685376
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000494864593781344,
      "loss": 1.4542,
      "theoretical_loss": 3.509294684885506,
      "tokens_seen": 1683750912
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004948445336008024,
      "loss": 1.3428,
      "theoretical_loss": 3.509283002912288,
      "tokens_seen": 1683816448
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004948244734202607,
      "loss": 1.4562,
      "theoretical_loss": 3.50927132152104,
      "tokens_seen": 1683881984
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004948044132397192,
      "loss": 1.5913,
      "theoretical_loss": 3.5092596407117105,
      "tokens_seen": 1683947520
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004947843530591776,
      "loss": 1.5066,
      "theoretical_loss": 3.509247960484247,
      "tokens_seen": 1684013056
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004947642928786359,
      "loss": 1.5073,
      "theoretical_loss": 3.509236280838599,
      "tokens_seen": 1684078592
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004947442326980943,
      "loss": 1.3078,
      "theoretical_loss": 3.509224601774714,
      "tokens_seen": 1684144128
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004947241725175527,
      "loss": 1.476,
      "theoretical_loss": 3.5092129232925413,
      "tokens_seen": 1684209664
    },
    {
      "epoch": 0.51,
      "objective/train/advantage_avg": 0.4935879707336426,
      "objective/train/docs_used": 615159,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.264041543006897,
      "objective/train/original_loss": 1.2640413045883179,
      "objective/train/theoretical_loss": 3.5092012453920285,
      "objective/train/tokens_used": 1704735200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24384334683418274,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050598382949829,
      "objective/train/weighted_lm_loss": 1.3278924226760864,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.5092012453920285,
      "tokens_seen": 1684275200
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004947041123370111,
      "loss": 1.3803,
      "theoretical_loss": 3.5092012453920285,
      "tokens_seen": 1684275200
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004946840521564695,
      "loss": 1.4228,
      "theoretical_loss": 3.5091895680731247,
      "tokens_seen": 1684340736
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004946639919759278,
      "loss": 1.3708,
      "theoretical_loss": 3.509177891335777,
      "tokens_seen": 1684406272
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004946439317953862,
      "loss": 1.4416,
      "theoretical_loss": 3.5091662151799357,
      "tokens_seen": 1684471808
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004946238716148446,
      "loss": 1.4408,
      "theoretical_loss": 3.5091545396055475,
      "tokens_seen": 1684537344
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004946038114343029,
      "loss": 1.4553,
      "theoretical_loss": 3.5091428646125618,
      "tokens_seen": 1684602880
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004945837512537613,
      "loss": 1.4324,
      "theoretical_loss": 3.509131190200926,
      "tokens_seen": 1684668416
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004945636910732196,
      "loss": 1.338,
      "theoretical_loss": 3.5091195163705904,
      "tokens_seen": 1684733952
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004945436308926781,
      "loss": 1.3911,
      "theoretical_loss": 3.5091078431215017,
      "tokens_seen": 1684799488
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004945235707121365,
      "loss": 1.4693,
      "theoretical_loss": 3.5090961704536094,
      "tokens_seen": 1684865024
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004945035105315948,
      "loss": 1.4488,
      "theoretical_loss": 3.5090844983668616,
      "tokens_seen": 1684930560
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004944834503510532,
      "loss": 1.3854,
      "theoretical_loss": 3.509072826861207,
      "tokens_seen": 1684996096
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004944633901705116,
      "loss": 1.2942,
      "theoretical_loss": 3.5090611559365925,
      "tokens_seen": 1685061632
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004944433299899699,
      "loss": 1.4115,
      "theoretical_loss": 3.509049485592969,
      "tokens_seen": 1685127168
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004944232698094283,
      "loss": 1.4256,
      "theoretical_loss": 3.5090378158302835,
      "tokens_seen": 1685192704
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004944032096288867,
      "loss": 1.3018,
      "theoretical_loss": 3.509026146648485,
      "tokens_seen": 1685258240
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000494383149448345,
      "loss": 1.5403,
      "theoretical_loss": 3.509014478047522,
      "tokens_seen": 1685323776
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004943630892678035,
      "loss": 1.4793,
      "theoretical_loss": 3.5090028100273427,
      "tokens_seen": 1685389312
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004943430290872618,
      "loss": 1.384,
      "theoretical_loss": 3.5089911425878952,
      "tokens_seen": 1685454848
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004943229689067202,
      "loss": 1.5354,
      "theoretical_loss": 3.508979475729129,
      "tokens_seen": 1685520384
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004943029087261786,
      "loss": 1.4675,
      "theoretical_loss": 3.5089678094509926,
      "tokens_seen": 1685585920
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004942828485456369,
      "loss": 1.4867,
      "theoretical_loss": 3.5089561437534336,
      "tokens_seen": 1685651456
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004942627883650953,
      "loss": 1.4755,
      "theoretical_loss": 3.5089444786364012,
      "tokens_seen": 1685716992
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004942427281845537,
      "loss": 1.4258,
      "theoretical_loss": 3.508932814099844,
      "tokens_seen": 1685782528
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000494222668004012,
      "loss": 1.3493,
      "theoretical_loss": 3.5089211501437103,
      "tokens_seen": 1685848064
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004942026078234704,
      "loss": 1.4084,
      "theoretical_loss": 3.5089094867679487,
      "tokens_seen": 1685913600
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004941825476429289,
      "loss": 1.4579,
      "theoretical_loss": 3.508897823972508,
      "tokens_seen": 1685979136
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004941624874623872,
      "loss": 1.3926,
      "theoretical_loss": 3.5088861617573364,
      "tokens_seen": 1686044672
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004941424272818456,
      "loss": 1.4121,
      "theoretical_loss": 3.5088745001223822,
      "tokens_seen": 1686110208
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004941223671013039,
      "loss": 1.5124,
      "theoretical_loss": 3.5088628390675947,
      "tokens_seen": 1686175744
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004941023069207623,
      "loss": 1.4711,
      "theoretical_loss": 3.5088511785929226,
      "tokens_seen": 1686241280
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004940822467402207,
      "loss": 1.4084,
      "theoretical_loss": 3.5088395186983137,
      "tokens_seen": 1686306816
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000494062186559679,
      "loss": 1.3918,
      "theoretical_loss": 3.508827859383717,
      "tokens_seen": 1686372352
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004940421263791374,
      "loss": 1.4027,
      "theoretical_loss": 3.508816200649081,
      "tokens_seen": 1686437888
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004940220661985958,
      "loss": 1.3242,
      "theoretical_loss": 3.508804542494355,
      "tokens_seen": 1686503424
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004940020060180542,
      "loss": 1.3776,
      "theoretical_loss": 3.5087928849194867,
      "tokens_seen": 1686568960
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004939819458375126,
      "loss": 1.4849,
      "theoretical_loss": 3.5087812279244246,
      "tokens_seen": 1686634496
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000493961885656971,
      "loss": 1.344,
      "theoretical_loss": 3.5087695715091183,
      "tokens_seen": 1686700032
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004939418254764293,
      "loss": 1.4318,
      "theoretical_loss": 3.508757915673516,
      "tokens_seen": 1686765568
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004939217652958877,
      "loss": 1.4725,
      "theoretical_loss": 3.508746260417566,
      "tokens_seen": 1686831104
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000493901705115346,
      "loss": 1.4738,
      "theoretical_loss": 3.5087346057412176,
      "tokens_seen": 1686896640
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004938816449348044,
      "loss": 1.5432,
      "theoretical_loss": 3.508722951644419,
      "tokens_seen": 1686962176
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004938615847542628,
      "loss": 1.6483,
      "theoretical_loss": 3.5087112981271185,
      "tokens_seen": 1687027712
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004938415245737211,
      "loss": 1.5548,
      "theoretical_loss": 3.5086996451892656,
      "tokens_seen": 1687093248
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004938214643931796,
      "loss": 1.3559,
      "theoretical_loss": 3.508687992830809,
      "tokens_seen": 1687158784
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000493801404212638,
      "loss": 1.5295,
      "theoretical_loss": 3.5086763410516966,
      "tokens_seen": 1687224320
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004937813440320963,
      "loss": 1.4473,
      "theoretical_loss": 3.5086646898518774,
      "tokens_seen": 1687289856
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004937612838515547,
      "loss": 1.6058,
      "theoretical_loss": 3.5086530392313007,
      "tokens_seen": 1687355392
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000493741223671013,
      "loss": 1.4609,
      "theoretical_loss": 3.5086413891899144,
      "tokens_seen": 1687420928
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004937211634904714,
      "loss": 1.4334,
      "theoretical_loss": 3.5086297397276676,
      "tokens_seen": 1687486464
    },
    {
      "epoch": 0.51,
      "objective/train/advantage_avg": 0.4926490783691406,
      "objective/train/docs_used": 616363,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4197611808776855,
      "objective/train/original_loss": 1.4197609424591064,
      "objective/train/theoretical_loss": 3.508618090844509,
      "objective/train/tokens_used": 1708012000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24316465854644775,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505011081695557,
      "objective/train/weighted_lm_loss": 1.4913710355758667,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.508618090844509,
      "tokens_seen": 1687552000
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004937011033099298,
      "loss": 1.4684,
      "theoretical_loss": 3.508618090844509,
      "tokens_seen": 1687552000
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004936810431293881,
      "loss": 1.4752,
      "theoretical_loss": 3.508606442540388,
      "tokens_seen": 1687617536
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004936609829488465,
      "loss": 1.5511,
      "theoretical_loss": 3.5085947948152514,
      "tokens_seen": 1687683072
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000493640922768305,
      "loss": 1.351,
      "theoretical_loss": 3.5085831476690497,
      "tokens_seen": 1687748608
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004936208625877633,
      "loss": 1.4351,
      "theoretical_loss": 3.508571501101731,
      "tokens_seen": 1687814144
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004936008024072217,
      "loss": 1.5154,
      "theoretical_loss": 3.508559855113244,
      "tokens_seen": 1687879680
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00049358074222668,
      "loss": 1.3973,
      "theoretical_loss": 3.5085482097035383,
      "tokens_seen": 1687945216
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004935606820461384,
      "loss": 1.4118,
      "theoretical_loss": 3.5085365648725615,
      "tokens_seen": 1688010752
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004935406218655968,
      "loss": 1.4428,
      "theoretical_loss": 3.5085249206202627,
      "tokens_seen": 1688076288
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004935205616850551,
      "loss": 1.4557,
      "theoretical_loss": 3.508513276946591,
      "tokens_seen": 1688141824
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004935005015045135,
      "loss": 1.4225,
      "theoretical_loss": 3.508501633851495,
      "tokens_seen": 1688207360
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004934804413239719,
      "loss": 1.3293,
      "theoretical_loss": 3.5084899913349235,
      "tokens_seen": 1688272896
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004934603811434303,
      "loss": 1.5194,
      "theoretical_loss": 3.5084783493968255,
      "tokens_seen": 1688338432
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004934403209628887,
      "loss": 1.3964,
      "theoretical_loss": 3.5084667080371497,
      "tokens_seen": 1688403968
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000493420260782347,
      "loss": 1.3251,
      "theoretical_loss": 3.5084550672558446,
      "tokens_seen": 1688469504
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004934002006018054,
      "loss": 1.3116,
      "theoretical_loss": 3.5084434270528595,
      "tokens_seen": 1688535040
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004933801404212638,
      "loss": 1.4179,
      "theoretical_loss": 3.5084317874281425,
      "tokens_seen": 1688600576
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004933600802407221,
      "loss": 1.5537,
      "theoretical_loss": 3.508420148381643,
      "tokens_seen": 1688666112
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004933400200601805,
      "loss": 1.4744,
      "theoretical_loss": 3.50840850991331,
      "tokens_seen": 1688731648
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004933199598796389,
      "loss": 1.305,
      "theoretical_loss": 3.5083968720230923,
      "tokens_seen": 1688797184
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004932998996990972,
      "loss": 1.4765,
      "theoretical_loss": 3.508385234710938,
      "tokens_seen": 1688862720
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004932798395185557,
      "loss": 1.2561,
      "theoretical_loss": 3.5083735979767967,
      "tokens_seen": 1688928256
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004932597793380141,
      "loss": 1.3955,
      "theoretical_loss": 3.5083619618206168,
      "tokens_seen": 1688993792
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004932397191574724,
      "loss": 1.4289,
      "theoretical_loss": 3.508350326242348,
      "tokens_seen": 1689059328
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004932196589769308,
      "loss": 1.4261,
      "theoretical_loss": 3.5083386912419385,
      "tokens_seen": 1689124864
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004931995987963893,
      "loss": 1.2548,
      "theoretical_loss": 3.5083270568193368,
      "tokens_seen": 1689190400
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004931795386158476,
      "loss": 1.3887,
      "theoretical_loss": 3.5083154229744924,
      "tokens_seen": 1689255936
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000493159478435306,
      "loss": 1.5305,
      "theoretical_loss": 3.5083037897073543,
      "tokens_seen": 1689321472
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004931394182547643,
      "loss": 1.4496,
      "theoretical_loss": 3.508292157017871,
      "tokens_seen": 1689387008
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004931193580742227,
      "loss": 1.6259,
      "theoretical_loss": 3.5082805249059916,
      "tokens_seen": 1689452544
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004930992978936811,
      "loss": 1.4803,
      "theoretical_loss": 3.5082688933716653,
      "tokens_seen": 1689518080
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004930792377131394,
      "loss": 1.2799,
      "theoretical_loss": 3.5082572624148405,
      "tokens_seen": 1689583616
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004930591775325978,
      "loss": 1.5568,
      "theoretical_loss": 3.5082456320354662,
      "tokens_seen": 1689649152
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004930391173520562,
      "loss": 1.5176,
      "theoretical_loss": 3.5082340022334915,
      "tokens_seen": 1689714688
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004930190571715146,
      "loss": 1.458,
      "theoretical_loss": 3.5082223730088655,
      "tokens_seen": 1689780224
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000492998996990973,
      "loss": 1.2974,
      "theoretical_loss": 3.508210744361537,
      "tokens_seen": 1689845760
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004929789368104313,
      "loss": 1.4916,
      "theoretical_loss": 3.508199116291455,
      "tokens_seen": 1689911296
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004929588766298897,
      "loss": 1.4742,
      "theoretical_loss": 3.5081874887985682,
      "tokens_seen": 1689976832
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004929388164493481,
      "loss": 1.5049,
      "theoretical_loss": 3.5081758618828256,
      "tokens_seen": 1690042368
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004929187562688064,
      "loss": 1.4785,
      "theoretical_loss": 3.5081642355441764,
      "tokens_seen": 1690107904
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004928986960882648,
      "loss": 1.3648,
      "theoretical_loss": 3.5081526097825697,
      "tokens_seen": 1690173440
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004928786359077232,
      "loss": 1.3769,
      "theoretical_loss": 3.5081409845979543,
      "tokens_seen": 1690238976
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004928585757271815,
      "loss": 1.4224,
      "theoretical_loss": 3.5081293599902788,
      "tokens_seen": 1690304512
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00049283851554664,
      "loss": 1.4509,
      "theoretical_loss": 3.508117735959493,
      "tokens_seen": 1690370048
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004928184553660984,
      "loss": 1.4101,
      "theoretical_loss": 3.5081061125055455,
      "tokens_seen": 1690435584
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004927983951855567,
      "loss": 1.4893,
      "theoretical_loss": 3.508094489628385,
      "tokens_seen": 1690501120
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004927783350050151,
      "loss": 1.4022,
      "theoretical_loss": 3.5080828673279614,
      "tokens_seen": 1690566656
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004927582748244734,
      "loss": 1.4552,
      "theoretical_loss": 3.508071245604223,
      "tokens_seen": 1690632192
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004927382146439318,
      "loss": 1.485,
      "theoretical_loss": 3.5080596244571183,
      "tokens_seen": 1690697728
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004927181544633902,
      "loss": 1.3271,
      "theoretical_loss": 3.5080480038865978,
      "tokens_seen": 1690763264
    },
    {
      "epoch": 0.51,
      "objective/train/advantage_avg": 0.49318400025367737,
      "objective/train/docs_used": 617433,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3830902576446533,
      "objective/train/original_loss": 1.3830904960632324,
      "objective/train/theoretical_loss": 3.5080363838926094,
      "objective/train/tokens_used": 1711288800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24402230978012085,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505589246749878,
      "objective/train/weighted_lm_loss": 1.4528931379318237,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5080363838926094,
      "tokens_seen": 1690828800
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004926980942828485,
      "loss": 1.4327,
      "theoretical_loss": 3.5080363838926094,
      "tokens_seen": 1690828800
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004926780341023069,
      "loss": 1.4641,
      "theoretical_loss": 3.5080247644751026,
      "tokens_seen": 1690894336
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004926579739217654,
      "loss": 1.397,
      "theoretical_loss": 3.5080131456340267,
      "tokens_seen": 1690959872
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004926379137412237,
      "loss": 1.3683,
      "theoretical_loss": 3.50800152736933,
      "tokens_seen": 1691025408
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004926178535606821,
      "loss": 1.4741,
      "theoretical_loss": 3.5079899096809624,
      "tokens_seen": 1691090944
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004925977933801404,
      "loss": 1.4814,
      "theoretical_loss": 3.507978292568872,
      "tokens_seen": 1691156480
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004925777331995988,
      "loss": 1.4436,
      "theoretical_loss": 3.507966676033009,
      "tokens_seen": 1691222016
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004925576730190572,
      "loss": 1.4216,
      "theoretical_loss": 3.507955060073322,
      "tokens_seen": 1691287552
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004925376128385155,
      "loss": 1.4068,
      "theoretical_loss": 3.5079434446897597,
      "tokens_seen": 1691353088
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004925175526579739,
      "loss": 1.468,
      "theoretical_loss": 3.507931829882272,
      "tokens_seen": 1691418624
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004924974924774323,
      "loss": 1.4336,
      "theoretical_loss": 3.5079202156508074,
      "tokens_seen": 1691484160
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004924774322968907,
      "loss": 1.4399,
      "theoretical_loss": 3.507908601995315,
      "tokens_seen": 1691549696
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004924573721163491,
      "loss": 1.4362,
      "theoretical_loss": 3.5078969889157445,
      "tokens_seen": 1691615232
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004924373119358075,
      "loss": 1.4094,
      "theoretical_loss": 3.507885376412044,
      "tokens_seen": 1691680768
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004924172517552658,
      "loss": 1.5391,
      "theoretical_loss": 3.507873764484164,
      "tokens_seen": 1691746304
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004923971915747242,
      "loss": 1.3219,
      "theoretical_loss": 3.5078621531320526,
      "tokens_seen": 1691811840
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004923771313941825,
      "loss": 1.4124,
      "theoretical_loss": 3.5078505423556594,
      "tokens_seen": 1691877376
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004923570712136409,
      "loss": 1.4751,
      "theoretical_loss": 3.5078389321549333,
      "tokens_seen": 1691942912
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004923370110330993,
      "loss": 1.4515,
      "theoretical_loss": 3.5078273225298235,
      "tokens_seen": 1692008448
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004923169508525576,
      "loss": 1.4261,
      "theoretical_loss": 3.5078157134802797,
      "tokens_seen": 1692073984
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004922968906720161,
      "loss": 1.5101,
      "theoretical_loss": 3.50780410500625,
      "tokens_seen": 1692139520
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004922768304914745,
      "loss": 1.5386,
      "theoretical_loss": 3.507792497107685,
      "tokens_seen": 1692205056
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004922567703109328,
      "loss": 1.4424,
      "theoretical_loss": 3.5077808897845326,
      "tokens_seen": 1692270592
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004922367101303912,
      "loss": 1.4269,
      "theoretical_loss": 3.5077692830367426,
      "tokens_seen": 1692336128
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004922166499498495,
      "loss": 1.4501,
      "theoretical_loss": 3.507757676864264,
      "tokens_seen": 1692401664
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004921965897693079,
      "loss": 1.4222,
      "theoretical_loss": 3.5077460712670465,
      "tokens_seen": 1692467200
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004921765295887663,
      "loss": 1.4066,
      "theoretical_loss": 3.5077344662450383,
      "tokens_seen": 1692532736
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004921564694082246,
      "loss": 1.3834,
      "theoretical_loss": 3.5077228617981895,
      "tokens_seen": 1692598272
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000492136409227683,
      "loss": 1.3279,
      "theoretical_loss": 3.5077112579264496,
      "tokens_seen": 1692663808
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004921163490471415,
      "loss": 1.4689,
      "theoretical_loss": 3.507699654629767,
      "tokens_seen": 1692729344
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004920962888665998,
      "loss": 1.4324,
      "theoretical_loss": 3.507688051908091,
      "tokens_seen": 1692794880
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004920762286860582,
      "loss": 1.4181,
      "theoretical_loss": 3.5076764497613713,
      "tokens_seen": 1692860416
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004920561685055166,
      "loss": 1.4127,
      "theoretical_loss": 3.507664848189557,
      "tokens_seen": 1692925952
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004920361083249749,
      "loss": 1.4484,
      "theoretical_loss": 3.5076532471925965,
      "tokens_seen": 1692991488
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004920160481444333,
      "loss": 1.5218,
      "theoretical_loss": 3.5076416467704408,
      "tokens_seen": 1693057024
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004919959879638916,
      "loss": 1.5343,
      "theoretical_loss": 3.5076300469230377,
      "tokens_seen": 1693122560
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00049197592778335,
      "loss": 1.4269,
      "theoretical_loss": 3.507618447650337,
      "tokens_seen": 1693188096
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004919558676028084,
      "loss": 1.544,
      "theoretical_loss": 3.507606848952289,
      "tokens_seen": 1693253632
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004919358074222668,
      "loss": 1.3975,
      "theoretical_loss": 3.507595250828841,
      "tokens_seen": 1693319168
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004919157472417252,
      "loss": 1.4624,
      "theoretical_loss": 3.5075836532799434,
      "tokens_seen": 1693384704
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004918956870611836,
      "loss": 1.3369,
      "theoretical_loss": 3.5075720563055457,
      "tokens_seen": 1693450240
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004918756268806419,
      "loss": 1.3583,
      "theoretical_loss": 3.507560459905597,
      "tokens_seen": 1693515776
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004918555667001003,
      "loss": 1.3065,
      "theoretical_loss": 3.5075488640800456,
      "tokens_seen": 1693581312
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004918355065195586,
      "loss": 1.4194,
      "theoretical_loss": 3.507537268828843,
      "tokens_seen": 1693646848
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000491815446339017,
      "loss": 1.4365,
      "theoretical_loss": 3.5075256741519363,
      "tokens_seen": 1693712384
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004917953861584754,
      "loss": 1.4635,
      "theoretical_loss": 3.507514080049276,
      "tokens_seen": 1693777920
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004917753259779337,
      "loss": 1.3263,
      "theoretical_loss": 3.5075024865208118,
      "tokens_seen": 1693843456
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004917552657973922,
      "loss": 1.4225,
      "theoretical_loss": 3.5074908935664917,
      "tokens_seen": 1693908992
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004917352056168506,
      "loss": 1.333,
      "theoretical_loss": 3.507479301186266,
      "tokens_seen": 1693974528
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004917151454363089,
      "loss": 1.4516,
      "theoretical_loss": 3.5074677093800846,
      "tokens_seen": 1694040064
    },
    {
      "epoch": 0.51,
      "objective/train/advantage_avg": 0.49125936627388,
      "objective/train/docs_used": 618718,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4581581354141235,
      "objective/train/original_loss": 1.4581581354141235,
      "objective/train/theoretical_loss": 3.507456118147896,
      "objective/train/tokens_used": 1714565600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24170486629009247,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503544807434082,
      "objective/train/weighted_lm_loss": 1.5313459634780884,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.507456118147896,
      "tokens_seen": 1694105600
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004916950852557673,
      "loss": 1.5339,
      "theoretical_loss": 3.507456118147896,
      "tokens_seen": 1694105600
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004916750250752258,
      "loss": 1.4774,
      "theoretical_loss": 3.5074445274896493,
      "tokens_seen": 1694171136
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004916549648946841,
      "loss": 1.4053,
      "theoretical_loss": 3.5074329374052944,
      "tokens_seen": 1694236672
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004916349047141425,
      "loss": 1.4528,
      "theoretical_loss": 3.507421347894781,
      "tokens_seen": 1694302208
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004916148445336008,
      "loss": 1.3356,
      "theoretical_loss": 3.5074097589580577,
      "tokens_seen": 1694367744
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004915947843530592,
      "loss": 1.3991,
      "theoretical_loss": 3.5073981705950743,
      "tokens_seen": 1694433280
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004915747241725176,
      "loss": 1.299,
      "theoretical_loss": 3.507386582805781,
      "tokens_seen": 1694498816
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004915546639919759,
      "loss": 1.3818,
      "theoretical_loss": 3.5073749955901254,
      "tokens_seen": 1694564352
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004915346038114343,
      "loss": 1.4881,
      "theoretical_loss": 3.5073634089480583,
      "tokens_seen": 1694629888
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004915145436308927,
      "loss": 1.453,
      "theoretical_loss": 3.507351822879529,
      "tokens_seen": 1694695424
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004914944834503511,
      "loss": 1.3331,
      "theoretical_loss": 3.5073402373844864,
      "tokens_seen": 1694760960
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004914744232698095,
      "loss": 1.3889,
      "theoretical_loss": 3.507328652462881,
      "tokens_seen": 1694826496
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004914543630892679,
      "loss": 1.4465,
      "theoretical_loss": 3.507317068114661,
      "tokens_seen": 1694892032
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004914343029087262,
      "loss": 1.4358,
      "theoretical_loss": 3.507305484339776,
      "tokens_seen": 1694957568
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004914142427281846,
      "loss": 1.4711,
      "theoretical_loss": 3.5072939011381763,
      "tokens_seen": 1695023104
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004913941825476429,
      "loss": 1.3763,
      "theoretical_loss": 3.5072823185098105,
      "tokens_seen": 1695088640
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004913741223671013,
      "loss": 1.4455,
      "theoretical_loss": 3.507270736454629,
      "tokens_seen": 1695154176
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004913540621865597,
      "loss": 1.37,
      "theoretical_loss": 3.50725915497258,
      "tokens_seen": 1695219712
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000491334002006018,
      "loss": 1.389,
      "theoretical_loss": 3.507247574063614,
      "tokens_seen": 1695285248
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004913139418254765,
      "loss": 1.4134,
      "theoretical_loss": 3.5072359937276802,
      "tokens_seen": 1695350784
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004912938816449349,
      "loss": 1.3137,
      "theoretical_loss": 3.507224413964728,
      "tokens_seen": 1695416320
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004912738214643932,
      "loss": 1.3861,
      "theoretical_loss": 3.507212834774707,
      "tokens_seen": 1695481856
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004912537612838516,
      "loss": 1.4006,
      "theoretical_loss": 3.5072012561575674,
      "tokens_seen": 1695547392
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00049123370110331,
      "loss": 1.2303,
      "theoretical_loss": 3.507189678113257,
      "tokens_seen": 1695612928
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004912136409227683,
      "loss": 1.4915,
      "theoretical_loss": 3.5071781006417266,
      "tokens_seen": 1695678464
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004911935807422267,
      "loss": 1.4992,
      "theoretical_loss": 3.507166523742926,
      "tokens_seen": 1695744000
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000491173520561685,
      "loss": 1.3899,
      "theoretical_loss": 3.5071549474168036,
      "tokens_seen": 1695809536
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004911534603811434,
      "loss": 1.3003,
      "theoretical_loss": 3.50714337166331,
      "tokens_seen": 1695875072
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004911334002006019,
      "loss": 1.5635,
      "theoretical_loss": 3.507131796482394,
      "tokens_seen": 1695940608
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004911133400200602,
      "loss": 1.3863,
      "theoretical_loss": 3.507120221874006,
      "tokens_seen": 1696006144
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004910932798395186,
      "loss": 1.5366,
      "theoretical_loss": 3.507108647838094,
      "tokens_seen": 1696071680
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000491073219658977,
      "loss": 1.3982,
      "theoretical_loss": 3.507097074374609,
      "tokens_seen": 1696137216
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004910531594784353,
      "loss": 1.3344,
      "theoretical_loss": 3.5070855014835,
      "tokens_seen": 1696202752
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004910330992978937,
      "loss": 1.4207,
      "theoretical_loss": 3.5070739291647173,
      "tokens_seen": 1696268288
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000491013039117352,
      "loss": 1.4176,
      "theoretical_loss": 3.507062357418209,
      "tokens_seen": 1696333824
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004909929789368104,
      "loss": 1.43,
      "theoretical_loss": 3.5070507862439264,
      "tokens_seen": 1696399360
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004909729187562688,
      "loss": 1.3976,
      "theoretical_loss": 3.507039215641818,
      "tokens_seen": 1696464896
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004909528585757272,
      "loss": 1.3627,
      "theoretical_loss": 3.507027645611834,
      "tokens_seen": 1696530432
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004909327983951856,
      "loss": 1.4453,
      "theoretical_loss": 3.5070160761539233,
      "tokens_seen": 1696595968
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000490912738214644,
      "loss": 1.4679,
      "theoretical_loss": 3.5070045072680367,
      "tokens_seen": 1696661504
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004908926780341023,
      "loss": 1.3961,
      "theoretical_loss": 3.506992938954122,
      "tokens_seen": 1696727040
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004908726178535607,
      "loss": 1.4629,
      "theoretical_loss": 3.5069813712121305,
      "tokens_seen": 1696792576
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000490852557673019,
      "loss": 1.3269,
      "theoretical_loss": 3.506969804042011,
      "tokens_seen": 1696858112
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004908324974924774,
      "loss": 1.3453,
      "theoretical_loss": 3.5069582374437136,
      "tokens_seen": 1696923648
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004908124373119358,
      "loss": 1.5245,
      "theoretical_loss": 3.5069466714171873,
      "tokens_seen": 1696989184
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004907923771313941,
      "loss": 1.4722,
      "theoretical_loss": 3.5069351059623823,
      "tokens_seen": 1697054720
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004907723169508526,
      "loss": 1.5122,
      "theoretical_loss": 3.5069235410792485,
      "tokens_seen": 1697120256
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000490752256770311,
      "loss": 1.3884,
      "theoretical_loss": 3.506911976767735,
      "tokens_seen": 1697185792
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004907321965897693,
      "loss": 1.3732,
      "theoretical_loss": 3.5069004130277914,
      "tokens_seen": 1697251328
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004907121364092277,
      "loss": 1.5141,
      "theoretical_loss": 3.5068888498593678,
      "tokens_seen": 1697316864
    },
    {
      "epoch": 0.51,
      "objective/train/advantage_avg": 0.49232470989227295,
      "objective/train/docs_used": 619912,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3043441772460938,
      "objective/train/original_loss": 1.3043444156646729,
      "objective/train/theoretical_loss": 3.506877287262414,
      "objective/train/tokens_used": 1717842400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24271300435066223,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504661798477173,
      "objective/train/weighted_lm_loss": 1.3700933456420898,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.506877287262414,
      "tokens_seen": 1697382400
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000490692076228686,
      "loss": 1.4618,
      "theoretical_loss": 3.506877287262414,
      "tokens_seen": 1697382400
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004906720160481444,
      "loss": 1.4557,
      "theoretical_loss": 3.5068657252368793,
      "tokens_seen": 1697447936
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004906519558676028,
      "loss": 1.316,
      "theoretical_loss": 3.5068541637827138,
      "tokens_seen": 1697513472
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004906318956870611,
      "loss": 1.4903,
      "theoretical_loss": 3.5068426028998667,
      "tokens_seen": 1697579008
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004906118355065195,
      "loss": 1.2932,
      "theoretical_loss": 3.5068310425882876,
      "tokens_seen": 1697644544
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000490591775325978,
      "loss": 1.4663,
      "theoretical_loss": 3.506819482847927,
      "tokens_seen": 1697710080
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004905717151454363,
      "loss": 1.4958,
      "theoretical_loss": 3.5068079236787346,
      "tokens_seen": 1697775616
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004905516549648947,
      "loss": 1.4743,
      "theoretical_loss": 3.5067963650806586,
      "tokens_seen": 1697841152
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004905315947843531,
      "loss": 1.3392,
      "theoretical_loss": 3.506784807053651,
      "tokens_seen": 1697906688
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004905115346038114,
      "loss": 1.4522,
      "theoretical_loss": 3.50677324959766,
      "tokens_seen": 1697972224
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004904914744232698,
      "loss": 1.4002,
      "theoretical_loss": 3.5067616927126357,
      "tokens_seen": 1698037760
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004904714142427281,
      "loss": 1.5004,
      "theoretical_loss": 3.5067501363985283,
      "tokens_seen": 1698103296
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004904513540621865,
      "loss": 1.4835,
      "theoretical_loss": 3.506738580655287,
      "tokens_seen": 1698168832
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004904312938816449,
      "loss": 1.2324,
      "theoretical_loss": 3.5067270254828613,
      "tokens_seen": 1698234368
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004904112337011033,
      "loss": 1.4047,
      "theoretical_loss": 3.506715470881202,
      "tokens_seen": 1698299904
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004903911735205617,
      "loss": 1.4392,
      "theoretical_loss": 3.506703916850258,
      "tokens_seen": 1698365440
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004903711133400201,
      "loss": 1.4841,
      "theoretical_loss": 3.5066923633899796,
      "tokens_seen": 1698430976
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004903510531594784,
      "loss": 1.4877,
      "theoretical_loss": 3.5066808105003164,
      "tokens_seen": 1698496512
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004903309929789368,
      "loss": 1.4065,
      "theoretical_loss": 3.506669258181218,
      "tokens_seen": 1698562048
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004903109327983952,
      "loss": 1.3602,
      "theoretical_loss": 3.5066577064326347,
      "tokens_seen": 1698627584
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004902908726178535,
      "loss": 1.3939,
      "theoretical_loss": 3.506646155254516,
      "tokens_seen": 1698693120
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004902708124373119,
      "loss": 1.4132,
      "theoretical_loss": 3.5066346046468118,
      "tokens_seen": 1698758656
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004902507522567703,
      "loss": 1.4426,
      "theoretical_loss": 3.506623054609472,
      "tokens_seen": 1698824192
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004902306920762287,
      "loss": 1.3798,
      "theoretical_loss": 3.5066115051424465,
      "tokens_seen": 1698889728
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004902106318956871,
      "loss": 1.4578,
      "theoretical_loss": 3.506599956245684,
      "tokens_seen": 1698955264
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004901905717151454,
      "loss": 1.3441,
      "theoretical_loss": 3.5065884079191365,
      "tokens_seen": 1699020800
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004901705115346038,
      "loss": 1.4552,
      "theoretical_loss": 3.5065768601627516,
      "tokens_seen": 1699086336
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004901504513540623,
      "loss": 1.3856,
      "theoretical_loss": 3.5065653129764813,
      "tokens_seen": 1699151872
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004901303911735206,
      "loss": 1.5297,
      "theoretical_loss": 3.5065537663602737,
      "tokens_seen": 1699217408
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.000490110330992979,
      "loss": 1.3942,
      "theoretical_loss": 3.5065422203140795,
      "tokens_seen": 1699282944
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004900902708124374,
      "loss": 1.3588,
      "theoretical_loss": 3.5065306748378484,
      "tokens_seen": 1699348480
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004900702106318957,
      "loss": 1.4334,
      "theoretical_loss": 3.5065191299315304,
      "tokens_seen": 1699414016
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0004900501504513541,
      "loss": 1.4365,
      "theoretical_loss": 3.506507585595075,
      "tokens_seen": 1699479552
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004900300902708124,
      "loss": 1.3826,
      "theoretical_loss": 3.506496041828433,
      "tokens_seen": 1699545088
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004900100300902708,
      "loss": 1.3907,
      "theoretical_loss": 3.5064844986315533,
      "tokens_seen": 1699610624
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004899899699097292,
      "loss": 1.3621,
      "theoretical_loss": 3.5064729560043864,
      "tokens_seen": 1699676160
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004899699097291876,
      "loss": 1.4966,
      "theoretical_loss": 3.506461413946882,
      "tokens_seen": 1699741696
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000489949849548646,
      "loss": 1.4258,
      "theoretical_loss": 3.5064498724589903,
      "tokens_seen": 1699807232
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004899297893681044,
      "loss": 1.3064,
      "theoretical_loss": 3.506438331540661,
      "tokens_seen": 1699872768
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004899097291875627,
      "loss": 1.3298,
      "theoretical_loss": 3.5064267911918434,
      "tokens_seen": 1699938304
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004898896690070211,
      "loss": 1.3332,
      "theoretical_loss": 3.5064152514124887,
      "tokens_seen": 1700003840
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004898696088264794,
      "loss": 1.4959,
      "theoretical_loss": 3.506403712202546,
      "tokens_seen": 1700069376
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004898495486459378,
      "loss": 1.4498,
      "theoretical_loss": 3.5063921735619656,
      "tokens_seen": 1700134912
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004898294884653962,
      "loss": 1.4023,
      "theoretical_loss": 3.5063806354906975,
      "tokens_seen": 1700200448
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004898094282848545,
      "loss": 1.4813,
      "theoretical_loss": 3.506369097988691,
      "tokens_seen": 1700265984
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000489789368104313,
      "loss": 1.4769,
      "theoretical_loss": 3.5063575610558964,
      "tokens_seen": 1700331520
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004897693079237714,
      "loss": 1.3912,
      "theoretical_loss": 3.506346024692265,
      "tokens_seen": 1700397056
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004897492477432297,
      "loss": 1.3306,
      "theoretical_loss": 3.5063344888977444,
      "tokens_seen": 1700462592
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004897291875626881,
      "loss": 1.4403,
      "theoretical_loss": 3.5063229536722864,
      "tokens_seen": 1700528128
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004897091273821465,
      "loss": 1.4293,
      "theoretical_loss": 3.506311419015841,
      "tokens_seen": 1700593664
    },
    {
      "epoch": 0.52,
      "objective/train/advantage_avg": 0.4946432411670685,
      "objective/train/docs_used": 621118,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.453629493713379,
      "objective/train/original_loss": 1.453629493713379,
      "objective/train/theoretical_loss": 3.5062998849283566,
      "objective/train/tokens_used": 1721119200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24500924348831177,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507099628448486,
      "objective/train/weighted_lm_loss": 1.5272786617279053,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5062998849283566,
      "tokens_seen": 1700659200
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004896890672016048,
      "loss": 1.39,
      "theoretical_loss": 3.5062998849283566,
      "tokens_seen": 1700659200
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004896690070210632,
      "loss": 1.4255,
      "theoretical_loss": 3.506288351409785,
      "tokens_seen": 1700724736
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004896489468405215,
      "loss": 1.4436,
      "theoretical_loss": 3.5062768184600754,
      "tokens_seen": 1700790272
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004896288866599799,
      "loss": 1.4219,
      "theoretical_loss": 3.506265286079177,
      "tokens_seen": 1700855808
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004896088264794384,
      "loss": 1.4146,
      "theoretical_loss": 3.5062537542670418,
      "tokens_seen": 1700921344
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004895887662988967,
      "loss": 1.3536,
      "theoretical_loss": 3.5062422230236185,
      "tokens_seen": 1700986880
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004895687061183551,
      "loss": 1.3681,
      "theoretical_loss": 3.5062306923488573,
      "tokens_seen": 1701052416
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004895486459378135,
      "loss": 1.365,
      "theoretical_loss": 3.5062191622427084,
      "tokens_seen": 1701117952
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004895285857572718,
      "loss": 1.5426,
      "theoretical_loss": 3.506207632705122,
      "tokens_seen": 1701183488
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004895085255767302,
      "loss": 1.5018,
      "theoretical_loss": 3.5061961037360474,
      "tokens_seen": 1701249024
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004894884653961885,
      "loss": 1.373,
      "theoretical_loss": 3.5061845753354355,
      "tokens_seen": 1701314560
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004894684052156469,
      "loss": 1.4028,
      "theoretical_loss": 3.5061730475032364,
      "tokens_seen": 1701380096
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004894483450351053,
      "loss": 1.4394,
      "theoretical_loss": 3.5061615202393996,
      "tokens_seen": 1701445632
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004894282848545637,
      "loss": 1.3924,
      "theoretical_loss": 3.506149993543876,
      "tokens_seen": 1701511168
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004894082246740221,
      "loss": 1.3303,
      "theoretical_loss": 3.5061384674166147,
      "tokens_seen": 1701576704
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004893881644934805,
      "loss": 1.3614,
      "theoretical_loss": 3.5061269418575662,
      "tokens_seen": 1701642240
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004893681043129388,
      "loss": 1.3414,
      "theoretical_loss": 3.506115416866681,
      "tokens_seen": 1701707776
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004893480441323972,
      "loss": 1.3569,
      "theoretical_loss": 3.506103892443909,
      "tokens_seen": 1701773312
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004893279839518556,
      "loss": 1.4378,
      "theoretical_loss": 3.5060923685892,
      "tokens_seen": 1701838848
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004893079237713139,
      "loss": 1.5044,
      "theoretical_loss": 3.5060808453025043,
      "tokens_seen": 1701904384
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004892878635907723,
      "loss": 1.419,
      "theoretical_loss": 3.5060693225837722,
      "tokens_seen": 1701969920
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004892678034102307,
      "loss": 1.4684,
      "theoretical_loss": 3.506057800432953,
      "tokens_seen": 1702035456
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004892477432296891,
      "loss": 1.518,
      "theoretical_loss": 3.5060462788499986,
      "tokens_seen": 1702100992
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004892276830491475,
      "loss": 1.4397,
      "theoretical_loss": 3.5060347578348576,
      "tokens_seen": 1702166528
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004892076228686058,
      "loss": 1.4371,
      "theoretical_loss": 3.5060232373874807,
      "tokens_seen": 1702232064
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004891875626880642,
      "loss": 1.3612,
      "theoretical_loss": 3.506011717507818,
      "tokens_seen": 1702297600
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004891675025075226,
      "loss": 1.426,
      "theoretical_loss": 3.50600019819582,
      "tokens_seen": 1702363136
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004891474423269809,
      "loss": 1.3757,
      "theoretical_loss": 3.5059886794514363,
      "tokens_seen": 1702428672
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004891273821464393,
      "loss": 1.3655,
      "theoretical_loss": 3.5059771612746173,
      "tokens_seen": 1702494208
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004891073219658976,
      "loss": 1.4512,
      "theoretical_loss": 3.5059656436653137,
      "tokens_seen": 1702559744
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004890872617853561,
      "loss": 1.4282,
      "theoretical_loss": 3.5059541266234744,
      "tokens_seen": 1702625280
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004890672016048145,
      "loss": 1.3953,
      "theoretical_loss": 3.5059426101490514,
      "tokens_seen": 1702690816
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004890471414242728,
      "loss": 1.4542,
      "theoretical_loss": 3.5059310942419932,
      "tokens_seen": 1702756352
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004890270812437312,
      "loss": 1.4311,
      "theoretical_loss": 3.505919578902251,
      "tokens_seen": 1702821888
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004890070210631896,
      "loss": 1.4321,
      "theoretical_loss": 3.505908064129775,
      "tokens_seen": 1702887424
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004889869608826479,
      "loss": 1.402,
      "theoretical_loss": 3.5058965499245147,
      "tokens_seen": 1702952960
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004889669007021063,
      "loss": 1.2992,
      "theoretical_loss": 3.505885036286421,
      "tokens_seen": 1703018496
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004889468405215647,
      "loss": 1.3889,
      "theoretical_loss": 3.505873523215444,
      "tokens_seen": 1703084032
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000488926780341023,
      "loss": 1.3503,
      "theoretical_loss": 3.5058620107115344,
      "tokens_seen": 1703149568
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004889067201604815,
      "loss": 1.4299,
      "theoretical_loss": 3.5058504987746413,
      "tokens_seen": 1703215104
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004888866599799398,
      "loss": 1.3284,
      "theoretical_loss": 3.505838987404716,
      "tokens_seen": 1703280640
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004888665997993982,
      "loss": 1.5219,
      "theoretical_loss": 3.5058274766017083,
      "tokens_seen": 1703346176
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004888465396188566,
      "loss": 1.3389,
      "theoretical_loss": 3.505815966365568,
      "tokens_seen": 1703411712
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004888264794383149,
      "loss": 1.4053,
      "theoretical_loss": 3.505804456696247,
      "tokens_seen": 1703477248
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004888064192577733,
      "loss": 1.5015,
      "theoretical_loss": 3.5057929475936938,
      "tokens_seen": 1703542784
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004887863590772317,
      "loss": 1.3507,
      "theoretical_loss": 3.5057814390578597,
      "tokens_seen": 1703608320
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048876629889669,
      "loss": 1.5345,
      "theoretical_loss": 3.5057699310886945,
      "tokens_seen": 1703673856
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004887462387161484,
      "loss": 1.4318,
      "theoretical_loss": 3.5057584236861485,
      "tokens_seen": 1703739392
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004887261785356069,
      "loss": 1.2807,
      "theoretical_loss": 3.505746916850173,
      "tokens_seen": 1703804928
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004887061183550652,
      "loss": 1.5655,
      "theoretical_loss": 3.5057354105807166,
      "tokens_seen": 1703870464
    },
    {
      "epoch": 0.52,
      "objective/train/advantage_avg": 0.4942810535430908,
      "objective/train/docs_used": 622235,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3802403211593628,
      "objective/train/original_loss": 1.3802404403686523,
      "objective/train/theoretical_loss": 3.505723904877731,
      "objective/train/tokens_used": 1724396000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24489769339561462,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506731271743774,
      "objective/train/weighted_lm_loss": 1.450076699256897,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.505723904877731,
      "tokens_seen": 1703936000
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004886860581745236,
      "loss": 1.4294,
      "theoretical_loss": 3.505723904877731,
      "tokens_seen": 1703936000
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004886659979939819,
      "loss": 1.3292,
      "theoretical_loss": 3.5057123997411663,
      "tokens_seen": 1704001536
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004886459378134403,
      "loss": 1.416,
      "theoretical_loss": 3.505700895170972,
      "tokens_seen": 1704067072
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004886258776328988,
      "loss": 1.3413,
      "theoretical_loss": 3.5056893911670994,
      "tokens_seen": 1704132608
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004886058174523571,
      "loss": 1.4447,
      "theoretical_loss": 3.5056778877294983,
      "tokens_seen": 1704198144
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004885857572718155,
      "loss": 1.4156,
      "theoretical_loss": 3.5056663848581193,
      "tokens_seen": 1704263680
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004885656970912739,
      "loss": 1.4122,
      "theoretical_loss": 3.5056548825529132,
      "tokens_seen": 1704329216
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004885456369107322,
      "loss": 1.4041,
      "theoretical_loss": 3.5056433808138294,
      "tokens_seen": 1704394752
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004885255767301906,
      "loss": 1.3962,
      "theoretical_loss": 3.5056318796408186,
      "tokens_seen": 1704460288
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000488505516549649,
      "loss": 1.4747,
      "theoretical_loss": 3.5056203790338314,
      "tokens_seen": 1704525824
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004884854563691073,
      "loss": 1.4171,
      "theoretical_loss": 3.505608878992818,
      "tokens_seen": 1704591360
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004884653961885657,
      "loss": 1.3931,
      "theoretical_loss": 3.505597379517729,
      "tokens_seen": 1704656896
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004884453360080241,
      "loss": 1.3696,
      "theoretical_loss": 3.505585880608515,
      "tokens_seen": 1704722432
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004884252758274825,
      "loss": 1.3775,
      "theoretical_loss": 3.5055743822651255,
      "tokens_seen": 1704787968
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004884052156469409,
      "loss": 1.4178,
      "theoretical_loss": 3.5055628844875115,
      "tokens_seen": 1704853504
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004883851554663992,
      "loss": 1.4725,
      "theoretical_loss": 3.5055513872756237,
      "tokens_seen": 1704919040
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004883650952858576,
      "loss": 1.3828,
      "theoretical_loss": 3.505539890629412,
      "tokens_seen": 1704984576
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000488345035105316,
      "loss": 1.4164,
      "theoretical_loss": 3.5055283945488274,
      "tokens_seen": 1705050112
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004883249749247743,
      "loss": 1.3616,
      "theoretical_loss": 3.5055168990338195,
      "tokens_seen": 1705115648
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004883049147442327,
      "loss": 1.401,
      "theoretical_loss": 3.5055054040843396,
      "tokens_seen": 1705181184
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004882848545636911,
      "loss": 1.5071,
      "theoretical_loss": 3.5054939097003377,
      "tokens_seen": 1705246720
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048826479438314946,
      "loss": 1.3424,
      "theoretical_loss": 3.505482415881764,
      "tokens_seen": 1705312256
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004882447342026078,
      "loss": 1.5249,
      "theoretical_loss": 3.5054709226285694,
      "tokens_seen": 1705377792
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048822467402206624,
      "loss": 1.4228,
      "theoretical_loss": 3.505459429940704,
      "tokens_seen": 1705443328
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004882046138415246,
      "loss": 1.4572,
      "theoretical_loss": 3.505447937818119,
      "tokens_seen": 1705508864
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048818455366098296,
      "loss": 1.5203,
      "theoretical_loss": 3.505436446260764,
      "tokens_seen": 1705574400
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004881644934804413,
      "loss": 1.5016,
      "theoretical_loss": 3.50542495526859,
      "tokens_seen": 1705639936
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048814443329989974,
      "loss": 1.3873,
      "theoretical_loss": 3.5054134648415474,
      "tokens_seen": 1705705472
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004881243731193581,
      "loss": 1.2809,
      "theoretical_loss": 3.5054019749795864,
      "tokens_seen": 1705771008
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048810431293881646,
      "loss": 1.4227,
      "theoretical_loss": 3.505390485682658,
      "tokens_seen": 1705836544
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004880842527582748,
      "loss": 1.6215,
      "theoretical_loss": 3.5053789969507125,
      "tokens_seen": 1705902080
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004880641925777332,
      "loss": 1.4926,
      "theoretical_loss": 3.5053675087836997,
      "tokens_seen": 1705967616
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004880441323971916,
      "loss": 1.3395,
      "theoretical_loss": 3.5053560211815715,
      "tokens_seen": 1706033152
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048802407221664997,
      "loss": 1.4123,
      "theoretical_loss": 3.5053445341442773,
      "tokens_seen": 1706098688
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048800401203610833,
      "loss": 1.4975,
      "theoretical_loss": 3.505333047671768,
      "tokens_seen": 1706164224
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004879839518555667,
      "loss": 1.4271,
      "theoretical_loss": 3.505321561763995,
      "tokens_seen": 1706229760
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004879638916750251,
      "loss": 1.3179,
      "theoretical_loss": 3.505310076420907,
      "tokens_seen": 1706295296
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048794383149448347,
      "loss": 1.3807,
      "theoretical_loss": 3.505298591642456,
      "tokens_seen": 1706360832
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048792377131394183,
      "loss": 1.4486,
      "theoretical_loss": 3.5052871074285923,
      "tokens_seen": 1706426368
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004879037111334002,
      "loss": 1.388,
      "theoretical_loss": 3.5052756237792657,
      "tokens_seen": 1706491904
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048788365095285856,
      "loss": 1.3577,
      "theoretical_loss": 3.505264140694428,
      "tokens_seen": 1706557440
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048786359077231697,
      "loss": 1.4898,
      "theoretical_loss": 3.5052526581740286,
      "tokens_seen": 1706622976
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048784353059177533,
      "loss": 1.4936,
      "theoretical_loss": 3.505241176218019,
      "tokens_seen": 1706688512
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004878234704112337,
      "loss": 1.3603,
      "theoretical_loss": 3.5052296948263493,
      "tokens_seen": 1706754048
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048780341023069206,
      "loss": 1.4123,
      "theoretical_loss": 3.5052182139989707,
      "tokens_seen": 1706819584
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004877833500501505,
      "loss": 1.4336,
      "theoretical_loss": 3.5052067337358324,
      "tokens_seen": 1706885120
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048776328986960884,
      "loss": 1.4822,
      "theoretical_loss": 3.5051952540368863,
      "tokens_seen": 1706950656
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004877432296890672,
      "loss": 1.4049,
      "theoretical_loss": 3.5051837749020827,
      "tokens_seen": 1707016192
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048772316950852556,
      "loss": 1.4582,
      "theoretical_loss": 3.5051722963313723,
      "tokens_seen": 1707081728
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004877031093279839,
      "loss": 1.355,
      "theoretical_loss": 3.5051608183247054,
      "tokens_seen": 1707147264
    },
    {
      "epoch": 0.52,
      "objective/train/advantage_avg": 0.4932672679424286,
      "objective/train/docs_used": 623332,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.500571846961975,
      "objective/train/original_loss": 1.5005717277526855,
      "objective/train/theoretical_loss": 3.5051493408820327,
      "objective/train/tokens_used": 1727672800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24388380348682404,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050566554069519,
      "objective/train/weighted_lm_loss": 1.5763435363769531,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5051493408820327,
      "tokens_seen": 1707212800
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048768304914744234,
      "loss": 1.5614,
      "theoretical_loss": 3.5051493408820327,
      "tokens_seen": 1707212800
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004876629889669007,
      "loss": 1.3185,
      "theoretical_loss": 3.505137864003305,
      "tokens_seen": 1707278336
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048764292878635907,
      "loss": 1.3366,
      "theoretical_loss": 3.505126387688473,
      "tokens_seen": 1707343872
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048762286860581743,
      "loss": 1.3382,
      "theoretical_loss": 3.505114911937487,
      "tokens_seen": 1707409408
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048760280842527584,
      "loss": 1.3546,
      "theoretical_loss": 3.5051034367502982,
      "tokens_seen": 1707474944
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004875827482447342,
      "loss": 1.3432,
      "theoretical_loss": 3.5050919621268566,
      "tokens_seen": 1707540480
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048756268806419257,
      "loss": 1.3874,
      "theoretical_loss": 3.5050804880671134,
      "tokens_seen": 1707606016
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048754262788365093,
      "loss": 1.3456,
      "theoretical_loss": 3.5050690145710193,
      "tokens_seen": 1707671552
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048752256770310935,
      "loss": 1.4659,
      "theoretical_loss": 3.5050575416385246,
      "tokens_seen": 1707737088
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004875025075225677,
      "loss": 1.3262,
      "theoretical_loss": 3.5050460692695804,
      "tokens_seen": 1707802624
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048748244734202607,
      "loss": 1.4753,
      "theoretical_loss": 3.505034597464137,
      "tokens_seen": 1707868160
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048746238716148443,
      "loss": 1.4352,
      "theoretical_loss": 3.505023126222145,
      "tokens_seen": 1707933696
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004874423269809428,
      "loss": 1.5062,
      "theoretical_loss": 3.5050116555435555,
      "tokens_seen": 1707999232
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004874222668004012,
      "loss": 1.4061,
      "theoretical_loss": 3.505000185428319,
      "tokens_seen": 1708064768
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004874022066198596,
      "loss": 1.3722,
      "theoretical_loss": 3.5049887158763866,
      "tokens_seen": 1708130304
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048738214643931794,
      "loss": 1.2463,
      "theoretical_loss": 3.5049772468877087,
      "tokens_seen": 1708195840
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004873620862587763,
      "loss": 1.4205,
      "theoretical_loss": 3.504965778462236,
      "tokens_seen": 1708261376
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004873420260782347,
      "loss": 1.3774,
      "theoretical_loss": 3.504954310599919,
      "tokens_seen": 1708326912
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004873219658976931,
      "loss": 1.3951,
      "theoretical_loss": 3.504942843300709,
      "tokens_seen": 1708392448
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048730190571715144,
      "loss": 1.3951,
      "theoretical_loss": 3.5049313765645564,
      "tokens_seen": 1708457984
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004872818455366098,
      "loss": 1.4191,
      "theoretical_loss": 3.504919910391412,
      "tokens_seen": 1708523520
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048726178535606816,
      "loss": 1.4368,
      "theoretical_loss": 3.504908444781227,
      "tokens_seen": 1708589056
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004872417251755266,
      "loss": 1.4083,
      "theoretical_loss": 3.5048969797339513,
      "tokens_seen": 1708654592
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000487221664994985,
      "loss": 1.37,
      "theoretical_loss": 3.5048855152495366,
      "tokens_seen": 1708720128
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048720160481444336,
      "loss": 1.3973,
      "theoretical_loss": 3.504874051327933,
      "tokens_seen": 1708785664
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004871815446339017,
      "loss": 1.58,
      "theoretical_loss": 3.5048625879690913,
      "tokens_seen": 1708851200
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048716148445336014,
      "loss": 1.4675,
      "theoretical_loss": 3.5048511251729626,
      "tokens_seen": 1708916736
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004871414242728185,
      "loss": 1.4341,
      "theoretical_loss": 3.5048396629394976,
      "tokens_seen": 1708982272
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048712136409227686,
      "loss": 1.3223,
      "theoretical_loss": 3.504828201268648,
      "tokens_seen": 1709047808
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004871013039117352,
      "loss": 1.4124,
      "theoretical_loss": 3.5048167401603623,
      "tokens_seen": 1709113344
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004870812437311936,
      "loss": 1.4531,
      "theoretical_loss": 3.504805279614594,
      "tokens_seen": 1709178880
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000487061183550652,
      "loss": 1.4842,
      "theoretical_loss": 3.504793819631292,
      "tokens_seen": 1709244416
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048704112337011037,
      "loss": 1.3766,
      "theoretical_loss": 3.504782360210408,
      "tokens_seen": 1709309952
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048702106318956873,
      "loss": 1.3676,
      "theoretical_loss": 3.5047709013518924,
      "tokens_seen": 1709375488
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004870010030090271,
      "loss": 1.6068,
      "theoretical_loss": 3.504759443055696,
      "tokens_seen": 1709441024
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004869809428284855,
      "loss": 1.4562,
      "theoretical_loss": 3.5047479853217705,
      "tokens_seen": 1709506560
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048696088264794387,
      "loss": 1.4169,
      "theoretical_loss": 3.504736528150066,
      "tokens_seen": 1709572096
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048694082246740223,
      "loss": 1.4276,
      "theoretical_loss": 3.5047250715405336,
      "tokens_seen": 1709637632
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004869207622868606,
      "loss": 1.3311,
      "theoretical_loss": 3.5047136154931238,
      "tokens_seen": 1709703168
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048690070210631896,
      "loss": 1.526,
      "theoretical_loss": 3.5047021600077883,
      "tokens_seen": 1709768704
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048688064192577737,
      "loss": 1.4849,
      "theoretical_loss": 3.504690705084477,
      "tokens_seen": 1709834240
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048686058174523573,
      "loss": 1.4028,
      "theoretical_loss": 3.5046792507231417,
      "tokens_seen": 1709899776
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004868405215646941,
      "loss": 1.3573,
      "theoretical_loss": 3.5046677969237328,
      "tokens_seen": 1709965312
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048682046138415246,
      "loss": 1.5071,
      "theoretical_loss": 3.504656343686201,
      "tokens_seen": 1710030848
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004868004012036109,
      "loss": 1.3577,
      "theoretical_loss": 3.5046448910104973,
      "tokens_seen": 1710096384
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048678034102306924,
      "loss": 1.4195,
      "theoretical_loss": 3.504633438896573,
      "tokens_seen": 1710161920
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004867602808425276,
      "loss": 1.3292,
      "theoretical_loss": 3.504621987344379,
      "tokens_seen": 1710227456
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048674022066198596,
      "loss": 1.4729,
      "theoretical_loss": 3.5046105363538658,
      "tokens_seen": 1710292992
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004867201604814443,
      "loss": 1.4243,
      "theoretical_loss": 3.5045990859249843,
      "tokens_seen": 1710358528
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048670010030090274,
      "loss": 1.4104,
      "theoretical_loss": 3.5045876360576855,
      "tokens_seen": 1710424064
    },
    {
      "epoch": 0.52,
      "objective/train/advantage_avg": 0.49364978075027466,
      "objective/train/docs_used": 624643,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4175605773925781,
      "objective/train/original_loss": 1.4175604581832886,
      "objective/train/theoretical_loss": 3.504576186751921,
      "objective/train/tokens_used": 1730949600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24400974810123444,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506054162979126,
      "objective/train/weighted_lm_loss": 1.4892531633377075,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.504576186751921,
      "tokens_seen": 1710489600
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004866800401203611,
      "loss": 1.4315,
      "theoretical_loss": 3.504576186751921,
      "tokens_seen": 1710489600
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048665997993981947,
      "loss": 1.4106,
      "theoretical_loss": 3.504564738007641,
      "tokens_seen": 1710555136
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048663991975927783,
      "loss": 1.4827,
      "theoretical_loss": 3.504553289824797,
      "tokens_seen": 1710620672
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048661985957873624,
      "loss": 1.3518,
      "theoretical_loss": 3.5045418422033396,
      "tokens_seen": 1710686208
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004865997993981946,
      "loss": 1.5339,
      "theoretical_loss": 3.5045303951432194,
      "tokens_seen": 1710751744
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048657973921765297,
      "loss": 1.3457,
      "theoretical_loss": 3.5045189486443884,
      "tokens_seen": 1710817280
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048655967903711133,
      "loss": 1.5314,
      "theoretical_loss": 3.5045075027067965,
      "tokens_seen": 1710882816
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048653961885656975,
      "loss": 1.257,
      "theoretical_loss": 3.5044960573303956,
      "tokens_seen": 1710948352
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004865195586760281,
      "loss": 1.5014,
      "theoretical_loss": 3.504484612515136,
      "tokens_seen": 1711013888
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048649949849548647,
      "loss": 1.4793,
      "theoretical_loss": 3.5044731682609687,
      "tokens_seen": 1711079424
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048647943831494483,
      "loss": 1.4499,
      "theoretical_loss": 3.5044617245678458,
      "tokens_seen": 1711144960
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004864593781344032,
      "loss": 1.4128,
      "theoretical_loss": 3.504450281435717,
      "tokens_seen": 1711210496
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004864393179538616,
      "loss": 1.4637,
      "theoretical_loss": 3.5044388388645333,
      "tokens_seen": 1711276032
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048641925777332,
      "loss": 1.4763,
      "theoretical_loss": 3.5044273968542465,
      "tokens_seen": 1711341568
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048639919759277834,
      "loss": 1.3419,
      "theoretical_loss": 3.5044159554048075,
      "tokens_seen": 1711407104
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004863791374122367,
      "loss": 1.3576,
      "theoretical_loss": 3.504404514516167,
      "tokens_seen": 1711472640
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004863590772316951,
      "loss": 1.2269,
      "theoretical_loss": 3.5043930741882763,
      "tokens_seen": 1711538176
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004863390170511535,
      "loss": 1.3803,
      "theoretical_loss": 3.5043816344210863,
      "tokens_seen": 1711603712
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048631895687061184,
      "loss": 1.4105,
      "theoretical_loss": 3.5043701952145483,
      "tokens_seen": 1711669248
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004862988966900702,
      "loss": 1.5428,
      "theoretical_loss": 3.504358756568613,
      "tokens_seen": 1711734784
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048627883650952857,
      "loss": 1.4075,
      "theoretical_loss": 3.504347318483232,
      "tokens_seen": 1711800320
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000486258776328987,
      "loss": 1.5528,
      "theoretical_loss": 3.5043358809583554,
      "tokens_seen": 1711865856
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048623871614844534,
      "loss": 1.37,
      "theoretical_loss": 3.504324443993935,
      "tokens_seen": 1711931392
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004862186559679037,
      "loss": 1.4789,
      "theoretical_loss": 3.504313007589922,
      "tokens_seen": 1711996928
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048619859578736207,
      "loss": 1.4827,
      "theoretical_loss": 3.504301571746267,
      "tokens_seen": 1712062464
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004861785356068205,
      "loss": 1.5571,
      "theoretical_loss": 3.5042901364629215,
      "tokens_seen": 1712128000
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048615847542627885,
      "loss": 1.3928,
      "theoretical_loss": 3.5042787017398362,
      "tokens_seen": 1712193536
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004861384152457372,
      "loss": 1.3945,
      "theoretical_loss": 3.504267267576963,
      "tokens_seen": 1712259072
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048611835506519557,
      "loss": 1.3997,
      "theoretical_loss": 3.504255833974252,
      "tokens_seen": 1712324608
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048609829488465393,
      "loss": 1.3041,
      "theoretical_loss": 3.504244400931655,
      "tokens_seen": 1712390144
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048607823470411235,
      "loss": 1.3711,
      "theoretical_loss": 3.5042329684491227,
      "tokens_seen": 1712455680
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004860581745235707,
      "loss": 1.4512,
      "theoretical_loss": 3.5042215365266065,
      "tokens_seen": 1712521216
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004860381143430291,
      "loss": 1.5281,
      "theoretical_loss": 3.5042101051640575,
      "tokens_seen": 1712586752
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048601805416248744,
      "loss": 1.5158,
      "theoretical_loss": 3.504198674361427,
      "tokens_seen": 1712652288
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048599799398194585,
      "loss": 1.4688,
      "theoretical_loss": 3.5041872441186657,
      "tokens_seen": 1712717824
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004859779338014042,
      "loss": 1.4887,
      "theoretical_loss": 3.504175814435725,
      "tokens_seen": 1712783360
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004859578736208626,
      "loss": 1.3968,
      "theoretical_loss": 3.5041643853125564,
      "tokens_seen": 1712848896
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048593781344032094,
      "loss": 1.4778,
      "theoretical_loss": 3.504152956749111,
      "tokens_seen": 1712914432
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004859177532597793,
      "loss": 1.4848,
      "theoretical_loss": 3.5041415287453397,
      "tokens_seen": 1712979968
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004858976930792377,
      "loss": 1.3748,
      "theoretical_loss": 3.504130101301193,
      "tokens_seen": 1713045504
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004858776328986961,
      "loss": 1.4239,
      "theoretical_loss": 3.504118674416623,
      "tokens_seen": 1713111040
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048585757271815444,
      "loss": 1.4365,
      "theoretical_loss": 3.5041072480915814,
      "tokens_seen": 1713176576
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004858375125376128,
      "loss": 1.421,
      "theoretical_loss": 3.504095822326018,
      "tokens_seen": 1713242112
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004858174523570712,
      "loss": 1.5395,
      "theoretical_loss": 3.504084397119885,
      "tokens_seen": 1713307648
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004857973921765296,
      "loss": 1.4426,
      "theoretical_loss": 3.504072972473133,
      "tokens_seen": 1713373184
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048577733199598795,
      "loss": 1.4631,
      "theoretical_loss": 3.504061548385714,
      "tokens_seen": 1713438720
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004857572718154463,
      "loss": 1.4025,
      "theoretical_loss": 3.504050124857579,
      "tokens_seen": 1713504256
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048573721163490467,
      "loss": 1.3516,
      "theoretical_loss": 3.5040387018886783,
      "tokens_seen": 1713569792
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048571715145436314,
      "loss": 1.3409,
      "theoretical_loss": 3.504027279478964,
      "tokens_seen": 1713635328
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004856970912738215,
      "loss": 1.5761,
      "theoretical_loss": 3.504015857628387,
      "tokens_seen": 1713700864
    },
    {
      "epoch": 0.52,
      "objective/train/advantage_avg": 0.4936310946941376,
      "objective/train/docs_used": 625897,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2754610776901245,
      "objective/train/original_loss": 1.275461196899414,
      "objective/train/theoretical_loss": 3.504004436336899,
      "objective/train/tokens_used": 1734226400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24397696554660797,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506035089492798,
      "objective/train/weighted_lm_loss": 1.3400050401687622,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.504004436336899,
      "tokens_seen": 1713766400
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048567703109327987,
      "loss": 1.3451,
      "theoretical_loss": 3.504004436336899,
      "tokens_seen": 1713766400
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048565697091273823,
      "loss": 1.3896,
      "theoretical_loss": 3.5039930156044514,
      "tokens_seen": 1713831936
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048563691073219664,
      "loss": 1.4191,
      "theoretical_loss": 3.5039815954309943,
      "tokens_seen": 1713897472
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000485616850551655,
      "loss": 1.4845,
      "theoretical_loss": 3.5039701758164803,
      "tokens_seen": 1713963008
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048559679037111337,
      "loss": 1.3024,
      "theoretical_loss": 3.503958756760859,
      "tokens_seen": 1714028544
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048557673019057173,
      "loss": 1.3507,
      "theoretical_loss": 3.503947338264084,
      "tokens_seen": 1714094080
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048555667001003015,
      "loss": 1.4979,
      "theoretical_loss": 3.5039359203261045,
      "tokens_seen": 1714159616
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004855366098294885,
      "loss": 1.3895,
      "theoretical_loss": 3.503924502946873,
      "tokens_seen": 1714225152
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048551654964894687,
      "loss": 1.3501,
      "theoretical_loss": 3.5039130861263406,
      "tokens_seen": 1714290688
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048549648946840523,
      "loss": 1.4702,
      "theoretical_loss": 3.503901669864458,
      "tokens_seen": 1714356224
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004854764292878636,
      "loss": 1.4024,
      "theoretical_loss": 3.5038902541611776,
      "tokens_seen": 1714421760
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000485456369107322,
      "loss": 1.4523,
      "theoretical_loss": 3.5038788390164495,
      "tokens_seen": 1714487296
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004854363089267804,
      "loss": 1.4859,
      "theoretical_loss": 3.5038674244302257,
      "tokens_seen": 1714552832
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048541624874623874,
      "loss": 1.4077,
      "theoretical_loss": 3.5038560104024574,
      "tokens_seen": 1714618368
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004853961885656971,
      "loss": 1.3379,
      "theoretical_loss": 3.503844596933096,
      "tokens_seen": 1714683904
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004853761283851555,
      "loss": 1.448,
      "theoretical_loss": 3.503833184022093,
      "tokens_seen": 1714749440
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004853560682046139,
      "loss": 1.5437,
      "theoretical_loss": 3.503821771669399,
      "tokens_seen": 1714814976
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048533600802407224,
      "loss": 1.5024,
      "theoretical_loss": 3.5038103598749664,
      "tokens_seen": 1714880512
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004853159478435306,
      "loss": 1.3678,
      "theoretical_loss": 3.503798948638746,
      "tokens_seen": 1714946048
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048529588766298897,
      "loss": 1.4216,
      "theoretical_loss": 3.5037875379606884,
      "tokens_seen": 1715011584
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004852758274824474,
      "loss": 1.4662,
      "theoretical_loss": 3.5037761278407467,
      "tokens_seen": 1715077120
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048525576730190574,
      "loss": 1.4747,
      "theoretical_loss": 3.503764718278871,
      "tokens_seen": 1715142656
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004852357071213641,
      "loss": 1.5574,
      "theoretical_loss": 3.503753309275013,
      "tokens_seen": 1715208192
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048521564694082247,
      "loss": 1.4552,
      "theoretical_loss": 3.503741900829124,
      "tokens_seen": 1715273728
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004851955867602809,
      "loss": 1.6023,
      "theoretical_loss": 3.503730492941156,
      "tokens_seen": 1715339264
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048517552657973925,
      "loss": 1.4359,
      "theoretical_loss": 3.5037190856110594,
      "tokens_seen": 1715404800
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004851554663991976,
      "loss": 1.4763,
      "theoretical_loss": 3.5037076788387864,
      "tokens_seen": 1715470336
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048513540621865597,
      "loss": 1.4489,
      "theoretical_loss": 3.503696272624288,
      "tokens_seen": 1715535872
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048511534603811433,
      "loss": 1.3843,
      "theoretical_loss": 3.503684866967516,
      "tokens_seen": 1715601408
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048509528585757275,
      "loss": 1.4325,
      "theoretical_loss": 3.503673461868421,
      "tokens_seen": 1715666944
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004850752256770311,
      "loss": 1.4423,
      "theoretical_loss": 3.503662057326956,
      "tokens_seen": 1715732480
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004850551654964895,
      "loss": 1.4286,
      "theoretical_loss": 3.5036506533430707,
      "tokens_seen": 1715798016
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048503510531594784,
      "loss": 1.4248,
      "theoretical_loss": 3.5036392499167173,
      "tokens_seen": 1715863552
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048501504513540625,
      "loss": 1.413,
      "theoretical_loss": 3.5036278470478477,
      "tokens_seen": 1715929088
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004849949849548646,
      "loss": 1.436,
      "theoretical_loss": 3.503616444736412,
      "tokens_seen": 1715994624
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000484974924774323,
      "loss": 1.4549,
      "theoretical_loss": 3.5036050429823637,
      "tokens_seen": 1716060160
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048495486459378134,
      "loss": 1.3563,
      "theoretical_loss": 3.5035936417856526,
      "tokens_seen": 1716125696
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004849348044132397,
      "loss": 1.4032,
      "theoretical_loss": 3.5035822411462307,
      "tokens_seen": 1716191232
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004849147442326981,
      "loss": 1.3675,
      "theoretical_loss": 3.5035708410640494,
      "tokens_seen": 1716256768
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004848946840521565,
      "loss": 1.4075,
      "theoretical_loss": 3.5035594415390605,
      "tokens_seen": 1716322304
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048487462387161484,
      "loss": 1.4314,
      "theoretical_loss": 3.503548042571215,
      "tokens_seen": 1716387840
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004848545636910732,
      "loss": 1.3926,
      "theoretical_loss": 3.503536644160465,
      "tokens_seen": 1716453376
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004848345035105316,
      "loss": 1.4602,
      "theoretical_loss": 3.5035252463067614,
      "tokens_seen": 1716518912
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048481444332999,
      "loss": 1.4434,
      "theoretical_loss": 3.503513849010056,
      "tokens_seen": 1716584448
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048479438314944835,
      "loss": 1.4578,
      "theoretical_loss": 3.5035024522703004,
      "tokens_seen": 1716649984
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004847743229689067,
      "loss": 1.3524,
      "theoretical_loss": 3.5034910560874457,
      "tokens_seen": 1716715520
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048475426278836507,
      "loss": 1.4388,
      "theoretical_loss": 3.503479660461444,
      "tokens_seen": 1716781056
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004847342026078235,
      "loss": 1.4494,
      "theoretical_loss": 3.5034682653922466,
      "tokens_seen": 1716846592
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048471414242728185,
      "loss": 1.3193,
      "theoretical_loss": 3.5034568708798046,
      "tokens_seen": 1716912128
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004846940822467402,
      "loss": 1.3691,
      "theoretical_loss": 3.5034454769240706,
      "tokens_seen": 1716977664
    },
    {
      "epoch": 0.52,
      "objective/train/advantage_avg": 0.4912497103214264,
      "objective/train/docs_used": 627098,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.333863615989685,
      "objective/train/original_loss": 1.3338637351989746,
      "objective/train/theoretical_loss": 3.5034340835249953,
      "objective/train/tokens_used": 1737503200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24193502962589264,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503548383712769,
      "objective/train/weighted_lm_loss": 1.4010065793991089,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5034340835249953,
      "tokens_seen": 1717043200
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004846740220661986,
      "loss": 1.3441,
      "theoretical_loss": 3.5034340835249953,
      "tokens_seen": 1717043200
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000484653961885657,
      "loss": 1.4599,
      "theoretical_loss": 3.5034226906825303,
      "tokens_seen": 1717108736
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048463390170511535,
      "loss": 1.4304,
      "theoretical_loss": 3.5034112983966277,
      "tokens_seen": 1717174272
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004846138415245737,
      "loss": 1.3379,
      "theoretical_loss": 3.5033999066672386,
      "tokens_seen": 1717239808
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004845937813440321,
      "loss": 1.4475,
      "theoretical_loss": 3.503388515494315,
      "tokens_seen": 1717305344
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048457372116349044,
      "loss": 1.4442,
      "theoretical_loss": 3.503377124877807,
      "tokens_seen": 1717370880
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048455366098294886,
      "loss": 1.409,
      "theoretical_loss": 3.5033657348176686,
      "tokens_seen": 1717436416
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004845336008024072,
      "loss": 1.3802,
      "theoretical_loss": 3.5033543453138494,
      "tokens_seen": 1717501952
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004845135406218656,
      "loss": 1.4193,
      "theoretical_loss": 3.5033429563663026,
      "tokens_seen": 1717567488
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048449348044132394,
      "loss": 1.4649,
      "theoretical_loss": 3.5033315679749784,
      "tokens_seen": 1717633024
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048447342026078236,
      "loss": 1.4134,
      "theoretical_loss": 3.503320180139829,
      "tokens_seen": 1717698560
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004844533600802407,
      "loss": 1.4918,
      "theoretical_loss": 3.503308792860806,
      "tokens_seen": 1717764096
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004844332998996991,
      "loss": 1.5062,
      "theoretical_loss": 3.503297406137861,
      "tokens_seen": 1717829632
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048441323971915745,
      "loss": 1.4696,
      "theoretical_loss": 3.503286019970946,
      "tokens_seen": 1717895168
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048439317953861586,
      "loss": 1.4392,
      "theoretical_loss": 3.503274634360012,
      "tokens_seen": 1717960704
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004843731193580742,
      "loss": 1.3401,
      "theoretical_loss": 3.503263249305011,
      "tokens_seen": 1718026240
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004843530591775326,
      "loss": 1.4782,
      "theoretical_loss": 3.5032518648058946,
      "tokens_seen": 1718091776
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048433299899699095,
      "loss": 1.3947,
      "theoretical_loss": 3.5032404808626145,
      "tokens_seen": 1718157312
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004843129388164493,
      "loss": 1.3781,
      "theoretical_loss": 3.503229097475122,
      "tokens_seen": 1718222848
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048429287863590773,
      "loss": 1.3341,
      "theoretical_loss": 3.5032177146433696,
      "tokens_seen": 1718288384
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004842728184553661,
      "loss": 1.4159,
      "theoretical_loss": 3.503206332367308,
      "tokens_seen": 1718353920
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048425275827482445,
      "loss": 1.3872,
      "theoretical_loss": 3.5031949506468893,
      "tokens_seen": 1718419456
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048423269809428287,
      "loss": 1.3404,
      "theoretical_loss": 3.5031835694820654,
      "tokens_seen": 1718484992
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004842126379137413,
      "loss": 1.447,
      "theoretical_loss": 3.503172188872788,
      "tokens_seen": 1718550528
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048419257773319965,
      "loss": 1.467,
      "theoretical_loss": 3.5031608088190085,
      "tokens_seen": 1718616064
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000484172517552658,
      "loss": 1.4296,
      "theoretical_loss": 3.5031494293206786,
      "tokens_seen": 1718681600
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048415245737211637,
      "loss": 1.4116,
      "theoretical_loss": 3.5031380503777503,
      "tokens_seen": 1718747136
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048413239719157473,
      "loss": 1.3253,
      "theoretical_loss": 3.503126671990175,
      "tokens_seen": 1718812672
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048411233701103315,
      "loss": 1.4139,
      "theoretical_loss": 3.5031152941579045,
      "tokens_seen": 1718878208
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004840922768304915,
      "loss": 1.3894,
      "theoretical_loss": 3.5031039168808906,
      "tokens_seen": 1718943744
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004840722166499499,
      "loss": 1.3476,
      "theoretical_loss": 3.503092540159085,
      "tokens_seen": 1719009280
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048405215646940824,
      "loss": 1.3462,
      "theoretical_loss": 3.50308116399244,
      "tokens_seen": 1719074816
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048403209628886665,
      "loss": 1.5106,
      "theoretical_loss": 3.5030697883809063,
      "tokens_seen": 1719140352
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000484012036108325,
      "loss": 1.5089,
      "theoretical_loss": 3.5030584133244362,
      "tokens_seen": 1719205888
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004839919759277834,
      "loss": 1.4515,
      "theoretical_loss": 3.5030470388229817,
      "tokens_seen": 1719271424
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048397191574724174,
      "loss": 1.3887,
      "theoretical_loss": 3.503035664876494,
      "tokens_seen": 1719336960
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004839518555667001,
      "loss": 1.4387,
      "theoretical_loss": 3.5030242914849254,
      "tokens_seen": 1719402496
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004839317953861585,
      "loss": 1.5575,
      "theoretical_loss": 3.5030129186482277,
      "tokens_seen": 1719468032
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004839117352056169,
      "loss": 1.4897,
      "theoretical_loss": 3.5030015463663515,
      "tokens_seen": 1719533568
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048389167502507524,
      "loss": 1.344,
      "theoretical_loss": 3.50299017463925,
      "tokens_seen": 1719599104
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004838716148445336,
      "loss": 1.4789,
      "theoretical_loss": 3.502978803466875,
      "tokens_seen": 1719664640
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000483851554663992,
      "loss": 1.5309,
      "theoretical_loss": 3.502967432849177,
      "tokens_seen": 1719730176
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004838314944834504,
      "loss": 1.4808,
      "theoretical_loss": 3.502956062786109,
      "tokens_seen": 1719795712
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048381143430290875,
      "loss": 1.4405,
      "theoretical_loss": 3.5029446932776223,
      "tokens_seen": 1719861248
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004837913741223671,
      "loss": 1.3852,
      "theoretical_loss": 3.502933324323669,
      "tokens_seen": 1719926784
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048377131394182547,
      "loss": 1.5443,
      "theoretical_loss": 3.502921955924201,
      "tokens_seen": 1719992320
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004837512537612839,
      "loss": 1.4123,
      "theoretical_loss": 3.5029105880791693,
      "tokens_seen": 1720057856
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048373119358074225,
      "loss": 1.4556,
      "theoretical_loss": 3.502899220788527,
      "tokens_seen": 1720123392
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004837111334002006,
      "loss": 1.461,
      "theoretical_loss": 3.502887854052225,
      "tokens_seen": 1720188928
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000483691073219659,
      "loss": 1.4142,
      "theoretical_loss": 3.502876487870215,
      "tokens_seen": 1720254464
    },
    {
      "epoch": 0.52,
      "objective/train/advantage_avg": 0.49109649658203125,
      "objective/train/docs_used": 628251,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5509767532348633,
      "objective/train/original_loss": 1.5509767532348633,
      "objective/train/theoretical_loss": 3.50286512224245,
      "objective/train/tokens_used": 1740780000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2420029491186142,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503398180007935,
      "objective/train/weighted_lm_loss": 1.6289955377578735,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.50286512224245,
      "tokens_seen": 1720320000
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004836710130391174,
      "loss": 1.505,
      "theoretical_loss": 3.50286512224245,
      "tokens_seen": 1720320000
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048365095285857575,
      "loss": 1.3923,
      "theoretical_loss": 3.5028537571688805,
      "tokens_seen": 1720385536
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004836308926780341,
      "loss": 1.4108,
      "theoretical_loss": 3.5028423926494594,
      "tokens_seen": 1720451072
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004836108324974925,
      "loss": 1.4644,
      "theoretical_loss": 3.502831028684138,
      "tokens_seen": 1720516608
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048359077231695084,
      "loss": 1.4488,
      "theoretical_loss": 3.5028196652728685,
      "tokens_seen": 1720582144
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048357071213640926,
      "loss": 1.4457,
      "theoretical_loss": 3.5028083024156027,
      "tokens_seen": 1720647680
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004835506519558676,
      "loss": 1.3513,
      "theoretical_loss": 3.5027969401122925,
      "tokens_seen": 1720713216
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000483530591775326,
      "loss": 1.4521,
      "theoretical_loss": 3.5027855783628894,
      "tokens_seen": 1720778752
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048351053159478434,
      "loss": 1.4114,
      "theoretical_loss": 3.502774217167346,
      "tokens_seen": 1720844288
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048349047141424276,
      "loss": 1.4492,
      "theoretical_loss": 3.5027628565256137,
      "tokens_seen": 1720909824
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004834704112337011,
      "loss": 1.3912,
      "theoretical_loss": 3.5027514964376447,
      "tokens_seen": 1720975360
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004834503510531595,
      "loss": 1.5046,
      "theoretical_loss": 3.50274013690339,
      "tokens_seen": 1721040896
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048343029087261785,
      "loss": 1.4842,
      "theoretical_loss": 3.5027287779228033,
      "tokens_seen": 1721106432
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048341023069207626,
      "loss": 1.4445,
      "theoretical_loss": 3.502717419495835,
      "tokens_seen": 1721171968
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004833901705115346,
      "loss": 1.381,
      "theoretical_loss": 3.502706061622438,
      "tokens_seen": 1721237504
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000483370110330993,
      "loss": 1.3932,
      "theoretical_loss": 3.502694704302564,
      "tokens_seen": 1721303040
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048335005015045135,
      "loss": 1.397,
      "theoretical_loss": 3.502683347536164,
      "tokens_seen": 1721368576
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004833299899699097,
      "loss": 1.4554,
      "theoretical_loss": 3.5026719913231914,
      "tokens_seen": 1721434112
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048330992978936813,
      "loss": 1.4379,
      "theoretical_loss": 3.502660635663597,
      "tokens_seen": 1721499648
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004832898696088265,
      "loss": 1.4746,
      "theoretical_loss": 3.5026492805573337,
      "tokens_seen": 1721565184
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048326980942828485,
      "loss": 1.42,
      "theoretical_loss": 3.502637926004353,
      "tokens_seen": 1721630720
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004832497492477432,
      "loss": 1.4971,
      "theoretical_loss": 3.5026265720046066,
      "tokens_seen": 1721696256
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048322968906720163,
      "loss": 1.408,
      "theoretical_loss": 3.5026152185580472,
      "tokens_seen": 1721761792
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048320962888666,
      "loss": 1.3928,
      "theoretical_loss": 3.5026038656646263,
      "tokens_seen": 1721827328
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048318956870611836,
      "loss": 1.3744,
      "theoretical_loss": 3.502592513324296,
      "tokens_seen": 1721892864
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004831695085255767,
      "loss": 1.5011,
      "theoretical_loss": 3.502581161537008,
      "tokens_seen": 1721958400
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004831494483450351,
      "loss": 1.412,
      "theoretical_loss": 3.502569810302715,
      "tokens_seen": 1722023936
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004831293881644935,
      "loss": 1.4747,
      "theoretical_loss": 3.502558459621369,
      "tokens_seen": 1722089472
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048310932798395186,
      "loss": 1.3841,
      "theoretical_loss": 3.502547109492921,
      "tokens_seen": 1722155008
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004830892678034102,
      "loss": 1.4604,
      "theoretical_loss": 3.5025357599173237,
      "tokens_seen": 1722220544
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004830692076228686,
      "loss": 1.3572,
      "theoretical_loss": 3.5025244108945293,
      "tokens_seen": 1722286080
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000483049147442327,
      "loss": 1.3558,
      "theoretical_loss": 3.50251306242449,
      "tokens_seen": 1722351616
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048302908726178536,
      "loss": 1.4006,
      "theoretical_loss": 3.502501714507157,
      "tokens_seen": 1722417152
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004830090270812437,
      "loss": 1.4969,
      "theoretical_loss": 3.5024903671424834,
      "tokens_seen": 1722482688
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004829889669007021,
      "loss": 1.3883,
      "theoretical_loss": 3.50247902033042,
      "tokens_seen": 1722548224
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048296890672016045,
      "loss": 1.4888,
      "theoretical_loss": 3.50246767407092,
      "tokens_seen": 1722613760
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048294884653961887,
      "loss": 1.459,
      "theoretical_loss": 3.5024563283639356,
      "tokens_seen": 1722679296
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048292878635907723,
      "loss": 1.4419,
      "theoretical_loss": 3.5024449832094176,
      "tokens_seen": 1722744832
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004829087261785356,
      "loss": 1.4525,
      "theoretical_loss": 3.502433638607319,
      "tokens_seen": 1722810368
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048288866599799395,
      "loss": 1.3867,
      "theoretical_loss": 3.502422294557592,
      "tokens_seen": 1722875904
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048286860581745237,
      "loss": 1.4631,
      "theoretical_loss": 3.502410951060188,
      "tokens_seen": 1722941440
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048284854563691073,
      "loss": 1.491,
      "theoretical_loss": 3.50239960811506,
      "tokens_seen": 1723006976
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004828284854563691,
      "loss": 1.42,
      "theoretical_loss": 3.502388265722159,
      "tokens_seen": 1723072512
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048280842527582746,
      "loss": 1.3997,
      "theoretical_loss": 3.502376923881438,
      "tokens_seen": 1723138048
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004827883650952858,
      "loss": 1.4316,
      "theoretical_loss": 3.502365582592849,
      "tokens_seen": 1723203584
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048276830491474423,
      "loss": 1.4343,
      "theoretical_loss": 3.5023542418563443,
      "tokens_seen": 1723269120
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004827482447342026,
      "loss": 1.5149,
      "theoretical_loss": 3.5023429016718755,
      "tokens_seen": 1723334656
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000482728184553661,
      "loss": 1.4096,
      "theoretical_loss": 3.5023315620393944,
      "tokens_seen": 1723400192
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004827081243731194,
      "loss": 1.4526,
      "theoretical_loss": 3.502320222958854,
      "tokens_seen": 1723465728
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004826880641925778,
      "loss": 1.5086,
      "theoretical_loss": 3.502308884430206,
      "tokens_seen": 1723531264
    },
    {
      "epoch": 0.52,
      "objective/train/advantage_avg": 0.49437761306762695,
      "objective/train/docs_used": 629447,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.478456974029541,
      "objective/train/original_loss": 1.478456974029541,
      "objective/train/theoretical_loss": 3.502297546453403,
      "objective/train/tokens_used": 1744056800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24472007155418396,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506818294525146,
      "objective/train/weighted_lm_loss": 1.5532763004302979,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.502297546453403,
      "tokens_seen": 1723596800
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048266800401203615,
      "loss": 1.4669,
      "theoretical_loss": 3.502297546453403,
      "tokens_seen": 1723596800
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004826479438314945,
      "loss": 1.4977,
      "theoretical_loss": 3.5022862090283966,
      "tokens_seen": 1723662336
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004826278836509529,
      "loss": 1.4508,
      "theoretical_loss": 3.502274872155139,
      "tokens_seen": 1723727872
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048260782347041124,
      "loss": 1.3122,
      "theoretical_loss": 3.502263535833583,
      "tokens_seen": 1723793408
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048258776328986966,
      "loss": 1.3923,
      "theoretical_loss": 3.5022522000636807,
      "tokens_seen": 1723858944
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000482567703109328,
      "loss": 1.4508,
      "theoretical_loss": 3.5022408648453833,
      "tokens_seen": 1723924480
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004825476429287864,
      "loss": 1.4175,
      "theoretical_loss": 3.502229530178644,
      "tokens_seen": 1723990016
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048252758274824474,
      "loss": 1.4804,
      "theoretical_loss": 3.5022181960634144,
      "tokens_seen": 1724055552
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048250752256770316,
      "loss": 1.4624,
      "theoretical_loss": 3.502206862499647,
      "tokens_seen": 1724121088
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004824874623871615,
      "loss": 1.3615,
      "theoretical_loss": 3.5021955294872944,
      "tokens_seen": 1724186624
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004824674022066199,
      "loss": 1.4478,
      "theoretical_loss": 3.502184197026308,
      "tokens_seen": 1724252160
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048244734202607825,
      "loss": 1.3735,
      "theoretical_loss": 3.5021728651166404,
      "tokens_seen": 1724317696
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048242728184553666,
      "loss": 1.4838,
      "theoretical_loss": 3.5021615337582435,
      "tokens_seen": 1724383232
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000482407221664995,
      "loss": 1.3833,
      "theoretical_loss": 3.50215020295107,
      "tokens_seen": 1724448768
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004823871614844534,
      "loss": 1.4693,
      "theoretical_loss": 3.502138872695072,
      "tokens_seen": 1724514304
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048236710130391175,
      "loss": 1.3846,
      "theoretical_loss": 3.502127542990202,
      "tokens_seen": 1724579840
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004823470411233701,
      "loss": 1.458,
      "theoretical_loss": 3.5021162138364117,
      "tokens_seen": 1724645376
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048232698094282853,
      "loss": 1.4091,
      "theoretical_loss": 3.5021048852336536,
      "tokens_seen": 1724710912
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004823069207622869,
      "loss": 1.6041,
      "theoretical_loss": 3.5020935571818805,
      "tokens_seen": 1724776448
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048228686058174525,
      "loss": 1.316,
      "theoretical_loss": 3.5020822296810437,
      "tokens_seen": 1724841984
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004822668004012036,
      "loss": 1.411,
      "theoretical_loss": 3.502070902731096,
      "tokens_seen": 1724907520
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048224674022066203,
      "loss": 1.5619,
      "theoretical_loss": 3.5020595763319897,
      "tokens_seen": 1724973056
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004822266800401204,
      "loss": 1.4409,
      "theoretical_loss": 3.5020482504836767,
      "tokens_seen": 1725038592
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048220661985957876,
      "loss": 1.452,
      "theoretical_loss": 3.50203692518611,
      "tokens_seen": 1725104128
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004821865596790371,
      "loss": 1.4642,
      "theoretical_loss": 3.502025600439241,
      "tokens_seen": 1725169664
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004821664994984955,
      "loss": 1.375,
      "theoretical_loss": 3.502014276243023,
      "tokens_seen": 1725235200
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004821464393179539,
      "loss": 1.4521,
      "theoretical_loss": 3.5020029525974077,
      "tokens_seen": 1725300736
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048212637913741226,
      "loss": 1.5543,
      "theoretical_loss": 3.5019916295023474,
      "tokens_seen": 1725366272
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004821063189568706,
      "loss": 1.5515,
      "theoretical_loss": 3.5019803069577944,
      "tokens_seen": 1725431808
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000482086258776329,
      "loss": 1.379,
      "theoretical_loss": 3.5019689849637015,
      "tokens_seen": 1725497344
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004820661985957874,
      "loss": 1.4555,
      "theoretical_loss": 3.5019576635200202,
      "tokens_seen": 1725562880
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048204613841524576,
      "loss": 1.4335,
      "theoretical_loss": 3.501946342626704,
      "tokens_seen": 1725628416
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004820260782347041,
      "loss": 1.4646,
      "theoretical_loss": 3.5019350222837042,
      "tokens_seen": 1725693952
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004820060180541625,
      "loss": 1.3429,
      "theoretical_loss": 3.5019237024909735,
      "tokens_seen": 1725759488
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048198595787362085,
      "loss": 1.4026,
      "theoretical_loss": 3.5019123832484644,
      "tokens_seen": 1725825024
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048196589769307927,
      "loss": 1.6515,
      "theoretical_loss": 3.5019010645561295,
      "tokens_seen": 1725890560
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048194583751253763,
      "loss": 1.4192,
      "theoretical_loss": 3.5018897464139203,
      "tokens_seen": 1725956096
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000481925777331996,
      "loss": 1.4294,
      "theoretical_loss": 3.5018784288217897,
      "tokens_seen": 1726021632
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048190571715145435,
      "loss": 1.4382,
      "theoretical_loss": 3.5018671117796902,
      "tokens_seen": 1726087168
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048188565697091277,
      "loss": 1.5653,
      "theoretical_loss": 3.501855795287574,
      "tokens_seen": 1726152704
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048186559679037113,
      "loss": 1.5051,
      "theoretical_loss": 3.501844479345394,
      "tokens_seen": 1726218240
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004818455366098295,
      "loss": 1.4609,
      "theoretical_loss": 3.5018331639531017,
      "tokens_seen": 1726283776
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048182547642928786,
      "loss": 1.4433,
      "theoretical_loss": 3.50182184911065,
      "tokens_seen": 1726349312
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004818054162487462,
      "loss": 1.4462,
      "theoretical_loss": 3.5018105348179915,
      "tokens_seen": 1726414848
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048178535606820463,
      "loss": 1.3765,
      "theoretical_loss": 3.5017992210750783,
      "tokens_seen": 1726480384
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000481765295887663,
      "loss": 1.5321,
      "theoretical_loss": 3.5017879078818623,
      "tokens_seen": 1726545920
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048174523570712136,
      "loss": 1.4416,
      "theoretical_loss": 3.5017765952382973,
      "tokens_seen": 1726611456
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004817251755265797,
      "loss": 1.4828,
      "theoretical_loss": 3.501765283144335,
      "tokens_seen": 1726676992
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048170511534603814,
      "loss": 1.481,
      "theoretical_loss": 3.5017539715999275,
      "tokens_seen": 1726742528
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004816850551654965,
      "loss": 1.4126,
      "theoretical_loss": 3.5017426606050277,
      "tokens_seen": 1726808064
    },
    {
      "epoch": 0.52,
      "objective/train/advantage_avg": 0.48893359303474426,
      "objective/train/docs_used": 630619,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4349052906036377,
      "objective/train/original_loss": 1.4349052906036377,
      "objective/train/theoretical_loss": 3.501731350159588,
      "objective/train/tokens_used": 1747333600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24070586264133453,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501171350479126,
      "objective/train/weighted_lm_loss": 1.5070555210113525,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.501731350159588,
      "tokens_seen": 1726873600
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048166499498495486,
      "loss": 1.5299,
      "theoretical_loss": 3.501731350159588,
      "tokens_seen": 1726873600
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004816449348044132,
      "loss": 1.4582,
      "theoretical_loss": 3.50172004026356,
      "tokens_seen": 1726939136
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004816248746238716,
      "loss": 1.3758,
      "theoretical_loss": 3.5017087309168975,
      "tokens_seen": 1727004672
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048160481444333,
      "loss": 1.3066,
      "theoretical_loss": 3.5016974221195527,
      "tokens_seen": 1727070208
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048158475426278837,
      "loss": 1.5561,
      "theoretical_loss": 3.501686113871478,
      "tokens_seen": 1727135744
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048156469408224673,
      "loss": 1.5168,
      "theoretical_loss": 3.501674806172625,
      "tokens_seen": 1727201280
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004815446339017051,
      "loss": 1.4879,
      "theoretical_loss": 3.501663499022947,
      "tokens_seen": 1727266816
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004815245737211635,
      "loss": 1.3821,
      "theoretical_loss": 3.5016521924223962,
      "tokens_seen": 1727332352
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048150451354062187,
      "loss": 1.3476,
      "theoretical_loss": 3.501640886370925,
      "tokens_seen": 1727397888
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048148445336008023,
      "loss": 1.4704,
      "theoretical_loss": 3.501629580868487,
      "tokens_seen": 1727463424
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004814643931795386,
      "loss": 1.4361,
      "theoretical_loss": 3.5016182759150336,
      "tokens_seen": 1727528960
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048144433299899696,
      "loss": 1.4957,
      "theoretical_loss": 3.501606971510517,
      "tokens_seen": 1727594496
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048142427281845537,
      "loss": 1.3972,
      "theoretical_loss": 3.5015956676548905,
      "tokens_seen": 1727660032
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048140421263791373,
      "loss": 1.3887,
      "theoretical_loss": 3.501584364348107,
      "tokens_seen": 1727725568
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004813841524573721,
      "loss": 1.4095,
      "theoretical_loss": 3.501573061590118,
      "tokens_seen": 1727791104
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048136409227683046,
      "loss": 1.404,
      "theoretical_loss": 3.5015617593808765,
      "tokens_seen": 1727856640
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004813440320962889,
      "loss": 1.4226,
      "theoretical_loss": 3.5015504577203354,
      "tokens_seen": 1727922176
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048132397191574724,
      "loss": 1.3703,
      "theoretical_loss": 3.5015391566084464,
      "tokens_seen": 1727987712
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004813039117352056,
      "loss": 1.4251,
      "theoretical_loss": 3.5015278560451626,
      "tokens_seen": 1728053248
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048128385155466396,
      "loss": 1.5577,
      "theoretical_loss": 3.5015165560304364,
      "tokens_seen": 1728118784
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004812637913741223,
      "loss": 1.4208,
      "theoretical_loss": 3.5015052565642213,
      "tokens_seen": 1728184320
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048124373119358074,
      "loss": 1.3837,
      "theoretical_loss": 3.5014939576464683,
      "tokens_seen": 1728249856
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048122367101303916,
      "loss": 1.3618,
      "theoretical_loss": 3.501482659277131,
      "tokens_seen": 1728315392
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004812036108324975,
      "loss": 1.481,
      "theoretical_loss": 3.5014713614561614,
      "tokens_seen": 1728380928
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004811835506519559,
      "loss": 1.5122,
      "theoretical_loss": 3.501460064183513,
      "tokens_seen": 1728446464
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004811634904714143,
      "loss": 1.4415,
      "theoretical_loss": 3.5014487674591375,
      "tokens_seen": 1728512000
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048114343029087266,
      "loss": 1.3485,
      "theoretical_loss": 3.501437471282988,
      "tokens_seen": 1728577536
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000481123370110331,
      "loss": 1.2895,
      "theoretical_loss": 3.5014261756550167,
      "tokens_seen": 1728643072
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004811033099297894,
      "loss": 1.4608,
      "theoretical_loss": 3.501414880575177,
      "tokens_seen": 1728708608
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004810832497492478,
      "loss": 1.4924,
      "theoretical_loss": 3.50140358604342,
      "tokens_seen": 1728774144
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048106318956870616,
      "loss": 1.2975,
      "theoretical_loss": 3.5013922920597,
      "tokens_seen": 1728839680
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004810431293881645,
      "loss": 1.4306,
      "theoretical_loss": 3.5013809986239686,
      "tokens_seen": 1728905216
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004810230692076229,
      "loss": 1.3499,
      "theoretical_loss": 3.5013697057361792,
      "tokens_seen": 1728970752
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048100300902708125,
      "loss": 1.4469,
      "theoretical_loss": 3.501358413396284,
      "tokens_seen": 1729036288
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048098294884653967,
      "loss": 1.4523,
      "theoretical_loss": 3.5013471216042356,
      "tokens_seen": 1729101824
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048096288866599803,
      "loss": 1.4394,
      "theoretical_loss": 3.5013358303599866,
      "tokens_seen": 1729167360
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004809428284854564,
      "loss": 1.2977,
      "theoretical_loss": 3.5013245396634898,
      "tokens_seen": 1729232896
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048092276830491475,
      "loss": 1.3849,
      "theoretical_loss": 3.5013132495146984,
      "tokens_seen": 1729298432
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048090270812437317,
      "loss": 1.3839,
      "theoretical_loss": 3.5013019599135635,
      "tokens_seen": 1729363968
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048088264794383153,
      "loss": 1.4241,
      "theoretical_loss": 3.5012906708600395,
      "tokens_seen": 1729429504
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004808625877632899,
      "loss": 1.4842,
      "theoretical_loss": 3.5012793823540784,
      "tokens_seen": 1729495040
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048084252758274826,
      "loss": 1.3914,
      "theoretical_loss": 3.501268094395633,
      "tokens_seen": 1729560576
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004808224674022066,
      "loss": 1.4702,
      "theoretical_loss": 3.501256806984656,
      "tokens_seen": 1729626112
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048080240722166503,
      "loss": 1.451,
      "theoretical_loss": 3.5012455201211,
      "tokens_seen": 1729691648
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004807823470411234,
      "loss": 1.4511,
      "theoretical_loss": 3.501234233804918,
      "tokens_seen": 1729757184
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048076228686058176,
      "loss": 1.3841,
      "theoretical_loss": 3.5012229480360615,
      "tokens_seen": 1729822720
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004807422266800401,
      "loss": 1.5082,
      "theoretical_loss": 3.5012116628144847,
      "tokens_seen": 1729888256
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048072216649949854,
      "loss": 1.3707,
      "theoretical_loss": 3.50120037814014,
      "tokens_seen": 1729953792
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004807021063189569,
      "loss": 1.2578,
      "theoretical_loss": 3.5011890940129797,
      "tokens_seen": 1730019328
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048068204613841526,
      "loss": 1.4235,
      "theoretical_loss": 3.501177810432957,
      "tokens_seen": 1730084864
    },
    {
      "epoch": 0.52,
      "objective/train/advantage_avg": 0.4888365864753723,
      "objective/train/docs_used": 631803,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.433842420578003,
      "objective/train/original_loss": 1.433842420578003,
      "objective/train/theoretical_loss": 3.5011665274000245,
      "objective/train/tokens_used": 1750610400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24063989520072937,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501067638397217,
      "objective/train/weighted_lm_loss": 1.5059046745300293,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.5011665274000245,
      "tokens_seen": 1730150400
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004806619859578736,
      "loss": 1.4598,
      "theoretical_loss": 3.5011665274000245,
      "tokens_seen": 1730150400
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000480641925777332,
      "loss": 1.5092,
      "theoretical_loss": 3.501155244914134,
      "tokens_seen": 1730215936
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004806218655967904,
      "loss": 1.507,
      "theoretical_loss": 3.50114396297524,
      "tokens_seen": 1730281472
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048060180541624877,
      "loss": 1.4668,
      "theoretical_loss": 3.5011326815832944,
      "tokens_seen": 1730347008
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048058174523570713,
      "loss": 1.4748,
      "theoretical_loss": 3.50112140073825,
      "tokens_seen": 1730412544
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004805616850551655,
      "loss": 1.4547,
      "theoretical_loss": 3.5011101204400594,
      "tokens_seen": 1730478080
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004805416248746239,
      "loss": 1.4289,
      "theoretical_loss": 3.501098840688675,
      "tokens_seen": 1730543616
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048052156469408227,
      "loss": 1.3187,
      "theoretical_loss": 3.501087561484051,
      "tokens_seen": 1730609152
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048050150451354063,
      "loss": 1.5323,
      "theoretical_loss": 3.501076282826139,
      "tokens_seen": 1730674688
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000480481444332999,
      "loss": 1.4953,
      "theoretical_loss": 3.501065004714892,
      "tokens_seen": 1730740224
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048046138415245736,
      "loss": 1.2776,
      "theoretical_loss": 3.501053727150263,
      "tokens_seen": 1730805760
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048044132397191577,
      "loss": 1.4925,
      "theoretical_loss": 3.5010424501322044,
      "tokens_seen": 1730871296
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048042126379137413,
      "loss": 1.4759,
      "theoretical_loss": 3.5010311736606696,
      "tokens_seen": 1730936832
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004804012036108325,
      "loss": 1.4114,
      "theoretical_loss": 3.501019897735611,
      "tokens_seen": 1731002368
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048038114343029086,
      "loss": 1.4146,
      "theoretical_loss": 3.501008622356982,
      "tokens_seen": 1731067904
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004803610832497493,
      "loss": 1.4855,
      "theoretical_loss": 3.5009973475247347,
      "tokens_seen": 1731133440
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048034102306920764,
      "loss": 1.5803,
      "theoretical_loss": 3.5009860732388223,
      "tokens_seen": 1731198976
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000480320962888666,
      "loss": 1.4257,
      "theoretical_loss": 3.5009747994991978,
      "tokens_seen": 1731264512
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048030090270812436,
      "loss": 1.441,
      "theoretical_loss": 3.5009635263058136,
      "tokens_seen": 1731330048
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004802808425275827,
      "loss": 1.4637,
      "theoretical_loss": 3.5009522536586233,
      "tokens_seen": 1731395584
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048026078234704114,
      "loss": 1.5105,
      "theoretical_loss": 3.5009409815575787,
      "tokens_seen": 1731461120
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004802407221664995,
      "loss": 1.4621,
      "theoretical_loss": 3.5009297100026338,
      "tokens_seen": 1731526656
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048022066198595786,
      "loss": 1.424,
      "theoretical_loss": 3.5009184389937404,
      "tokens_seen": 1731592192
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048020060180541623,
      "loss": 1.3351,
      "theoretical_loss": 3.500907168530852,
      "tokens_seen": 1731657728
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048018054162487464,
      "loss": 1.377,
      "theoretical_loss": 3.500895898613922,
      "tokens_seen": 1731723264
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.000480160481444333,
      "loss": 1.3786,
      "theoretical_loss": 3.5008846292429023,
      "tokens_seen": 1731788800
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048014042126379137,
      "loss": 1.4566,
      "theoretical_loss": 3.500873360417746,
      "tokens_seen": 1731854336
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048012036108324973,
      "loss": 1.5382,
      "theoretical_loss": 3.500862092138407,
      "tokens_seen": 1731919872
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048010030090270815,
      "loss": 1.487,
      "theoretical_loss": 3.5008508244048366,
      "tokens_seen": 1731985408
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004800802407221665,
      "loss": 1.389,
      "theoretical_loss": 3.5008395572169886,
      "tokens_seen": 1732050944
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048006018054162487,
      "loss": 1.4717,
      "theoretical_loss": 3.500828290574816,
      "tokens_seen": 1732116480
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048004012036108323,
      "loss": 1.4215,
      "theoretical_loss": 3.5008170244782715,
      "tokens_seen": 1732182016
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004800200601805416,
      "loss": 1.4224,
      "theoretical_loss": 3.5008057589273083,
      "tokens_seen": 1732247552
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00048,
      "loss": 1.4199,
      "theoretical_loss": 3.500794493921879,
      "tokens_seen": 1732313088
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0004799799398194584,
      "loss": 1.3875,
      "theoretical_loss": 3.5007832294619368,
      "tokens_seen": 1732378624
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.00047995987963891674,
      "loss": 1.3567,
      "theoretical_loss": 3.5007719655474343,
      "tokens_seen": 1732444160
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004799398194583751,
      "loss": 1.4078,
      "theoretical_loss": 3.500760702178325,
      "tokens_seen": 1732509696
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004799197592778335,
      "loss": 1.3841,
      "theoretical_loss": 3.5007494393545615,
      "tokens_seen": 1732575232
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004798996990972919,
      "loss": 1.3867,
      "theoretical_loss": 3.5007381770760966,
      "tokens_seen": 1732640768
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047987963891675024,
      "loss": 1.3313,
      "theoretical_loss": 3.5007269153428835,
      "tokens_seen": 1732706304
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004798595787362086,
      "loss": 1.3158,
      "theoretical_loss": 3.5007156541548756,
      "tokens_seen": 1732771840
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047983951855566696,
      "loss": 1.4064,
      "theoretical_loss": 3.5007043935120254,
      "tokens_seen": 1732837376
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004798194583751254,
      "loss": 1.5059,
      "theoretical_loss": 3.500693133414286,
      "tokens_seen": 1732902912
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047979939819458374,
      "loss": 1.4252,
      "theoretical_loss": 3.50068187386161,
      "tokens_seen": 1732968448
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004797793380140421,
      "loss": 1.5245,
      "theoretical_loss": 3.5006706148539513,
      "tokens_seen": 1733033984
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047975927783350047,
      "loss": 1.5415,
      "theoretical_loss": 3.500659356391262,
      "tokens_seen": 1733099520
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004797392176529589,
      "loss": 1.4182,
      "theoretical_loss": 3.500648098473495,
      "tokens_seen": 1733165056
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004797191574724173,
      "loss": 1.5603,
      "theoretical_loss": 3.5006368411006052,
      "tokens_seen": 1733230592
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047969909729187566,
      "loss": 1.4291,
      "theoretical_loss": 3.500625584272543,
      "tokens_seen": 1733296128
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000479679037111334,
      "loss": 1.3877,
      "theoretical_loss": 3.5006143279892634,
      "tokens_seen": 1733361664
    },
    {
      "epoch": 0.53,
      "objective/train/advantage_avg": 0.49312207102775574,
      "objective/train/docs_used": 632954,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3543633222579956,
      "objective/train/original_loss": 1.3543634414672852,
      "objective/train/theoretical_loss": 3.500603072250718,
      "objective/train/tokens_used": 1753887200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2438594400882721,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505518913269043,
      "objective/train/weighted_lm_loss": 1.4227088689804077,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.500603072250718,
      "tokens_seen": 1733427200
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004796589769307924,
      "loss": 1.4242,
      "theoretical_loss": 3.500603072250718,
      "tokens_seen": 1733427200
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004796389167502508,
      "loss": 1.4074,
      "theoretical_loss": 3.5005918170568613,
      "tokens_seen": 1733492736
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047961885656970917,
      "loss": 1.436,
      "theoretical_loss": 3.5005805624076456,
      "tokens_seen": 1733558272
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047959879638916753,
      "loss": 1.3825,
      "theoretical_loss": 3.5005693083030236,
      "tokens_seen": 1733623808
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004795787362086259,
      "loss": 1.4022,
      "theoretical_loss": 3.500558054742949,
      "tokens_seen": 1733689344
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004795586760280843,
      "loss": 1.3948,
      "theoretical_loss": 3.500546801727374,
      "tokens_seen": 1733754880
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047953861584754267,
      "loss": 1.4996,
      "theoretical_loss": 3.500535549256253,
      "tokens_seen": 1733820416
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047951855566700103,
      "loss": 1.3312,
      "theoretical_loss": 3.5005242973295383,
      "tokens_seen": 1733885952
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004794984954864594,
      "loss": 1.4438,
      "theoretical_loss": 3.5005130459471827,
      "tokens_seen": 1733951488
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047947843530591776,
      "loss": 1.445,
      "theoretical_loss": 3.50050179510914,
      "tokens_seen": 1734017024
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047945837512537617,
      "loss": 1.3383,
      "theoretical_loss": 3.5004905448153627,
      "tokens_seen": 1734082560
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047943831494483453,
      "loss": 1.5082,
      "theoretical_loss": 3.500479295065804,
      "tokens_seen": 1734148096
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004794182547642929,
      "loss": 1.3421,
      "theoretical_loss": 3.5004680458604174,
      "tokens_seen": 1734213632
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047939819458375126,
      "loss": 1.4394,
      "theoretical_loss": 3.500456797199156,
      "tokens_seen": 1734279168
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004793781344032097,
      "loss": 1.5356,
      "theoretical_loss": 3.500445549081972,
      "tokens_seen": 1734344704
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047935807422266804,
      "loss": 1.3647,
      "theoretical_loss": 3.50043430150882,
      "tokens_seen": 1734410240
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004793380140421264,
      "loss": 1.4034,
      "theoretical_loss": 3.5004230544796515,
      "tokens_seen": 1734475776
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047931795386158476,
      "loss": 1.469,
      "theoretical_loss": 3.5004118079944213,
      "tokens_seen": 1734541312
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004792978936810431,
      "loss": 1.4651,
      "theoretical_loss": 3.500400562053081,
      "tokens_seen": 1734606848
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047927783350050154,
      "loss": 1.4526,
      "theoretical_loss": 3.5003893166555846,
      "tokens_seen": 1734672384
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004792577733199599,
      "loss": 1.3831,
      "theoretical_loss": 3.5003780718018858,
      "tokens_seen": 1734737920
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047923771313941827,
      "loss": 1.3926,
      "theoretical_loss": 3.500366827491936,
      "tokens_seen": 1734803456
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047921765295887663,
      "loss": 1.55,
      "theoretical_loss": 3.50035558372569,
      "tokens_seen": 1734868992
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047919759277833504,
      "loss": 1.5615,
      "theoretical_loss": 3.5003443405031005,
      "tokens_seen": 1734934528
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004791775325977934,
      "loss": 1.4147,
      "theoretical_loss": 3.5003330978241207,
      "tokens_seen": 1735000064
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047915747241725177,
      "loss": 1.4303,
      "theoretical_loss": 3.5003218556887035,
      "tokens_seen": 1735065600
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047915747241725177,
      "loss": 1.4609,
      "theoretical_loss": 3.5003106140968026,
      "tokens_seen": 1735131136
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047913741223671013,
      "loss": 1.4339,
      "theoretical_loss": 3.50029937304837,
      "tokens_seen": 1735196672
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047911735205616855,
      "loss": 1.4078,
      "theoretical_loss": 3.5002881325433606,
      "tokens_seen": 1735262208
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004790972918756269,
      "loss": 1.5127,
      "theoretical_loss": 3.5002768925817263,
      "tokens_seen": 1735327744
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047907723169508527,
      "loss": 1.4622,
      "theoretical_loss": 3.5002656531634213,
      "tokens_seen": 1735393280
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047905717151454363,
      "loss": 1.4652,
      "theoretical_loss": 3.5002544142883982,
      "tokens_seen": 1735458816
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000479037111334002,
      "loss": 1.4637,
      "theoretical_loss": 3.50024317595661,
      "tokens_seen": 1735524352
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004790170511534604,
      "loss": 1.4363,
      "theoretical_loss": 3.5002319381680103,
      "tokens_seen": 1735589888
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004789969909729188,
      "loss": 1.3872,
      "theoretical_loss": 3.5002207009225526,
      "tokens_seen": 1735655424
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047897693079237714,
      "loss": 1.3879,
      "theoretical_loss": 3.5002094642201897,
      "tokens_seen": 1735720960
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004789568706118355,
      "loss": 1.4103,
      "theoretical_loss": 3.500198228060875,
      "tokens_seen": 1735786496
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004789368104312939,
      "loss": 1.4511,
      "theoretical_loss": 3.500186992444562,
      "tokens_seen": 1735852032
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004789167502507523,
      "loss": 1.4226,
      "theoretical_loss": 3.5001757573712036,
      "tokens_seen": 1735917568
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047889669007021064,
      "loss": 1.5224,
      "theoretical_loss": 3.5001645228407527,
      "tokens_seen": 1735983104
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000478876629889669,
      "loss": 1.4416,
      "theoretical_loss": 3.5001532888531637,
      "tokens_seen": 1736048640
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047885656970912736,
      "loss": 1.3797,
      "theoretical_loss": 3.5001420554083884,
      "tokens_seen": 1736114176
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004788365095285858,
      "loss": 1.4673,
      "theoretical_loss": 3.5001308225063816,
      "tokens_seen": 1736179712
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047881644934804414,
      "loss": 1.4792,
      "theoretical_loss": 3.500119590147096,
      "tokens_seen": 1736245248
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004787963891675025,
      "loss": 1.3293,
      "theoretical_loss": 3.5001083583304844,
      "tokens_seen": 1736310784
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047877632898696087,
      "loss": 1.423,
      "theoretical_loss": 3.5000971270565007,
      "tokens_seen": 1736376320
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004787562688064193,
      "loss": 1.482,
      "theoretical_loss": 3.500085896325098,
      "tokens_seen": 1736441856
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047873620862587765,
      "loss": 1.3762,
      "theoretical_loss": 3.5000746661362294,
      "tokens_seen": 1736507392
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000478716148445336,
      "loss": 1.3661,
      "theoretical_loss": 3.500063436489848,
      "tokens_seen": 1736572928
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047869608826479437,
      "loss": 1.4538,
      "theoretical_loss": 3.5000522073859086,
      "tokens_seen": 1736638464
    },
    {
      "epoch": 0.53,
      "objective/train/advantage_avg": 0.4934524595737457,
      "objective/train/docs_used": 634251,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3750983476638794,
      "objective/train/original_loss": 1.3750982284545898,
      "objective/train/theoretical_loss": 3.5000409788243623,
      "objective/train/tokens_used": 1757164000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24383777379989624,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505847930908203,
      "objective/train/weighted_lm_loss": 1.4447449445724487,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.5000409788243623,
      "tokens_seen": 1736704000
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047867602808425273,
      "loss": 1.5186,
      "theoretical_loss": 3.5000409788243623,
      "tokens_seen": 1736704000
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047865596790371115,
      "loss": 1.4377,
      "theoretical_loss": 3.5000297508051643,
      "tokens_seen": 1736769536
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004786359077231695,
      "loss": 1.4836,
      "theoretical_loss": 3.5000185233282677,
      "tokens_seen": 1736835072
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004786158475426279,
      "loss": 1.4579,
      "theoretical_loss": 3.5000072963936244,
      "tokens_seen": 1736900608
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047859578736208624,
      "loss": 1.4975,
      "theoretical_loss": 3.4999960700011896,
      "tokens_seen": 1736966144
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047857572718154465,
      "loss": 1.4081,
      "theoretical_loss": 3.4999848441509154,
      "tokens_seen": 1737031680
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000478555667001003,
      "loss": 1.4808,
      "theoretical_loss": 3.4999736188427555,
      "tokens_seen": 1737097216
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004785356068204614,
      "loss": 1.4165,
      "theoretical_loss": 3.4999623940766638,
      "tokens_seen": 1737162752
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047851554663991974,
      "loss": 1.486,
      "theoretical_loss": 3.4999511698525927,
      "tokens_seen": 1737228288
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004784954864593781,
      "loss": 1.4848,
      "theoretical_loss": 3.499939946170496,
      "tokens_seen": 1737293824
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004784754262788365,
      "loss": 1.4513,
      "theoretical_loss": 3.4999287230303278,
      "tokens_seen": 1737359360
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004784553660982949,
      "loss": 1.4285,
      "theoretical_loss": 3.49991750043204,
      "tokens_seen": 1737424896
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047843530591775324,
      "loss": 1.3694,
      "theoretical_loss": 3.4999062783755877,
      "tokens_seen": 1737490432
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004784152457372116,
      "loss": 1.2821,
      "theoretical_loss": 3.4998950568609235,
      "tokens_seen": 1737555968
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047839518555667,
      "loss": 1.3618,
      "theoretical_loss": 3.4998838358880002,
      "tokens_seen": 1737621504
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004783751253761284,
      "loss": 1.4636,
      "theoretical_loss": 3.499872615456772,
      "tokens_seen": 1737687040
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047835506519558675,
      "loss": 1.4899,
      "theoretical_loss": 3.4998613955671924,
      "tokens_seen": 1737752576
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004783350050150451,
      "loss": 1.501,
      "theoretical_loss": 3.4998501762192142,
      "tokens_seen": 1737818112
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047831494483450347,
      "loss": 1.3871,
      "theoretical_loss": 3.4998389574127913,
      "tokens_seen": 1737883648
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004782948846539619,
      "loss": 1.3348,
      "theoretical_loss": 3.499827739147877,
      "tokens_seen": 1737949184
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047827482447342025,
      "loss": 1.4258,
      "theoretical_loss": 3.4998165214244246,
      "tokens_seen": 1738014720
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004782547642928786,
      "loss": 1.4934,
      "theoretical_loss": 3.499805304242388,
      "tokens_seen": 1738080256
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000478234704112337,
      "loss": 1.4083,
      "theoretical_loss": 3.4997940876017206,
      "tokens_seen": 1738145792
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047821464393179544,
      "loss": 1.4159,
      "theoretical_loss": 3.4997828715023753,
      "tokens_seen": 1738211328
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004781945837512538,
      "loss": 1.4379,
      "theoretical_loss": 3.499771655944306,
      "tokens_seen": 1738276864
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047817452357071217,
      "loss": 1.4561,
      "theoretical_loss": 3.4997604409274663,
      "tokens_seen": 1738342400
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047815446339017053,
      "loss": 1.3568,
      "theoretical_loss": 3.499749226451809,
      "tokens_seen": 1738407936
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047813440320962895,
      "loss": 1.4175,
      "theoretical_loss": 3.4997380125172883,
      "tokens_seen": 1738473472
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004781143430290873,
      "loss": 1.4167,
      "theoretical_loss": 3.4997267991238568,
      "tokens_seen": 1738539008
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047809428284854567,
      "loss": 1.3629,
      "theoretical_loss": 3.4997155862714697,
      "tokens_seen": 1738604544
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047807422266800403,
      "loss": 1.3765,
      "theoretical_loss": 3.4997043739600784,
      "tokens_seen": 1738670080
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004780541624874624,
      "loss": 1.3525,
      "theoretical_loss": 3.4996931621896383,
      "tokens_seen": 1738735616
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004780341023069208,
      "loss": 1.4359,
      "theoretical_loss": 3.4996819509601016,
      "tokens_seen": 1738801152
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004780140421263792,
      "loss": 1.3781,
      "theoretical_loss": 3.4996707402714224,
      "tokens_seen": 1738866688
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047799398194583754,
      "loss": 1.3393,
      "theoretical_loss": 3.4996595301235542,
      "tokens_seen": 1738932224
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004779739217652959,
      "loss": 1.4032,
      "theoretical_loss": 3.4996483205164504,
      "tokens_seen": 1738997760
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004779538615847543,
      "loss": 1.4482,
      "theoretical_loss": 3.499637111450064,
      "tokens_seen": 1739063296
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004779338014042127,
      "loss": 1.5219,
      "theoretical_loss": 3.4996259029243495,
      "tokens_seen": 1739128832
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047791374122367104,
      "loss": 1.4889,
      "theoretical_loss": 3.49961469493926,
      "tokens_seen": 1739194368
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004778936810431294,
      "loss": 1.4327,
      "theoretical_loss": 3.4996034874947495,
      "tokens_seen": 1739259904
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047787362086258776,
      "loss": 1.4362,
      "theoretical_loss": 3.499592280590771,
      "tokens_seen": 1739325440
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004778535606820462,
      "loss": 1.4656,
      "theoretical_loss": 3.4995810742272777,
      "tokens_seen": 1739390976
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047783350050150454,
      "loss": 1.5404,
      "theoretical_loss": 3.4995698684042242,
      "tokens_seen": 1739456512
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004778134403209629,
      "loss": 1.4411,
      "theoretical_loss": 3.4995586631215634,
      "tokens_seen": 1739522048
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047779338014042127,
      "loss": 1.4506,
      "theoretical_loss": 3.499547458379249,
      "tokens_seen": 1739587584
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004777733199598797,
      "loss": 1.3575,
      "theoretical_loss": 3.499536254177235,
      "tokens_seen": 1739653120
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047775325977933805,
      "loss": 1.5226,
      "theoretical_loss": 3.499525050515474,
      "tokens_seen": 1739718656
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004777331995987964,
      "loss": 1.4354,
      "theoretical_loss": 3.4995138473939207,
      "tokens_seen": 1739784192
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047771313941825477,
      "loss": 1.4164,
      "theoretical_loss": 3.4995026448125284,
      "tokens_seen": 1739849728
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047769307923771313,
      "loss": 1.3588,
      "theoretical_loss": 3.49949144277125,
      "tokens_seen": 1739915264
    },
    {
      "epoch": 0.53,
      "objective/train/advantage_avg": 0.49501553177833557,
      "objective/train/docs_used": 635535,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3129806518554688,
      "objective/train/original_loss": 1.3129806518554688,
      "objective/train/theoretical_loss": 3.49948024127004,
      "objective/train/tokens_used": 1760440800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24523977935314178,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050748348236084,
      "objective/train/weighted_lm_loss": 1.379601240158081,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 3.49948024127004,
      "tokens_seen": 1739980800
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047767301905717155,
      "loss": 1.374,
      "theoretical_loss": 3.49948024127004,
      "tokens_seen": 1739980800
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004776529588766299,
      "loss": 1.4295,
      "theoretical_loss": 3.4994690403088518,
      "tokens_seen": 1740046336
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004776328986960883,
      "loss": 1.3822,
      "theoretical_loss": 3.499457839887639,
      "tokens_seen": 1740111872
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047761283851554664,
      "loss": 1.4385,
      "theoretical_loss": 3.4994466400063544,
      "tokens_seen": 1740177408
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047759277833500505,
      "loss": 1.3912,
      "theoretical_loss": 3.4994354406649535,
      "tokens_seen": 1740242944
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004775727181544634,
      "loss": 1.4385,
      "theoretical_loss": 3.499424241863388,
      "tokens_seen": 1740308480
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004775526579739218,
      "loss": 1.3298,
      "theoretical_loss": 3.4994130436016126,
      "tokens_seen": 1740374016
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047753259779338014,
      "loss": 1.3834,
      "theoretical_loss": 3.499401845879581,
      "tokens_seen": 1740439552
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004775125376128385,
      "loss": 1.5109,
      "theoretical_loss": 3.499390648697246,
      "tokens_seen": 1740505088
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004774924774322969,
      "loss": 1.4313,
      "theoretical_loss": 3.4993794520545625,
      "tokens_seen": 1740570624
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004774724172517553,
      "loss": 1.417,
      "theoretical_loss": 3.4993682559514836,
      "tokens_seen": 1740636160
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047745235707121364,
      "loss": 1.4225,
      "theoretical_loss": 3.4993570603879625,
      "tokens_seen": 1740701696
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000477432296890672,
      "loss": 1.4633,
      "theoretical_loss": 3.4993458653639538,
      "tokens_seen": 1740767232
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004774122367101304,
      "loss": 1.3666,
      "theoretical_loss": 3.49933467087941,
      "tokens_seen": 1740832768
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004773921765295888,
      "loss": 1.456,
      "theoretical_loss": 3.4993234769342862,
      "tokens_seen": 1740898304
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047737211634904715,
      "loss": 1.3917,
      "theoretical_loss": 3.4993122835285346,
      "tokens_seen": 1740963840
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004773520561685055,
      "loss": 1.3723,
      "theoretical_loss": 3.4993010906621107,
      "tokens_seen": 1741029376
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047733199598796387,
      "loss": 1.3755,
      "theoretical_loss": 3.4992898983349665,
      "tokens_seen": 1741094912
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004773119358074223,
      "loss": 1.4409,
      "theoretical_loss": 3.4992787065470567,
      "tokens_seen": 1741160448
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047729187562688065,
      "loss": 1.296,
      "theoretical_loss": 3.499267515298335,
      "tokens_seen": 1741225984
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000477271815446339,
      "loss": 1.474,
      "theoretical_loss": 3.499256324588755,
      "tokens_seen": 1741291520
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004772517552657974,
      "loss": 1.3656,
      "theoretical_loss": 3.4992451344182696,
      "tokens_seen": 1741357056
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004772316950852558,
      "loss": 1.4674,
      "theoretical_loss": 3.499233944786834,
      "tokens_seen": 1741422592
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047721163490471415,
      "loss": 1.2863,
      "theoretical_loss": 3.499222755694401,
      "tokens_seen": 1741488128
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004771915747241725,
      "loss": 1.4963,
      "theoretical_loss": 3.4992115671409243,
      "tokens_seen": 1741553664
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004771715145436309,
      "loss": 1.3721,
      "theoretical_loss": 3.499200379126358,
      "tokens_seen": 1741619200
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047715145436308924,
      "loss": 1.4822,
      "theoretical_loss": 3.499189191650656,
      "tokens_seen": 1741684736
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047713139418254766,
      "loss": 1.4564,
      "theoretical_loss": 3.499178004713772,
      "tokens_seen": 1741750272
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000477111334002006,
      "loss": 1.3576,
      "theoretical_loss": 3.4991668183156595,
      "tokens_seen": 1741815808
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004770912738214644,
      "loss": 1.385,
      "theoretical_loss": 3.4991556324562723,
      "tokens_seen": 1741881344
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047707121364092274,
      "loss": 1.47,
      "theoretical_loss": 3.499144447135565,
      "tokens_seen": 1741946880
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047705115346038116,
      "loss": 1.4619,
      "theoretical_loss": 3.4991332623534896,
      "tokens_seen": 1742012416
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004770310932798395,
      "loss": 1.5953,
      "theoretical_loss": 3.499122078110002,
      "tokens_seen": 1742077952
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004770110330992979,
      "loss": 1.4248,
      "theoretical_loss": 3.499110894405055,
      "tokens_seen": 1742143488
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047699097291875625,
      "loss": 1.4412,
      "theoretical_loss": 3.499099711238602,
      "tokens_seen": 1742209024
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047697091273821466,
      "loss": 1.418,
      "theoretical_loss": 3.4990885286105975,
      "tokens_seen": 1742274560
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000476950852557673,
      "loss": 1.3347,
      "theoretical_loss": 3.499077346520995,
      "tokens_seen": 1742340096
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004769307923771314,
      "loss": 1.4649,
      "theoretical_loss": 3.499066164969748,
      "tokens_seen": 1742405632
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047691073219658975,
      "loss": 1.3787,
      "theoretical_loss": 3.4990549839568112,
      "tokens_seen": 1742471168
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004768906720160481,
      "loss": 1.4026,
      "theoretical_loss": 3.499043803482138,
      "tokens_seen": 1742536704
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047687061183550653,
      "loss": 1.4178,
      "theoretical_loss": 3.4990326235456823,
      "tokens_seen": 1742602240
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004768505516549649,
      "loss": 1.6135,
      "theoretical_loss": 3.4990214441473975,
      "tokens_seen": 1742667776
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047683049147442325,
      "loss": 1.4084,
      "theoretical_loss": 3.4990102652872377,
      "tokens_seen": 1742733312
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004768104312938816,
      "loss": 1.5312,
      "theoretical_loss": 3.4989990869651573,
      "tokens_seen": 1742798848
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047679037111334003,
      "loss": 1.4594,
      "theoretical_loss": 3.4989879091811096,
      "tokens_seen": 1742864384
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004767703109327984,
      "loss": 1.5007,
      "theoretical_loss": 3.498976731935049,
      "tokens_seen": 1742929920
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047675025075225676,
      "loss": 1.4016,
      "theoretical_loss": 3.4989655552269285,
      "tokens_seen": 1742995456
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004767301905717151,
      "loss": 1.4836,
      "theoretical_loss": 3.498954379056703,
      "tokens_seen": 1743060992
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047671013039117353,
      "loss": 1.4085,
      "theoretical_loss": 3.498943203424326,
      "tokens_seen": 1743126528
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047669007021063195,
      "loss": 1.5632,
      "theoretical_loss": 3.4989320283297505,
      "tokens_seen": 1743192064
    },
    {
      "epoch": 0.53,
      "objective/train/advantage_avg": 0.49678072333335876,
      "objective/train/docs_used": 636817,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3850852251052856,
      "objective/train/original_loss": 1.385085105895996,
      "objective/train/theoretical_loss": 3.4989208537729315,
      "objective/train/tokens_used": 1763717600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24693624675273895,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0509334802627563,
      "objective/train/weighted_lm_loss": 1.455576777458191,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4989208537729315,
      "tokens_seen": 1743257600
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004766700100300903,
      "loss": 1.4383,
      "theoretical_loss": 3.4989208537729315,
      "tokens_seen": 1743257600
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004766499498495487,
      "loss": 1.4605,
      "theoretical_loss": 3.498909679753823,
      "tokens_seen": 1743323136
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047662988966900704,
      "loss": 1.5183,
      "theoretical_loss": 3.4988985062723783,
      "tokens_seen": 1743388672
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047660982948846545,
      "loss": 1.4269,
      "theoretical_loss": 3.4988873333285513,
      "tokens_seen": 1743454208
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004765897693079238,
      "loss": 1.4222,
      "theoretical_loss": 3.498876160922296,
      "tokens_seen": 1743519744
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004765697091273822,
      "loss": 1.406,
      "theoretical_loss": 3.4988649890535672,
      "tokens_seen": 1743585280
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047654964894684054,
      "loss": 1.4653,
      "theoretical_loss": 3.498853817722318,
      "tokens_seen": 1743650816
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004765295887662989,
      "loss": 1.4579,
      "theoretical_loss": 3.4988426469285017,
      "tokens_seen": 1743716352
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004765095285857573,
      "loss": 1.4447,
      "theoretical_loss": 3.4988314766720734,
      "tokens_seen": 1743781888
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004764894684052157,
      "loss": 1.3005,
      "theoretical_loss": 3.498820306952987,
      "tokens_seen": 1743847424
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047646940822467404,
      "loss": 1.5676,
      "theoretical_loss": 3.498809137771196,
      "tokens_seen": 1743912960
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004764493480441324,
      "loss": 1.4861,
      "theoretical_loss": 3.498797969126654,
      "tokens_seen": 1743978496
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004764292878635908,
      "loss": 1.4899,
      "theoretical_loss": 3.4987868010193157,
      "tokens_seen": 1744044032
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004764092276830492,
      "loss": 1.4341,
      "theoretical_loss": 3.498775633449135,
      "tokens_seen": 1744109568
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047638916750250755,
      "loss": 1.3299,
      "theoretical_loss": 3.4987644664160658,
      "tokens_seen": 1744175104
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004763691073219659,
      "loss": 1.3036,
      "theoretical_loss": 3.498753299920062,
      "tokens_seen": 1744240640
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047634904714142427,
      "loss": 1.3108,
      "theoretical_loss": 3.4987421339610774,
      "tokens_seen": 1744306176
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004763289869608827,
      "loss": 1.4281,
      "theoretical_loss": 3.498730968539066,
      "tokens_seen": 1744371712
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047630892678034105,
      "loss": 1.4249,
      "theoretical_loss": 3.4987198036539824,
      "tokens_seen": 1744437248
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004762888665997994,
      "loss": 1.3837,
      "theoretical_loss": 3.49870863930578,
      "tokens_seen": 1744502784
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004762688064192578,
      "loss": 1.4296,
      "theoretical_loss": 3.498697475494413,
      "tokens_seen": 1744568320
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004762487462387162,
      "loss": 1.3915,
      "theoretical_loss": 3.498686312219836,
      "tokens_seen": 1744633856
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047622868605817455,
      "loss": 1.4719,
      "theoretical_loss": 3.4986751494820014,
      "tokens_seen": 1744699392
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004762086258776329,
      "loss": 1.5206,
      "theoretical_loss": 3.498663987280865,
      "tokens_seen": 1744764928
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004761885656970913,
      "loss": 1.3718,
      "theoretical_loss": 3.49865282561638,
      "tokens_seen": 1744830464
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047616850551654964,
      "loss": 1.4373,
      "theoretical_loss": 3.498641664488501,
      "tokens_seen": 1744896000
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047614844533600806,
      "loss": 1.6221,
      "theoretical_loss": 3.498630503897181,
      "tokens_seen": 1744961536
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004761283851554664,
      "loss": 1.3519,
      "theoretical_loss": 3.4986193438423747,
      "tokens_seen": 1745027072
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004761083249749248,
      "loss": 1.406,
      "theoretical_loss": 3.498608184324037,
      "tokens_seen": 1745092608
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047608826479438314,
      "loss": 1.359,
      "theoretical_loss": 3.4985970253421206,
      "tokens_seen": 1745158144
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047606820461384156,
      "loss": 1.541,
      "theoretical_loss": 3.4985858668965797,
      "tokens_seen": 1745223680
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004760481444332999,
      "loss": 1.4408,
      "theoretical_loss": 3.498574708987369,
      "tokens_seen": 1745289216
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004760280842527583,
      "loss": 1.4593,
      "theoretical_loss": 3.498563551614443,
      "tokens_seen": 1745354752
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047600802407221665,
      "loss": 1.4604,
      "theoretical_loss": 3.498552394777754,
      "tokens_seen": 1745420288
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047598796389167506,
      "loss": 1.3203,
      "theoretical_loss": 3.498541238477258,
      "tokens_seen": 1745485824
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004759679037111334,
      "loss": 1.5424,
      "theoretical_loss": 3.498530082712908,
      "tokens_seen": 1745551360
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004759478435305918,
      "loss": 1.3751,
      "theoretical_loss": 3.4985189274846586,
      "tokens_seen": 1745616896
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047592778335005015,
      "loss": 1.5033,
      "theoretical_loss": 3.498507772792464,
      "tokens_seen": 1745682432
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004759077231695085,
      "loss": 1.4632,
      "theoretical_loss": 3.4984966186362776,
      "tokens_seen": 1745747968
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047588766298896693,
      "loss": 1.481,
      "theoretical_loss": 3.498485465016054,
      "tokens_seen": 1745813504
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004758676028084253,
      "loss": 1.4673,
      "theoretical_loss": 3.498474311931748,
      "tokens_seen": 1745879040
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047584754262788365,
      "loss": 1.4695,
      "theoretical_loss": 3.498463159383312,
      "tokens_seen": 1745944576
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000475827482447342,
      "loss": 1.4061,
      "theoretical_loss": 3.498452007370702,
      "tokens_seen": 1746010112
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047580742226680043,
      "loss": 1.5522,
      "theoretical_loss": 3.498440855893871,
      "tokens_seen": 1746075648
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004757873620862588,
      "loss": 1.4521,
      "theoretical_loss": 3.4984297049527737,
      "tokens_seen": 1746141184
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047576730190571716,
      "loss": 1.4822,
      "theoretical_loss": 3.498418554547364,
      "tokens_seen": 1746206720
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004757472417251755,
      "loss": 1.4133,
      "theoretical_loss": 3.498407404677596,
      "tokens_seen": 1746272256
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004757271815446339,
      "loss": 1.5241,
      "theoretical_loss": 3.4983962553434242,
      "tokens_seen": 1746337792
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004757071213640923,
      "loss": 1.358,
      "theoretical_loss": 3.498385106544802,
      "tokens_seen": 1746403328
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047568706118355066,
      "loss": 1.2773,
      "theoretical_loss": 3.498373958281685,
      "tokens_seen": 1746468864
    },
    {
      "epoch": 0.53,
      "objective/train/advantage_avg": 0.49545732140541077,
      "objective/train/docs_used": 637932,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4124590158462524,
      "objective/train/original_loss": 1.412458896636963,
      "objective/train/theoretical_loss": 3.498362810554026,
      "objective/train/tokens_used": 1766994400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2457018792629242,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507949590682983,
      "objective/train/weighted_lm_loss": 1.484118938446045,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.498362810554026,
      "tokens_seen": 1746534400
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000475667001003009,
      "loss": 1.3852,
      "theoretical_loss": 3.498362810554026,
      "tokens_seen": 1746534400
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004756469408224674,
      "loss": 1.4955,
      "theoretical_loss": 3.4983516633617797,
      "tokens_seen": 1746599936
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004756268806419258,
      "loss": 1.3769,
      "theoretical_loss": 3.4983405167049004,
      "tokens_seen": 1746665472
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047560682046138416,
      "loss": 1.4656,
      "theoretical_loss": 3.4983293705833423,
      "tokens_seen": 1746731008
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004755867602808425,
      "loss": 1.3941,
      "theoretical_loss": 3.4983182249970595,
      "tokens_seen": 1746796544
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004755667001003009,
      "loss": 1.4333,
      "theoretical_loss": 3.498307079946006,
      "tokens_seen": 1746862080
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047554663991975925,
      "loss": 1.4532,
      "theoretical_loss": 3.4982959354301366,
      "tokens_seen": 1746927616
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047552657973921766,
      "loss": 1.351,
      "theoretical_loss": 3.4982847914494046,
      "tokens_seen": 1746993152
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047550651955867603,
      "loss": 1.3598,
      "theoretical_loss": 3.4982736480037655,
      "tokens_seen": 1747058688
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004754864593781344,
      "loss": 1.2945,
      "theoretical_loss": 3.498262505093172,
      "tokens_seen": 1747124224
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047546639919759275,
      "loss": 1.3811,
      "theoretical_loss": 3.49825136271758,
      "tokens_seen": 1747189760
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047544633901705117,
      "loss": 1.3927,
      "theoretical_loss": 3.4982402208769425,
      "tokens_seen": 1747255296
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047542627883650953,
      "loss": 1.4874,
      "theoretical_loss": 3.4982290795712148,
      "tokens_seen": 1747320832
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004754062186559679,
      "loss": 1.4025,
      "theoretical_loss": 3.49821793880035,
      "tokens_seen": 1747386368
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047538615847542625,
      "loss": 1.562,
      "theoretical_loss": 3.498206798564303,
      "tokens_seen": 1747451904
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004753660982948846,
      "loss": 1.4412,
      "theoretical_loss": 3.4981956588630276,
      "tokens_seen": 1747517440
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047534603811434303,
      "loss": 1.5521,
      "theoretical_loss": 3.498184519696479,
      "tokens_seen": 1747582976
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004753259779338014,
      "loss": 1.3141,
      "theoretical_loss": 3.4981733810646105,
      "tokens_seen": 1747648512
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047530591775325976,
      "loss": 1.392,
      "theoretical_loss": 3.498162242967377,
      "tokens_seen": 1747714048
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004752858575727181,
      "loss": 1.4631,
      "theoretical_loss": 3.498151105404733,
      "tokens_seen": 1747779584
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047526579739217654,
      "loss": 1.386,
      "theoretical_loss": 3.498139968376632,
      "tokens_seen": 1747845120
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004752457372116349,
      "loss": 1.4252,
      "theoretical_loss": 3.4981288318830286,
      "tokens_seen": 1747910656
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047522567703109326,
      "loss": 1.5116,
      "theoretical_loss": 3.4981176959238773,
      "tokens_seen": 1747976192
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004752056168505517,
      "loss": 1.3285,
      "theoretical_loss": 3.4981065604991324,
      "tokens_seen": 1748041728
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047518555667001004,
      "loss": 1.5125,
      "theoretical_loss": 3.4980954256087484,
      "tokens_seen": 1748107264
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047516549648946846,
      "loss": 1.454,
      "theoretical_loss": 3.498084291252679,
      "tokens_seen": 1748172800
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004751454363089268,
      "loss": 1.3172,
      "theoretical_loss": 3.4980731574308788,
      "tokens_seen": 1748238336
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004751253761283852,
      "loss": 1.4176,
      "theoretical_loss": 3.4980620241433025,
      "tokens_seen": 1748303872
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047510531594784354,
      "loss": 1.3743,
      "theoretical_loss": 3.498050891389904,
      "tokens_seen": 1748369408
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047508525576730196,
      "loss": 1.2444,
      "theoretical_loss": 3.498039759170638,
      "tokens_seen": 1748434944
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004750651955867603,
      "loss": 1.399,
      "theoretical_loss": 3.498028627485459,
      "tokens_seen": 1748500480
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004750451354062187,
      "loss": 1.3285,
      "theoretical_loss": 3.4980174963343202,
      "tokens_seen": 1748566016
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047502507522567705,
      "loss": 1.4013,
      "theoretical_loss": 3.4980063657171776,
      "tokens_seen": 1748631552
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047500501504513546,
      "loss": 1.4476,
      "theoretical_loss": 3.4979952356339847,
      "tokens_seen": 1748697088
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004749849548645938,
      "loss": 1.5078,
      "theoretical_loss": 3.497984106084696,
      "tokens_seen": 1748762624
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004749648946840522,
      "loss": 1.4015,
      "theoretical_loss": 3.4979729770692654,
      "tokens_seen": 1748828160
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047494483450351055,
      "loss": 1.5007,
      "theoretical_loss": 3.4979618485876482,
      "tokens_seen": 1748893696
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004749247743229689,
      "loss": 1.5076,
      "theoretical_loss": 3.497950720639798,
      "tokens_seen": 1748959232
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047490471414242733,
      "loss": 1.5163,
      "theoretical_loss": 3.4979395932256696,
      "tokens_seen": 1749024768
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004748846539618857,
      "loss": 1.4352,
      "theoretical_loss": 3.4979284663452175,
      "tokens_seen": 1749090304
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047486459378134405,
      "loss": 1.425,
      "theoretical_loss": 3.497917339998396,
      "tokens_seen": 1749155840
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004748445336008024,
      "loss": 1.3482,
      "theoretical_loss": 3.497906214185159,
      "tokens_seen": 1749221376
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047482447342026083,
      "loss": 1.4009,
      "theoretical_loss": 3.4978950889054614,
      "tokens_seen": 1749286912
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004748044132397192,
      "loss": 1.3661,
      "theoretical_loss": 3.497883964159258,
      "tokens_seen": 1749352448
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047478435305917756,
      "loss": 1.383,
      "theoretical_loss": 3.4978728399465027,
      "tokens_seen": 1749417984
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004747642928786359,
      "loss": 1.4536,
      "theoretical_loss": 3.49786171626715,
      "tokens_seen": 1749483520
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004747442326980943,
      "loss": 1.4755,
      "theoretical_loss": 3.4978505931211545,
      "tokens_seen": 1749549056
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004747241725175527,
      "loss": 1.4268,
      "theoretical_loss": 3.4978394705084703,
      "tokens_seen": 1749614592
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047470411233701106,
      "loss": 1.5151,
      "theoretical_loss": 3.497828348429053,
      "tokens_seen": 1749680128
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004746840521564694,
      "loss": 1.3435,
      "theoretical_loss": 3.497817226882855,
      "tokens_seen": 1749745664
    },
    {
      "epoch": 0.53,
      "objective/train/advantage_avg": 0.4915338456630707,
      "objective/train/docs_used": 639149,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3481122255325317,
      "objective/train/original_loss": 1.3481122255325317,
      "objective/train/theoretical_loss": 3.4978061058698326,
      "objective/train/tokens_used": 1770271200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2421993911266327,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503846406936646,
      "objective/train/weighted_lm_loss": 1.4161285161972046,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4978061058698326,
      "tokens_seen": 1749811200
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004746639919759278,
      "loss": 1.4579,
      "theoretical_loss": 3.4978061058698326,
      "tokens_seen": 1749811200
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004746439317953862,
      "loss": 1.399,
      "theoretical_loss": 3.4977949853899393,
      "tokens_seen": 1749876736
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047462387161484456,
      "loss": 1.4611,
      "theoretical_loss": 3.49778386544313,
      "tokens_seen": 1749942272
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004746038114343029,
      "loss": 1.5121,
      "theoretical_loss": 3.4977727460293595,
      "tokens_seen": 1750007808
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004745837512537613,
      "loss": 1.5872,
      "theoretical_loss": 3.4977616271485816,
      "tokens_seen": 1750073344
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047456369107321965,
      "loss": 1.4976,
      "theoretical_loss": 3.497750508800751,
      "tokens_seen": 1750138880
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047454363089267807,
      "loss": 1.4479,
      "theoretical_loss": 3.497739390985822,
      "tokens_seen": 1750204416
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047452357071213643,
      "loss": 1.5364,
      "theoretical_loss": 3.49772827370375,
      "tokens_seen": 1750269952
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004745035105315948,
      "loss": 1.2249,
      "theoretical_loss": 3.4977171569544883,
      "tokens_seen": 1750335488
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047448345035105315,
      "loss": 1.3389,
      "theoretical_loss": 3.497706040737992,
      "tokens_seen": 1750401024
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047446339017051157,
      "loss": 1.4018,
      "theoretical_loss": 3.497694925054216,
      "tokens_seen": 1750466560
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047444332998996993,
      "loss": 1.3638,
      "theoretical_loss": 3.4976838099031142,
      "tokens_seen": 1750532096
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004744232698094283,
      "loss": 1.4563,
      "theoretical_loss": 3.497672695284641,
      "tokens_seen": 1750597632
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047440320962888666,
      "loss": 1.437,
      "theoretical_loss": 3.497661581198752,
      "tokens_seen": 1750663168
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000474383149448345,
      "loss": 1.3788,
      "theoretical_loss": 3.4976504676454008,
      "tokens_seen": 1750728704
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047436308926780343,
      "loss": 1.5196,
      "theoretical_loss": 3.4976393546245426,
      "tokens_seen": 1750794240
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004743430290872618,
      "loss": 1.341,
      "theoretical_loss": 3.497628242136131,
      "tokens_seen": 1750859776
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047432296890672016,
      "loss": 1.3186,
      "theoretical_loss": 3.497617130180121,
      "tokens_seen": 1750925312
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004743029087261785,
      "loss": 1.4427,
      "theoretical_loss": 3.4976060187564677,
      "tokens_seen": 1750990848
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047428284854563694,
      "loss": 1.376,
      "theoretical_loss": 3.497594907865125,
      "tokens_seen": 1751056384
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004742627883650953,
      "loss": 1.4619,
      "theoretical_loss": 3.497583797506048,
      "tokens_seen": 1751121920
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047424272818455366,
      "loss": 1.4647,
      "theoretical_loss": 3.497572687679191,
      "tokens_seen": 1751187456
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000474222668004012,
      "loss": 1.3781,
      "theoretical_loss": 3.4975615783845084,
      "tokens_seen": 1751252992
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004742026078234704,
      "loss": 1.4512,
      "theoretical_loss": 3.497550469621955,
      "tokens_seen": 1751318528
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004741825476429288,
      "loss": 1.5216,
      "theoretical_loss": 3.4975393613914854,
      "tokens_seen": 1751384064
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047416248746238716,
      "loss": 1.4418,
      "theoretical_loss": 3.4975282536930545,
      "tokens_seen": 1751449600
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004741424272818455,
      "loss": 1.4189,
      "theoretical_loss": 3.497517146526617,
      "tokens_seen": 1751515136
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004741223671013039,
      "loss": 1.4485,
      "theoretical_loss": 3.4975060398921265,
      "tokens_seen": 1751580672
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004741023069207623,
      "loss": 1.4175,
      "theoretical_loss": 3.497494933789538,
      "tokens_seen": 1751646208
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047408224674022067,
      "loss": 1.4987,
      "theoretical_loss": 3.4974838282188068,
      "tokens_seen": 1751711744
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047406218655967903,
      "loss": 1.4266,
      "theoretical_loss": 3.4974727231798868,
      "tokens_seen": 1751777280
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004740421263791374,
      "loss": 1.5751,
      "theoretical_loss": 3.4974616186727334,
      "tokens_seen": 1751842816
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047402206619859575,
      "loss": 1.3883,
      "theoretical_loss": 3.4974505146973005,
      "tokens_seen": 1751908352
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047400200601805417,
      "loss": 1.461,
      "theoretical_loss": 3.497439411253543,
      "tokens_seen": 1751973888
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047398194583751253,
      "loss": 1.5315,
      "theoretical_loss": 3.497428308341416,
      "tokens_seen": 1752039424
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004739618856569709,
      "loss": 1.4702,
      "theoretical_loss": 3.4974172059608737,
      "tokens_seen": 1752104960
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047394182547642926,
      "loss": 1.5046,
      "theoretical_loss": 3.497406104111871,
      "tokens_seen": 1752170496
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004739217652958877,
      "loss": 1.4895,
      "theoretical_loss": 3.497395002794362,
      "tokens_seen": 1752236032
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047390170511534604,
      "loss": 1.5828,
      "theoretical_loss": 3.497383902008302,
      "tokens_seen": 1752301568
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004738816449348044,
      "loss": 1.3858,
      "theoretical_loss": 3.497372801753645,
      "tokens_seen": 1752367104
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047386158475426276,
      "loss": 1.4255,
      "theoretical_loss": 3.4973617020303465,
      "tokens_seen": 1752432640
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004738415245737211,
      "loss": 1.435,
      "theoretical_loss": 3.497350602838361,
      "tokens_seen": 1752498176
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047382146439317954,
      "loss": 1.4291,
      "theoretical_loss": 3.497339504177643,
      "tokens_seen": 1752563712
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004738014042126379,
      "loss": 1.5687,
      "theoretical_loss": 3.4973284060481475,
      "tokens_seen": 1752629248
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047378134403209626,
      "loss": 1.4117,
      "theoretical_loss": 3.497317308449828,
      "tokens_seen": 1752694784
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004737612838515546,
      "loss": 1.3499,
      "theoretical_loss": 3.497306211382641,
      "tokens_seen": 1752760320
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047374122367101304,
      "loss": 1.3041,
      "theoretical_loss": 3.4972951148465405,
      "tokens_seen": 1752825856
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047372116349047146,
      "loss": 1.4058,
      "theoretical_loss": 3.497284018841481,
      "tokens_seen": 1752891392
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004737011033099298,
      "loss": 1.3829,
      "theoretical_loss": 3.4972729233674174,
      "tokens_seen": 1752956928
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004736810431293882,
      "loss": 1.3878,
      "theoretical_loss": 3.497261828424304,
      "tokens_seen": 1753022464
    },
    {
      "epoch": 0.53,
      "objective/train/advantage_avg": 0.49423301219940186,
      "objective/train/docs_used": 640338,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3288896083831787,
      "objective/train/original_loss": 1.3288898468017578,
      "objective/train/theoretical_loss": 3.4972507340120966,
      "objective/train/tokens_used": 1773548000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24473364651203156,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506675243377686,
      "objective/train/weighted_lm_loss": 1.396265983581543,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.4972507340120966,
      "tokens_seen": 1753088000
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004736609829488466,
      "loss": 1.4012,
      "theoretical_loss": 3.4972507340120966,
      "tokens_seen": 1753088000
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047364092276830496,
      "loss": 1.4086,
      "theoretical_loss": 3.497239640130749,
      "tokens_seen": 1753153536
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004736208625877633,
      "loss": 1.4462,
      "theoretical_loss": 3.4972285467802164,
      "tokens_seen": 1753219072
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004736008024072217,
      "loss": 1.4663,
      "theoretical_loss": 3.4972174539604532,
      "tokens_seen": 1753284608
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047358074222668005,
      "loss": 1.467,
      "theoretical_loss": 3.497206361671414,
      "tokens_seen": 1753350144
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047356068204613847,
      "loss": 1.4076,
      "theoretical_loss": 3.4971952699130546,
      "tokens_seen": 1753415680
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047354062186559683,
      "loss": 1.5304,
      "theoretical_loss": 3.4971841786853295,
      "tokens_seen": 1753481216
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004735205616850552,
      "loss": 1.4637,
      "theoretical_loss": 3.497173087988192,
      "tokens_seen": 1753546752
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047350050150451355,
      "loss": 1.4344,
      "theoretical_loss": 3.497161997821599,
      "tokens_seen": 1753612288
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047348044132397197,
      "loss": 1.4955,
      "theoretical_loss": 3.4971509081855037,
      "tokens_seen": 1753677824
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047346038114343033,
      "loss": 1.4598,
      "theoretical_loss": 3.4971398190798615,
      "tokens_seen": 1753743360
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004734403209628887,
      "loss": 1.4941,
      "theoretical_loss": 3.4971287305046275,
      "tokens_seen": 1753808896
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047342026078234706,
      "loss": 1.3917,
      "theoretical_loss": 3.4971176424597563,
      "tokens_seen": 1753874432
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004734002006018054,
      "loss": 1.3792,
      "theoretical_loss": 3.4971065549452023,
      "tokens_seen": 1753939968
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047338014042126383,
      "loss": 1.4299,
      "theoretical_loss": 3.4970954679609214,
      "tokens_seen": 1754005504
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004733600802407222,
      "loss": 1.4584,
      "theoretical_loss": 3.497084381506867,
      "tokens_seen": 1754071040
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047334002006018056,
      "loss": 1.5587,
      "theoretical_loss": 3.497073295582995,
      "tokens_seen": 1754136576
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004733199598796389,
      "loss": 1.6032,
      "theoretical_loss": 3.4970622101892594,
      "tokens_seen": 1754202112
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047329989969909734,
      "loss": 1.3842,
      "theoretical_loss": 3.4970511253256156,
      "tokens_seen": 1754267648
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004732798395185557,
      "loss": 1.507,
      "theoretical_loss": 3.497040040992019,
      "tokens_seen": 1754333184
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047325977933801406,
      "loss": 1.5107,
      "theoretical_loss": 3.497028957188423,
      "tokens_seen": 1754398720
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004732397191574724,
      "loss": 1.4833,
      "theoretical_loss": 3.4970178739147837,
      "tokens_seen": 1754464256
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004732196589769308,
      "loss": 1.455,
      "theoretical_loss": 3.4970067911710556,
      "tokens_seen": 1754529792
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004731995987963892,
      "loss": 1.5494,
      "theoretical_loss": 3.496995708957193,
      "tokens_seen": 1754595328
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047317953861584757,
      "loss": 1.5096,
      "theoretical_loss": 3.496984627273152,
      "tokens_seen": 1754660864
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047315947843530593,
      "loss": 1.4945,
      "theoretical_loss": 3.496973546118886,
      "tokens_seen": 1754726400
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004731394182547643,
      "loss": 1.4843,
      "theoretical_loss": 3.4969624654943514,
      "tokens_seen": 1754791936
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004731193580742227,
      "loss": 1.4687,
      "theoretical_loss": 3.496951385399502,
      "tokens_seen": 1754857472
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047309929789368107,
      "loss": 1.4129,
      "theoretical_loss": 3.4969403058342925,
      "tokens_seen": 1754923008
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047307923771313943,
      "loss": 1.4316,
      "theoretical_loss": 3.496929226798679,
      "tokens_seen": 1754988544
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004730591775325978,
      "loss": 1.4567,
      "theoretical_loss": 3.496918148292616,
      "tokens_seen": 1755054080
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047303911735205615,
      "loss": 1.3952,
      "theoretical_loss": 3.496907070316058,
      "tokens_seen": 1755119616
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047301905717151457,
      "loss": 1.528,
      "theoretical_loss": 3.49689599286896,
      "tokens_seen": 1755185152
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047299899699097293,
      "loss": 1.4615,
      "theoretical_loss": 3.496884915951277,
      "tokens_seen": 1755250688
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004729789368104313,
      "loss": 1.4382,
      "theoretical_loss": 3.496873839562964,
      "tokens_seen": 1755316224
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047295887662988966,
      "loss": 1.4431,
      "theoretical_loss": 3.4968627637039758,
      "tokens_seen": 1755381760
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004729388164493481,
      "loss": 1.4665,
      "theoretical_loss": 3.4968516883742673,
      "tokens_seen": 1755447296
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047291875626880644,
      "loss": 1.4307,
      "theoretical_loss": 3.496840613573794,
      "tokens_seen": 1755512832
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004728986960882648,
      "loss": 1.3908,
      "theoretical_loss": 3.49682953930251,
      "tokens_seen": 1755578368
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047287863590772316,
      "loss": 1.4672,
      "theoretical_loss": 3.4968184655603713,
      "tokens_seen": 1755643904
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004728585757271815,
      "loss": 1.4347,
      "theoretical_loss": 3.496807392347332,
      "tokens_seen": 1755709440
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047283851554663994,
      "loss": 1.4351,
      "theoretical_loss": 3.4967963196633476,
      "tokens_seen": 1755774976
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004728184553660983,
      "loss": 1.345,
      "theoretical_loss": 3.4967852475083725,
      "tokens_seen": 1755840512
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047279839518555666,
      "loss": 1.4467,
      "theoretical_loss": 3.4967741758823623,
      "tokens_seen": 1755906048
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000472778335005015,
      "loss": 1.4025,
      "theoretical_loss": 3.496763104785271,
      "tokens_seen": 1755971584
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047275827482447344,
      "loss": 1.4642,
      "theoretical_loss": 3.496752034217055,
      "tokens_seen": 1756037120
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004727382146439318,
      "loss": 1.487,
      "theoretical_loss": 3.4967409641776683,
      "tokens_seen": 1756102656
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047271815446339017,
      "loss": 1.3436,
      "theoretical_loss": 3.4967298946670664,
      "tokens_seen": 1756168192
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047269809428284853,
      "loss": 1.4331,
      "theoretical_loss": 3.496718825685204,
      "tokens_seen": 1756233728
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047267803410230695,
      "loss": 1.472,
      "theoretical_loss": 3.496707757232036,
      "tokens_seen": 1756299264
    },
    {
      "epoch": 0.53,
      "objective/train/advantage_avg": 0.4940010607242584,
      "objective/train/docs_used": 641471,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3080871105194092,
      "objective/train/original_loss": 1.30808687210083,
      "objective/train/theoretical_loss": 3.4966966893075178,
      "objective/train/tokens_used": 1776824800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24430374801158905,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506421327590942,
      "objective/train/weighted_lm_loss": 1.3743795156478882,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.4966966893075178,
      "tokens_seen": 1756364800
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004726579739217653,
      "loss": 1.3372,
      "theoretical_loss": 3.4966966893075178,
      "tokens_seen": 1756364800
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047263791374122367,
      "loss": 1.4536,
      "theoretical_loss": 3.4966856219116043,
      "tokens_seen": 1756430336
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047261785356068203,
      "loss": 1.5869,
      "theoretical_loss": 3.4966745550442506,
      "tokens_seen": 1756495872
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004725977933801404,
      "loss": 1.3397,
      "theoretical_loss": 3.4966634887054115,
      "tokens_seen": 1756561408
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004725777331995988,
      "loss": 1.4741,
      "theoretical_loss": 3.496652422895042,
      "tokens_seen": 1756626944
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004725576730190572,
      "loss": 1.4281,
      "theoretical_loss": 3.4966413576130977,
      "tokens_seen": 1756692480
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047253761283851554,
      "loss": 1.3755,
      "theoretical_loss": 3.496630292859533,
      "tokens_seen": 1756758016
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004725175526579739,
      "loss": 1.4521,
      "theoretical_loss": 3.4966192286343034,
      "tokens_seen": 1756823552
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004724974924774323,
      "loss": 1.4656,
      "theoretical_loss": 3.496608164937364,
      "tokens_seen": 1756889088
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004724774322968907,
      "loss": 1.4736,
      "theoretical_loss": 3.4965971017686694,
      "tokens_seen": 1756954624
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047245737211634904,
      "loss": 1.5682,
      "theoretical_loss": 3.496586039128175,
      "tokens_seen": 1757020160
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004724373119358074,
      "loss": 1.4298,
      "theoretical_loss": 3.496574977015836,
      "tokens_seen": 1757085696
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047241725175526576,
      "loss": 1.386,
      "theoretical_loss": 3.496563915431607,
      "tokens_seen": 1757151232
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004723971915747242,
      "loss": 1.4495,
      "theoretical_loss": 3.4965528543754436,
      "tokens_seen": 1757216768
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047237713139418254,
      "loss": 1.4675,
      "theoretical_loss": 3.4965417938473005,
      "tokens_seen": 1757282304
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004723570712136409,
      "loss": 1.576,
      "theoretical_loss": 3.496530733847133,
      "tokens_seen": 1757347840
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047233701103309927,
      "loss": 1.3759,
      "theoretical_loss": 3.4965196743748965,
      "tokens_seen": 1757413376
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004723169508525577,
      "loss": 1.5296,
      "theoretical_loss": 3.496508615430546,
      "tokens_seen": 1757478912
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047229689067201605,
      "loss": 1.4203,
      "theoretical_loss": 3.4964975570140364,
      "tokens_seen": 1757544448
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004722768304914744,
      "loss": 1.4814,
      "theoretical_loss": 3.4964864991253224,
      "tokens_seen": 1757609984
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047225677031093277,
      "loss": 1.4539,
      "theoretical_loss": 3.49647544176436,
      "tokens_seen": 1757675520
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047223671013039113,
      "loss": 1.5235,
      "theoretical_loss": 3.496464384931104,
      "tokens_seen": 1757741056
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004722166499498496,
      "loss": 1.4446,
      "theoretical_loss": 3.4964533286255093,
      "tokens_seen": 1757806592
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047219658976930797,
      "loss": 1.4637,
      "theoretical_loss": 3.4964422728475313,
      "tokens_seen": 1757872128
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047217652958876633,
      "loss": 1.6392,
      "theoretical_loss": 3.4964312175971246,
      "tokens_seen": 1757937664
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004721564694082247,
      "loss": 1.5368,
      "theoretical_loss": 3.4964201628742457,
      "tokens_seen": 1758003200
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004721364092276831,
      "loss": 1.5013,
      "theoretical_loss": 3.496409108678849,
      "tokens_seen": 1758068736
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047211634904714147,
      "loss": 1.4434,
      "theoretical_loss": 3.4963980550108884,
      "tokens_seen": 1758134272
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047209628886659983,
      "loss": 1.5095,
      "theoretical_loss": 3.496387001870321,
      "tokens_seen": 1758199808
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004720762286860582,
      "loss": 1.5236,
      "theoretical_loss": 3.4963759492571014,
      "tokens_seen": 1758265344
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047205616850551656,
      "loss": 1.4539,
      "theoretical_loss": 3.4963648971711843,
      "tokens_seen": 1758330880
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047203610832497497,
      "loss": 1.4444,
      "theoretical_loss": 3.496353845612526,
      "tokens_seen": 1758396416
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047201604814443333,
      "loss": 1.4256,
      "theoretical_loss": 3.49634279458108,
      "tokens_seen": 1758461952
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004719959879638917,
      "loss": 1.4937,
      "theoretical_loss": 3.4963317440768025,
      "tokens_seen": 1758527488
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047197592778335006,
      "loss": 1.3803,
      "theoretical_loss": 3.4963206940996487,
      "tokens_seen": 1758593024
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004719558676028085,
      "loss": 1.4373,
      "theoretical_loss": 3.496309644649574,
      "tokens_seen": 1758658560
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047193580742226684,
      "loss": 1.3752,
      "theoretical_loss": 3.4962985957265333,
      "tokens_seen": 1758724096
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004719157472417252,
      "loss": 1.4666,
      "theoretical_loss": 3.4962875473304815,
      "tokens_seen": 1758789632
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047189568706118356,
      "loss": 1.3727,
      "theoretical_loss": 3.4962764994613744,
      "tokens_seen": 1758855168
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004718756268806419,
      "loss": 1.3146,
      "theoretical_loss": 3.496265452119167,
      "tokens_seen": 1758920704
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047185556670010034,
      "loss": 1.488,
      "theoretical_loss": 3.496254405303815,
      "tokens_seen": 1758986240
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004718355065195587,
      "loss": 1.4627,
      "theoretical_loss": 3.496243359015273,
      "tokens_seen": 1759051776
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047181544633901706,
      "loss": 1.4525,
      "theoretical_loss": 3.496232313253496,
      "tokens_seen": 1759117312
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004717953861584754,
      "loss": 1.5446,
      "theoretical_loss": 3.4962212680184406,
      "tokens_seen": 1759182848
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047177532597793384,
      "loss": 1.3703,
      "theoretical_loss": 3.4962102233100607,
      "tokens_seen": 1759248384
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004717552657973922,
      "loss": 1.5026,
      "theoretical_loss": 3.496199179128312,
      "tokens_seen": 1759313920
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047173520561685057,
      "loss": 1.4686,
      "theoretical_loss": 3.49618813547315,
      "tokens_seen": 1759379456
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047171514543630893,
      "loss": 1.5656,
      "theoretical_loss": 3.4961770923445297,
      "tokens_seen": 1759444992
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047169508525576735,
      "loss": 1.489,
      "theoretical_loss": 3.4961660497424063,
      "tokens_seen": 1759510528
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004716750250752257,
      "loss": 1.5597,
      "theoretical_loss": 3.4961550076667356,
      "tokens_seen": 1759576064
    },
    {
      "epoch": 0.53,
      "objective/train/advantage_avg": 0.4943464696407318,
      "objective/train/docs_used": 642679,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3202155828475952,
      "objective/train/original_loss": 1.3202155828475952,
      "objective/train/theoretical_loss": 3.4961439661174727,
      "objective/train/tokens_used": 1780101600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24511238932609558,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506807565689087,
      "objective/train/weighted_lm_loss": 1.3870290517807007,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4961439661174727,
      "tokens_seen": 1759641600
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047165496489468407,
      "loss": 1.4607,
      "theoretical_loss": 3.4961439661174727,
      "tokens_seen": 1759641600
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047163490471414243,
      "loss": 1.4233,
      "theoretical_loss": 3.4961329250945727,
      "tokens_seen": 1759707136
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004716148445336008,
      "loss": 1.374,
      "theoretical_loss": 3.496121884597991,
      "tokens_seen": 1759772672
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004715947843530592,
      "loss": 1.446,
      "theoretical_loss": 3.4961108446276823,
      "tokens_seen": 1759838208
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004715747241725176,
      "loss": 1.4658,
      "theoretical_loss": 3.4960998051836034,
      "tokens_seen": 1759903744
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047155466399197594,
      "loss": 1.3623,
      "theoretical_loss": 3.496088766265708,
      "tokens_seen": 1759969280
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004715346038114343,
      "loss": 1.5037,
      "theoretical_loss": 3.4960777278739528,
      "tokens_seen": 1760034816
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004715145436308927,
      "loss": 1.5057,
      "theoretical_loss": 3.496066690008292,
      "tokens_seen": 1760100352
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004714944834503511,
      "loss": 1.524,
      "theoretical_loss": 3.496055652668682,
      "tokens_seen": 1760165888
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047147442326980944,
      "loss": 1.5135,
      "theoretical_loss": 3.4960446158550775,
      "tokens_seen": 1760231424
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004714543630892678,
      "loss": 1.4373,
      "theoretical_loss": 3.4960335795674338,
      "tokens_seen": 1760296960
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047143430290872616,
      "loss": 1.4382,
      "theoretical_loss": 3.496022543805706,
      "tokens_seen": 1760362496
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004714142427281846,
      "loss": 1.4732,
      "theoretical_loss": 3.4960115085698504,
      "tokens_seen": 1760428032
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047139418254764294,
      "loss": 1.4877,
      "theoretical_loss": 3.496000473859821,
      "tokens_seen": 1760493568
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004713741223671013,
      "loss": 1.4332,
      "theoretical_loss": 3.495989439675575,
      "tokens_seen": 1760559104
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047135406218655967,
      "loss": 1.4672,
      "theoretical_loss": 3.495978406017066,
      "tokens_seen": 1760624640
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004713340020060181,
      "loss": 1.3574,
      "theoretical_loss": 3.495967372884251,
      "tokens_seen": 1760690176
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047131394182547645,
      "loss": 1.4897,
      "theoretical_loss": 3.4959563402770835,
      "tokens_seen": 1760755712
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004712938816449348,
      "loss": 1.4979,
      "theoretical_loss": 3.4959453081955205,
      "tokens_seen": 1760821248
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047127382146439317,
      "loss": 1.4289,
      "theoretical_loss": 3.495934276639517,
      "tokens_seen": 1760886784
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047125376128385153,
      "loss": 1.4471,
      "theoretical_loss": 3.4959232456090277,
      "tokens_seen": 1760952320
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047123370110330995,
      "loss": 1.4752,
      "theoretical_loss": 3.495912215104009,
      "tokens_seen": 1761017856
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004712136409227683,
      "loss": 1.3504,
      "theoretical_loss": 3.495901185124416,
      "tokens_seen": 1761083392
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004711935807422267,
      "loss": 1.3647,
      "theoretical_loss": 3.4958901556702036,
      "tokens_seen": 1761148928
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047117352056168504,
      "loss": 1.4386,
      "theoretical_loss": 3.4958791267413276,
      "tokens_seen": 1761214464
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047115346038114345,
      "loss": 1.4239,
      "theoretical_loss": 3.495868098337744,
      "tokens_seen": 1761280000
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004711334002006018,
      "loss": 1.524,
      "theoretical_loss": 3.4958570704594067,
      "tokens_seen": 1761345536
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004711133400200602,
      "loss": 1.4955,
      "theoretical_loss": 3.4958460431062726,
      "tokens_seen": 1761411072
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047109327983951854,
      "loss": 1.4215,
      "theoretical_loss": 3.4958350162782965,
      "tokens_seen": 1761476608
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004710732196589769,
      "loss": 1.4312,
      "theoretical_loss": 3.4958239899754346,
      "tokens_seen": 1761542144
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004710531594784353,
      "loss": 1.3265,
      "theoretical_loss": 3.495812964197641,
      "tokens_seen": 1761607680
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004710330992978937,
      "loss": 1.4972,
      "theoretical_loss": 3.4958019389448727,
      "tokens_seen": 1761673216
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047101303911735204,
      "loss": 1.422,
      "theoretical_loss": 3.4957909142170838,
      "tokens_seen": 1761738752
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004709929789368104,
      "loss": 1.4414,
      "theoretical_loss": 3.495779890014231,
      "tokens_seen": 1761804288
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004709729187562688,
      "loss": 1.4367,
      "theoretical_loss": 3.4957688663362685,
      "tokens_seen": 1761869824
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004709528585757272,
      "loss": 1.5303,
      "theoretical_loss": 3.495757843183153,
      "tokens_seen": 1761935360
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047093279839518555,
      "loss": 1.5408,
      "theoretical_loss": 3.4957468205548388,
      "tokens_seen": 1762000896
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004709127382146439,
      "loss": 1.4517,
      "theoretical_loss": 3.4957357984512827,
      "tokens_seen": 1762066432
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047089267803410227,
      "loss": 1.4122,
      "theoretical_loss": 3.495724776872439,
      "tokens_seen": 1762131968
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004708726178535607,
      "loss": 1.4947,
      "theoretical_loss": 3.495713755818264,
      "tokens_seen": 1762197504
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047085255767301905,
      "loss": 1.4411,
      "theoretical_loss": 3.495702735288713,
      "tokens_seen": 1762263040
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004708324974924774,
      "loss": 1.5203,
      "theoretical_loss": 3.4956917152837415,
      "tokens_seen": 1762328576
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004708124373119358,
      "loss": 1.4214,
      "theoretical_loss": 3.4956806958033044,
      "tokens_seen": 1762394112
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004707923771313942,
      "loss": 1.4859,
      "theoretical_loss": 3.4956696768473585,
      "tokens_seen": 1762459648
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047077231695085255,
      "loss": 1.427,
      "theoretical_loss": 3.4956586584158584,
      "tokens_seen": 1762525184
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004707522567703109,
      "loss": 1.3253,
      "theoretical_loss": 3.4956476405087598,
      "tokens_seen": 1762590720
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004707321965897693,
      "loss": 1.3861,
      "theoretical_loss": 3.4956366231260185,
      "tokens_seen": 1762656256
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047071213640922775,
      "loss": 1.4902,
      "theoretical_loss": 3.4956256062675903,
      "tokens_seen": 1762721792
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004706920762286861,
      "loss": 1.3649,
      "theoretical_loss": 3.49561458993343,
      "tokens_seen": 1762787328
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047067201604814447,
      "loss": 1.445,
      "theoretical_loss": 3.4956035741234937,
      "tokens_seen": 1762852864
    },
    {
      "epoch": 0.53,
      "objective/train/advantage_avg": 0.49608179926872253,
      "objective/train/docs_used": 643830,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1808991432189941,
      "objective/train/original_loss": 1.1808991432189941,
      "objective/train/theoretical_loss": 3.4955925588377363,
      "objective/train/tokens_used": 1783378400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24635344743728638,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508607625961304,
      "objective/train/weighted_lm_loss": 1.2407828569412231,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4955925588377363,
      "tokens_seen": 1762918400
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047065195586760283,
      "loss": 1.343,
      "theoretical_loss": 3.4955925588377363,
      "tokens_seen": 1762918400
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004706318956870612,
      "loss": 1.5593,
      "theoretical_loss": 3.4955815440761144,
      "tokens_seen": 1762983936
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004706118355065196,
      "loss": 1.6091,
      "theoretical_loss": 3.4955705298385826,
      "tokens_seen": 1763049472
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000470591775325978,
      "loss": 1.351,
      "theoretical_loss": 3.495559516125097,
      "tokens_seen": 1763115008
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047057171514543634,
      "loss": 1.5235,
      "theoretical_loss": 3.4955485029356135,
      "tokens_seen": 1763180544
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004705516549648947,
      "loss": 1.3823,
      "theoretical_loss": 3.4955374902700864,
      "tokens_seen": 1763246080
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004705315947843531,
      "loss": 1.485,
      "theoretical_loss": 3.4955264781284727,
      "tokens_seen": 1763311616
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004705115346038115,
      "loss": 1.5096,
      "theoretical_loss": 3.495515466510728,
      "tokens_seen": 1763377152
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047049147442326984,
      "loss": 1.3731,
      "theoretical_loss": 3.495504455416807,
      "tokens_seen": 1763442688
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004704714142427282,
      "loss": 1.4657,
      "theoretical_loss": 3.495493444846666,
      "tokens_seen": 1763508224
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047045135406218656,
      "loss": 1.3872,
      "theoretical_loss": 3.4954824348002598,
      "tokens_seen": 1763573760
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.000470431293881645,
      "loss": 1.309,
      "theoretical_loss": 3.495471425277545,
      "tokens_seen": 1763639296
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047041123370110334,
      "loss": 1.4154,
      "theoretical_loss": 3.495460416278477,
      "tokens_seen": 1763704832
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004703911735205617,
      "loss": 1.4116,
      "theoretical_loss": 3.4954494078030107,
      "tokens_seen": 1763770368
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047037111334002007,
      "loss": 1.4092,
      "theoretical_loss": 3.4954383998511025,
      "tokens_seen": 1763835904
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004703510531594785,
      "loss": 1.4979,
      "theoretical_loss": 3.495427392422708,
      "tokens_seen": 1763901440
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047033099297893685,
      "loss": 1.4,
      "theoretical_loss": 3.4954163855177827,
      "tokens_seen": 1763966976
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004703109327983952,
      "loss": 1.4291,
      "theoretical_loss": 3.495405379136282,
      "tokens_seen": 1764032512
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047029087261785357,
      "loss": 1.4003,
      "theoretical_loss": 3.495394373278162,
      "tokens_seen": 1764098048
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047027081243731193,
      "loss": 1.5184,
      "theoretical_loss": 3.495383367943378,
      "tokens_seen": 1764163584
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047025075225677035,
      "loss": 1.4515,
      "theoretical_loss": 3.495372363131886,
      "tokens_seen": 1764229120
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004702306920762287,
      "loss": 1.4331,
      "theoretical_loss": 3.4953613588436414,
      "tokens_seen": 1764294656
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004702106318956871,
      "loss": 1.3404,
      "theoretical_loss": 3.4953503550786005,
      "tokens_seen": 1764360192
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047019057171514544,
      "loss": 1.357,
      "theoretical_loss": 3.4953393518367175,
      "tokens_seen": 1764425728
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047017051153460385,
      "loss": 1.3733,
      "theoretical_loss": 3.4953283491179503,
      "tokens_seen": 1764491264
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004701504513540622,
      "loss": 1.3493,
      "theoretical_loss": 3.4953173469222527,
      "tokens_seen": 1764556800
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004701303911735206,
      "loss": 1.514,
      "theoretical_loss": 3.495306345249581,
      "tokens_seen": 1764622336
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047011033099297894,
      "loss": 1.5303,
      "theoretical_loss": 3.4952953440998917,
      "tokens_seen": 1764687872
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004700902708124373,
      "loss": 1.3616,
      "theoretical_loss": 3.4952843434731395,
      "tokens_seen": 1764753408
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004700702106318957,
      "loss": 1.5221,
      "theoretical_loss": 3.4952733433692806,
      "tokens_seen": 1764818944
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004700501504513541,
      "loss": 1.3901,
      "theoretical_loss": 3.4952623437882706,
      "tokens_seen": 1764884480
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00047003009027081244,
      "loss": 1.4621,
      "theoretical_loss": 3.495251344730065,
      "tokens_seen": 1764950016
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004700100300902708,
      "loss": 1.4057,
      "theoretical_loss": 3.49524034619462,
      "tokens_seen": 1765015552
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004699899699097292,
      "loss": 1.4766,
      "theoretical_loss": 3.495229348181891,
      "tokens_seen": 1765081088
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004699699097291876,
      "loss": 1.4424,
      "theoretical_loss": 3.495218350691834,
      "tokens_seen": 1765146624
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00046994984954864595,
      "loss": 1.5282,
      "theoretical_loss": 3.4952073537244046,
      "tokens_seen": 1765212160
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004699297893681043,
      "loss": 1.4963,
      "theoretical_loss": 3.495196357279559,
      "tokens_seen": 1765277696
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00046990972918756267,
      "loss": 1.4854,
      "theoretical_loss": 3.495185361357252,
      "tokens_seen": 1765343232
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0004698896690070211,
      "loss": 1.4547,
      "theoretical_loss": 3.49517436595744,
      "tokens_seen": 1765408768
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00046986960882647945,
      "loss": 1.4828,
      "theoretical_loss": 3.4951633710800785,
      "tokens_seen": 1765474304
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004698495486459378,
      "loss": 1.5065,
      "theoretical_loss": 3.495152376725124,
      "tokens_seen": 1765539840
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004698294884653962,
      "loss": 1.3543,
      "theoretical_loss": 3.4951413828925317,
      "tokens_seen": 1765605376
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004698094282848546,
      "loss": 1.3343,
      "theoretical_loss": 3.4951303895822576,
      "tokens_seen": 1765670912
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046978936810431295,
      "loss": 1.345,
      "theoretical_loss": 3.495119396794257,
      "tokens_seen": 1765736448
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004697693079237713,
      "loss": 1.4591,
      "theoretical_loss": 3.4951084045284864,
      "tokens_seen": 1765801984
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004697492477432297,
      "loss": 1.4101,
      "theoretical_loss": 3.495097412784901,
      "tokens_seen": 1765867520
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046972918756268804,
      "loss": 1.5042,
      "theoretical_loss": 3.4950864215634567,
      "tokens_seen": 1765933056
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046970912738214646,
      "loss": 1.3923,
      "theoretical_loss": 3.4950754308641105,
      "tokens_seen": 1765998592
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004696890672016048,
      "loss": 1.4719,
      "theoretical_loss": 3.495064440686816,
      "tokens_seen": 1766064128
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004696690070210632,
      "loss": 1.4826,
      "theoretical_loss": 3.495053451031531,
      "tokens_seen": 1766129664
    },
    {
      "epoch": 0.54,
      "objective/train/advantage_avg": 0.4962384104728699,
      "objective/train/docs_used": 644977,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5751638412475586,
      "objective/train/original_loss": 1.5751638412475586,
      "objective/train/theoretical_loss": 3.495042461898211,
      "objective/train/tokens_used": 1786655200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2464943528175354,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508770942687988,
      "objective/train/weighted_lm_loss": 1.6552430391311646,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.495042461898211,
      "tokens_seen": 1766195200
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046964894684052154,
      "loss": 1.5175,
      "theoretical_loss": 3.495042461898211,
      "tokens_seen": 1766195200
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046962888665997996,
      "loss": 1.2921,
      "theoretical_loss": 3.4950314732868106,
      "tokens_seen": 1766260736
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004696088264794383,
      "loss": 1.5126,
      "theoretical_loss": 3.495020485197287,
      "tokens_seen": 1766326272
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004695887662988967,
      "loss": 1.417,
      "theoretical_loss": 3.495009497629596,
      "tokens_seen": 1766391808
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046956870611835505,
      "loss": 1.4489,
      "theoretical_loss": 3.4949985105836925,
      "tokens_seen": 1766457344
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046954864593781346,
      "loss": 1.493,
      "theoretical_loss": 3.4949875240595327,
      "tokens_seen": 1766522880
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004695285857572718,
      "loss": 1.4706,
      "theoretical_loss": 3.494976538057073,
      "tokens_seen": 1766588416
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004695085255767302,
      "loss": 1.5036,
      "theoretical_loss": 3.494965552576269,
      "tokens_seen": 1766653952
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046948846539618855,
      "loss": 1.4416,
      "theoretical_loss": 3.494954567617076,
      "tokens_seen": 1766719488
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004694684052156469,
      "loss": 1.5068,
      "theoretical_loss": 3.4949435831794506,
      "tokens_seen": 1766785024
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004694483450351053,
      "loss": 1.4338,
      "theoretical_loss": 3.4949325992633486,
      "tokens_seen": 1766850560
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004694282848545637,
      "loss": 1.3907,
      "theoretical_loss": 3.4949216158687264,
      "tokens_seen": 1766916096
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046940822467402205,
      "loss": 1.4669,
      "theoretical_loss": 3.4949106329955386,
      "tokens_seen": 1766981632
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004693881644934804,
      "loss": 1.5718,
      "theoretical_loss": 3.4948996506437418,
      "tokens_seen": 1767047168
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046936810431293883,
      "loss": 1.4149,
      "theoretical_loss": 3.4948886688132923,
      "tokens_seen": 1767112704
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004693480441323972,
      "loss": 1.4701,
      "theoretical_loss": 3.494877687504145,
      "tokens_seen": 1767178240
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046932798395185555,
      "loss": 1.4195,
      "theoretical_loss": 3.494866706716257,
      "tokens_seen": 1767243776
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004693079237713139,
      "loss": 1.476,
      "theoretical_loss": 3.4948557264495834,
      "tokens_seen": 1767309312
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004692878635907723,
      "loss": 1.4522,
      "theoretical_loss": 3.4948447467040804,
      "tokens_seen": 1767374848
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004692678034102307,
      "loss": 1.5082,
      "theoretical_loss": 3.494833767479704,
      "tokens_seen": 1767440384
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046924774322968906,
      "loss": 1.338,
      "theoretical_loss": 3.49482278877641,
      "tokens_seen": 1767505920
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004692276830491474,
      "loss": 1.5317,
      "theoretical_loss": 3.494811810594155,
      "tokens_seen": 1767571456
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046920762286860584,
      "loss": 1.3883,
      "theoretical_loss": 3.494800832932894,
      "tokens_seen": 1767636992
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046918756268806425,
      "loss": 1.435,
      "theoretical_loss": 3.494789855792583,
      "tokens_seen": 1767702528
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004691675025075226,
      "loss": 1.4411,
      "theoretical_loss": 3.494778879173179,
      "tokens_seen": 1767768064
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000469147442326981,
      "loss": 1.2385,
      "theoretical_loss": 3.494767903074637,
      "tokens_seen": 1767833600
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046912738214643934,
      "loss": 1.38,
      "theoretical_loss": 3.494756927496913,
      "tokens_seen": 1767899136
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004691073219658977,
      "loss": 1.4167,
      "theoretical_loss": 3.4947459524399633,
      "tokens_seen": 1767964672
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004690872617853561,
      "loss": 1.4024,
      "theoretical_loss": 3.494734977903744,
      "tokens_seen": 1768030208
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004690672016048145,
      "loss": 1.4373,
      "theoretical_loss": 3.494724003888211,
      "tokens_seen": 1768095744
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046904714142427284,
      "loss": 1.3996,
      "theoretical_loss": 3.49471303039332,
      "tokens_seen": 1768161280
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004690270812437312,
      "loss": 1.5781,
      "theoretical_loss": 3.494702057419027,
      "tokens_seen": 1768226816
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004690070210631896,
      "loss": 1.4911,
      "theoretical_loss": 3.4946910849652886,
      "tokens_seen": 1768292352
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000468986960882648,
      "loss": 1.3316,
      "theoretical_loss": 3.494680113032061,
      "tokens_seen": 1768357888
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046896690070210635,
      "loss": 1.461,
      "theoretical_loss": 3.4946691416192985,
      "tokens_seen": 1768423424
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004689468405215647,
      "loss": 1.5073,
      "theoretical_loss": 3.494658170726959,
      "tokens_seen": 1768488960
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046892678034102307,
      "loss": 1.4402,
      "theoretical_loss": 3.494647200354998,
      "tokens_seen": 1768554496
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004689067201604815,
      "loss": 1.3311,
      "theoretical_loss": 3.494636230503371,
      "tokens_seen": 1768620032
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046888665997993985,
      "loss": 1.4385,
      "theoretical_loss": 3.4946252611720348,
      "tokens_seen": 1768685568
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004688665997993982,
      "loss": 1.4398,
      "theoretical_loss": 3.4946142923609447,
      "tokens_seen": 1768751104
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004688465396188566,
      "loss": 1.4593,
      "theoretical_loss": 3.494603324070057,
      "tokens_seen": 1768816640
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000468826479438315,
      "loss": 1.4774,
      "theoretical_loss": 3.4945923562993277,
      "tokens_seen": 1768882176
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046880641925777335,
      "loss": 1.4711,
      "theoretical_loss": 3.494581389048714,
      "tokens_seen": 1768947712
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004687863590772317,
      "loss": 1.424,
      "theoretical_loss": 3.49457042231817,
      "tokens_seen": 1769013248
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004687662988966901,
      "loss": 1.536,
      "theoretical_loss": 3.494559456107653,
      "tokens_seen": 1769078784
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046874623871614844,
      "loss": 1.4787,
      "theoretical_loss": 3.4945484904171193,
      "tokens_seen": 1769144320
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046872617853560686,
      "loss": 1.3152,
      "theoretical_loss": 3.494537525246524,
      "tokens_seen": 1769209856
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004687061183550652,
      "loss": 1.3653,
      "theoretical_loss": 3.4945265605958236,
      "tokens_seen": 1769275392
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004686860581745236,
      "loss": 1.3759,
      "theoretical_loss": 3.4945155964649746,
      "tokens_seen": 1769340928
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046866599799398194,
      "loss": 1.3237,
      "theoretical_loss": 3.494504632853933,
      "tokens_seen": 1769406464
    },
    {
      "epoch": 0.54,
      "objective/train/advantage_avg": 0.49283674359321594,
      "objective/train/docs_used": 646001,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.267704725265503,
      "objective/train/original_loss": 1.267704725265503,
      "objective/train/theoretical_loss": 3.4944936697626545,
      "objective/train/tokens_used": 1789932000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24428628385066986,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505253076553345,
      "objective/train/weighted_lm_loss": 1.3315951824188232,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.7046881318092346,
      "theoretical_loss": 3.4944936697626545,
      "tokens_seen": 1769472000
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046864593781344036,
      "loss": 1.3416,
      "theoretical_loss": 3.4944936697626545,
      "tokens_seen": 1769472000
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004686258776328987,
      "loss": 1.4673,
      "theoretical_loss": 3.4944827071910955,
      "tokens_seen": 1769537536
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004686058174523571,
      "loss": 1.4219,
      "theoretical_loss": 3.4944717451392115,
      "tokens_seen": 1769603072
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046858575727181545,
      "loss": 1.3888,
      "theoretical_loss": 3.49446078360696,
      "tokens_seen": 1769668608
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004685656970912738,
      "loss": 1.3566,
      "theoretical_loss": 3.4944498225942953,
      "tokens_seen": 1769734144
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004685456369107322,
      "loss": 1.535,
      "theoretical_loss": 3.494438862101175,
      "tokens_seen": 1769799680
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004685255767301906,
      "loss": 1.423,
      "theoretical_loss": 3.494427902127555,
      "tokens_seen": 1769865216
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046850551654964895,
      "loss": 1.4095,
      "theoretical_loss": 3.494416942673391,
      "tokens_seen": 1769930752
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004684854563691073,
      "loss": 1.4226,
      "theoretical_loss": 3.4944059837386394,
      "tokens_seen": 1769996288
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046846539618856573,
      "loss": 1.5514,
      "theoretical_loss": 3.4943950253232563,
      "tokens_seen": 1770061824
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004684453360080241,
      "loss": 1.344,
      "theoretical_loss": 3.4943840674271973,
      "tokens_seen": 1770127360
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046842527582748245,
      "loss": 1.5257,
      "theoretical_loss": 3.49437311005042,
      "tokens_seen": 1770192896
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004684052156469408,
      "loss": 1.4904,
      "theoretical_loss": 3.494362153192879,
      "tokens_seen": 1770258432
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046838515546639923,
      "loss": 1.4065,
      "theoretical_loss": 3.4943511968545313,
      "tokens_seen": 1770323968
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004683650952858576,
      "loss": 1.4659,
      "theoretical_loss": 3.4943402410353332,
      "tokens_seen": 1770389504
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046834503510531595,
      "loss": 1.3372,
      "theoretical_loss": 3.4943292857352404,
      "tokens_seen": 1770455040
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004683249749247743,
      "loss": 1.5732,
      "theoretical_loss": 3.494318330954209,
      "tokens_seen": 1770520576
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004683049147442327,
      "loss": 1.401,
      "theoretical_loss": 3.4943073766921957,
      "tokens_seen": 1770586112
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004682848545636911,
      "loss": 1.5298,
      "theoretical_loss": 3.494296422949157,
      "tokens_seen": 1770651648
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046826479438314946,
      "loss": 1.4413,
      "theoretical_loss": 3.4942854697250483,
      "tokens_seen": 1770717184
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004682447342026078,
      "loss": 1.3535,
      "theoretical_loss": 3.494274517019826,
      "tokens_seen": 1770782720
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004682246740220662,
      "loss": 1.4702,
      "theoretical_loss": 3.494263564833447,
      "tokens_seen": 1770848256
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004682046138415246,
      "loss": 1.4583,
      "theoretical_loss": 3.494252613165866,
      "tokens_seen": 1770913792
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046818455366098296,
      "loss": 1.4531,
      "theoretical_loss": 3.4942416620170405,
      "tokens_seen": 1770979328
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004681644934804413,
      "loss": 1.4309,
      "theoretical_loss": 3.494230711386926,
      "tokens_seen": 1771044864
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004681444332998997,
      "loss": 1.3819,
      "theoretical_loss": 3.49421976127548,
      "tokens_seen": 1771110400
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046812437311935805,
      "loss": 1.4025,
      "theoretical_loss": 3.4942088116826575,
      "tokens_seen": 1771175936
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046810431293881646,
      "loss": 1.4211,
      "theoretical_loss": 3.4941978626084156,
      "tokens_seen": 1771241472
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004680842527582748,
      "loss": 1.4399,
      "theoretical_loss": 3.4941869140527095,
      "tokens_seen": 1771307008
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004680641925777332,
      "loss": 1.4266,
      "theoretical_loss": 3.4941759660154963,
      "tokens_seen": 1771372544
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046804413239719155,
      "loss": 1.4456,
      "theoretical_loss": 3.494165018496732,
      "tokens_seen": 1771438080
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046802407221664997,
      "loss": 1.4266,
      "theoretical_loss": 3.494154071496373,
      "tokens_seen": 1771503616
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046800401203610833,
      "loss": 1.3893,
      "theoretical_loss": 3.494143125014375,
      "tokens_seen": 1771569152
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004679839518555667,
      "loss": 1.479,
      "theoretical_loss": 3.4941321790506956,
      "tokens_seen": 1771634688
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046796389167502505,
      "loss": 1.482,
      "theoretical_loss": 3.4941212336052896,
      "tokens_seen": 1771700224
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004679438314944834,
      "loss": 1.4358,
      "theoretical_loss": 3.4941102886781135,
      "tokens_seen": 1771765760
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046792377131394183,
      "loss": 1.5119,
      "theoretical_loss": 3.4940993442691246,
      "tokens_seen": 1771831296
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004679037111334002,
      "loss": 1.4324,
      "theoretical_loss": 3.4940884003782786,
      "tokens_seen": 1771896832
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046788365095285856,
      "loss": 1.4597,
      "theoretical_loss": 3.494077457005532,
      "tokens_seen": 1771962368
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004678635907723169,
      "loss": 1.4655,
      "theoretical_loss": 3.4940665141508407,
      "tokens_seen": 1772027904
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046784353059177534,
      "loss": 1.4378,
      "theoretical_loss": 3.4940555718141613,
      "tokens_seen": 1772093440
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004678234704112337,
      "loss": 1.4083,
      "theoretical_loss": 3.49404462999545,
      "tokens_seen": 1772158976
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046780341023069206,
      "loss": 1.441,
      "theoretical_loss": 3.494033688694663,
      "tokens_seen": 1772224512
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004677833500501504,
      "loss": 1.3724,
      "theoretical_loss": 3.494022747911757,
      "tokens_seen": 1772290048
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004677632898696088,
      "loss": 1.5025,
      "theoretical_loss": 3.4940118076466886,
      "tokens_seen": 1772355584
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004677432296890672,
      "loss": 1.4923,
      "theoretical_loss": 3.4940008678994134,
      "tokens_seen": 1772421120
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046772316950852556,
      "loss": 1.4009,
      "theoretical_loss": 3.4939899286698877,
      "tokens_seen": 1772486656
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000467703109327984,
      "loss": 1.4591,
      "theoretical_loss": 3.4939789899580687,
      "tokens_seen": 1772552192
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046768304914744234,
      "loss": 1.3652,
      "theoretical_loss": 3.493968051763912,
      "tokens_seen": 1772617728
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046766298896690076,
      "loss": 1.4331,
      "theoretical_loss": 3.4939571140873746,
      "tokens_seen": 1772683264
    },
    {
      "epoch": 0.54,
      "objective/train/advantage_avg": 0.49378469586372375,
      "objective/train/docs_used": 647211,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5280696153640747,
      "objective/train/original_loss": 1.5280694961547852,
      "objective/train/theoretical_loss": 3.493946176928412,
      "objective/train/tokens_used": 1793208800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24423588812351227,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506200790405273,
      "objective/train/weighted_lm_loss": 1.605360507965088,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.493946176928412,
      "tokens_seen": 1772748800
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004676429287863591,
      "loss": 1.5161,
      "theoretical_loss": 3.493946176928412,
      "tokens_seen": 1772748800
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004676228686058175,
      "loss": 1.4212,
      "theoretical_loss": 3.493935240286982,
      "tokens_seen": 1772814336
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046760280842527585,
      "loss": 1.3565,
      "theoretical_loss": 3.4939243041630395,
      "tokens_seen": 1772879872
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004675827482447342,
      "loss": 1.4072,
      "theoretical_loss": 3.493913368556541,
      "tokens_seen": 1772945408
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004675626880641926,
      "loss": 1.5295,
      "theoretical_loss": 3.493902433467444,
      "tokens_seen": 1773010944
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000467542627883651,
      "loss": 1.5212,
      "theoretical_loss": 3.493891498895704,
      "tokens_seen": 1773076480
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046752256770310935,
      "loss": 1.337,
      "theoretical_loss": 3.4938805648412776,
      "tokens_seen": 1773142016
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004675025075225677,
      "loss": 1.5188,
      "theoretical_loss": 3.493869631304121,
      "tokens_seen": 1773207552
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046748244734202613,
      "loss": 1.4198,
      "theoretical_loss": 3.4938586982841917,
      "tokens_seen": 1773273088
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004674623871614845,
      "loss": 1.3265,
      "theoretical_loss": 3.4938477657814446,
      "tokens_seen": 1773338624
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046744232698094285,
      "loss": 1.5276,
      "theoretical_loss": 3.4938368337958368,
      "tokens_seen": 1773404160
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004674222668004012,
      "loss": 1.5029,
      "theoretical_loss": 3.493825902327325,
      "tokens_seen": 1773469696
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046740220661985963,
      "loss": 1.2295,
      "theoretical_loss": 3.4938149713758655,
      "tokens_seen": 1773535232
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000467382146439318,
      "loss": 1.4347,
      "theoretical_loss": 3.493804040941414,
      "tokens_seen": 1773600768
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046736208625877636,
      "loss": 1.4387,
      "theoretical_loss": 3.493793111023928,
      "tokens_seen": 1773666304
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004673420260782347,
      "loss": 1.3888,
      "theoretical_loss": 3.493782181623364,
      "tokens_seen": 1773731840
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004673219658976931,
      "loss": 1.4221,
      "theoretical_loss": 3.493771252739677,
      "tokens_seen": 1773797376
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004673019057171515,
      "loss": 1.4246,
      "theoretical_loss": 3.493760324372825,
      "tokens_seen": 1773862912
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046728184553660986,
      "loss": 1.3863,
      "theoretical_loss": 3.4937493965227633,
      "tokens_seen": 1773928448
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004672617853560682,
      "loss": 1.4861,
      "theoretical_loss": 3.4937384691894495,
      "tokens_seen": 1773993984
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004672417251755266,
      "loss": 1.416,
      "theoretical_loss": 3.49372754237284,
      "tokens_seen": 1774059520
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000467221664994985,
      "loss": 1.2899,
      "theoretical_loss": 3.4937166160728896,
      "tokens_seen": 1774125056
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046720160481444336,
      "loss": 1.3357,
      "theoretical_loss": 3.4937056902895565,
      "tokens_seen": 1774190592
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004671815446339017,
      "loss": 1.524,
      "theoretical_loss": 3.4936947650227967,
      "tokens_seen": 1774256128
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004671614844533601,
      "loss": 1.3736,
      "theoretical_loss": 3.493683840272567,
      "tokens_seen": 1774321664
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046714142427281845,
      "loss": 1.4515,
      "theoretical_loss": 3.4936729160388236,
      "tokens_seen": 1774387200
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046712136409227686,
      "loss": 1.412,
      "theoretical_loss": 3.4936619923215226,
      "tokens_seen": 1774452736
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004671013039117352,
      "loss": 1.3239,
      "theoretical_loss": 3.493651069120621,
      "tokens_seen": 1774518272
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004670812437311936,
      "loss": 1.3164,
      "theoretical_loss": 3.493640146436076,
      "tokens_seen": 1774583808
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046706118355065195,
      "loss": 1.4044,
      "theoretical_loss": 3.493629224267842,
      "tokens_seen": 1774649344
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046704112337011037,
      "loss": 1.4261,
      "theoretical_loss": 3.493618302615878,
      "tokens_seen": 1774714880
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046702106318956873,
      "loss": 1.4212,
      "theoretical_loss": 3.4936073814801385,
      "tokens_seen": 1774780416
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004670010030090271,
      "loss": 1.4662,
      "theoretical_loss": 3.4935964608605814,
      "tokens_seen": 1774845952
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046698094282848545,
      "loss": 1.4163,
      "theoretical_loss": 3.493585540757163,
      "tokens_seen": 1774911488
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004669608826479438,
      "loss": 1.3798,
      "theoretical_loss": 3.4935746211698393,
      "tokens_seen": 1774977024
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046694082246740223,
      "loss": 1.3697,
      "theoretical_loss": 3.4935637020985673,
      "tokens_seen": 1775042560
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004669207622868606,
      "loss": 1.3859,
      "theoretical_loss": 3.4935527835433033,
      "tokens_seen": 1775108096
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046690070210631896,
      "loss": 1.4167,
      "theoretical_loss": 3.4935418655040045,
      "tokens_seen": 1775173632
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004668806419257773,
      "loss": 1.4932,
      "theoretical_loss": 3.4935309479806262,
      "tokens_seen": 1775239168
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046686058174523574,
      "loss": 1.4056,
      "theoretical_loss": 3.4935200309731265,
      "tokens_seen": 1775304704
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004668405215646941,
      "loss": 1.502,
      "theoretical_loss": 3.493509114481461,
      "tokens_seen": 1775370240
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046682046138415246,
      "loss": 1.5137,
      "theoretical_loss": 3.493498198505587,
      "tokens_seen": 1775435776
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004668004012036108,
      "loss": 1.5157,
      "theoretical_loss": 3.49348728304546,
      "tokens_seen": 1775501312
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004667803410230692,
      "loss": 1.3804,
      "theoretical_loss": 3.493476368101037,
      "tokens_seen": 1775566848
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004667602808425276,
      "loss": 1.3536,
      "theoretical_loss": 3.4934654536722753,
      "tokens_seen": 1775632384
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046674022066198596,
      "loss": 1.314,
      "theoretical_loss": 3.4934545397591314,
      "tokens_seen": 1775697920
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004667201604814443,
      "loss": 1.4255,
      "theoretical_loss": 3.493443626361561,
      "tokens_seen": 1775763456
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004667001003009027,
      "loss": 1.3921,
      "theoretical_loss": 3.493432713479521,
      "tokens_seen": 1775828992
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004666800401203611,
      "loss": 1.4503,
      "theoretical_loss": 3.493421801112969,
      "tokens_seen": 1775894528
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046665997993981947,
      "loss": 1.4055,
      "theoretical_loss": 3.4934108892618605,
      "tokens_seen": 1775960064
    },
    {
      "epoch": 0.54,
      "objective/train/advantage_avg": 0.4946158230304718,
      "objective/train/docs_used": 648417,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.360501766204834,
      "objective/train/original_loss": 1.360502004623413,
      "objective/train/theoretical_loss": 3.4933999779261526,
      "objective/train/tokens_used": 1796485600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24500499665737152,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507071018218994,
      "objective/train/weighted_lm_loss": 1.4293471574783325,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4933999779261526,
      "tokens_seen": 1776025600
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046663991975927783,
      "loss": 1.4991,
      "theoretical_loss": 3.4933999779261526,
      "tokens_seen": 1776025600
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004666198595787362,
      "loss": 1.3865,
      "theoretical_loss": 3.4933890671058023,
      "tokens_seen": 1776091136
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046659979939819455,
      "loss": 1.3186,
      "theoretical_loss": 3.493378156800765,
      "tokens_seen": 1776156672
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046657973921765297,
      "loss": 1.4036,
      "theoretical_loss": 3.493367247010999,
      "tokens_seen": 1776222208
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046655967903711133,
      "loss": 1.6044,
      "theoretical_loss": 3.4933563377364596,
      "tokens_seen": 1776287744
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004665396188565697,
      "loss": 1.3387,
      "theoretical_loss": 3.493345428977104,
      "tokens_seen": 1776353280
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046651955867602806,
      "loss": 1.3434,
      "theoretical_loss": 3.4933345207328896,
      "tokens_seen": 1776418816
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004664994984954865,
      "loss": 1.4816,
      "theoretical_loss": 3.4933236130037715,
      "tokens_seen": 1776484352
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046647943831494484,
      "loss": 1.464,
      "theoretical_loss": 3.493312705789708,
      "tokens_seen": 1776549888
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004664593781344032,
      "loss": 1.4419,
      "theoretical_loss": 3.493301799090654,
      "tokens_seen": 1776615424
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046643931795386156,
      "loss": 1.4034,
      "theoretical_loss": 3.493290892906568,
      "tokens_seen": 1776680960
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004664192577733199,
      "loss": 1.545,
      "theoretical_loss": 3.4932799872374054,
      "tokens_seen": 1776746496
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046639919759277834,
      "loss": 1.408,
      "theoretical_loss": 3.493269082083123,
      "tokens_seen": 1776812032
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004663791374122367,
      "loss": 1.463,
      "theoretical_loss": 3.4932581774436793,
      "tokens_seen": 1776877568
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046635907723169506,
      "loss": 1.3951,
      "theoretical_loss": 3.493247273319028,
      "tokens_seen": 1776943104
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004663390170511534,
      "loss": 1.492,
      "theoretical_loss": 3.4932363697091287,
      "tokens_seen": 1777008640
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046631895687061184,
      "loss": 1.5297,
      "theoretical_loss": 3.4932254666139357,
      "tokens_seen": 1777074176
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004662988966900702,
      "loss": 1.4165,
      "theoretical_loss": 3.4932145640334076,
      "tokens_seen": 1777139712
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046627883650952857,
      "loss": 1.3146,
      "theoretical_loss": 3.4932036619674998,
      "tokens_seen": 1777205248
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046625877632898693,
      "loss": 1.3651,
      "theoretical_loss": 3.4931927604161697,
      "tokens_seen": 1777270784
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046623871614844535,
      "loss": 1.2909,
      "theoretical_loss": 3.493181859379374,
      "tokens_seen": 1777336320
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004662186559679037,
      "loss": 1.2857,
      "theoretical_loss": 3.4931709588570694,
      "tokens_seen": 1777401856
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004661985957873621,
      "loss": 1.5406,
      "theoretical_loss": 3.4931600588492127,
      "tokens_seen": 1777467392
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004661785356068205,
      "loss": 1.4027,
      "theoretical_loss": 3.4931491593557604,
      "tokens_seen": 1777532928
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046615847542627885,
      "loss": 1.4357,
      "theoretical_loss": 3.4931382603766696,
      "tokens_seen": 1777598464
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046613841524573727,
      "loss": 1.445,
      "theoretical_loss": 3.493127361911897,
      "tokens_seen": 1777664000
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046611835506519563,
      "loss": 1.2908,
      "theoretical_loss": 3.493116463961399,
      "tokens_seen": 1777729536
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000466098294884654,
      "loss": 1.3255,
      "theoretical_loss": 3.493105566525133,
      "tokens_seen": 1777795072
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046607823470411235,
      "loss": 1.4241,
      "theoretical_loss": 3.493094669603055,
      "tokens_seen": 1777860608
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046605817452357077,
      "loss": 1.3757,
      "theoretical_loss": 3.4930837731951225,
      "tokens_seen": 1777926144
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046603811434302913,
      "loss": 1.3506,
      "theoretical_loss": 3.493072877301292,
      "tokens_seen": 1777991680
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004660180541624875,
      "loss": 1.4744,
      "theoretical_loss": 3.49306198192152,
      "tokens_seen": 1778057216
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046599799398194586,
      "loss": 1.4725,
      "theoretical_loss": 3.493051087055764,
      "tokens_seen": 1778122752
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004659779338014042,
      "loss": 1.4269,
      "theoretical_loss": 3.49304019270398,
      "tokens_seen": 1778188288
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046595787362086263,
      "loss": 1.4223,
      "theoretical_loss": 3.4930292988661256,
      "tokens_seen": 1778253824
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000465937813440321,
      "loss": 1.5009,
      "theoretical_loss": 3.4930184055421565,
      "tokens_seen": 1778319360
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046591775325977936,
      "loss": 1.519,
      "theoretical_loss": 3.493007512732031,
      "tokens_seen": 1778384896
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004658976930792377,
      "loss": 1.3877,
      "theoretical_loss": 3.4929966204357052,
      "tokens_seen": 1778450432
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046587763289869614,
      "loss": 1.4569,
      "theoretical_loss": 3.492985728653135,
      "tokens_seen": 1778515968
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004658575727181545,
      "loss": 1.4016,
      "theoretical_loss": 3.4929748373842795,
      "tokens_seen": 1778581504
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046583751253761286,
      "loss": 1.541,
      "theoretical_loss": 3.4929639466290934,
      "tokens_seen": 1778647040
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004658174523570712,
      "loss": 1.4869,
      "theoretical_loss": 3.4929530563875346,
      "tokens_seen": 1778712576
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004657973921765296,
      "loss": 1.4159,
      "theoretical_loss": 3.4929421666595593,
      "tokens_seen": 1778778112
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000465777331995988,
      "loss": 1.3596,
      "theoretical_loss": 3.492931277445125,
      "tokens_seen": 1778843648
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046575727181544636,
      "loss": 1.3772,
      "theoretical_loss": 3.492920388744188,
      "tokens_seen": 1778909184
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004657372116349047,
      "loss": 1.4165,
      "theoretical_loss": 3.4929095005567063,
      "tokens_seen": 1778974720
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004657171514543631,
      "loss": 1.3947,
      "theoretical_loss": 3.492898612882635,
      "tokens_seen": 1779040256
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004656970912738215,
      "loss": 1.4609,
      "theoretical_loss": 3.492887725721933,
      "tokens_seen": 1779105792
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046567703109327987,
      "loss": 1.4842,
      "theoretical_loss": 3.4928768390745555,
      "tokens_seen": 1779171328
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046565697091273823,
      "loss": 1.4349,
      "theoretical_loss": 3.49286595294046,
      "tokens_seen": 1779236864
    },
    {
      "epoch": 0.54,
      "objective/train/advantage_avg": 0.49288085103034973,
      "objective/train/docs_used": 649543,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4467459917068481,
      "objective/train/original_loss": 1.4467461109161377,
      "objective/train/theoretical_loss": 3.4928550673196033,
      "objective/train/tokens_used": 1799762400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2433895766735077,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050525426864624,
      "objective/train/weighted_lm_loss": 1.5197176933288574,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4928550673196033,
      "tokens_seen": 1779302400
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004656369107321966,
      "loss": 1.4481,
      "theoretical_loss": 3.4928550673196033,
      "tokens_seen": 1779302400
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046561685055165495,
      "loss": 1.4385,
      "theoretical_loss": 3.4928441822119423,
      "tokens_seen": 1779367936
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046559679037111337,
      "loss": 1.3894,
      "theoretical_loss": 3.4928332976174348,
      "tokens_seen": 1779433472
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046557673019057173,
      "loss": 1.3354,
      "theoretical_loss": 3.492822413536036,
      "tokens_seen": 1779499008
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004655566700100301,
      "loss": 1.3498,
      "theoretical_loss": 3.492811529967704,
      "tokens_seen": 1779564544
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046553660982948846,
      "loss": 1.434,
      "theoretical_loss": 3.4928006469123956,
      "tokens_seen": 1779630080
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004655165496489469,
      "loss": 1.5103,
      "theoretical_loss": 3.492789764370068,
      "tokens_seen": 1779695616
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046549648946840524,
      "loss": 1.3493,
      "theoretical_loss": 3.4927788823406773,
      "tokens_seen": 1779761152
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004654764292878636,
      "loss": 1.4182,
      "theoretical_loss": 3.4927680008241806,
      "tokens_seen": 1779826688
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046545636910732196,
      "loss": 1.4381,
      "theoretical_loss": 3.492757119820536,
      "tokens_seen": 1779892224
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004654363089267803,
      "loss": 1.5572,
      "theoretical_loss": 3.4927462393296986,
      "tokens_seen": 1779957760
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046541624874623874,
      "loss": 1.4996,
      "theoretical_loss": 3.4927353593516264,
      "tokens_seen": 1780023296
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004653961885656971,
      "loss": 1.6517,
      "theoretical_loss": 3.4927244798862764,
      "tokens_seen": 1780088832
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046537612838515546,
      "loss": 1.4355,
      "theoretical_loss": 3.4927136009336057,
      "tokens_seen": 1780154368
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004653560682046138,
      "loss": 1.3798,
      "theoretical_loss": 3.492702722493571,
      "tokens_seen": 1780219904
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046533600802407224,
      "loss": 1.4656,
      "theoretical_loss": 3.492691844566129,
      "tokens_seen": 1780285440
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004653159478435306,
      "loss": 1.2804,
      "theoretical_loss": 3.492680967151237,
      "tokens_seen": 1780350976
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046529588766298897,
      "loss": 1.454,
      "theoretical_loss": 3.492670090248852,
      "tokens_seen": 1780416512
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046527582748244733,
      "loss": 1.3653,
      "theoretical_loss": 3.4926592138589307,
      "tokens_seen": 1780482048
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046525576730190575,
      "loss": 1.3695,
      "theoretical_loss": 3.4926483379814304,
      "tokens_seen": 1780547584
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004652357071213641,
      "loss": 1.3604,
      "theoretical_loss": 3.492637462616308,
      "tokens_seen": 1780613120
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046521564694082247,
      "loss": 1.4252,
      "theoretical_loss": 3.4926265877635205,
      "tokens_seen": 1780678656
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046519558676028083,
      "loss": 1.42,
      "theoretical_loss": 3.4926157134230253,
      "tokens_seen": 1780744192
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004651755265797392,
      "loss": 1.418,
      "theoretical_loss": 3.4926048395947786,
      "tokens_seen": 1780809728
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004651554663991976,
      "loss": 1.4767,
      "theoretical_loss": 3.4925939662787377,
      "tokens_seen": 1780875264
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000465135406218656,
      "loss": 1.4349,
      "theoretical_loss": 3.4925830934748605,
      "tokens_seen": 1780940800
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046511534603811434,
      "loss": 1.47,
      "theoretical_loss": 3.4925722211831025,
      "tokens_seen": 1781006336
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004650952858575727,
      "loss": 1.4186,
      "theoretical_loss": 3.492561349403422,
      "tokens_seen": 1781071872
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004650752256770311,
      "loss": 1.2961,
      "theoretical_loss": 3.4925504781357755,
      "tokens_seen": 1781137408
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004650551654964895,
      "loss": 1.5587,
      "theoretical_loss": 3.4925396073801203,
      "tokens_seen": 1781202944
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046503510531594784,
      "loss": 1.4237,
      "theoretical_loss": 3.4925287371364124,
      "tokens_seen": 1781268480
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004650150451354062,
      "loss": 1.2428,
      "theoretical_loss": 3.4925178674046107,
      "tokens_seen": 1781334016
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046499498495486456,
      "loss": 1.3501,
      "theoretical_loss": 3.4925069981846715,
      "tokens_seen": 1781399552
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000464974924774323,
      "loss": 1.5214,
      "theoretical_loss": 3.4924961294765504,
      "tokens_seen": 1781465088
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046495486459378134,
      "loss": 1.381,
      "theoretical_loss": 3.4924852612802066,
      "tokens_seen": 1781530624
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004649348044132397,
      "loss": 1.5084,
      "theoretical_loss": 3.4924743935955966,
      "tokens_seen": 1781596160
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046491474423269807,
      "loss": 1.3962,
      "theoretical_loss": 3.492463526422677,
      "tokens_seen": 1781661696
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004648946840521565,
      "loss": 1.5002,
      "theoretical_loss": 3.4924526597614047,
      "tokens_seen": 1781727232
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046487462387161485,
      "loss": 1.3296,
      "theoretical_loss": 3.4924417936117376,
      "tokens_seen": 1781792768
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004648545636910732,
      "loss": 1.4392,
      "theoretical_loss": 3.492430927973632,
      "tokens_seen": 1781858304
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046483450351053157,
      "loss": 1.4668,
      "theoretical_loss": 3.492420062847045,
      "tokens_seen": 1781923840
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046481444332998993,
      "loss": 1.4358,
      "theoretical_loss": 3.492409198231935,
      "tokens_seen": 1781989376
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046479438314944835,
      "loss": 1.5226,
      "theoretical_loss": 3.492398334128258,
      "tokens_seen": 1782054912
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004647743229689067,
      "loss": 1.4921,
      "theoretical_loss": 3.492387470535971,
      "tokens_seen": 1782120448
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004647542627883651,
      "loss": 1.3456,
      "theoretical_loss": 3.4923766074550313,
      "tokens_seen": 1782185984
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046473420260782344,
      "loss": 1.3224,
      "theoretical_loss": 3.4923657448853964,
      "tokens_seen": 1782251520
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004647141424272819,
      "loss": 1.3353,
      "theoretical_loss": 3.492354882827023,
      "tokens_seen": 1782317056
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046469408224674027,
      "loss": 1.4128,
      "theoretical_loss": 3.492344021279869,
      "tokens_seen": 1782382592
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046467402206619863,
      "loss": 1.449,
      "theoretical_loss": 3.4923331602438905,
      "tokens_seen": 1782448128
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000464653961885657,
      "loss": 1.4891,
      "theoretical_loss": 3.4923222997190457,
      "tokens_seen": 1782513664
    },
    {
      "epoch": 0.54,
      "objective/train/advantage_avg": 0.4926307499408722,
      "objective/train/docs_used": 650724,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5393986701965332,
      "objective/train/original_loss": 1.5393986701965332,
      "objective/train/theoretical_loss": 3.4923114397052903,
      "objective/train/tokens_used": 1803039200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24310098588466644,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504989624023438,
      "objective/train/weighted_lm_loss": 1.6169780492782593,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4923114397052903,
      "tokens_seen": 1782579200
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046463390170511535,
      "loss": 1.5135,
      "theoretical_loss": 3.4923114397052903,
      "tokens_seen": 1782579200
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046461384152457377,
      "loss": 1.4106,
      "theoretical_loss": 3.492300580202583,
      "tokens_seen": 1782644736
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046459378134403213,
      "loss": 1.463,
      "theoretical_loss": 3.49228972121088,
      "tokens_seen": 1782710272
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004645737211634905,
      "loss": 1.4388,
      "theoretical_loss": 3.4922788627301387,
      "tokens_seen": 1782775808
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046455366098294886,
      "loss": 1.434,
      "theoretical_loss": 3.4922680047603167,
      "tokens_seen": 1782841344
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004645336008024073,
      "loss": 1.4892,
      "theoretical_loss": 3.4922571473013706,
      "tokens_seen": 1782906880
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046451354062186564,
      "loss": 1.4889,
      "theoretical_loss": 3.492246290353258,
      "tokens_seen": 1782972416
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000464493480441324,
      "loss": 1.5439,
      "theoretical_loss": 3.492235433915936,
      "tokens_seen": 1783037952
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046447342026078236,
      "loss": 1.4393,
      "theoretical_loss": 3.4922245779893615,
      "tokens_seen": 1783103488
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004644533600802407,
      "loss": 1.5048,
      "theoretical_loss": 3.492213722573492,
      "tokens_seen": 1783169024
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046443329989969914,
      "loss": 1.4777,
      "theoretical_loss": 3.4922028676682846,
      "tokens_seen": 1783234560
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004644132397191575,
      "loss": 1.3805,
      "theoretical_loss": 3.4921920132736965,
      "tokens_seen": 1783300096
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046439317953861586,
      "loss": 1.6295,
      "theoretical_loss": 3.492181159389685,
      "tokens_seen": 1783365632
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004643731193580742,
      "loss": 1.4358,
      "theoretical_loss": 3.492170306016207,
      "tokens_seen": 1783431168
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046435305917753264,
      "loss": 1.4342,
      "theoretical_loss": 3.49215945315322,
      "tokens_seen": 1783496704
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000464332998996991,
      "loss": 1.448,
      "theoretical_loss": 3.4921486008006815,
      "tokens_seen": 1783562240
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046431293881644937,
      "loss": 1.4918,
      "theoretical_loss": 3.4921377489585486,
      "tokens_seen": 1783627776
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046429287863590773,
      "loss": 1.413,
      "theoretical_loss": 3.492126897626778,
      "tokens_seen": 1783693312
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046427281845536615,
      "loss": 1.4103,
      "theoretical_loss": 3.492116046805328,
      "tokens_seen": 1783758848
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004642527582748245,
      "loss": 1.4426,
      "theoretical_loss": 3.4921051964941543,
      "tokens_seen": 1783824384
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046423269809428287,
      "loss": 1.4768,
      "theoretical_loss": 3.4920943466932153,
      "tokens_seen": 1783889920
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046421263791374123,
      "loss": 1.3986,
      "theoretical_loss": 3.4920834974024686,
      "tokens_seen": 1783955456
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004641925777331996,
      "loss": 1.4389,
      "theoretical_loss": 3.4920726486218703,
      "tokens_seen": 1784020992
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000464172517552658,
      "loss": 1.4616,
      "theoretical_loss": 3.4920618003513786,
      "tokens_seen": 1784086528
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004641524573721164,
      "loss": 1.3451,
      "theoretical_loss": 3.49205095259095,
      "tokens_seen": 1784152064
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046413239719157474,
      "loss": 1.3915,
      "theoretical_loss": 3.4920401053405428,
      "tokens_seen": 1784217600
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004641123370110331,
      "loss": 1.4034,
      "theoretical_loss": 3.4920292586001134,
      "tokens_seen": 1784283136
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004640922768304915,
      "loss": 1.472,
      "theoretical_loss": 3.492018412369619,
      "tokens_seen": 1784348672
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004640722166499499,
      "loss": 1.4121,
      "theoretical_loss": 3.492007566649018,
      "tokens_seen": 1784414208
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046405215646940824,
      "loss": 1.4751,
      "theoretical_loss": 3.491996721438267,
      "tokens_seen": 1784479744
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004640320962888666,
      "loss": 1.3104,
      "theoretical_loss": 3.491985876737323,
      "tokens_seen": 1784545280
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046401203610832496,
      "loss": 1.3998,
      "theoretical_loss": 3.4919750325461436,
      "tokens_seen": 1784610816
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004639919759277834,
      "loss": 1.4414,
      "theoretical_loss": 3.491964188864686,
      "tokens_seen": 1784676352
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046397191574724174,
      "loss": 1.4219,
      "theoretical_loss": 3.491953345692908,
      "tokens_seen": 1784741888
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004639518555667001,
      "loss": 1.6043,
      "theoretical_loss": 3.4919425030307667,
      "tokens_seen": 1784807424
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046393179538615847,
      "loss": 1.4351,
      "theoretical_loss": 3.491931660878219,
      "tokens_seen": 1784872960
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004639117352056169,
      "loss": 1.5782,
      "theoretical_loss": 3.491920819235223,
      "tokens_seen": 1784938496
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046389167502507525,
      "loss": 1.4583,
      "theoretical_loss": 3.4919099781017353,
      "tokens_seen": 1785004032
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004638716148445336,
      "loss": 1.5275,
      "theoretical_loss": 3.4918991374777137,
      "tokens_seen": 1785069568
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046385155466399197,
      "loss": 1.5572,
      "theoretical_loss": 3.491888297363116,
      "tokens_seen": 1785135104
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046383149448345033,
      "loss": 1.4113,
      "theoretical_loss": 3.491877457757898,
      "tokens_seen": 1785200640
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046381143430290875,
      "loss": 1.3714,
      "theoretical_loss": 3.4918666186620184,
      "tokens_seen": 1785266176
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004637913741223671,
      "loss": 1.4474,
      "theoretical_loss": 3.491855780075434,
      "tokens_seen": 1785331712
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004637713139418255,
      "loss": 1.3771,
      "theoretical_loss": 3.491844941998103,
      "tokens_seen": 1785397248
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046375125376128384,
      "loss": 1.4465,
      "theoretical_loss": 3.491834104429982,
      "tokens_seen": 1785462784
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046373119358074225,
      "loss": 1.445,
      "theoretical_loss": 3.491823267371028,
      "tokens_seen": 1785528320
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004637111334002006,
      "loss": 1.3018,
      "theoretical_loss": 3.4918124308211995,
      "tokens_seen": 1785593856
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000463691073219659,
      "loss": 1.4219,
      "theoretical_loss": 3.4918015947804535,
      "tokens_seen": 1785659392
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046367101303911734,
      "loss": 1.3935,
      "theoretical_loss": 3.491790759248747,
      "tokens_seen": 1785724928
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004636509528585757,
      "loss": 1.3458,
      "theoretical_loss": 3.4917799242260372,
      "tokens_seen": 1785790464
    },
    {
      "epoch": 0.54,
      "objective/train/advantage_avg": 0.4935075044631958,
      "objective/train/docs_used": 652000,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5572582483291626,
      "objective/train/original_loss": 1.557258129119873,
      "objective/train/theoretical_loss": 3.4917690897122826,
      "objective/train/tokens_used": 1806316000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2444930225610733,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050593614578247,
      "objective/train/weighted_lm_loss": 1.6361744403839111,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4917690897122826,
      "tokens_seen": 1785856000
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004636308926780341,
      "loss": 1.5509,
      "theoretical_loss": 3.4917690897122826,
      "tokens_seen": 1785856000
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004636108324974925,
      "loss": 1.4316,
      "theoretical_loss": 3.49175825570744,
      "tokens_seen": 1785921536
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046359077231695084,
      "loss": 1.5212,
      "theoretical_loss": 3.491747422211467,
      "tokens_seen": 1785987072
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004635707121364092,
      "loss": 1.3831,
      "theoretical_loss": 3.4917365892243204,
      "tokens_seen": 1786052608
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004635506519558676,
      "loss": 1.4122,
      "theoretical_loss": 3.491725756745958,
      "tokens_seen": 1786118144
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000463530591775326,
      "loss": 1.4652,
      "theoretical_loss": 3.4917149247763377,
      "tokens_seen": 1786183680
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046351053159478434,
      "loss": 1.3724,
      "theoretical_loss": 3.491704093315416,
      "tokens_seen": 1786249216
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004634904714142427,
      "loss": 1.4842,
      "theoretical_loss": 3.4916932623631514,
      "tokens_seen": 1786314752
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046347041123370107,
      "loss": 1.327,
      "theoretical_loss": 3.491682431919501,
      "tokens_seen": 1786380288
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004634503510531595,
      "loss": 1.4982,
      "theoretical_loss": 3.4916716019844216,
      "tokens_seen": 1786445824
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046343029087261785,
      "loss": 1.4283,
      "theoretical_loss": 3.4916607725578714,
      "tokens_seen": 1786511360
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004634102306920762,
      "loss": 1.469,
      "theoretical_loss": 3.491649943639808,
      "tokens_seen": 1786576896
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046339017051153457,
      "loss": 1.3691,
      "theoretical_loss": 3.491639115230188,
      "tokens_seen": 1786642432
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000463370110330993,
      "loss": 1.292,
      "theoretical_loss": 3.49162828732897,
      "tokens_seen": 1786707968
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046335005015045135,
      "loss": 1.5666,
      "theoretical_loss": 3.4916174599361103,
      "tokens_seen": 1786773504
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004633299899699097,
      "loss": 1.594,
      "theoretical_loss": 3.491606633051567,
      "tokens_seen": 1786839040
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004633099297893681,
      "loss": 1.5104,
      "theoretical_loss": 3.4915958066752983,
      "tokens_seen": 1786904576
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046328986960882644,
      "loss": 1.426,
      "theoretical_loss": 3.49158498080726,
      "tokens_seen": 1786970112
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046326980942828485,
      "loss": 1.4444,
      "theoretical_loss": 3.491574155447411,
      "tokens_seen": 1787035648
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004632497492477432,
      "loss": 1.3807,
      "theoretical_loss": 3.4915633305957083,
      "tokens_seen": 1787101184
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004632296890672016,
      "loss": 1.4756,
      "theoretical_loss": 3.4915525062521096,
      "tokens_seen": 1787166720
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046320962888666,
      "loss": 1.4193,
      "theoretical_loss": 3.4915416824165724,
      "tokens_seen": 1787232256
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004631895687061184,
      "loss": 1.3898,
      "theoretical_loss": 3.491530859089054,
      "tokens_seen": 1787297792
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004631695085255768,
      "loss": 1.4626,
      "theoretical_loss": 3.491520036269512,
      "tokens_seen": 1787363328
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046314944834503514,
      "loss": 1.4529,
      "theoretical_loss": 3.4915092139579036,
      "tokens_seen": 1787428864
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004631293881644935,
      "loss": 1.2308,
      "theoretical_loss": 3.4914983921541873,
      "tokens_seen": 1787494400
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004631093279839519,
      "loss": 1.4001,
      "theoretical_loss": 3.49148757085832,
      "tokens_seen": 1787559936
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004630892678034103,
      "loss": 1.6014,
      "theoretical_loss": 3.4914767500702597,
      "tokens_seen": 1787625472
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046306920762286864,
      "loss": 1.4562,
      "theoretical_loss": 3.491465929789963,
      "tokens_seen": 1787691008
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000463049147442327,
      "loss": 1.4322,
      "theoretical_loss": 3.4914551100173883,
      "tokens_seen": 1787756544
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046302908726178536,
      "loss": 1.4941,
      "theoretical_loss": 3.4914442907524927,
      "tokens_seen": 1787822080
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004630090270812438,
      "loss": 1.4385,
      "theoretical_loss": 3.491433471995234,
      "tokens_seen": 1787887616
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046298896690070214,
      "loss": 1.3256,
      "theoretical_loss": 3.4914226537455697,
      "tokens_seen": 1787953152
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004629689067201605,
      "loss": 1.4738,
      "theoretical_loss": 3.491411836003458,
      "tokens_seen": 1788018688
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046294884653961887,
      "loss": 1.3437,
      "theoretical_loss": 3.4914010187688556,
      "tokens_seen": 1788084224
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004629287863590773,
      "loss": 1.4113,
      "theoretical_loss": 3.4913902020417202,
      "tokens_seen": 1788149760
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046290872617853565,
      "loss": 1.4333,
      "theoretical_loss": 3.49137938582201,
      "tokens_seen": 1788215296
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000462888665997994,
      "loss": 1.466,
      "theoretical_loss": 3.4913685701096817,
      "tokens_seen": 1788280832
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046286860581745237,
      "loss": 1.4154,
      "theoretical_loss": 3.4913577549046937,
      "tokens_seen": 1788346368
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046284854563691073,
      "loss": 1.3735,
      "theoretical_loss": 3.4913469402070034,
      "tokens_seen": 1788411904
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046282848545636915,
      "loss": 1.4715,
      "theoretical_loss": 3.4913361260165683,
      "tokens_seen": 1788477440
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004628084252758275,
      "loss": 1.3792,
      "theoretical_loss": 3.491325312333346,
      "tokens_seen": 1788542976
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004627883650952859,
      "loss": 1.3532,
      "theoretical_loss": 3.491314499157294,
      "tokens_seen": 1788608512
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046276830491474424,
      "loss": 1.39,
      "theoretical_loss": 3.4913036864883704,
      "tokens_seen": 1788674048
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046274824473420265,
      "loss": 1.4386,
      "theoretical_loss": 3.4912928743265326,
      "tokens_seen": 1788739584
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000462728184553661,
      "loss": 1.4289,
      "theoretical_loss": 3.491282062671738,
      "tokens_seen": 1788805120
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004627081243731194,
      "loss": 1.3767,
      "theoretical_loss": 3.491271251523945,
      "tokens_seen": 1788870656
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046268806419257774,
      "loss": 1.4063,
      "theoretical_loss": 3.49126044088311,
      "tokens_seen": 1788936192
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004626680040120361,
      "loss": 1.4862,
      "theoretical_loss": 3.491249630749192,
      "tokens_seen": 1789001728
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004626479438314945,
      "loss": 1.4023,
      "theoretical_loss": 3.4912388211221472,
      "tokens_seen": 1789067264
    },
    {
      "epoch": 0.54,
      "objective/train/advantage_avg": 0.4920370578765869,
      "objective/train/docs_used": 653179,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.452174186706543,
      "objective/train/original_loss": 1.452174186706543,
      "objective/train/theoretical_loss": 3.491228012001935,
      "objective/train/tokens_used": 1809592800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24276697635650635,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504378080368042,
      "objective/train/weighted_lm_loss": 1.5254582166671753,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.491228012001935,
      "tokens_seen": 1789132800
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004626278836509529,
      "loss": 1.3981,
      "theoretical_loss": 3.491228012001935,
      "tokens_seen": 1789132800
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046260782347041124,
      "loss": 1.4205,
      "theoretical_loss": 3.4912172033885116,
      "tokens_seen": 1789198336
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004625877632898696,
      "loss": 1.3698,
      "theoretical_loss": 3.4912063952818353,
      "tokens_seen": 1789263872
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000462567703109328,
      "loss": 1.4544,
      "theoretical_loss": 3.491195587681864,
      "tokens_seen": 1789329408
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004625476429287864,
      "loss": 1.4255,
      "theoretical_loss": 3.4911847805885547,
      "tokens_seen": 1789394944
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046252758274824475,
      "loss": 1.4564,
      "theoretical_loss": 3.4911739740018657,
      "tokens_seen": 1789460480
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004625075225677031,
      "loss": 1.3313,
      "theoretical_loss": 3.4911631679217545,
      "tokens_seen": 1789526016
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046248746238716147,
      "loss": 1.4928,
      "theoretical_loss": 3.4911523623481786,
      "tokens_seen": 1789591552
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004624674022066199,
      "loss": 1.4005,
      "theoretical_loss": 3.491141557281096,
      "tokens_seen": 1789657088
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046244734202607825,
      "loss": 1.5564,
      "theoretical_loss": 3.4911307527204647,
      "tokens_seen": 1789722624
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004624272818455366,
      "loss": 1.4563,
      "theoretical_loss": 3.491119948666242,
      "tokens_seen": 1789788160
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000462407221664995,
      "loss": 1.4477,
      "theoretical_loss": 3.4911091451183847,
      "tokens_seen": 1789853696
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004623871614844534,
      "loss": 1.559,
      "theoretical_loss": 3.491098342076852,
      "tokens_seen": 1789919232
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046236710130391175,
      "loss": 1.3675,
      "theoretical_loss": 3.491087539541602,
      "tokens_seen": 1789984768
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004623470411233701,
      "loss": 1.4894,
      "theoretical_loss": 3.4910767375125906,
      "tokens_seen": 1790050304
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004623269809428285,
      "loss": 1.3222,
      "theoretical_loss": 3.491065935989777,
      "tokens_seen": 1790115840
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046230692076228684,
      "loss": 1.453,
      "theoretical_loss": 3.4910551349731183,
      "tokens_seen": 1790181376
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046228686058174525,
      "loss": 1.4955,
      "theoretical_loss": 3.491044334462572,
      "tokens_seen": 1790246912
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004622668004012036,
      "loss": 1.2844,
      "theoretical_loss": 3.4910335344580967,
      "tokens_seen": 1790312448
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000462246740220662,
      "loss": 1.3132,
      "theoretical_loss": 3.4910227349596497,
      "tokens_seen": 1790377984
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046222668004012034,
      "loss": 1.5109,
      "theoretical_loss": 3.4910119359671885,
      "tokens_seen": 1790443520
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046220661985957876,
      "loss": 1.3889,
      "theoretical_loss": 3.4910011374806715,
      "tokens_seen": 1790509056
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004621865596790371,
      "loss": 1.4466,
      "theoretical_loss": 3.490990339500056,
      "tokens_seen": 1790574592
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004621664994984955,
      "loss": 1.3487,
      "theoretical_loss": 3.4909795420253,
      "tokens_seen": 1790640128
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046214643931795384,
      "loss": 1.2529,
      "theoretical_loss": 3.490968745056361,
      "tokens_seen": 1790705664
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004621263791374122,
      "loss": 1.4207,
      "theoretical_loss": 3.4909579485931976,
      "tokens_seen": 1790771200
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004621063189568706,
      "loss": 1.4926,
      "theoretical_loss": 3.4909471526357665,
      "tokens_seen": 1790836736
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000462086258776329,
      "loss": 1.3709,
      "theoretical_loss": 3.490936357184026,
      "tokens_seen": 1790902272
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046206619859578735,
      "loss": 1.3839,
      "theoretical_loss": 3.4909255622379343,
      "tokens_seen": 1790967808
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004620461384152457,
      "loss": 1.5271,
      "theoretical_loss": 3.490914767797449,
      "tokens_seen": 1791033344
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004620260782347041,
      "loss": 1.4146,
      "theoretical_loss": 3.490903973862527,
      "tokens_seen": 1791098880
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004620060180541625,
      "loss": 1.3875,
      "theoretical_loss": 3.4908931804331274,
      "tokens_seen": 1791164416
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046198595787362085,
      "loss": 1.3672,
      "theoretical_loss": 3.490882387509207,
      "tokens_seen": 1791229952
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004619658976930792,
      "loss": 1.5195,
      "theoretical_loss": 3.4908715950907245,
      "tokens_seen": 1791295488
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046194583751253763,
      "loss": 1.399,
      "theoretical_loss": 3.490860803177638,
      "tokens_seen": 1791361024
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000461925777331996,
      "loss": 1.5037,
      "theoretical_loss": 3.490850011769904,
      "tokens_seen": 1791426560
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046190571715145435,
      "loss": 1.389,
      "theoretical_loss": 3.490839220867481,
      "tokens_seen": 1791492096
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004618856569709127,
      "loss": 1.4145,
      "theoretical_loss": 3.4908284304703274,
      "tokens_seen": 1791557632
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004618655967903711,
      "loss": 1.3984,
      "theoretical_loss": 3.4908176405784004,
      "tokens_seen": 1791623168
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004618455366098295,
      "loss": 1.2228,
      "theoretical_loss": 3.490806851191658,
      "tokens_seen": 1791688704
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046182547642928786,
      "loss": 1.4279,
      "theoretical_loss": 3.490796062310058,
      "tokens_seen": 1791754240
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004618054162487462,
      "loss": 1.3874,
      "theoretical_loss": 3.4907852739335588,
      "tokens_seen": 1791819776
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004617853560682046,
      "loss": 1.4112,
      "theoretical_loss": 3.4907744860621177,
      "tokens_seen": 1791885312
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000461765295887663,
      "loss": 1.4824,
      "theoretical_loss": 3.4907636986956927,
      "tokens_seen": 1791950848
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046174523570712136,
      "loss": 1.353,
      "theoretical_loss": 3.4907529118342415,
      "tokens_seen": 1792016384
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004617251755265797,
      "loss": 1.3802,
      "theoretical_loss": 3.4907421254777233,
      "tokens_seen": 1792081920
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046170511534603814,
      "loss": 1.4359,
      "theoretical_loss": 3.4907313396260937,
      "tokens_seen": 1792147456
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004616850551654965,
      "loss": 1.3553,
      "theoretical_loss": 3.4907205542793127,
      "tokens_seen": 1792212992
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004616649949849549,
      "loss": 1.447,
      "theoretical_loss": 3.490709769437337,
      "tokens_seen": 1792278528
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004616449348044133,
      "loss": 1.3887,
      "theoretical_loss": 3.490698985100125,
      "tokens_seen": 1792344064
    },
    {
      "epoch": 0.54,
      "objective/train/advantage_avg": 0.4924985468387604,
      "objective/train/docs_used": 654198,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1999764442443848,
      "objective/train/original_loss": 1.1999763250350952,
      "objective/train/theoretical_loss": 3.4906882012676346,
      "objective/train/tokens_used": 1812869600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24291373789310455,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504847764968872,
      "objective/train/weighted_lm_loss": 1.2607429027557373,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4906882012676346,
      "tokens_seen": 1792409600
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046162487462387164,
      "loss": 1.279,
      "theoretical_loss": 3.4906882012676346,
      "tokens_seen": 1792409600
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046160481444333,
      "loss": 1.3867,
      "theoretical_loss": 3.4906774179398234,
      "tokens_seen": 1792475136
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004615847542627884,
      "loss": 1.3699,
      "theoretical_loss": 3.49066663511665,
      "tokens_seen": 1792540672
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004615646940822468,
      "loss": 1.3465,
      "theoretical_loss": 3.4906558527980716,
      "tokens_seen": 1792606208
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046154463390170515,
      "loss": 1.489,
      "theoretical_loss": 3.490645070984046,
      "tokens_seen": 1792671744
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004615245737211635,
      "loss": 1.4868,
      "theoretical_loss": 3.4906342896745324,
      "tokens_seen": 1792737280
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046150451354062187,
      "loss": 1.3733,
      "theoretical_loss": 3.4906235088694872,
      "tokens_seen": 1792802816
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004614844533600803,
      "loss": 1.3176,
      "theoretical_loss": 3.4906127285688697,
      "tokens_seen": 1792868352
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046146439317953865,
      "loss": 1.4765,
      "theoretical_loss": 3.490601948772637,
      "tokens_seen": 1792933888
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000461444332998997,
      "loss": 1.4326,
      "theoretical_loss": 3.490591169480747,
      "tokens_seen": 1792999424
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004614242728184554,
      "loss": 1.3982,
      "theoretical_loss": 3.4905803906931587,
      "tokens_seen": 1793064960
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004614042126379138,
      "loss": 1.3649,
      "theoretical_loss": 3.490569612409829,
      "tokens_seen": 1793130496
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046138415245737215,
      "loss": 1.378,
      "theoretical_loss": 3.490558834630716,
      "tokens_seen": 1793196032
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004613640922768305,
      "loss": 1.2665,
      "theoretical_loss": 3.4905480573557783,
      "tokens_seen": 1793261568
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004613440320962889,
      "loss": 1.3966,
      "theoretical_loss": 3.4905372805849737,
      "tokens_seen": 1793327104
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046132397191574724,
      "loss": 1.4093,
      "theoretical_loss": 3.490526504318259,
      "tokens_seen": 1793392640
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046130391173520566,
      "loss": 1.4992,
      "theoretical_loss": 3.4905157285555948,
      "tokens_seen": 1793458176
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000461283851554664,
      "loss": 1.3545,
      "theoretical_loss": 3.490504953296936,
      "tokens_seen": 1793523712
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004612637913741224,
      "loss": 1.1828,
      "theoretical_loss": 3.490494178542243,
      "tokens_seen": 1793589248
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046124373119358074,
      "loss": 1.3025,
      "theoretical_loss": 3.490483404291473,
      "tokens_seen": 1793654784
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046122367101303916,
      "loss": 1.3862,
      "theoretical_loss": 3.490472630544584,
      "tokens_seen": 1793720320
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004612036108324975,
      "loss": 1.4753,
      "theoretical_loss": 3.4904618573015336,
      "tokens_seen": 1793785856
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004611835506519559,
      "loss": 1.3097,
      "theoretical_loss": 3.4904510845622805,
      "tokens_seen": 1793851392
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046116349047141425,
      "loss": 1.4684,
      "theoretical_loss": 3.490440312326782,
      "tokens_seen": 1793916928
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004611434302908726,
      "loss": 1.5144,
      "theoretical_loss": 3.490429540594997,
      "tokens_seen": 1793982464
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000461123370110331,
      "loss": 1.363,
      "theoretical_loss": 3.490418769366883,
      "tokens_seen": 1794048000
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004611033099297894,
      "loss": 1.3875,
      "theoretical_loss": 3.490407998642399,
      "tokens_seen": 1794113536
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046108324974924775,
      "loss": 1.3701,
      "theoretical_loss": 3.4903972284215015,
      "tokens_seen": 1794179072
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004610631895687061,
      "loss": 1.4412,
      "theoretical_loss": 3.4903864587041493,
      "tokens_seen": 1794244608
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004610431293881645,
      "loss": 1.5223,
      "theoretical_loss": 3.4903756894903006,
      "tokens_seen": 1794310144
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004610230692076229,
      "loss": 1.3999,
      "theoretical_loss": 3.4903649207799137,
      "tokens_seen": 1794375680
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046100300902708125,
      "loss": 1.3425,
      "theoretical_loss": 3.4903541525729462,
      "tokens_seen": 1794441216
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004609829488465396,
      "loss": 1.3944,
      "theoretical_loss": 3.490343384869356,
      "tokens_seen": 1794506752
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046096288866599803,
      "loss": 1.5517,
      "theoretical_loss": 3.4903326176691016,
      "tokens_seen": 1794572288
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004609428284854564,
      "loss": 1.4421,
      "theoretical_loss": 3.4903218509721414,
      "tokens_seen": 1794637824
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046092276830491475,
      "loss": 1.3257,
      "theoretical_loss": 3.4903110847784324,
      "tokens_seen": 1794703360
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004609027081243731,
      "loss": 1.472,
      "theoretical_loss": 3.4903003190879343,
      "tokens_seen": 1794768896
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004608826479438315,
      "loss": 1.4304,
      "theoretical_loss": 3.4902895539006034,
      "tokens_seen": 1794834432
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004608625877632899,
      "loss": 1.3698,
      "theoretical_loss": 3.4902787892163993,
      "tokens_seen": 1794899968
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046084252758274826,
      "loss": 1.4145,
      "theoretical_loss": 3.4902680250352796,
      "tokens_seen": 1794965504
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004608224674022066,
      "loss": 1.4467,
      "theoretical_loss": 3.490257261357202,
      "tokens_seen": 1795031040
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000460802407221665,
      "loss": 1.4216,
      "theoretical_loss": 3.490246498182125,
      "tokens_seen": 1795096576
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004607823470411234,
      "loss": 1.3937,
      "theoretical_loss": 3.490235735510007,
      "tokens_seen": 1795162112
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046076228686058176,
      "loss": 1.4592,
      "theoretical_loss": 3.4902249733408057,
      "tokens_seen": 1795227648
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004607422266800401,
      "loss": 1.4947,
      "theoretical_loss": 3.490214211674479,
      "tokens_seen": 1795293184
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004607221664994985,
      "loss": 1.4538,
      "theoretical_loss": 3.490203450510986,
      "tokens_seen": 1795358720
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046070210631895685,
      "loss": 1.3924,
      "theoretical_loss": 3.490192689850284,
      "tokens_seen": 1795424256
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046068204613841526,
      "loss": 1.489,
      "theoretical_loss": 3.4901819296923318,
      "tokens_seen": 1795489792
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004606619859578736,
      "loss": 1.4002,
      "theoretical_loss": 3.4901711700370868,
      "tokens_seen": 1795555328
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000460641925777332,
      "loss": 1.5174,
      "theoretical_loss": 3.490160410884508,
      "tokens_seen": 1795620864
    },
    {
      "epoch": 0.54,
      "objective/train/advantage_avg": 0.49078354239463806,
      "objective/train/docs_used": 655419,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.37962007522583,
      "objective/train/original_loss": 1.379619836807251,
      "objective/train/theoretical_loss": 3.4901496522345528,
      "objective/train/tokens_used": 1816146400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24142871797084808,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503056049346924,
      "objective/train/weighted_lm_loss": 1.4489210844039917,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4901496522345528,
      "tokens_seen": 1795686400
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046062186559679035,
      "loss": 1.3863,
      "theoretical_loss": 3.4901496522345528,
      "tokens_seen": 1795686400
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046060180541624877,
      "loss": 1.4049,
      "theoretical_loss": 3.49013889408718,
      "tokens_seen": 1795751936
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046058174523570713,
      "loss": 1.4051,
      "theoretical_loss": 3.4901281364423475,
      "tokens_seen": 1795817472
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004605616850551655,
      "loss": 1.3652,
      "theoretical_loss": 3.4901173793000133,
      "tokens_seen": 1795883008
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046054162487462385,
      "loss": 1.5175,
      "theoretical_loss": 3.490106622660136,
      "tokens_seen": 1795948544
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004605215646940822,
      "loss": 1.5281,
      "theoretical_loss": 3.4900958665226733,
      "tokens_seen": 1796014080
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046050150451354063,
      "loss": 1.402,
      "theoretical_loss": 3.490085110887583,
      "tokens_seen": 1796079616
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000460481444332999,
      "loss": 1.3744,
      "theoretical_loss": 3.4900743557548255,
      "tokens_seen": 1796145152
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046046138415245736,
      "loss": 1.4586,
      "theoretical_loss": 3.4900636011243567,
      "tokens_seen": 1796210688
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004604413239719157,
      "loss": 1.4738,
      "theoretical_loss": 3.490052846996136,
      "tokens_seen": 1796276224
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046042126379137414,
      "loss": 1.2883,
      "theoretical_loss": 3.4900420933701204,
      "tokens_seen": 1796341760
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004604012036108325,
      "loss": 1.5811,
      "theoretical_loss": 3.49003134024627,
      "tokens_seen": 1796407296
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046038114343029086,
      "loss": 1.5001,
      "theoretical_loss": 3.4900205876245414,
      "tokens_seen": 1796472832
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004603610832497492,
      "loss": 1.4258,
      "theoretical_loss": 3.4900098355048934,
      "tokens_seen": 1796538368
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004603410230692076,
      "loss": 1.6231,
      "theoretical_loss": 3.4899990838872847,
      "tokens_seen": 1796603904
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000460320962888666,
      "loss": 1.416,
      "theoretical_loss": 3.4899883327716728,
      "tokens_seen": 1796669440
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046030090270812436,
      "loss": 1.388,
      "theoretical_loss": 3.4899775821580166,
      "tokens_seen": 1796734976
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004602808425275827,
      "loss": 1.5025,
      "theoretical_loss": 3.4899668320462736,
      "tokens_seen": 1796800512
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004602607823470411,
      "loss": 1.4362,
      "theoretical_loss": 3.489956082436403,
      "tokens_seen": 1796866048
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004602407221664995,
      "loss": 1.3162,
      "theoretical_loss": 3.489945333328362,
      "tokens_seen": 1796931584
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046022066198595787,
      "loss": 1.473,
      "theoretical_loss": 3.4899345847221097,
      "tokens_seen": 1796997120
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004602006018054163,
      "loss": 1.291,
      "theoretical_loss": 3.489923836617604,
      "tokens_seen": 1797062656
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046018054162487465,
      "loss": 1.5049,
      "theoretical_loss": 3.4899130890148036,
      "tokens_seen": 1797128192
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000460160481444333,
      "loss": 1.363,
      "theoretical_loss": 3.489902341913666,
      "tokens_seen": 1797193728
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004601404212637914,
      "loss": 1.3504,
      "theoretical_loss": 3.4898915953141505,
      "tokens_seen": 1797259264
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004601203610832498,
      "loss": 1.3478,
      "theoretical_loss": 3.489880849216215,
      "tokens_seen": 1797324800
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046010030090270815,
      "loss": 1.3949,
      "theoretical_loss": 3.489870103619817,
      "tokens_seen": 1797390336
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004600802407221665,
      "loss": 1.3639,
      "theoretical_loss": 3.4898593585249156,
      "tokens_seen": 1797455872
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046006018054162493,
      "loss": 1.5097,
      "theoretical_loss": 3.4898486139314695,
      "tokens_seen": 1797521408
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004600401203610833,
      "loss": 1.5408,
      "theoretical_loss": 3.489837869839436,
      "tokens_seen": 1797586944
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046002006018054165,
      "loss": 1.4227,
      "theoretical_loss": 3.489827126248774,
      "tokens_seen": 1797652480
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00046,
      "loss": 1.4072,
      "theoretical_loss": 3.489816383159442,
      "tokens_seen": 1797718016
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00045997993981945843,
      "loss": 1.43,
      "theoretical_loss": 3.489805640571398,
      "tokens_seen": 1797783552
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004599598796389168,
      "loss": 1.2978,
      "theoretical_loss": 3.4897948984846003,
      "tokens_seen": 1797849088
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00045993981945837515,
      "loss": 1.4764,
      "theoretical_loss": 3.489784156899008,
      "tokens_seen": 1797914624
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004599197592778335,
      "loss": 1.3967,
      "theoretical_loss": 3.489773415814578,
      "tokens_seen": 1797980160
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004598996990972919,
      "loss": 1.3524,
      "theoretical_loss": 3.48976267523127,
      "tokens_seen": 1798045696
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004598796389167503,
      "loss": 1.4043,
      "theoretical_loss": 3.489751935149041,
      "tokens_seen": 1798111232
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00045985957873620866,
      "loss": 1.343,
      "theoretical_loss": 3.4897411955678512,
      "tokens_seen": 1798176768
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.000459839518555667,
      "loss": 1.4634,
      "theoretical_loss": 3.4897304564876572,
      "tokens_seen": 1798242304
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004598194583751254,
      "loss": 1.4284,
      "theoretical_loss": 3.4897197179084185,
      "tokens_seen": 1798307840
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0004597993981945838,
      "loss": 1.4857,
      "theoretical_loss": 3.489708979830093,
      "tokens_seen": 1798373376
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00045977933801404216,
      "loss": 1.4208,
      "theoretical_loss": 3.489698242252639,
      "tokens_seen": 1798438912
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004597592778335005,
      "loss": 1.4867,
      "theoretical_loss": 3.4896875051760152,
      "tokens_seen": 1798504448
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004597392176529589,
      "loss": 1.4195,
      "theoretical_loss": 3.48967676860018,
      "tokens_seen": 1798569984
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045971915747241725,
      "loss": 1.3359,
      "theoretical_loss": 3.4896660325250917,
      "tokens_seen": 1798635520
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045969909729187566,
      "loss": 1.4788,
      "theoretical_loss": 3.4896552969507084,
      "tokens_seen": 1798701056
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000459679037111334,
      "loss": 1.4281,
      "theoretical_loss": 3.4896445618769887,
      "tokens_seen": 1798766592
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004596589769307924,
      "loss": 1.5906,
      "theoretical_loss": 3.4896338273038916,
      "tokens_seen": 1798832128
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045963891675025075,
      "loss": 1.4569,
      "theoretical_loss": 3.4896230932313745,
      "tokens_seen": 1798897664
    },
    {
      "epoch": 0.55,
      "objective/train/advantage_avg": 0.49270281195640564,
      "objective/train/docs_used": 656735,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.338792324066162,
      "objective/train/original_loss": 1.338792085647583,
      "objective/train/theoretical_loss": 3.4896123596593966,
      "objective/train/tokens_used": 1819423200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24320612847805023,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050506591796875,
      "objective/train/weighted_lm_loss": 1.406286358833313,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4896123596593966,
      "tokens_seen": 1798963200
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045961885656970917,
      "loss": 1.3641,
      "theoretical_loss": 3.4896123596593966,
      "tokens_seen": 1798963200
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045959879638916753,
      "loss": 1.3573,
      "theoretical_loss": 3.4896016265879153,
      "tokens_seen": 1799028736
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004595787362086259,
      "loss": 1.3471,
      "theoretical_loss": 3.4895908940168905,
      "tokens_seen": 1799094272
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045955867602808425,
      "loss": 1.4733,
      "theoretical_loss": 3.4895801619462796,
      "tokens_seen": 1799159808
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004595386158475426,
      "loss": 1.4229,
      "theoretical_loss": 3.489569430376042,
      "tokens_seen": 1799225344
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045951855566700103,
      "loss": 1.3044,
      "theoretical_loss": 3.4895586993061345,
      "tokens_seen": 1799290880
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004594984954864594,
      "loss": 1.5923,
      "theoretical_loss": 3.489547968736517,
      "tokens_seen": 1799356416
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045947843530591776,
      "loss": 1.4199,
      "theoretical_loss": 3.4895372386671477,
      "tokens_seen": 1799421952
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004594583751253761,
      "loss": 1.3729,
      "theoretical_loss": 3.489526509097985,
      "tokens_seen": 1799487488
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045943831494483454,
      "loss": 1.4585,
      "theoretical_loss": 3.489515780028986,
      "tokens_seen": 1799553024
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004594182547642929,
      "loss": 1.5515,
      "theoretical_loss": 3.4895050514601116,
      "tokens_seen": 1799618560
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045939819458375126,
      "loss": 1.4409,
      "theoretical_loss": 3.4894943233913187,
      "tokens_seen": 1799684096
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004593781344032096,
      "loss": 1.4308,
      "theoretical_loss": 3.4894835958225663,
      "tokens_seen": 1799749632
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000459358074222668,
      "loss": 1.4071,
      "theoretical_loss": 3.489472868753813,
      "tokens_seen": 1799815168
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004593380140421264,
      "loss": 1.4101,
      "theoretical_loss": 3.4894621421850163,
      "tokens_seen": 1799880704
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045931795386158476,
      "loss": 1.564,
      "theoretical_loss": 3.4894514161161356,
      "tokens_seen": 1799946240
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004592978936810431,
      "loss": 1.3908,
      "theoretical_loss": 3.48944069054713,
      "tokens_seen": 1800011776
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004592778335005015,
      "loss": 1.3926,
      "theoretical_loss": 3.489429965477956,
      "tokens_seen": 1800077312
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004592577733199599,
      "loss": 1.3675,
      "theoretical_loss": 3.489419240908574,
      "tokens_seen": 1800142848
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045923771313941827,
      "loss": 1.4585,
      "theoretical_loss": 3.489408516838942,
      "tokens_seen": 1800208384
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045921765295887663,
      "loss": 1.3783,
      "theoretical_loss": 3.489397793269018,
      "tokens_seen": 1800273920
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000459197592778335,
      "loss": 1.5658,
      "theoretical_loss": 3.4893870701987613,
      "tokens_seen": 1800339456
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045917753259779335,
      "loss": 1.4468,
      "theoretical_loss": 3.48937634762813,
      "tokens_seen": 1800404992
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045915747241725177,
      "loss": 1.3292,
      "theoretical_loss": 3.489365625557082,
      "tokens_seen": 1800470528
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045913741223671013,
      "loss": 1.4072,
      "theoretical_loss": 3.489354903985577,
      "tokens_seen": 1800536064
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004591173520561685,
      "loss": 1.4172,
      "theoretical_loss": 3.489344182913573,
      "tokens_seen": 1800601600
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045909729187562686,
      "loss": 1.4337,
      "theoretical_loss": 3.489333462341029,
      "tokens_seen": 1800667136
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004590772316950853,
      "loss": 1.5475,
      "theoretical_loss": 3.4893227422679027,
      "tokens_seen": 1800732672
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045905717151454364,
      "loss": 1.4548,
      "theoretical_loss": 3.489312022694153,
      "tokens_seen": 1800798208
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000459037111334002,
      "loss": 1.2855,
      "theoretical_loss": 3.4893013036197384,
      "tokens_seen": 1800863744
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045901705115346036,
      "loss": 1.394,
      "theoretical_loss": 3.4892905850446185,
      "tokens_seen": 1800929280
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004589969909729187,
      "loss": 1.5414,
      "theoretical_loss": 3.4892798669687504,
      "tokens_seen": 1800994816
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045897693079237714,
      "loss": 1.2435,
      "theoretical_loss": 3.4892691493920935,
      "tokens_seen": 1801060352
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004589568706118355,
      "loss": 1.3987,
      "theoretical_loss": 3.489258432314606,
      "tokens_seen": 1801125888
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045893681043129386,
      "loss": 1.4099,
      "theoretical_loss": 3.489247715736247,
      "tokens_seen": 1801191424
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004589167502507522,
      "loss": 1.4031,
      "theoretical_loss": 3.4892369996569745,
      "tokens_seen": 1801256960
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045889669007021064,
      "loss": 1.5983,
      "theoretical_loss": 3.4892262840767474,
      "tokens_seen": 1801322496
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000458876629889669,
      "loss": 1.4159,
      "theoretical_loss": 3.489215568995524,
      "tokens_seen": 1801388032
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045885656970912737,
      "loss": 1.3439,
      "theoretical_loss": 3.489204854413264,
      "tokens_seen": 1801453568
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045883650952858573,
      "loss": 1.4165,
      "theoretical_loss": 3.489194140329925,
      "tokens_seen": 1801519104
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045881644934804415,
      "loss": 1.3784,
      "theoretical_loss": 3.4891834267454653,
      "tokens_seen": 1801584640
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004587963891675025,
      "loss": 1.2502,
      "theoretical_loss": 3.489172713659844,
      "tokens_seen": 1801650176
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045877632898696087,
      "loss": 1.436,
      "theoretical_loss": 3.48916200107302,
      "tokens_seen": 1801715712
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045875626880641923,
      "loss": 1.4137,
      "theoretical_loss": 3.489151288984952,
      "tokens_seen": 1801781248
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004587362086258776,
      "loss": 1.3803,
      "theoretical_loss": 3.4891405773955984,
      "tokens_seen": 1801846784
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000458716148445336,
      "loss": 1.3065,
      "theoretical_loss": 3.4891298663049177,
      "tokens_seen": 1801912320
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004586960882647944,
      "loss": 1.4245,
      "theoretical_loss": 3.489119155712868,
      "tokens_seen": 1801977856
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004586760280842528,
      "loss": 1.3183,
      "theoretical_loss": 3.4891084456194092,
      "tokens_seen": 1802043392
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045865596790371115,
      "loss": 1.3693,
      "theoretical_loss": 3.4890977360244992,
      "tokens_seen": 1802108928
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045863590772316957,
      "loss": 1.4779,
      "theoretical_loss": 3.4890870269280967,
      "tokens_seen": 1802174464
    },
    {
      "epoch": 0.55,
      "objective/train/advantage_avg": 0.4883981943130493,
      "objective/train/docs_used": 657910,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.441488265991211,
      "objective/train/original_loss": 1.441488265991211,
      "objective/train/theoretical_loss": 3.4890763183301607,
      "objective/train/tokens_used": 1822700000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24002443253993988,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500599145889282,
      "objective/train/weighted_lm_loss": 1.5139132738113403,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4890763183301607,
      "tokens_seen": 1802240000
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045861584754262793,
      "loss": 1.4992,
      "theoretical_loss": 3.4890763183301607,
      "tokens_seen": 1802240000
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004585957873620863,
      "loss": 1.4154,
      "theoretical_loss": 3.4890656102306496,
      "tokens_seen": 1802305536
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045857572718154465,
      "loss": 1.4706,
      "theoretical_loss": 3.489054902629522,
      "tokens_seen": 1802371072
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000458555667001003,
      "loss": 1.382,
      "theoretical_loss": 3.489044195526737,
      "tokens_seen": 1802436608
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045853560682046143,
      "loss": 1.4784,
      "theoretical_loss": 3.489033488922253,
      "tokens_seen": 1802502144
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004585155466399198,
      "loss": 1.4119,
      "theoretical_loss": 3.4890227828160283,
      "tokens_seen": 1802567680
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045849548645937816,
      "loss": 1.4578,
      "theoretical_loss": 3.489012077208022,
      "tokens_seen": 1802633216
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004584754262788365,
      "loss": 1.3873,
      "theoretical_loss": 3.489001372098193,
      "tokens_seen": 1802698752
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045845536609829494,
      "loss": 1.32,
      "theoretical_loss": 3.4889906674865,
      "tokens_seen": 1802764288
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004584353059177533,
      "loss": 1.5426,
      "theoretical_loss": 3.4889799633729015,
      "tokens_seen": 1802829824
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045841524573721166,
      "loss": 1.5095,
      "theoretical_loss": 3.488969259757356,
      "tokens_seen": 1802895360
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045839518555667,
      "loss": 1.402,
      "theoretical_loss": 3.4889585566398225,
      "tokens_seen": 1802960896
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004583751253761284,
      "loss": 1.4971,
      "theoretical_loss": 3.48894785402026,
      "tokens_seen": 1803026432
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004583550651955868,
      "loss": 1.4526,
      "theoretical_loss": 3.488937151898627,
      "tokens_seen": 1803091968
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045833500501504516,
      "loss": 1.3749,
      "theoretical_loss": 3.4889264502748816,
      "tokens_seen": 1803157504
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004583149448345035,
      "loss": 1.3896,
      "theoretical_loss": 3.4889157491489833,
      "tokens_seen": 1803223040
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004582948846539619,
      "loss": 1.3935,
      "theoretical_loss": 3.4889050485208912,
      "tokens_seen": 1803288576
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004582748244734203,
      "loss": 1.3698,
      "theoretical_loss": 3.488894348390563,
      "tokens_seen": 1803354112
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045825476429287867,
      "loss": 1.3907,
      "theoretical_loss": 3.488883648757958,
      "tokens_seen": 1803419648
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045823470411233703,
      "loss": 1.5747,
      "theoretical_loss": 3.4888729496230346,
      "tokens_seen": 1803485184
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004582146439317954,
      "loss": 1.3374,
      "theoretical_loss": 3.4888622509857523,
      "tokens_seen": 1803550720
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045819458375125375,
      "loss": 1.2647,
      "theoretical_loss": 3.4888515528460693,
      "tokens_seen": 1803616256
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045817452357071217,
      "loss": 1.3031,
      "theoretical_loss": 3.4888408552039447,
      "tokens_seen": 1803681792
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045815446339017053,
      "loss": 1.4117,
      "theoretical_loss": 3.4888301580593373,
      "tokens_seen": 1803747328
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004581344032096289,
      "loss": 1.4219,
      "theoretical_loss": 3.488819461412205,
      "tokens_seen": 1803812864
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045811434302908726,
      "loss": 1.5162,
      "theoretical_loss": 3.488808765262508,
      "tokens_seen": 1803878400
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004580942828485457,
      "loss": 1.4294,
      "theoretical_loss": 3.488798069610204,
      "tokens_seen": 1803943936
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045807422266800404,
      "loss": 1.4668,
      "theoretical_loss": 3.4887873744552524,
      "tokens_seen": 1804009472
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004580541624874624,
      "loss": 1.4349,
      "theoretical_loss": 3.4887766797976116,
      "tokens_seen": 1804075008
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045803410230692076,
      "loss": 1.5221,
      "theoretical_loss": 3.488765985637241,
      "tokens_seen": 1804140544
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004580140421263791,
      "loss": 1.4046,
      "theoretical_loss": 3.4887552919740985,
      "tokens_seen": 1804206080
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045799398194583754,
      "loss": 1.5436,
      "theoretical_loss": 3.488744598808144,
      "tokens_seen": 1804271616
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004579739217652959,
      "loss": 1.3556,
      "theoretical_loss": 3.488733906139336,
      "tokens_seen": 1804337152
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045795386158475426,
      "loss": 1.4892,
      "theoretical_loss": 3.4887232139676323,
      "tokens_seen": 1804402688
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004579338014042126,
      "loss": 1.52,
      "theoretical_loss": 3.488712522292993,
      "tokens_seen": 1804468224
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045791374122367104,
      "loss": 1.4204,
      "theoretical_loss": 3.4887018311153764,
      "tokens_seen": 1804533760
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004578936810431294,
      "loss": 1.4324,
      "theoretical_loss": 3.4886911404347414,
      "tokens_seen": 1804599296
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045787362086258777,
      "loss": 1.4206,
      "theoretical_loss": 3.4886804502510467,
      "tokens_seen": 1804664832
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045785356068204613,
      "loss": 1.4573,
      "theoretical_loss": 3.4886697605642514,
      "tokens_seen": 1804730368
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045783350050150455,
      "loss": 1.4143,
      "theoretical_loss": 3.4886590713743146,
      "tokens_seen": 1804795904
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004578134403209629,
      "loss": 1.2944,
      "theoretical_loss": 3.4886483826811947,
      "tokens_seen": 1804861440
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045779338014042127,
      "loss": 1.3411,
      "theoretical_loss": 3.4886376944848507,
      "tokens_seen": 1804926976
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045777331995987963,
      "loss": 1.417,
      "theoretical_loss": 3.488627006785242,
      "tokens_seen": 1804992512
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000457753259779338,
      "loss": 1.4207,
      "theoretical_loss": 3.4886163195823263,
      "tokens_seen": 1805058048
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004577331995987964,
      "loss": 1.3997,
      "theoretical_loss": 3.4886056328760633,
      "tokens_seen": 1805123584
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004577131394182548,
      "loss": 1.5403,
      "theoretical_loss": 3.488594946666412,
      "tokens_seen": 1805189120
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045769307923771314,
      "loss": 1.364,
      "theoretical_loss": 3.488584260953331,
      "tokens_seen": 1805254656
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004576730190571715,
      "loss": 1.4215,
      "theoretical_loss": 3.4885735757367793,
      "tokens_seen": 1805320192
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004576529588766299,
      "loss": 1.4524,
      "theoretical_loss": 3.4885628910167155,
      "tokens_seen": 1805385728
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004576328986960883,
      "loss": 1.4806,
      "theoretical_loss": 3.488552206793099,
      "tokens_seen": 1805451264
    },
    {
      "epoch": 0.55,
      "objective/train/advantage_avg": 0.49104854464530945,
      "objective/train/docs_used": 658936,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.207532525062561,
      "objective/train/original_loss": 1.2075326442718506,
      "objective/train/theoretical_loss": 3.4885415230658885,
      "objective/train/tokens_used": 1825976800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2421344369649887,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503356456756592,
      "objective/train/weighted_lm_loss": 1.268287181854248,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4885415230658885,
      "tokens_seen": 1805516800
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045761283851554664,
      "loss": 1.4531,
      "theoretical_loss": 3.4885415230658885,
      "tokens_seen": 1805516800
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000457592778335005,
      "loss": 1.4926,
      "theoretical_loss": 3.4885308398350428,
      "tokens_seen": 1805582336
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045757271815446336,
      "loss": 1.3976,
      "theoretical_loss": 3.488520157100521,
      "tokens_seen": 1805647872
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004575526579739218,
      "loss": 1.5406,
      "theoretical_loss": 3.4885094748622816,
      "tokens_seen": 1805713408
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045753259779338014,
      "loss": 1.3486,
      "theoretical_loss": 3.4884987931202844,
      "tokens_seen": 1805778944
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004575125376128385,
      "loss": 1.5157,
      "theoretical_loss": 3.4884881118744877,
      "tokens_seen": 1805844480
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045749247743229687,
      "loss": 1.4313,
      "theoretical_loss": 3.4884774311248505,
      "tokens_seen": 1805910016
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004574724172517553,
      "loss": 1.4674,
      "theoretical_loss": 3.488466750871332,
      "tokens_seen": 1805975552
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045745235707121364,
      "loss": 1.3992,
      "theoretical_loss": 3.488456071113891,
      "tokens_seen": 1806041088
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000457432296890672,
      "loss": 1.3768,
      "theoretical_loss": 3.488445391852486,
      "tokens_seen": 1806106624
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045741223671013037,
      "loss": 1.3374,
      "theoretical_loss": 3.4884347130870768,
      "tokens_seen": 1806172160
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045739217652958873,
      "loss": 1.3087,
      "theoretical_loss": 3.488424034817622,
      "tokens_seen": 1806237696
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045737211634904715,
      "loss": 1.3847,
      "theoretical_loss": 3.48841335704408,
      "tokens_seen": 1806303232
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004573520561685055,
      "loss": 1.3958,
      "theoretical_loss": 3.488402679766411,
      "tokens_seen": 1806368768
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045733199598796387,
      "loss": 1.3972,
      "theoretical_loss": 3.4883920029845727,
      "tokens_seen": 1806434304
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045731193580742223,
      "loss": 1.4213,
      "theoretical_loss": 3.4883813266985246,
      "tokens_seen": 1806499840
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045729187562688065,
      "loss": 1.4131,
      "theoretical_loss": 3.4883706509082266,
      "tokens_seen": 1806565376
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000457271815446339,
      "loss": 1.3642,
      "theoretical_loss": 3.488359975613636,
      "tokens_seen": 1806630912
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004572517552657974,
      "loss": 1.5124,
      "theoretical_loss": 3.488349300814713,
      "tokens_seen": 1806696448
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045723169508525574,
      "loss": 1.3705,
      "theoretical_loss": 3.4883386265114167,
      "tokens_seen": 1806761984
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004572116349047141,
      "loss": 1.4204,
      "theoretical_loss": 3.488327952703705,
      "tokens_seen": 1806827520
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045719157472417257,
      "loss": 1.4397,
      "theoretical_loss": 3.4883172793915387,
      "tokens_seen": 1806893056
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045717151454363093,
      "loss": 1.4853,
      "theoretical_loss": 3.4883066065748745,
      "tokens_seen": 1806958592
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004571514543630893,
      "loss": 1.3957,
      "theoretical_loss": 3.4882959342536735,
      "tokens_seen": 1807024128
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045713139418254766,
      "loss": 1.5057,
      "theoretical_loss": 3.4882852624278935,
      "tokens_seen": 1807089664
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004571113340020061,
      "loss": 1.4493,
      "theoretical_loss": 3.488274591097494,
      "tokens_seen": 1807155200
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045709127382146444,
      "loss": 1.4784,
      "theoretical_loss": 3.488263920262434,
      "tokens_seen": 1807220736
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004570712136409228,
      "loss": 1.4483,
      "theoretical_loss": 3.4882532499226726,
      "tokens_seen": 1807286272
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045705115346038116,
      "loss": 1.3833,
      "theoretical_loss": 3.4882425800781687,
      "tokens_seen": 1807351808
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004570310932798395,
      "loss": 1.4511,
      "theoretical_loss": 3.4882319107288815,
      "tokens_seen": 1807417344
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045701103309929794,
      "loss": 1.4029,
      "theoretical_loss": 3.4882212418747693,
      "tokens_seen": 1807482880
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004569909729187563,
      "loss": 1.3667,
      "theoretical_loss": 3.4882105735157927,
      "tokens_seen": 1807548416
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045697091273821466,
      "loss": 1.3202,
      "theoretical_loss": 3.4881999056519097,
      "tokens_seen": 1807613952
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000456950852557673,
      "loss": 1.4295,
      "theoretical_loss": 3.4881892382830797,
      "tokens_seen": 1807679488
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045693079237713144,
      "loss": 1.5607,
      "theoretical_loss": 3.4881785714092617,
      "tokens_seen": 1807745024
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004569107321965898,
      "loss": 1.4451,
      "theoretical_loss": 3.488167905030414,
      "tokens_seen": 1807810560
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045689067201604817,
      "loss": 1.5176,
      "theoretical_loss": 3.4881572391464974,
      "tokens_seen": 1807876096
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045687061183550653,
      "loss": 1.4817,
      "theoretical_loss": 3.4881465737574695,
      "tokens_seen": 1807941632
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045685055165496495,
      "loss": 1.3793,
      "theoretical_loss": 3.48813590886329,
      "tokens_seen": 1808007168
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004568304914744233,
      "loss": 1.4244,
      "theoretical_loss": 3.488125244463918,
      "tokens_seen": 1808072704
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045681043129388167,
      "loss": 1.4164,
      "theoretical_loss": 3.488114580559313,
      "tokens_seen": 1808138240
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045679037111334003,
      "loss": 1.3844,
      "theoretical_loss": 3.488103917149433,
      "tokens_seen": 1808203776
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004567703109327984,
      "loss": 1.4618,
      "theoretical_loss": 3.488093254234238,
      "tokens_seen": 1808269312
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004567502507522568,
      "loss": 1.3503,
      "theoretical_loss": 3.4880825918136873,
      "tokens_seen": 1808334848
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004567301905717152,
      "loss": 1.3803,
      "theoretical_loss": 3.488071929887739,
      "tokens_seen": 1808400384
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045671013039117354,
      "loss": 1.5575,
      "theoretical_loss": 3.4880612684563532,
      "tokens_seen": 1808465920
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004566900702106319,
      "loss": 1.3589,
      "theoretical_loss": 3.488050607519489,
      "tokens_seen": 1808531456
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004566700100300903,
      "loss": 1.5692,
      "theoretical_loss": 3.488039947077105,
      "tokens_seen": 1808596992
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004566499498495487,
      "loss": 1.3734,
      "theoretical_loss": 3.4880292871291605,
      "tokens_seen": 1808662528
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045662988966900704,
      "loss": 1.4268,
      "theoretical_loss": 3.488018627675615,
      "tokens_seen": 1808728064
    },
    {
      "epoch": 0.55,
      "objective/train/advantage_avg": 0.4919125735759735,
      "objective/train/docs_used": 660069,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3428508043289185,
      "objective/train/original_loss": 1.342850685119629,
      "objective/train/theoretical_loss": 3.4880079687164267,
      "objective/train/tokens_used": 1829253600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24292032420635223,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504261255264282,
      "objective/train/weighted_lm_loss": 1.4104918241500854,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.4880079687164267,
      "tokens_seen": 1808793600
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004566098294884654,
      "loss": 1.4223,
      "theoretical_loss": 3.4880079687164267,
      "tokens_seen": 1808793600
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045658976930792376,
      "loss": 1.4562,
      "theoretical_loss": 3.4879973102515565,
      "tokens_seen": 1808859136
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004565697091273822,
      "loss": 1.4159,
      "theoretical_loss": 3.487986652280962,
      "tokens_seen": 1808924672
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045654964894684054,
      "loss": 1.3381,
      "theoretical_loss": 3.4879759948046027,
      "tokens_seen": 1808990208
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004565295887662989,
      "loss": 1.4134,
      "theoretical_loss": 3.4879653378224384,
      "tokens_seen": 1809055744
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045650952858575727,
      "loss": 1.4093,
      "theoretical_loss": 3.487954681334428,
      "tokens_seen": 1809121280
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004564894684052157,
      "loss": 1.4602,
      "theoretical_loss": 3.4879440253405303,
      "tokens_seen": 1809186816
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045646940822467405,
      "loss": 1.4746,
      "theoretical_loss": 3.4879333698407047,
      "tokens_seen": 1809252352
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004564493480441324,
      "loss": 1.4329,
      "theoretical_loss": 3.4879227148349106,
      "tokens_seen": 1809317888
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045642928786359077,
      "loss": 1.3918,
      "theoretical_loss": 3.487912060323107,
      "tokens_seen": 1809383424
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045640922768304913,
      "loss": 1.3532,
      "theoretical_loss": 3.4879014063052534,
      "tokens_seen": 1809448960
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045638916750250755,
      "loss": 1.3189,
      "theoretical_loss": 3.4878907527813086,
      "tokens_seen": 1809514496
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004563691073219659,
      "loss": 1.4351,
      "theoretical_loss": 3.487880099751232,
      "tokens_seen": 1809580032
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045634904714142427,
      "loss": 1.4076,
      "theoretical_loss": 3.4878694472149823,
      "tokens_seen": 1809645568
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045632898696088263,
      "loss": 1.3565,
      "theoretical_loss": 3.48785879517252,
      "tokens_seen": 1809711104
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045630892678034105,
      "loss": 1.3128,
      "theoretical_loss": 3.487848143623803,
      "tokens_seen": 1809776640
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004562888665997994,
      "loss": 1.2607,
      "theoretical_loss": 3.487837492568792,
      "tokens_seen": 1809842176
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004562688064192578,
      "loss": 1.4924,
      "theoretical_loss": 3.4878268420074443,
      "tokens_seen": 1809907712
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045624874623871614,
      "loss": 1.4523,
      "theoretical_loss": 3.4878161919397206,
      "tokens_seen": 1809973248
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004562286860581745,
      "loss": 1.4244,
      "theoretical_loss": 3.48780554236558,
      "tokens_seen": 1810038784
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004562086258776329,
      "loss": 1.3968,
      "theoretical_loss": 3.487794893284981,
      "tokens_seen": 1810104320
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004561885656970913,
      "loss": 1.3675,
      "theoretical_loss": 3.4877842446978833,
      "tokens_seen": 1810169856
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045616850551654964,
      "loss": 1.3782,
      "theoretical_loss": 3.4877735966042467,
      "tokens_seen": 1810235392
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000456148445336008,
      "loss": 1.4728,
      "theoretical_loss": 3.48776294900403,
      "tokens_seen": 1810300928
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004561283851554664,
      "loss": 1.4582,
      "theoretical_loss": 3.487752301897192,
      "tokens_seen": 1810366464
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004561083249749248,
      "loss": 1.5162,
      "theoretical_loss": 3.487741655283693,
      "tokens_seen": 1810432000
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045608826479438314,
      "loss": 1.3994,
      "theoretical_loss": 3.487731009163491,
      "tokens_seen": 1810497536
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004560682046138415,
      "loss": 1.3827,
      "theoretical_loss": 3.487720363536546,
      "tokens_seen": 1810563072
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045604814443329987,
      "loss": 1.5065,
      "theoretical_loss": 3.487709718402818,
      "tokens_seen": 1810628608
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004560280842527583,
      "loss": 1.4334,
      "theoretical_loss": 3.4876990737622657,
      "tokens_seen": 1810694144
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045600802407221665,
      "loss": 1.3763,
      "theoretical_loss": 3.4876884296148476,
      "tokens_seen": 1810759680
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000455987963891675,
      "loss": 1.403,
      "theoretical_loss": 3.487677785960524,
      "tokens_seen": 1810825216
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045596790371113337,
      "loss": 1.3833,
      "theoretical_loss": 3.4876671427992543,
      "tokens_seen": 1810890752
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004559478435305918,
      "loss": 1.3778,
      "theoretical_loss": 3.4876565001309965,
      "tokens_seen": 1810956288
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045592778335005015,
      "loss": 1.4101,
      "theoretical_loss": 3.4876458579557115,
      "tokens_seen": 1811021824
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004559077231695085,
      "loss": 1.371,
      "theoretical_loss": 3.4876352162733584,
      "tokens_seen": 1811087360
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004558876629889669,
      "loss": 1.3881,
      "theoretical_loss": 3.4876245750838955,
      "tokens_seen": 1811152896
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045586760280842524,
      "loss": 1.4261,
      "theoretical_loss": 3.4876139343872827,
      "tokens_seen": 1811218432
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045584754262788365,
      "loss": 1.3957,
      "theoretical_loss": 3.48760329418348,
      "tokens_seen": 1811283968
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000455827482447342,
      "loss": 1.495,
      "theoretical_loss": 3.4875926544724454,
      "tokens_seen": 1811349504
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004558074222668004,
      "loss": 1.4155,
      "theoretical_loss": 3.48758201525414,
      "tokens_seen": 1811415040
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045578736208625874,
      "loss": 1.4293,
      "theoretical_loss": 3.4875713765285212,
      "tokens_seen": 1811480576
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045576730190571716,
      "loss": 1.3968,
      "theoretical_loss": 3.4875607382955494,
      "tokens_seen": 1811546112
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004557472417251755,
      "loss": 1.3304,
      "theoretical_loss": 3.4875501005551843,
      "tokens_seen": 1811611648
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004557271815446339,
      "loss": 1.4051,
      "theoretical_loss": 3.487539463307385,
      "tokens_seen": 1811677184
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045570712136409224,
      "loss": 1.358,
      "theoretical_loss": 3.4875288265521105,
      "tokens_seen": 1811742720
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004556870611835507,
      "loss": 1.4451,
      "theoretical_loss": 3.48751819028932,
      "tokens_seen": 1811808256
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004556670010030091,
      "loss": 1.3451,
      "theoretical_loss": 3.4875075545189738,
      "tokens_seen": 1811873792
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045564694082246744,
      "loss": 1.4794,
      "theoretical_loss": 3.4874969192410306,
      "tokens_seen": 1811939328
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004556268806419258,
      "loss": 1.399,
      "theoretical_loss": 3.48748628445545,
      "tokens_seen": 1812004864
    },
    {
      "epoch": 0.55,
      "objective/train/advantage_avg": 0.49013379216194153,
      "objective/train/docs_used": 661289,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.272716760635376,
      "objective/train/original_loss": 1.2727165222167969,
      "objective/train/theoretical_loss": 3.4874756501621915,
      "objective/train/tokens_used": 1832530400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.240818053483963,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502376556396484,
      "objective/train/weighted_lm_loss": 1.3364663124084473,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4874756501621915,
      "tokens_seen": 1812070400
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045560682046138416,
      "loss": 1.3861,
      "theoretical_loss": 3.4874756501621915,
      "tokens_seen": 1812070400
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004555867602808426,
      "loss": 1.3579,
      "theoretical_loss": 3.487465016361214,
      "tokens_seen": 1812135936
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045556670010030094,
      "loss": 1.4325,
      "theoretical_loss": 3.4874543830524782,
      "tokens_seen": 1812201472
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004555466399197593,
      "loss": 1.3204,
      "theoretical_loss": 3.4874437502359417,
      "tokens_seen": 1812267008
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045552657973921767,
      "loss": 1.3834,
      "theoretical_loss": 3.487433117911565,
      "tokens_seen": 1812332544
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004555065195586761,
      "loss": 1.4378,
      "theoretical_loss": 3.4874224860793075,
      "tokens_seen": 1812398080
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045548645937813445,
      "loss": 1.579,
      "theoretical_loss": 3.487411854739128,
      "tokens_seen": 1812463616
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004554663991975928,
      "loss": 1.4583,
      "theoretical_loss": 3.4874012238909877,
      "tokens_seen": 1812529152
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045544633901705117,
      "loss": 1.425,
      "theoretical_loss": 3.4873905935348435,
      "tokens_seen": 1812594688
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045542627883650953,
      "loss": 1.4435,
      "theoretical_loss": 3.487379963670657,
      "tokens_seen": 1812660224
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045540621865596795,
      "loss": 1.4036,
      "theoretical_loss": 3.487369334298386,
      "tokens_seen": 1812725760
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004553861584754263,
      "loss": 1.3706,
      "theoretical_loss": 3.487358705417991,
      "tokens_seen": 1812791296
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004553660982948847,
      "loss": 1.3746,
      "theoretical_loss": 3.4873480770294307,
      "tokens_seen": 1812856832
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045534603811434304,
      "loss": 1.3235,
      "theoretical_loss": 3.4873374491326654,
      "tokens_seen": 1812922368
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045532597793380145,
      "loss": 1.5098,
      "theoretical_loss": 3.4873268217276543,
      "tokens_seen": 1812987904
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004553059177532598,
      "loss": 1.4917,
      "theoretical_loss": 3.4873161948143565,
      "tokens_seen": 1813053440
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004552858575727182,
      "loss": 1.5342,
      "theoretical_loss": 3.487305568392732,
      "tokens_seen": 1813118976
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045526579739217654,
      "loss": 1.3548,
      "theoretical_loss": 3.48729494246274,
      "tokens_seen": 1813184512
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004552457372116349,
      "loss": 1.3662,
      "theoretical_loss": 3.48728431702434,
      "tokens_seen": 1813250048
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004552256770310933,
      "loss": 1.3692,
      "theoretical_loss": 3.4872736920774914,
      "tokens_seen": 1813315584
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004552056168505517,
      "loss": 1.2166,
      "theoretical_loss": 3.4872630676221537,
      "tokens_seen": 1813381120
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045518555667001004,
      "loss": 1.4851,
      "theoretical_loss": 3.487252443658287,
      "tokens_seen": 1813446656
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004551654964894684,
      "loss": 1.4967,
      "theoretical_loss": 3.4872418201858495,
      "tokens_seen": 1813512192
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004551454363089268,
      "loss": 1.3982,
      "theoretical_loss": 3.487231197204802,
      "tokens_seen": 1813577728
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004551253761283852,
      "loss": 1.494,
      "theoretical_loss": 3.487220574715103,
      "tokens_seen": 1813643264
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045510531594784354,
      "loss": 1.4247,
      "theoretical_loss": 3.487209952716713,
      "tokens_seen": 1813708800
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004550852557673019,
      "loss": 1.3907,
      "theoretical_loss": 3.487199331209591,
      "tokens_seen": 1813774336
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045506519558676027,
      "loss": 1.3922,
      "theoretical_loss": 3.4871887101936965,
      "tokens_seen": 1813839872
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004550451354062187,
      "loss": 1.514,
      "theoretical_loss": 3.487178089668989,
      "tokens_seen": 1813905408
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045502507522567705,
      "loss": 1.4643,
      "theoretical_loss": 3.4871674696354287,
      "tokens_seen": 1813970944
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004550050150451354,
      "loss": 1.4886,
      "theoretical_loss": 3.487156850092974,
      "tokens_seen": 1814036480
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045498495486459377,
      "loss": 1.5752,
      "theoretical_loss": 3.487146231041585,
      "tokens_seen": 1814102016
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004549648946840522,
      "loss": 1.4014,
      "theoretical_loss": 3.4871356124812216,
      "tokens_seen": 1814167552
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045494483450351055,
      "loss": 1.5086,
      "theoretical_loss": 3.487124994411843,
      "tokens_seen": 1814233088
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004549247743229689,
      "loss": 1.4735,
      "theoretical_loss": 3.487114376833409,
      "tokens_seen": 1814298624
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004549047141424273,
      "loss": 1.4914,
      "theoretical_loss": 3.487103759745879,
      "tokens_seen": 1814364160
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045488465396188564,
      "loss": 1.3167,
      "theoretical_loss": 3.4870931431492123,
      "tokens_seen": 1814429696
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045486459378134405,
      "loss": 1.4363,
      "theoretical_loss": 3.487082527043369,
      "tokens_seen": 1814495232
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004548445336008024,
      "loss": 1.3753,
      "theoretical_loss": 3.487071911428308,
      "tokens_seen": 1814560768
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004548244734202608,
      "loss": 1.324,
      "theoretical_loss": 3.48706129630399,
      "tokens_seen": 1814626304
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045480441323971914,
      "loss": 1.4625,
      "theoretical_loss": 3.4870506816703735,
      "tokens_seen": 1814691840
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045478435305917756,
      "loss": 1.3119,
      "theoretical_loss": 3.4870400675274182,
      "tokens_seen": 1814757376
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004547642928786359,
      "loss": 1.345,
      "theoretical_loss": 3.487029453875085,
      "tokens_seen": 1814822912
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004547442326980943,
      "loss": 1.3943,
      "theoretical_loss": 3.4870188407133313,
      "tokens_seen": 1814888448
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045472417251755264,
      "loss": 1.4707,
      "theoretical_loss": 3.487008228042119,
      "tokens_seen": 1814953984
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000454704112337011,
      "loss": 1.4705,
      "theoretical_loss": 3.486997615861406,
      "tokens_seen": 1815019520
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004546840521564694,
      "loss": 1.5098,
      "theoretical_loss": 3.4869870041711524,
      "tokens_seen": 1815085056
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004546639919759278,
      "loss": 1.4464,
      "theoretical_loss": 3.486976392971318,
      "tokens_seen": 1815150592
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045464393179538615,
      "loss": 1.3373,
      "theoretical_loss": 3.486965782261863,
      "tokens_seen": 1815216128
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004546238716148445,
      "loss": 1.495,
      "theoretical_loss": 3.486955172042746,
      "tokens_seen": 1815281664
    },
    {
      "epoch": 0.55,
      "objective/train/advantage_avg": 0.49541038274765015,
      "objective/train/docs_used": 662349,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2784487009048462,
      "objective/train/original_loss": 1.2784487009048462,
      "objective/train/theoretical_loss": 3.4869445623139272,
      "objective/train/tokens_used": 1835807200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24577026069164276,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050790548324585,
      "objective/train/weighted_lm_loss": 1.3433135747909546,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4869445623139272,
      "tokens_seen": 1815347200
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004546038114343029,
      "loss": 1.382,
      "theoretical_loss": 3.4869445623139272,
      "tokens_seen": 1815347200
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004545837512537613,
      "loss": 1.4849,
      "theoretical_loss": 3.486933953075366,
      "tokens_seen": 1815412736
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045456369107321965,
      "loss": 1.3985,
      "theoretical_loss": 3.4869233443270224,
      "tokens_seen": 1815478272
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000454543630892678,
      "loss": 1.4493,
      "theoretical_loss": 3.486912736068856,
      "tokens_seen": 1815543808
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045452357071213643,
      "loss": 1.5936,
      "theoretical_loss": 3.4869021283008257,
      "tokens_seen": 1815609344
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004545035105315948,
      "loss": 1.4492,
      "theoretical_loss": 3.4868915210228923,
      "tokens_seen": 1815674880
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045448345035105315,
      "loss": 1.4682,
      "theoretical_loss": 3.4868809142350146,
      "tokens_seen": 1815740416
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004544633901705115,
      "loss": 1.2861,
      "theoretical_loss": 3.486870307937153,
      "tokens_seen": 1815805952
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004544433299899699,
      "loss": 1.3378,
      "theoretical_loss": 3.4868597021292658,
      "tokens_seen": 1815871488
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004544232698094283,
      "loss": 1.4191,
      "theoretical_loss": 3.4868490968113144,
      "tokens_seen": 1815937024
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045440320962888666,
      "loss": 1.471,
      "theoretical_loss": 3.486838491983258,
      "tokens_seen": 1816002560
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000454383149448345,
      "loss": 1.4237,
      "theoretical_loss": 3.4868278876450556,
      "tokens_seen": 1816068096
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004543630892678034,
      "loss": 1.5768,
      "theoretical_loss": 3.4868172837966673,
      "tokens_seen": 1816133632
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004543430290872618,
      "loss": 1.4477,
      "theoretical_loss": 3.486806680438053,
      "tokens_seen": 1816199168
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045432296890672016,
      "loss": 1.3949,
      "theoretical_loss": 3.4867960775691724,
      "tokens_seen": 1816264704
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004543029087261785,
      "loss": 1.3467,
      "theoretical_loss": 3.4867854751899845,
      "tokens_seen": 1816330240
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004542828485456369,
      "loss": 1.3242,
      "theoretical_loss": 3.48677487330045,
      "tokens_seen": 1816395776
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045426278836509525,
      "loss": 1.383,
      "theoretical_loss": 3.486764271900528,
      "tokens_seen": 1816461312
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045424272818455366,
      "loss": 1.3945,
      "theoretical_loss": 3.4867536709901787,
      "tokens_seen": 1816526848
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000454222668004012,
      "loss": 1.4389,
      "theoretical_loss": 3.4867430705693616,
      "tokens_seen": 1816592384
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045420260782347044,
      "loss": 1.4548,
      "theoretical_loss": 3.486732470638036,
      "tokens_seen": 1816657920
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004541825476429288,
      "loss": 1.4064,
      "theoretical_loss": 3.486721871196162,
      "tokens_seen": 1816723456
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004541624874623872,
      "loss": 1.3535,
      "theoretical_loss": 3.4867112722436997,
      "tokens_seen": 1816788992
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004541424272818456,
      "loss": 1.4639,
      "theoretical_loss": 3.486700673780608,
      "tokens_seen": 1816854528
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045412236710130395,
      "loss": 1.4552,
      "theoretical_loss": 3.4866900758068478,
      "tokens_seen": 1816920064
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004541023069207623,
      "loss": 1.4136,
      "theoretical_loss": 3.486679478322378,
      "tokens_seen": 1816985600
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045408224674022067,
      "loss": 1.386,
      "theoretical_loss": 3.4866688813271587,
      "tokens_seen": 1817051136
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004540621865596791,
      "loss": 1.3945,
      "theoretical_loss": 3.486658284821149,
      "tokens_seen": 1817116672
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045404212637913745,
      "loss": 1.3716,
      "theoretical_loss": 3.4866476888043096,
      "tokens_seen": 1817182208
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004540220661985958,
      "loss": 1.3843,
      "theoretical_loss": 3.4866370932765998,
      "tokens_seen": 1817247744
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045400200601805417,
      "loss": 1.4969,
      "theoretical_loss": 3.48662649823798,
      "tokens_seen": 1817313280
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004539819458375126,
      "loss": 1.3392,
      "theoretical_loss": 3.486615903688409,
      "tokens_seen": 1817378816
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045396188565697095,
      "loss": 1.4281,
      "theoretical_loss": 3.486605309627847,
      "tokens_seen": 1817444352
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004539418254764293,
      "loss": 1.4375,
      "theoretical_loss": 3.486594716056254,
      "tokens_seen": 1817509888
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004539217652958877,
      "loss": 1.386,
      "theoretical_loss": 3.4865841229735897,
      "tokens_seen": 1817575424
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045390170511534604,
      "loss": 1.5057,
      "theoretical_loss": 3.4865735303798138,
      "tokens_seen": 1817640960
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045388164493480445,
      "loss": 1.4252,
      "theoretical_loss": 3.4865629382748864,
      "tokens_seen": 1817706496
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004538615847542628,
      "loss": 1.5444,
      "theoretical_loss": 3.486552346658767,
      "tokens_seen": 1817772032
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004538415245737212,
      "loss": 1.4233,
      "theoretical_loss": 3.486541755531415,
      "tokens_seen": 1817837568
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045382146439317954,
      "loss": 1.4398,
      "theoretical_loss": 3.4865311648927912,
      "tokens_seen": 1817903104
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045380140421263796,
      "loss": 1.3986,
      "theoretical_loss": 3.486520574742855,
      "tokens_seen": 1817968640
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004537813440320963,
      "loss": 1.3073,
      "theoretical_loss": 3.4865099850815664,
      "tokens_seen": 1818034176
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004537612838515547,
      "loss": 1.4583,
      "theoretical_loss": 3.4864993959088846,
      "tokens_seen": 1818099712
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045374122367101304,
      "loss": 1.4728,
      "theoretical_loss": 3.48648880722477,
      "tokens_seen": 1818165248
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004537211634904714,
      "loss": 1.2998,
      "theoretical_loss": 3.486478219029183,
      "tokens_seen": 1818230784
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004537011033099298,
      "loss": 1.3605,
      "theoretical_loss": 3.486467631322082,
      "tokens_seen": 1818296320
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004536810431293882,
      "loss": 1.4294,
      "theoretical_loss": 3.486457044103428,
      "tokens_seen": 1818361856
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045366098294884655,
      "loss": 1.4777,
      "theoretical_loss": 3.4864464573731806,
      "tokens_seen": 1818427392
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004536409227683049,
      "loss": 1.5071,
      "theoretical_loss": 3.4864358711312997,
      "tokens_seen": 1818492928
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004536208625877633,
      "loss": 1.3137,
      "theoretical_loss": 3.486425285377744,
      "tokens_seen": 1818558464
    },
    {
      "epoch": 0.55,
      "objective/train/advantage_avg": 0.4934084415435791,
      "objective/train/docs_used": 663450,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3484067916870117,
      "objective/train/original_loss": 1.3484067916870117,
      "objective/train/theoretical_loss": 3.486414700112476,
      "objective/train/tokens_used": 1839084000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24387194216251373,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050580620765686,
      "objective/train/weighted_lm_loss": 1.4165771007537842,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.486414700112476,
      "tokens_seen": 1818624000
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004536008024072217,
      "loss": 1.4283,
      "theoretical_loss": 3.486414700112476,
      "tokens_seen": 1818624000
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045358074222668005,
      "loss": 1.4579,
      "theoretical_loss": 3.486404115335453,
      "tokens_seen": 1818689536
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004535606820461384,
      "loss": 1.4475,
      "theoretical_loss": 3.4863935310466365,
      "tokens_seen": 1818755072
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045354062186559683,
      "loss": 1.4116,
      "theoretical_loss": 3.486382947245985,
      "tokens_seen": 1818820608
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004535205616850552,
      "loss": 1.3915,
      "theoretical_loss": 3.4863723639334596,
      "tokens_seen": 1818886144
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045350050150451355,
      "loss": 1.4627,
      "theoretical_loss": 3.48636178110902,
      "tokens_seen": 1818951680
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004534804413239719,
      "loss": 1.4391,
      "theoretical_loss": 3.486351198772626,
      "tokens_seen": 1819017216
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004534603811434303,
      "loss": 1.3659,
      "theoretical_loss": 3.4863406169242372,
      "tokens_seen": 1819082752
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004534403209628887,
      "loss": 1.3687,
      "theoretical_loss": 3.486330035563814,
      "tokens_seen": 1819148288
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045342026078234706,
      "loss": 1.4387,
      "theoretical_loss": 3.486319454691315,
      "tokens_seen": 1819213824
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004534002006018054,
      "loss": 1.4456,
      "theoretical_loss": 3.4863088743067023,
      "tokens_seen": 1819279360
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004533801404212638,
      "loss": 1.4833,
      "theoretical_loss": 3.4862982944099343,
      "tokens_seen": 1819344896
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004533600802407222,
      "loss": 1.47,
      "theoretical_loss": 3.4862877150009712,
      "tokens_seen": 1819410432
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045334002006018056,
      "loss": 1.4566,
      "theoretical_loss": 3.4862771360797735,
      "tokens_seen": 1819475968
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004533199598796389,
      "loss": 1.4543,
      "theoretical_loss": 3.4862665576463003,
      "tokens_seen": 1819541504
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004532998996990973,
      "loss": 1.3987,
      "theoretical_loss": 3.486255979700512,
      "tokens_seen": 1819607040
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045327983951855565,
      "loss": 1.3894,
      "theoretical_loss": 3.4862454022423686,
      "tokens_seen": 1819672576
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045325977933801406,
      "loss": 1.4487,
      "theoretical_loss": 3.48623482527183,
      "tokens_seen": 1819738112
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004532397191574724,
      "loss": 1.3696,
      "theoretical_loss": 3.4862242487888566,
      "tokens_seen": 1819803648
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004532196589769308,
      "loss": 1.3579,
      "theoretical_loss": 3.486213672793408,
      "tokens_seen": 1819869184
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045319959879638915,
      "loss": 1.5555,
      "theoretical_loss": 3.486203097285443,
      "tokens_seen": 1819934720
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045317953861584757,
      "loss": 1.3463,
      "theoretical_loss": 3.4861925222649233,
      "tokens_seen": 1820000256
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045315947843530593,
      "loss": 1.3868,
      "theoretical_loss": 3.486181947731808,
      "tokens_seen": 1820065792
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004531394182547643,
      "loss": 1.5215,
      "theoretical_loss": 3.486171373686058,
      "tokens_seen": 1820131328
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045311935807422265,
      "loss": 1.3516,
      "theoretical_loss": 3.4861608001276316,
      "tokens_seen": 1820196864
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000453099297893681,
      "loss": 1.2981,
      "theoretical_loss": 3.4861502270564904,
      "tokens_seen": 1820262400
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045307923771313943,
      "loss": 1.5148,
      "theoretical_loss": 3.486139654472594,
      "tokens_seen": 1820327936
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004530591775325978,
      "loss": 1.2555,
      "theoretical_loss": 3.486129082375902,
      "tokens_seen": 1820393472
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045303911735205616,
      "loss": 1.4457,
      "theoretical_loss": 3.4861185107663744,
      "tokens_seen": 1820459008
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004530190571715145,
      "loss": 1.4214,
      "theoretical_loss": 3.486107939643972,
      "tokens_seen": 1820524544
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045299899699097294,
      "loss": 1.5125,
      "theoretical_loss": 3.486097369008654,
      "tokens_seen": 1820590080
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004529789368104313,
      "loss": 1.6293,
      "theoretical_loss": 3.4860867988603803,
      "tokens_seen": 1820655616
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045295887662988966,
      "loss": 1.4122,
      "theoretical_loss": 3.486076229199112,
      "tokens_seen": 1820721152
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000452938816449348,
      "loss": 1.4208,
      "theoretical_loss": 3.486065660024808,
      "tokens_seen": 1820786688
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004529187562688064,
      "loss": 1.364,
      "theoretical_loss": 3.4860550913374286,
      "tokens_seen": 1820852224
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004528986960882648,
      "loss": 1.4394,
      "theoretical_loss": 3.4860445231369344,
      "tokens_seen": 1820917760
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045287863590772316,
      "loss": 1.4564,
      "theoretical_loss": 3.4860339554232853,
      "tokens_seen": 1820983296
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004528585757271815,
      "loss": 1.4152,
      "theoretical_loss": 3.4860233881964406,
      "tokens_seen": 1821048832
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004528385155466399,
      "loss": 1.5798,
      "theoretical_loss": 3.4860128214563613,
      "tokens_seen": 1821114368
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004528184553660983,
      "loss": 1.4505,
      "theoretical_loss": 3.4860022552030068,
      "tokens_seen": 1821179904
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045279839518555667,
      "loss": 1.4596,
      "theoretical_loss": 3.4859916894363376,
      "tokens_seen": 1821245440
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045277833500501503,
      "loss": 1.4343,
      "theoretical_loss": 3.4859811241563134,
      "tokens_seen": 1821310976
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004527582748244734,
      "loss": 1.5355,
      "theoretical_loss": 3.4859705593628947,
      "tokens_seen": 1821376512
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045273821464393175,
      "loss": 1.3985,
      "theoretical_loss": 3.4859599950560414,
      "tokens_seen": 1821442048
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045271815446339017,
      "loss": 1.4998,
      "theoretical_loss": 3.485949431235713,
      "tokens_seen": 1821507584
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004526980942828486,
      "loss": 1.4268,
      "theoretical_loss": 3.4859388679018704,
      "tokens_seen": 1821573120
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045267803410230695,
      "loss": 1.2958,
      "theoretical_loss": 3.4859283050544736,
      "tokens_seen": 1821638656
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004526579739217653,
      "loss": 1.3374,
      "theoretical_loss": 3.4859177426934824,
      "tokens_seen": 1821704192
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004526379137412237,
      "loss": 1.468,
      "theoretical_loss": 3.485907180818857,
      "tokens_seen": 1821769728
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004526178535606821,
      "loss": 1.4645,
      "theoretical_loss": 3.485896619430558,
      "tokens_seen": 1821835264
    },
    {
      "epoch": 0.55,
      "objective/train/advantage_avg": 0.49363186955451965,
      "objective/train/docs_used": 664649,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2212663888931274,
      "objective/train/original_loss": 1.221266269683838,
      "objective/train/theoretical_loss": 3.4858860585285445,
      "objective/train/tokens_used": 1842360800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24396121501922607,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506035089492798,
      "objective/train/weighted_lm_loss": 1.28294837474823,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4858860585285445,
      "tokens_seen": 1821900800
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045259779338014045,
      "loss": 1.3625,
      "theoretical_loss": 3.4858860585285445,
      "tokens_seen": 1821900800
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004525777331995988,
      "loss": 1.3411,
      "theoretical_loss": 3.485875498112777,
      "tokens_seen": 1821966336
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045255767301905723,
      "loss": 1.4909,
      "theoretical_loss": 3.4858649381832163,
      "tokens_seen": 1822031872
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004525376128385156,
      "loss": 1.487,
      "theoretical_loss": 3.4858543787398215,
      "tokens_seen": 1822097408
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045251755265797395,
      "loss": 1.4769,
      "theoretical_loss": 3.485843819782554,
      "tokens_seen": 1822162944
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004524974924774323,
      "loss": 1.5933,
      "theoretical_loss": 3.4858332613113725,
      "tokens_seen": 1822228480
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004524774322968907,
      "loss": 1.5323,
      "theoretical_loss": 3.485822703326238,
      "tokens_seen": 1822294016
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004524573721163491,
      "loss": 1.3894,
      "theoretical_loss": 3.4858121458271105,
      "tokens_seen": 1822359552
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045243731193580746,
      "loss": 1.4511,
      "theoretical_loss": 3.48580158881395,
      "tokens_seen": 1822425088
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004524172517552658,
      "loss": 1.4506,
      "theoretical_loss": 3.485791032286717,
      "tokens_seen": 1822490624
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004523971915747242,
      "loss": 1.4395,
      "theoretical_loss": 3.4857804762453712,
      "tokens_seen": 1822556160
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004523771313941826,
      "loss": 1.297,
      "theoretical_loss": 3.4857699206898736,
      "tokens_seen": 1822621696
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045235707121364096,
      "loss": 1.5453,
      "theoretical_loss": 3.4857593656201833,
      "tokens_seen": 1822687232
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004523370110330993,
      "loss": 1.5105,
      "theoretical_loss": 3.4857488110362613,
      "tokens_seen": 1822752768
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004523169508525577,
      "loss": 1.4264,
      "theoretical_loss": 3.485738256938067,
      "tokens_seen": 1822818304
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045229689067201605,
      "loss": 1.399,
      "theoretical_loss": 3.485727703325561,
      "tokens_seen": 1822883840
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045227683049147446,
      "loss": 1.4855,
      "theoretical_loss": 3.4857171501987034,
      "tokens_seen": 1822949376
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004522567703109328,
      "loss": 1.4354,
      "theoretical_loss": 3.485706597557455,
      "tokens_seen": 1823014912
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004522367101303912,
      "loss": 1.436,
      "theoretical_loss": 3.4856960454017756,
      "tokens_seen": 1823080448
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045221664994984955,
      "loss": 1.3178,
      "theoretical_loss": 3.4856854937316246,
      "tokens_seen": 1823145984
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045219658976930797,
      "loss": 1.3599,
      "theoretical_loss": 3.4856749425469635,
      "tokens_seen": 1823211520
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045217652958876633,
      "loss": 1.5689,
      "theoretical_loss": 3.4856643918477515,
      "tokens_seen": 1823277056
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004521564694082247,
      "loss": 1.505,
      "theoretical_loss": 3.4856538416339493,
      "tokens_seen": 1823342592
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045213640922768305,
      "loss": 1.5338,
      "theoretical_loss": 3.485643291905517,
      "tokens_seen": 1823408128
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004521163490471414,
      "loss": 1.512,
      "theoretical_loss": 3.485632742662416,
      "tokens_seen": 1823473664
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045209628886659983,
      "loss": 1.3813,
      "theoretical_loss": 3.4856221939046037,
      "tokens_seen": 1823539200
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004520762286860582,
      "loss": 1.5276,
      "theoretical_loss": 3.4856116456320434,
      "tokens_seen": 1823604736
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045205616850551656,
      "loss": 1.4973,
      "theoretical_loss": 3.485601097844693,
      "tokens_seen": 1823670272
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004520361083249749,
      "loss": 1.383,
      "theoretical_loss": 3.4855905505425144,
      "tokens_seen": 1823735808
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045201604814443334,
      "loss": 1.485,
      "theoretical_loss": 3.4855800037254667,
      "tokens_seen": 1823801344
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004519959879638917,
      "loss": 1.4545,
      "theoretical_loss": 3.4855694573935114,
      "tokens_seen": 1823866880
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045197592778335006,
      "loss": 1.3938,
      "theoretical_loss": 3.4855589115466072,
      "tokens_seen": 1823932416
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004519558676028084,
      "loss": 1.3752,
      "theoretical_loss": 3.485548366184716,
      "tokens_seen": 1823997952
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004519358074222668,
      "loss": 1.4686,
      "theoretical_loss": 3.4855378213077963,
      "tokens_seen": 1824063488
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004519157472417252,
      "loss": 1.4686,
      "theoretical_loss": 3.4855272769158097,
      "tokens_seen": 1824129024
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045189568706118356,
      "loss": 1.4512,
      "theoretical_loss": 3.4855167330087165,
      "tokens_seen": 1824194560
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004518756268806419,
      "loss": 1.5004,
      "theoretical_loss": 3.4855061895864763,
      "tokens_seen": 1824260096
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004518555667001003,
      "loss": 1.2201,
      "theoretical_loss": 3.485495646649049,
      "tokens_seen": 1824325632
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004518355065195587,
      "loss": 1.4852,
      "theoretical_loss": 3.485485104196396,
      "tokens_seen": 1824391168
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045181544633901707,
      "loss": 1.5728,
      "theoretical_loss": 3.4854745622284775,
      "tokens_seen": 1824456704
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045179538615847543,
      "loss": 1.4404,
      "theoretical_loss": 3.485464020745253,
      "tokens_seen": 1824522240
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004517753259779338,
      "loss": 1.4747,
      "theoretical_loss": 3.485453479746684,
      "tokens_seen": 1824587776
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045175526579739215,
      "loss": 1.36,
      "theoretical_loss": 3.4854429392327293,
      "tokens_seen": 1824653312
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045173520561685057,
      "loss": 1.4249,
      "theoretical_loss": 3.48543239920335,
      "tokens_seen": 1824718848
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045171514543630893,
      "loss": 1.2848,
      "theoretical_loss": 3.4854218596585067,
      "tokens_seen": 1824784384
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004516950852557673,
      "loss": 1.4283,
      "theoretical_loss": 3.4854113205981596,
      "tokens_seen": 1824849920
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045167502507522566,
      "loss": 1.342,
      "theoretical_loss": 3.4854007820222686,
      "tokens_seen": 1824915456
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045165496489468407,
      "loss": 1.4544,
      "theoretical_loss": 3.485390243930794,
      "tokens_seen": 1824980992
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045163490471414244,
      "loss": 1.3774,
      "theoretical_loss": 3.485379706323697,
      "tokens_seen": 1825046528
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004516148445336008,
      "loss": 1.3494,
      "theoretical_loss": 3.485369169200937,
      "tokens_seen": 1825112064
    },
    {
      "epoch": 0.55,
      "objective/train/advantage_avg": 0.4947434961795807,
      "objective/train/docs_used": 665640,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2047829627990723,
      "objective/train/original_loss": 1.204782485961914,
      "objective/train/theoretical_loss": 3.4853586325624746,
      "objective/train/tokens_used": 1845637600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24499619007110596,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507198572158813,
      "objective/train/weighted_lm_loss": 1.2659132480621338,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4853586325624746,
      "tokens_seen": 1825177600
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045159478435305916,
      "loss": 1.4066,
      "theoretical_loss": 3.4853586325624746,
      "tokens_seen": 1825177600
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004515747241725175,
      "loss": 1.4689,
      "theoretical_loss": 3.485348096408271,
      "tokens_seen": 1825243136
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045155466399197594,
      "loss": 1.4937,
      "theoretical_loss": 3.4853375607382846,
      "tokens_seen": 1825308672
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004515346038114343,
      "loss": 1.5348,
      "theoretical_loss": 3.4853270255524778,
      "tokens_seen": 1825374208
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045151454363089266,
      "loss": 1.4724,
      "theoretical_loss": 3.48531649085081,
      "tokens_seen": 1825439744
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.000451494483450351,
      "loss": 1.3027,
      "theoretical_loss": 3.485305956633242,
      "tokens_seen": 1825505280
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045147442326980944,
      "loss": 1.4097,
      "theoretical_loss": 3.4852954228997337,
      "tokens_seen": 1825570816
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004514543630892678,
      "loss": 1.4166,
      "theoretical_loss": 3.4852848896502455,
      "tokens_seen": 1825636352
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045143430290872617,
      "loss": 1.4036,
      "theoretical_loss": 3.485274356884738,
      "tokens_seen": 1825701888
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045141424272818453,
      "loss": 1.4068,
      "theoretical_loss": 3.4852638246031717,
      "tokens_seen": 1825767424
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045139418254764294,
      "loss": 1.4888,
      "theoretical_loss": 3.485253292805507,
      "tokens_seen": 1825832960
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004513741223671013,
      "loss": 1.43,
      "theoretical_loss": 3.485242761491704,
      "tokens_seen": 1825898496
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045135406218655967,
      "loss": 1.4684,
      "theoretical_loss": 3.4852322306617234,
      "tokens_seen": 1825964032
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045133400200601803,
      "loss": 1.492,
      "theoretical_loss": 3.4852217003155253,
      "tokens_seen": 1826029568
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004513139418254764,
      "loss": 1.4218,
      "theoretical_loss": 3.48521117045307,
      "tokens_seen": 1826095104
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004512938816449348,
      "loss": 1.3799,
      "theoretical_loss": 3.485200641074319,
      "tokens_seen": 1826160640
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045127382146439317,
      "loss": 1.4523,
      "theoretical_loss": 3.4851901121792315,
      "tokens_seen": 1826226176
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045125376128385153,
      "loss": 1.5856,
      "theoretical_loss": 3.485179583767768,
      "tokens_seen": 1826291712
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004512337011033099,
      "loss": 1.4232,
      "theoretical_loss": 3.4851690558398896,
      "tokens_seen": 1826357248
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004512136409227683,
      "loss": 1.39,
      "theoretical_loss": 3.4851585283955564,
      "tokens_seen": 1826422784
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045119358074222673,
      "loss": 1.3803,
      "theoretical_loss": 3.485148001434729,
      "tokens_seen": 1826488320
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004511735205616851,
      "loss": 1.3899,
      "theoretical_loss": 3.4851374749573676,
      "tokens_seen": 1826553856
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045115346038114345,
      "loss": 1.381,
      "theoretical_loss": 3.4851269489634324,
      "tokens_seen": 1826619392
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004511334002006018,
      "loss": 1.4899,
      "theoretical_loss": 3.485116423452885,
      "tokens_seen": 1826684928
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045111334002006023,
      "loss": 1.2072,
      "theoretical_loss": 3.4851058984256844,
      "tokens_seen": 1826750464
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004510932798395186,
      "loss": 1.3601,
      "theoretical_loss": 3.485095373881792,
      "tokens_seen": 1826816000
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045107321965897696,
      "loss": 1.4341,
      "theoretical_loss": 3.4850848498211677,
      "tokens_seen": 1826881536
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004510531594784353,
      "loss": 1.3354,
      "theoretical_loss": 3.485074326243773,
      "tokens_seen": 1826947072
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045103309929789374,
      "loss": 1.5102,
      "theoretical_loss": 3.485063803149567,
      "tokens_seen": 1827012608
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004510130391173521,
      "loss": 1.3619,
      "theoretical_loss": 3.4850532805385113,
      "tokens_seen": 1827078144
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045099297893681046,
      "loss": 1.3302,
      "theoretical_loss": 3.4850427584105654,
      "tokens_seen": 1827143680
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004509729187562688,
      "loss": 1.4131,
      "theoretical_loss": 3.485032236765691,
      "tokens_seen": 1827209216
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004509528585757272,
      "loss": 1.5051,
      "theoretical_loss": 3.4850217156038474,
      "tokens_seen": 1827274752
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004509327983951856,
      "loss": 1.4336,
      "theoretical_loss": 3.4850111949249953,
      "tokens_seen": 1827340288
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045091273821464396,
      "loss": 1.2744,
      "theoretical_loss": 3.4850006747290965,
      "tokens_seen": 1827405824
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004508926780341023,
      "loss": 1.3675,
      "theoretical_loss": 3.48499015501611,
      "tokens_seen": 1827471360
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004508726178535607,
      "loss": 1.5272,
      "theoretical_loss": 3.484979635785997,
      "tokens_seen": 1827536896
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004508525576730191,
      "loss": 1.4906,
      "theoretical_loss": 3.484969117038718,
      "tokens_seen": 1827602432
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045083249749247747,
      "loss": 1.4794,
      "theoretical_loss": 3.4849585987742326,
      "tokens_seen": 1827667968
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045081243731193583,
      "loss": 1.5345,
      "theoretical_loss": 3.4849480809925026,
      "tokens_seen": 1827733504
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004507923771313942,
      "loss": 1.5179,
      "theoretical_loss": 3.4849375636934887,
      "tokens_seen": 1827799040
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045077231695085255,
      "loss": 1.4482,
      "theoretical_loss": 3.48492704687715,
      "tokens_seen": 1827864576
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045075225677031097,
      "loss": 1.4113,
      "theoretical_loss": 3.4849165305434484,
      "tokens_seen": 1827930112
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045073219658976933,
      "loss": 1.3872,
      "theoretical_loss": 3.484906014692344,
      "tokens_seen": 1827995648
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004507121364092277,
      "loss": 1.2747,
      "theoretical_loss": 3.4848954993237964,
      "tokens_seen": 1828061184
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045069207622868606,
      "loss": 1.4858,
      "theoretical_loss": 3.484884984437768,
      "tokens_seen": 1828126720
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004506720160481445,
      "loss": 1.3918,
      "theoretical_loss": 3.4848744700342174,
      "tokens_seen": 1828192256
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045065195586760284,
      "loss": 1.3793,
      "theoretical_loss": 3.4848639561131067,
      "tokens_seen": 1828257792
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004506318956870612,
      "loss": 1.53,
      "theoretical_loss": 3.484853442674396,
      "tokens_seen": 1828323328
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045061183550651956,
      "loss": 1.4382,
      "theoretical_loss": 3.484842929718045,
      "tokens_seen": 1828388864
    },
    {
      "epoch": 0.55,
      "objective/train/advantage_avg": 0.4918673634529114,
      "objective/train/docs_used": 666768,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.267907977104187,
      "objective/train/original_loss": 1.2679076194763184,
      "objective/train/theoretical_loss": 3.484832417244016,
      "objective/train/tokens_used": 1848914400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.242312490940094,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050418496131897,
      "objective/train/weighted_lm_loss": 1.3318935632705688,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.484832417244016,
      "tokens_seen": 1828454400
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004505917753259779,
      "loss": 1.4361,
      "theoretical_loss": 3.484832417244016,
      "tokens_seen": 1828454400
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045057171514543634,
      "loss": 1.3863,
      "theoretical_loss": 3.4848219052522684,
      "tokens_seen": 1828519936
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004505516549648947,
      "loss": 1.4104,
      "theoretical_loss": 3.484811393742763,
      "tokens_seen": 1828585472
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045053159478435306,
      "loss": 1.4766,
      "theoretical_loss": 3.48480088271546,
      "tokens_seen": 1828651008
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004505115346038114,
      "loss": 1.3747,
      "theoretical_loss": 3.484790372170321,
      "tokens_seen": 1828716544
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045049147442326984,
      "loss": 1.3144,
      "theoretical_loss": 3.484779862107306,
      "tokens_seen": 1828782080
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004504714142427282,
      "loss": 1.4223,
      "theoretical_loss": 3.4847693525263757,
      "tokens_seen": 1828847616
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045045135406218657,
      "loss": 1.3264,
      "theoretical_loss": 3.4847588434274908,
      "tokens_seen": 1828913152
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045043129388164493,
      "loss": 1.4845,
      "theoretical_loss": 3.484748334810611,
      "tokens_seen": 1828978688
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045041123370110334,
      "loss": 1.4659,
      "theoretical_loss": 3.4847378266756985,
      "tokens_seen": 1829044224
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004503911735205617,
      "loss": 1.45,
      "theoretical_loss": 3.4847273190227126,
      "tokens_seen": 1829109760
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045037111334002007,
      "loss": 1.2664,
      "theoretical_loss": 3.484716811851615,
      "tokens_seen": 1829175296
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045035105315947843,
      "loss": 1.4224,
      "theoretical_loss": 3.484706305162365,
      "tokens_seen": 1829240832
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004503309929789368,
      "loss": 1.3564,
      "theoretical_loss": 3.4846957989549248,
      "tokens_seen": 1829306368
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004503109327983952,
      "loss": 1.3507,
      "theoretical_loss": 3.484685293229254,
      "tokens_seen": 1829371904
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045029087261785357,
      "loss": 1.4647,
      "theoretical_loss": 3.4846747879853135,
      "tokens_seen": 1829437440
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045027081243731193,
      "loss": 1.4015,
      "theoretical_loss": 3.484664283223064,
      "tokens_seen": 1829502976
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004502507522567703,
      "loss": 1.5223,
      "theoretical_loss": 3.484653778942466,
      "tokens_seen": 1829568512
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004502306920762287,
      "loss": 1.4548,
      "theoretical_loss": 3.4846432751434806,
      "tokens_seen": 1829634048
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004502106318956871,
      "loss": 1.3793,
      "theoretical_loss": 3.484632771826068,
      "tokens_seen": 1829699584
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045019057171514544,
      "loss": 1.388,
      "theoretical_loss": 3.484622268990189,
      "tokens_seen": 1829765120
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004501705115346038,
      "loss": 1.3458,
      "theoretical_loss": 3.4846117666358047,
      "tokens_seen": 1829830656
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045015045135406216,
      "loss": 1.6047,
      "theoretical_loss": 3.4846012647628752,
      "tokens_seen": 1829896192
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004501303911735206,
      "loss": 1.4365,
      "theoretical_loss": 3.484590763371361,
      "tokens_seen": 1829961728
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045011033099297894,
      "loss": 1.3515,
      "theoretical_loss": 3.4845802624612237,
      "tokens_seen": 1830027264
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004500902708124373,
      "loss": 1.3986,
      "theoretical_loss": 3.4845697620324234,
      "tokens_seen": 1830092800
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045007021063189567,
      "loss": 1.3621,
      "theoretical_loss": 3.484559262084921,
      "tokens_seen": 1830158336
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004500501504513541,
      "loss": 1.3859,
      "theoretical_loss": 3.484548762618677,
      "tokens_seen": 1830223872
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00045003009027081244,
      "loss": 1.467,
      "theoretical_loss": 3.484538263633652,
      "tokens_seen": 1830289408
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004500100300902708,
      "loss": 1.494,
      "theoretical_loss": 3.484527765129807,
      "tokens_seen": 1830354944
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00044998996990972917,
      "loss": 1.4331,
      "theoretical_loss": 3.4845172671071025,
      "tokens_seen": 1830420480
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00044996990972918753,
      "loss": 1.4522,
      "theoretical_loss": 3.4845067695655,
      "tokens_seen": 1830486016
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00044994984954864595,
      "loss": 1.3603,
      "theoretical_loss": 3.484496272504959,
      "tokens_seen": 1830551552
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004499297893681043,
      "loss": 1.4616,
      "theoretical_loss": 3.4844857759254406,
      "tokens_seen": 1830617088
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00044990972918756267,
      "loss": 1.2793,
      "theoretical_loss": 3.4844752798269063,
      "tokens_seen": 1830682624
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00044988966900702103,
      "loss": 1.4821,
      "theoretical_loss": 3.484464784209316,
      "tokens_seen": 1830748160
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00044986960882647945,
      "loss": 1.3701,
      "theoretical_loss": 3.484454289072631,
      "tokens_seen": 1830813696
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004498495486459378,
      "loss": 1.416,
      "theoretical_loss": 3.484443794416812,
      "tokens_seen": 1830879232
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004498294884653962,
      "loss": 1.3544,
      "theoretical_loss": 3.484433300241819,
      "tokens_seen": 1830944768
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00044980942828485454,
      "loss": 1.5869,
      "theoretical_loss": 3.484422806547614,
      "tokens_seen": 1831010304
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004497893681043129,
      "loss": 1.3942,
      "theoretical_loss": 3.4844123133341567,
      "tokens_seen": 1831075840
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004497693079237713,
      "loss": 1.5099,
      "theoretical_loss": 3.4844018206014082,
      "tokens_seen": 1831141376
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004497492477432297,
      "loss": 1.3692,
      "theoretical_loss": 3.4843913283493295,
      "tokens_seen": 1831206912
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00044972918756268804,
      "loss": 1.3669,
      "theoretical_loss": 3.4843808365778814,
      "tokens_seen": 1831272448
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004497091273821464,
      "loss": 1.3351,
      "theoretical_loss": 3.484370345287024,
      "tokens_seen": 1831337984
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0004496890672016049,
      "loss": 1.4218,
      "theoretical_loss": 3.4843598544767187,
      "tokens_seen": 1831403520
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.00044966900702106324,
      "loss": 1.3807,
      "theoretical_loss": 3.4843493641469263,
      "tokens_seen": 1831469056
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004496489468405216,
      "loss": 1.4085,
      "theoretical_loss": 3.484338874297608,
      "tokens_seen": 1831534592
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044962888665997996,
      "loss": 1.3038,
      "theoretical_loss": 3.484328384928723,
      "tokens_seen": 1831600128
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004496088264794383,
      "loss": 1.4332,
      "theoretical_loss": 3.484317896040234,
      "tokens_seen": 1831665664
    },
    {
      "epoch": 0.56,
      "objective/train/advantage_avg": 0.49268364906311035,
      "objective/train/docs_used": 668070,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2882417440414429,
      "objective/train/original_loss": 1.2882417440414429,
      "objective/train/theoretical_loss": 3.484307407632101,
      "objective/train/tokens_used": 1852191200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24309706687927246,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050504207611084,
      "objective/train/weighted_lm_loss": 1.3532958030700684,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.484307407632101,
      "tokens_seen": 1831731200
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044958876629889674,
      "loss": 1.4058,
      "theoretical_loss": 3.484307407632101,
      "tokens_seen": 1831731200
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004495687061183551,
      "loss": 1.3402,
      "theoretical_loss": 3.484296919704284,
      "tokens_seen": 1831796736
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044954864593781346,
      "loss": 1.4225,
      "theoretical_loss": 3.484286432256745,
      "tokens_seen": 1831862272
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004495285857572718,
      "loss": 1.3745,
      "theoretical_loss": 3.484275945289445,
      "tokens_seen": 1831927808
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044950852557673024,
      "loss": 1.5259,
      "theoretical_loss": 3.484265458802344,
      "tokens_seen": 1831993344
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004494884653961886,
      "loss": 1.4529,
      "theoretical_loss": 3.4842549727954033,
      "tokens_seen": 1832058880
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044946840521564697,
      "loss": 1.3471,
      "theoretical_loss": 3.484244487268583,
      "tokens_seen": 1832124416
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044944834503510533,
      "loss": 1.3935,
      "theoretical_loss": 3.4842340022218448,
      "tokens_seen": 1832189952
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044942828485456375,
      "loss": 1.4581,
      "theoretical_loss": 3.4842235176551495,
      "tokens_seen": 1832255488
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004494082246740221,
      "loss": 1.4071,
      "theoretical_loss": 3.4842130335684574,
      "tokens_seen": 1832321024
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044938816449348047,
      "loss": 1.4483,
      "theoretical_loss": 3.4842025499617297,
      "tokens_seen": 1832386560
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044936810431293883,
      "loss": 1.4681,
      "theoretical_loss": 3.4841920668349275,
      "tokens_seen": 1832452096
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004493480441323972,
      "loss": 1.4495,
      "theoretical_loss": 3.4841815841880113,
      "tokens_seen": 1832517632
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004493279839518556,
      "loss": 1.5782,
      "theoretical_loss": 3.484171102020942,
      "tokens_seen": 1832583168
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044930792377131397,
      "loss": 1.4373,
      "theoretical_loss": 3.4841606203336806,
      "tokens_seen": 1832648704
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044928786359077234,
      "loss": 1.4258,
      "theoretical_loss": 3.4841501391261875,
      "tokens_seen": 1832714240
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004492678034102307,
      "loss": 1.3418,
      "theoretical_loss": 3.4841396583984245,
      "tokens_seen": 1832779776
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004492477432296891,
      "loss": 1.4231,
      "theoretical_loss": 3.484129178150352,
      "tokens_seen": 1832845312
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004492276830491475,
      "loss": 1.3207,
      "theoretical_loss": 3.4841186983819306,
      "tokens_seen": 1832910848
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044920762286860584,
      "loss": 1.3608,
      "theoretical_loss": 3.484108219093122,
      "tokens_seen": 1832976384
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004491875626880642,
      "loss": 1.3208,
      "theoretical_loss": 3.484097740283886,
      "tokens_seen": 1833041920
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044916750250752256,
      "loss": 1.4989,
      "theoretical_loss": 3.4840872619541843,
      "tokens_seen": 1833107456
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000449147442326981,
      "loss": 1.464,
      "theoretical_loss": 3.4840767841039777,
      "tokens_seen": 1833172992
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044912738214643934,
      "loss": 1.4396,
      "theoretical_loss": 3.484066306733227,
      "tokens_seen": 1833238528
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004491073219658977,
      "loss": 1.3563,
      "theoretical_loss": 3.4840558298418935,
      "tokens_seen": 1833304064
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044908726178535607,
      "loss": 1.514,
      "theoretical_loss": 3.484045353429937,
      "tokens_seen": 1833369600
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004490672016048145,
      "loss": 1.4784,
      "theoretical_loss": 3.48403487749732,
      "tokens_seen": 1833435136
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044904714142427284,
      "loss": 1.4553,
      "theoretical_loss": 3.4840244020440023,
      "tokens_seen": 1833500672
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004490270812437312,
      "loss": 1.451,
      "theoretical_loss": 3.4840139270699453,
      "tokens_seen": 1833566208
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044900702106318957,
      "loss": 1.4992,
      "theoretical_loss": 3.4840034525751093,
      "tokens_seen": 1833631744
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044898696088264793,
      "loss": 1.4706,
      "theoretical_loss": 3.4839929785594563,
      "tokens_seen": 1833697280
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044896690070210635,
      "loss": 1.4473,
      "theoretical_loss": 3.483982505022947,
      "tokens_seen": 1833762816
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004489468405215647,
      "loss": 1.2317,
      "theoretical_loss": 3.483972031965542,
      "tokens_seen": 1833828352
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044892678034102307,
      "loss": 1.5038,
      "theoretical_loss": 3.4839615593872018,
      "tokens_seen": 1833893888
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044890672016048143,
      "loss": 1.4343,
      "theoretical_loss": 3.4839510872878883,
      "tokens_seen": 1833959424
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044888665997993985,
      "loss": 1.4488,
      "theoretical_loss": 3.483940615667562,
      "tokens_seen": 1834024960
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004488665997993982,
      "loss": 1.4008,
      "theoretical_loss": 3.483930144526184,
      "tokens_seen": 1834090496
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004488465396188566,
      "loss": 1.3806,
      "theoretical_loss": 3.4839196738637153,
      "tokens_seen": 1834156032
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044882647943831494,
      "loss": 1.3743,
      "theoretical_loss": 3.483909203680117,
      "tokens_seen": 1834221568
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004488064192577733,
      "loss": 1.3298,
      "theoretical_loss": 3.4838987339753498,
      "tokens_seen": 1834287104
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004487863590772317,
      "loss": 1.4941,
      "theoretical_loss": 3.4838882647493747,
      "tokens_seen": 1834352640
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004487662988966901,
      "loss": 1.4822,
      "theoretical_loss": 3.4838777960021527,
      "tokens_seen": 1834418176
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044874623871614844,
      "loss": 1.4407,
      "theoretical_loss": 3.483867327733645,
      "tokens_seen": 1834483712
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004487261785356068,
      "loss": 1.3894,
      "theoretical_loss": 3.483856859943813,
      "tokens_seen": 1834549248
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004487061183550652,
      "loss": 1.3091,
      "theoretical_loss": 3.483846392632617,
      "tokens_seen": 1834614784
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004486860581745236,
      "loss": 1.3825,
      "theoretical_loss": 3.483835925800018,
      "tokens_seen": 1834680320
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044866599799398194,
      "loss": 1.5243,
      "theoretical_loss": 3.4838254594459777,
      "tokens_seen": 1834745856
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004486459378134403,
      "loss": 1.4479,
      "theoretical_loss": 3.4838149935704563,
      "tokens_seen": 1834811392
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044862587763289867,
      "loss": 1.4465,
      "theoretical_loss": 3.4838045281734153,
      "tokens_seen": 1834876928
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004486058174523571,
      "loss": 1.4369,
      "theoretical_loss": 3.483794063254816,
      "tokens_seen": 1834942464
    },
    {
      "epoch": 0.56,
      "objective/train/advantage_avg": 0.4931764602661133,
      "objective/train/docs_used": 669245,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3102281093597412,
      "objective/train/original_loss": 1.3102281093597412,
      "objective/train/theoretical_loss": 3.483783598814619,
      "objective/train/tokens_used": 1855468000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24374912679195404,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505567789077759,
      "objective/train/weighted_lm_loss": 1.3763251304626465,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.483783598814619,
      "tokens_seen": 1835008000
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044858575727181545,
      "loss": 1.3384,
      "theoretical_loss": 3.483783598814619,
      "tokens_seen": 1835008000
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004485656970912738,
      "loss": 1.3951,
      "theoretical_loss": 3.4837731348527856,
      "tokens_seen": 1835073536
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044854563691073217,
      "loss": 1.396,
      "theoretical_loss": 3.4837626713692766,
      "tokens_seen": 1835139072
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004485255767301906,
      "loss": 1.4854,
      "theoretical_loss": 3.483752208364053,
      "tokens_seen": 1835204608
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044850551654964895,
      "loss": 1.4304,
      "theoretical_loss": 3.4837417458370767,
      "tokens_seen": 1835270144
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004484854563691073,
      "loss": 1.4448,
      "theoretical_loss": 3.483731283788307,
      "tokens_seen": 1835335680
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004484653961885657,
      "loss": 1.3549,
      "theoretical_loss": 3.483720822217707,
      "tokens_seen": 1835401216
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044844533600802404,
      "loss": 1.4694,
      "theoretical_loss": 3.4837103611252367,
      "tokens_seen": 1835466752
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044842527582748245,
      "loss": 1.3082,
      "theoretical_loss": 3.483699900510857,
      "tokens_seen": 1835532288
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004484052156469408,
      "loss": 1.4539,
      "theoretical_loss": 3.4836894403745298,
      "tokens_seen": 1835597824
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004483851554663992,
      "loss": 1.5229,
      "theoretical_loss": 3.483678980716215,
      "tokens_seen": 1835663360
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044836509528585754,
      "loss": 1.4044,
      "theoretical_loss": 3.4836685215358747,
      "tokens_seen": 1835728896
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044834503510531596,
      "loss": 1.3583,
      "theoretical_loss": 3.4836580628334697,
      "tokens_seen": 1835794432
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004483249749247743,
      "loss": 1.4833,
      "theoretical_loss": 3.483647604608961,
      "tokens_seen": 1835859968
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004483049147442327,
      "loss": 1.4445,
      "theoretical_loss": 3.4836371468623097,
      "tokens_seen": 1835925504
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044828485456369104,
      "loss": 1.4627,
      "theoretical_loss": 3.4836266895934775,
      "tokens_seen": 1835991040
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004482647943831494,
      "loss": 1.5875,
      "theoretical_loss": 3.4836162328024245,
      "tokens_seen": 1836056576
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004482447342026078,
      "loss": 1.4324,
      "theoretical_loss": 3.4836057764891124,
      "tokens_seen": 1836122112
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004482246740220662,
      "loss": 1.4296,
      "theoretical_loss": 3.483595320653502,
      "tokens_seen": 1836187648
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044820461384152455,
      "loss": 1.4072,
      "theoretical_loss": 3.4835848652955557,
      "tokens_seen": 1836253184
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044818455366098296,
      "loss": 1.3931,
      "theoretical_loss": 3.4835744104152324,
      "tokens_seen": 1836318720
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004481644934804414,
      "loss": 1.5208,
      "theoretical_loss": 3.483563956012495,
      "tokens_seen": 1836384256
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044814443329989974,
      "loss": 1.4749,
      "theoretical_loss": 3.483553502087304,
      "tokens_seen": 1836449792
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004481243731193581,
      "loss": 1.3959,
      "theoretical_loss": 3.4835430486396204,
      "tokens_seen": 1836515328
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044810431293881647,
      "loss": 1.3998,
      "theoretical_loss": 3.483532595669406,
      "tokens_seen": 1836580864
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004480842527582749,
      "loss": 1.3346,
      "theoretical_loss": 3.4835221431766215,
      "tokens_seen": 1836646400
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044806419257773324,
      "loss": 1.2604,
      "theoretical_loss": 3.483511691161228,
      "tokens_seen": 1836711936
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004480441323971916,
      "loss": 1.4689,
      "theoretical_loss": 3.4835012396231866,
      "tokens_seen": 1836777472
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044802407221664997,
      "loss": 1.3439,
      "theoretical_loss": 3.4834907885624586,
      "tokens_seen": 1836843008
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044800401203610833,
      "loss": 1.3019,
      "theoretical_loss": 3.483480337979005,
      "tokens_seen": 1836908544
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044798395185556675,
      "loss": 1.4673,
      "theoretical_loss": 3.483469887872787,
      "tokens_seen": 1836974080
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004479638916750251,
      "loss": 1.313,
      "theoretical_loss": 3.483459438243767,
      "tokens_seen": 1837039616
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044794383149448347,
      "loss": 1.3635,
      "theoretical_loss": 3.4834489890919045,
      "tokens_seen": 1837105152
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044792377131394183,
      "loss": 1.3698,
      "theoretical_loss": 3.4834385404171613,
      "tokens_seen": 1837170688
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044790371113340025,
      "loss": 1.3977,
      "theoretical_loss": 3.4834280922194987,
      "tokens_seen": 1837236224
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004478836509528586,
      "loss": 1.3711,
      "theoretical_loss": 3.4834176444988776,
      "tokens_seen": 1837301760
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000447863590772317,
      "loss": 1.4397,
      "theoretical_loss": 3.48340719725526,
      "tokens_seen": 1837367296
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044784353059177534,
      "loss": 1.4198,
      "theoretical_loss": 3.483396750488606,
      "tokens_seen": 1837432832
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004478234704112337,
      "loss": 1.4486,
      "theoretical_loss": 3.4833863041988775,
      "tokens_seen": 1837498368
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004478034102306921,
      "loss": 1.5241,
      "theoretical_loss": 3.4833758583860357,
      "tokens_seen": 1837563904
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004477833500501505,
      "loss": 1.3831,
      "theoretical_loss": 3.4833654130500413,
      "tokens_seen": 1837629440
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044776328986960884,
      "loss": 1.3401,
      "theoretical_loss": 3.4833549681908567,
      "tokens_seen": 1837694976
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004477432296890672,
      "loss": 1.3559,
      "theoretical_loss": 3.4833445238084417,
      "tokens_seen": 1837760512
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004477231695085256,
      "loss": 1.4455,
      "theoretical_loss": 3.4833340799027583,
      "tokens_seen": 1837826048
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000447703109327984,
      "loss": 1.3635,
      "theoretical_loss": 3.4833236364737674,
      "tokens_seen": 1837891584
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044768304914744234,
      "loss": 1.2877,
      "theoretical_loss": 3.4833131935214308,
      "tokens_seen": 1837957120
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004476629889669007,
      "loss": 1.423,
      "theoretical_loss": 3.4833027510457093,
      "tokens_seen": 1838022656
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044764292878635907,
      "loss": 1.4793,
      "theoretical_loss": 3.483292309046565,
      "tokens_seen": 1838088192
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004476228686058175,
      "loss": 1.4294,
      "theoretical_loss": 3.4832818675239574,
      "tokens_seen": 1838153728
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044760280842527585,
      "loss": 1.3883,
      "theoretical_loss": 3.483271426477849,
      "tokens_seen": 1838219264
    },
    {
      "epoch": 0.56,
      "objective/train/advantage_avg": 0.4916801452636719,
      "objective/train/docs_used": 670495,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3920093774795532,
      "objective/train/original_loss": 1.3920094966888428,
      "objective/train/theoretical_loss": 3.483260985908201,
      "objective/train/tokens_used": 1858744800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24229270219802856,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050399899482727,
      "objective/train/weighted_lm_loss": 1.4621683359146118,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.483260985908201,
      "tokens_seen": 1838284800
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004475827482447342,
      "loss": 1.4885,
      "theoretical_loss": 3.483260985908201,
      "tokens_seen": 1838284800
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044756268806419257,
      "loss": 1.3794,
      "theoretical_loss": 3.4832505458149745,
      "tokens_seen": 1838350336
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000447542627883651,
      "loss": 1.4499,
      "theoretical_loss": 3.483240106198131,
      "tokens_seen": 1838415872
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044752256770310935,
      "loss": 1.4972,
      "theoretical_loss": 3.483229667057632,
      "tokens_seen": 1838481408
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004475025075225677,
      "loss": 1.3841,
      "theoretical_loss": 3.4832192283934376,
      "tokens_seen": 1838546944
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004474824473420261,
      "loss": 1.4235,
      "theoretical_loss": 3.48320879020551,
      "tokens_seen": 1838612480
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044746238716148444,
      "loss": 1.5651,
      "theoretical_loss": 3.483198352493811,
      "tokens_seen": 1838678016
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044744232698094285,
      "loss": 1.4622,
      "theoretical_loss": 3.4831879152583003,
      "tokens_seen": 1838743552
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004474222668004012,
      "loss": 1.5011,
      "theoretical_loss": 3.483177478498941,
      "tokens_seen": 1838809088
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004474022066198596,
      "loss": 1.5362,
      "theoretical_loss": 3.483167042215693,
      "tokens_seen": 1838874624
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044738214643931794,
      "loss": 1.4551,
      "theoretical_loss": 3.4831566064085187,
      "tokens_seen": 1838940160
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044736208625877636,
      "loss": 1.4331,
      "theoretical_loss": 3.4831461710773786,
      "tokens_seen": 1839005696
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004473420260782347,
      "loss": 1.3322,
      "theoretical_loss": 3.483135736222234,
      "tokens_seen": 1839071232
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004473219658976931,
      "loss": 1.4435,
      "theoretical_loss": 3.4831253018430477,
      "tokens_seen": 1839136768
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044730190571715144,
      "loss": 1.4254,
      "theoretical_loss": 3.483114867939779,
      "tokens_seen": 1839202304
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004472818455366098,
      "loss": 1.327,
      "theoretical_loss": 3.4831044345123905,
      "tokens_seen": 1839267840
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004472617853560682,
      "loss": 1.4402,
      "theoretical_loss": 3.4830940015608425,
      "tokens_seen": 1839333376
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004472417251755266,
      "loss": 1.418,
      "theoretical_loss": 3.483083569085098,
      "tokens_seen": 1839398912
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044722166499498495,
      "loss": 1.4199,
      "theoretical_loss": 3.4830731370851167,
      "tokens_seen": 1839464448
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004472016048144433,
      "loss": 1.3178,
      "theoretical_loss": 3.4830627055608607,
      "tokens_seen": 1839529984
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004471815446339017,
      "loss": 1.393,
      "theoretical_loss": 3.483052274512292,
      "tokens_seen": 1839595520
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004471614844533601,
      "loss": 1.438,
      "theoretical_loss": 3.48304184393937,
      "tokens_seen": 1839661056
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044714142427281845,
      "loss": 1.4346,
      "theoretical_loss": 3.483031413842058,
      "tokens_seen": 1839726592
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004471213640922768,
      "loss": 1.4553,
      "theoretical_loss": 3.4830209842203166,
      "tokens_seen": 1839792128
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044710130391173523,
      "loss": 1.4614,
      "theoretical_loss": 3.4830105550741077,
      "tokens_seen": 1839857664
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004470812437311936,
      "loss": 1.4538,
      "theoretical_loss": 3.4830001264033914,
      "tokens_seen": 1839923200
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044706118355065195,
      "loss": 1.4177,
      "theoretical_loss": 3.4829896982081303,
      "tokens_seen": 1839988736
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004470411233701103,
      "loss": 1.4597,
      "theoretical_loss": 3.4829792704882854,
      "tokens_seen": 1840054272
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004470210631895687,
      "loss": 1.5584,
      "theoretical_loss": 3.4829688432438184,
      "tokens_seen": 1840119808
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004470010030090271,
      "loss": 1.4369,
      "theoretical_loss": 3.4829584164746903,
      "tokens_seen": 1840185344
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044698094282848546,
      "loss": 1.3525,
      "theoretical_loss": 3.4829479901808624,
      "tokens_seen": 1840250880
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004469608826479438,
      "loss": 1.385,
      "theoretical_loss": 3.482937564362296,
      "tokens_seen": 1840316416
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004469408224674022,
      "loss": 1.4141,
      "theoretical_loss": 3.482927139018954,
      "tokens_seen": 1840381952
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004469207622868606,
      "loss": 1.3929,
      "theoretical_loss": 3.4829167141507957,
      "tokens_seen": 1840447488
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044690070210631896,
      "loss": 1.3503,
      "theoretical_loss": 3.482906289757784,
      "tokens_seen": 1840513024
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004468806419257773,
      "loss": 1.5057,
      "theoretical_loss": 3.482895865839879,
      "tokens_seen": 1840578560
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004468605817452357,
      "loss": 1.3895,
      "theoretical_loss": 3.4828854423970435,
      "tokens_seen": 1840644096
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044684052156469405,
      "loss": 1.4119,
      "theoretical_loss": 3.482875019429238,
      "tokens_seen": 1840709632
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044682046138415246,
      "loss": 1.481,
      "theoretical_loss": 3.482864596936425,
      "tokens_seen": 1840775168
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004468004012036108,
      "loss": 1.3656,
      "theoretical_loss": 3.4828541749185646,
      "tokens_seen": 1840840704
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004467803410230692,
      "loss": 1.3226,
      "theoretical_loss": 3.482843753375619,
      "tokens_seen": 1840906240
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044676028084252755,
      "loss": 1.5245,
      "theoretical_loss": 3.4828333323075498,
      "tokens_seen": 1840971776
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044674022066198597,
      "loss": 1.4776,
      "theoretical_loss": 3.482822911714318,
      "tokens_seen": 1841037312
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044672016048144433,
      "loss": 1.5417,
      "theoretical_loss": 3.4828124915958854,
      "tokens_seen": 1841102848
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004467001003009027,
      "loss": 1.4848,
      "theoretical_loss": 3.4828020719522135,
      "tokens_seen": 1841168384
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004466800401203611,
      "loss": 1.3505,
      "theoretical_loss": 3.482791652783263,
      "tokens_seen": 1841233920
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044665997993981947,
      "loss": 1.2993,
      "theoretical_loss": 3.4827812340889963,
      "tokens_seen": 1841299456
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004466399197592779,
      "loss": 1.4013,
      "theoretical_loss": 3.482770815869375,
      "tokens_seen": 1841364992
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044661985957873625,
      "loss": 1.483,
      "theoretical_loss": 3.4827603981243596,
      "tokens_seen": 1841430528
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004465997993981946,
      "loss": 1.4048,
      "theoretical_loss": 3.4827499808539124,
      "tokens_seen": 1841496064
    },
    {
      "epoch": 0.56,
      "objective/train/advantage_avg": 0.4878895580768585,
      "objective/train/docs_used": 671743,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5395152568817139,
      "objective/train/original_loss": 1.5395153760910034,
      "objective/train/theoretical_loss": 3.4827395640579946,
      "objective/train/tokens_used": 1862021600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23950281739234924,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500065088272095,
      "objective/train/weighted_lm_loss": 1.6167365312576294,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4827395640579946,
      "tokens_seen": 1841561600
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044657973921765297,
      "loss": 1.4255,
      "theoretical_loss": 3.4827395640579946,
      "tokens_seen": 1841561600
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004465596790371114,
      "loss": 1.3664,
      "theoretical_loss": 3.482729147736568,
      "tokens_seen": 1841627136
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044653961885656975,
      "loss": 1.3327,
      "theoretical_loss": 3.482718731889593,
      "tokens_seen": 1841692672
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004465195586760281,
      "loss": 1.3542,
      "theoretical_loss": 3.4827083165170327,
      "tokens_seen": 1841758208
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004464994984954865,
      "loss": 1.3734,
      "theoretical_loss": 3.4826979016188475,
      "tokens_seen": 1841823744
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044647943831494484,
      "loss": 1.338,
      "theoretical_loss": 3.4826874871949993,
      "tokens_seen": 1841889280
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044645937813440325,
      "loss": 1.452,
      "theoretical_loss": 3.4826770732454495,
      "tokens_seen": 1841954816
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004464393179538616,
      "loss": 1.5108,
      "theoretical_loss": 3.4826666597701603,
      "tokens_seen": 1842020352
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044641925777332,
      "loss": 1.4337,
      "theoretical_loss": 3.482656246769092,
      "tokens_seen": 1842085888
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044639919759277834,
      "loss": 1.4058,
      "theoretical_loss": 3.4826458342422066,
      "tokens_seen": 1842151424
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044637913741223676,
      "loss": 1.4872,
      "theoretical_loss": 3.4826354221894666,
      "tokens_seen": 1842216960
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004463590772316951,
      "loss": 1.4063,
      "theoretical_loss": 3.4826250106108323,
      "tokens_seen": 1842282496
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004463590772316951,
      "loss": 1.3863,
      "theoretical_loss": 3.482614599506266,
      "tokens_seen": 1842348032
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004463390170511535,
      "loss": 1.4507,
      "theoretical_loss": 3.4826041888757286,
      "tokens_seen": 1842413568
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044631895687061184,
      "loss": 1.3528,
      "theoretical_loss": 3.482593778719182,
      "tokens_seen": 1842479104
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004462988966900702,
      "loss": 1.5505,
      "theoretical_loss": 3.482583369036588,
      "tokens_seen": 1842544640
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004462788365095286,
      "loss": 1.4113,
      "theoretical_loss": 3.482572959827908,
      "tokens_seen": 1842610176
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000446258776328987,
      "loss": 1.587,
      "theoretical_loss": 3.4825625510931033,
      "tokens_seen": 1842675712
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044623871614844535,
      "loss": 1.4184,
      "theoretical_loss": 3.4825521428321355,
      "tokens_seen": 1842741248
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004462186559679037,
      "loss": 1.3573,
      "theoretical_loss": 3.4825417350449666,
      "tokens_seen": 1842806784
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004461985957873621,
      "loss": 1.4385,
      "theoretical_loss": 3.482531327731558,
      "tokens_seen": 1842872320
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004461785356068205,
      "loss": 1.4348,
      "theoretical_loss": 3.4825209208918713,
      "tokens_seen": 1842937856
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044615847542627885,
      "loss": 1.3783,
      "theoretical_loss": 3.4825105145258677,
      "tokens_seen": 1843003392
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004461384152457372,
      "loss": 1.4283,
      "theoretical_loss": 3.4825001086335092,
      "tokens_seen": 1843068928
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044611835506519563,
      "loss": 1.4528,
      "theoretical_loss": 3.4824897032147577,
      "tokens_seen": 1843134464
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000446098294884654,
      "loss": 1.3787,
      "theoretical_loss": 3.482479298269574,
      "tokens_seen": 1843200000
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044607823470411235,
      "loss": 1.4227,
      "theoretical_loss": 3.4824688937979205,
      "tokens_seen": 1843265536
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004460581745235707,
      "loss": 1.3747,
      "theoretical_loss": 3.4824584897997584,
      "tokens_seen": 1843331072
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004460381143430291,
      "loss": 1.4215,
      "theoretical_loss": 3.482448086275049,
      "tokens_seen": 1843396608
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004460180541624875,
      "loss": 1.4556,
      "theoretical_loss": 3.482437683223755,
      "tokens_seen": 1843462144
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044599799398194586,
      "loss": 1.2638,
      "theoretical_loss": 3.4824272806458367,
      "tokens_seen": 1843527680
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004459779338014042,
      "loss": 1.2752,
      "theoretical_loss": 3.4824168785412564,
      "tokens_seen": 1843593216
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004459578736208626,
      "loss": 1.4368,
      "theoretical_loss": 3.4824064769099756,
      "tokens_seen": 1843658752
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000445937813440321,
      "loss": 1.3987,
      "theoretical_loss": 3.4823960757519563,
      "tokens_seen": 1843724288
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044591775325977936,
      "loss": 1.3594,
      "theoretical_loss": 3.4823856750671602,
      "tokens_seen": 1843789824
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004458976930792377,
      "loss": 1.3961,
      "theoretical_loss": 3.4823752748555483,
      "tokens_seen": 1843855360
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004458776328986961,
      "loss": 1.3695,
      "theoretical_loss": 3.4823648751170824,
      "tokens_seen": 1843920896
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044585757271815445,
      "loss": 1.4205,
      "theoretical_loss": 3.4823544758517246,
      "tokens_seen": 1843986432
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044583751253761286,
      "loss": 1.3265,
      "theoretical_loss": 3.4823440770594365,
      "tokens_seen": 1844051968
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004458174523570712,
      "loss": 1.3767,
      "theoretical_loss": 3.4823336787401793,
      "tokens_seen": 1844117504
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004457973921765296,
      "loss": 1.4055,
      "theoretical_loss": 3.482323280893915,
      "tokens_seen": 1844183040
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044577733199598795,
      "loss": 1.4216,
      "theoretical_loss": 3.482312883520605,
      "tokens_seen": 1844248576
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044575727181544637,
      "loss": 1.4393,
      "theoretical_loss": 3.4823024866202115,
      "tokens_seen": 1844314112
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044573721163490473,
      "loss": 1.279,
      "theoretical_loss": 3.4822920901926957,
      "tokens_seen": 1844379648
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004457171514543631,
      "loss": 1.402,
      "theoretical_loss": 3.4822816942380195,
      "tokens_seen": 1844445184
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044569709127382145,
      "loss": 1.2247,
      "theoretical_loss": 3.482271298756145,
      "tokens_seen": 1844510720
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004456770310932798,
      "loss": 1.3041,
      "theoretical_loss": 3.482260903747033,
      "tokens_seen": 1844576256
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044565697091273823,
      "loss": 1.6129,
      "theoretical_loss": 3.4822505092106457,
      "tokens_seen": 1844641792
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004456369107321966,
      "loss": 1.4207,
      "theoretical_loss": 3.4822401151469453,
      "tokens_seen": 1844707328
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044561685055165496,
      "loss": 1.3871,
      "theoretical_loss": 3.4822297215558926,
      "tokens_seen": 1844772864
    },
    {
      "epoch": 0.56,
      "objective/train/advantage_avg": 0.49138695001602173,
      "objective/train/docs_used": 672951,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.388164758682251,
      "objective/train/original_loss": 1.3881645202636719,
      "objective/train/theoretical_loss": 3.4822193284374503,
      "objective/train/tokens_used": 1865298400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2425222545862198,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050371527671814,
      "objective/train/weighted_lm_loss": 1.4581269025802612,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.7788007855415344,
      "theoretical_loss": 3.4822193284374503,
      "tokens_seen": 1844838400
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004455967903711133,
      "loss": 1.5201,
      "theoretical_loss": 3.4822193284374503,
      "tokens_seen": 1844838400
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044557673019057173,
      "loss": 1.4516,
      "theoretical_loss": 3.4822089357915784,
      "tokens_seen": 1844903936
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004455566700100301,
      "loss": 1.4629,
      "theoretical_loss": 3.4821985436182405,
      "tokens_seen": 1844969472
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044553660982948846,
      "loss": 1.445,
      "theoretical_loss": 3.4821881519173976,
      "tokens_seen": 1845035008
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004455165496489468,
      "loss": 1.3134,
      "theoretical_loss": 3.4821777606890114,
      "tokens_seen": 1845100544
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004454964894684052,
      "loss": 1.2506,
      "theoretical_loss": 3.4821673699330438,
      "tokens_seen": 1845166080
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004454764292878636,
      "loss": 1.4159,
      "theoretical_loss": 3.4821569796494565,
      "tokens_seen": 1845231616
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044545636910732196,
      "loss": 1.4421,
      "theoretical_loss": 3.482146589838211,
      "tokens_seen": 1845297152
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004454363089267803,
      "loss": 1.3012,
      "theoretical_loss": 3.4821362004992693,
      "tokens_seen": 1845362688
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004454162487462387,
      "loss": 1.4387,
      "theoretical_loss": 3.482125811632593,
      "tokens_seen": 1845428224
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004453961885656971,
      "loss": 1.2967,
      "theoretical_loss": 3.482115423238144,
      "tokens_seen": 1845493760
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044537612838515547,
      "loss": 1.4412,
      "theoretical_loss": 3.482105035315884,
      "tokens_seen": 1845559296
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044535606820461383,
      "loss": 1.4942,
      "theoretical_loss": 3.4820946478657744,
      "tokens_seen": 1845624832
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004453360080240722,
      "loss": 1.4992,
      "theoretical_loss": 3.482084260887778,
      "tokens_seen": 1845690368
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044531594784353055,
      "loss": 1.471,
      "theoretical_loss": 3.4820738743818556,
      "tokens_seen": 1845755904
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044529588766298897,
      "loss": 1.4691,
      "theoretical_loss": 3.4820634883479693,
      "tokens_seen": 1845821440
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044527582748244733,
      "loss": 1.452,
      "theoretical_loss": 3.4820531027860815,
      "tokens_seen": 1845886976
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004452557673019057,
      "loss": 1.3199,
      "theoretical_loss": 3.4820427176961526,
      "tokens_seen": 1845952512
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044523570712136406,
      "loss": 1.5346,
      "theoretical_loss": 3.4820323330781457,
      "tokens_seen": 1846018048
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044521564694082247,
      "loss": 1.3314,
      "theoretical_loss": 3.482021948932022,
      "tokens_seen": 1846083584
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044519558676028083,
      "loss": 1.4505,
      "theoretical_loss": 3.482011565257743,
      "tokens_seen": 1846149120
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044517552657973925,
      "loss": 1.4573,
      "theoretical_loss": 3.4820011820552716,
      "tokens_seen": 1846214656
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004451554663991976,
      "loss": 1.3779,
      "theoretical_loss": 3.481990799324568,
      "tokens_seen": 1846280192
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044513540621865603,
      "loss": 1.3091,
      "theoretical_loss": 3.481980417065596,
      "tokens_seen": 1846345728
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004451153460381144,
      "loss": 1.4098,
      "theoretical_loss": 3.481970035278316,
      "tokens_seen": 1846411264
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044509528585757275,
      "loss": 1.5736,
      "theoretical_loss": 3.48195965396269,
      "tokens_seen": 1846476800
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004450752256770311,
      "loss": 1.39,
      "theoretical_loss": 3.4819492731186807,
      "tokens_seen": 1846542336
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004450551654964895,
      "loss": 1.3693,
      "theoretical_loss": 3.4819388927462485,
      "tokens_seen": 1846607872
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004450351053159479,
      "loss": 1.3667,
      "theoretical_loss": 3.4819285128453563,
      "tokens_seen": 1846673408
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044501504513540626,
      "loss": 1.4109,
      "theoretical_loss": 3.4819181334159657,
      "tokens_seen": 1846738944
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004449949849548646,
      "loss": 1.3805,
      "theoretical_loss": 3.481907754458039,
      "tokens_seen": 1846804480
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000444974924774323,
      "loss": 1.5005,
      "theoretical_loss": 3.4818973759715366,
      "tokens_seen": 1846870016
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004449548645937814,
      "loss": 1.3926,
      "theoretical_loss": 3.4818869979564218,
      "tokens_seen": 1846935552
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044493480441323976,
      "loss": 1.3699,
      "theoretical_loss": 3.481876620412656,
      "tokens_seen": 1847001088
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004449147442326981,
      "loss": 1.402,
      "theoretical_loss": 3.4818662433402014,
      "tokens_seen": 1847066624
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004448946840521565,
      "loss": 1.4413,
      "theoretical_loss": 3.4818558667390196,
      "tokens_seen": 1847132160
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044487462387161485,
      "loss": 1.4602,
      "theoretical_loss": 3.481845490609072,
      "tokens_seen": 1847197696
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044485456369107326,
      "loss": 1.3369,
      "theoretical_loss": 3.4818351149503206,
      "tokens_seen": 1847263232
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004448345035105316,
      "loss": 1.4411,
      "theoretical_loss": 3.4818247397627284,
      "tokens_seen": 1847328768
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044481444332999,
      "loss": 1.3265,
      "theoretical_loss": 3.4818143650462563,
      "tokens_seen": 1847394304
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044479438314944835,
      "loss": 1.3277,
      "theoretical_loss": 3.481803990800866,
      "tokens_seen": 1847459840
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044477432296890677,
      "loss": 1.4376,
      "theoretical_loss": 3.4817936170265202,
      "tokens_seen": 1847525376
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044475426278836513,
      "loss": 1.432,
      "theoretical_loss": 3.48178324372318,
      "tokens_seen": 1847590912
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004447342026078235,
      "loss": 1.2985,
      "theoretical_loss": 3.481772870890808,
      "tokens_seen": 1847656448
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044471414242728185,
      "loss": 1.3904,
      "theoretical_loss": 3.481762498529365,
      "tokens_seen": 1847721984
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004446940822467402,
      "loss": 1.4075,
      "theoretical_loss": 3.4817521266388147,
      "tokens_seen": 1847787520
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044467402206619863,
      "loss": 1.2821,
      "theoretical_loss": 3.481741755219118,
      "tokens_seen": 1847853056
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000444653961885657,
      "loss": 1.5241,
      "theoretical_loss": 3.4817313842702364,
      "tokens_seen": 1847918592
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044463390170511536,
      "loss": 1.4767,
      "theoretical_loss": 3.4817210137921326,
      "tokens_seen": 1847984128
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004446138415245737,
      "loss": 1.4423,
      "theoretical_loss": 3.4817106437847682,
      "tokens_seen": 1848049664
    },
    {
      "epoch": 0.56,
      "objective/train/advantage_avg": 0.4928208887577057,
      "objective/train/docs_used": 673981,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3690332174301147,
      "objective/train/original_loss": 1.3690329790115356,
      "objective/train/theoretical_loss": 3.481700274248105,
      "objective/train/tokens_used": 1868575200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24323411285877228,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505186319351196,
      "objective/train/weighted_lm_loss": 1.4381954669952393,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.481700274248105,
      "tokens_seen": 1848115200
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044459378134403214,
      "loss": 1.4438,
      "theoretical_loss": 3.481700274248105,
      "tokens_seen": 1848115200
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004445737211634905,
      "loss": 1.4361,
      "theoretical_loss": 3.4816899051821055,
      "tokens_seen": 1848180736
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044455366098294886,
      "loss": 1.5101,
      "theoretical_loss": 3.481679536586731,
      "tokens_seen": 1848246272
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004445336008024072,
      "loss": 1.4272,
      "theoretical_loss": 3.4816691684619436,
      "tokens_seen": 1848311808
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004445135406218656,
      "loss": 1.5047,
      "theoretical_loss": 3.481658800807706,
      "tokens_seen": 1848377344
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000444493480441324,
      "loss": 1.6086,
      "theoretical_loss": 3.4816484336239784,
      "tokens_seen": 1848442880
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044447342026078236,
      "loss": 1.4912,
      "theoretical_loss": 3.4816380669107247,
      "tokens_seen": 1848508416
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004444533600802407,
      "loss": 1.3777,
      "theoretical_loss": 3.481627700667906,
      "tokens_seen": 1848573952
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004444332998996991,
      "loss": 1.3733,
      "theoretical_loss": 3.4816173348954846,
      "tokens_seen": 1848639488
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004444132397191575,
      "loss": 1.4991,
      "theoretical_loss": 3.4816069695934218,
      "tokens_seen": 1848705024
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044439317953861587,
      "loss": 1.4517,
      "theoretical_loss": 3.48159660476168,
      "tokens_seen": 1848770560
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044437311935807423,
      "loss": 1.4724,
      "theoretical_loss": 3.4815862404002216,
      "tokens_seen": 1848836096
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004443530591775326,
      "loss": 1.3501,
      "theoretical_loss": 3.481575876509008,
      "tokens_seen": 1848901632
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044433299899699095,
      "loss": 1.562,
      "theoretical_loss": 3.481565513088001,
      "tokens_seen": 1848967168
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044431293881644937,
      "loss": 1.4763,
      "theoretical_loss": 3.4815551501371633,
      "tokens_seen": 1849032704
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044429287863590773,
      "loss": 1.5255,
      "theoretical_loss": 3.481544787656457,
      "tokens_seen": 1849098240
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004442728184553661,
      "loss": 1.4547,
      "theoretical_loss": 3.4815344256458434,
      "tokens_seen": 1849163776
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044425275827482446,
      "loss": 1.5212,
      "theoretical_loss": 3.481524064105285,
      "tokens_seen": 1849229312
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044423269809428287,
      "loss": 1.4803,
      "theoretical_loss": 3.481513703034743,
      "tokens_seen": 1849294848
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044421263791374123,
      "loss": 1.3757,
      "theoretical_loss": 3.4815033424341806,
      "tokens_seen": 1849360384
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004441925777331996,
      "loss": 1.4354,
      "theoretical_loss": 3.4814929823035596,
      "tokens_seen": 1849425920
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044417251755265796,
      "loss": 1.4176,
      "theoretical_loss": 3.481482622642841,
      "tokens_seen": 1849491456
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004441524573721163,
      "loss": 1.533,
      "theoretical_loss": 3.4814722634519883,
      "tokens_seen": 1849556992
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044413239719157474,
      "loss": 1.4406,
      "theoretical_loss": 3.4814619047309625,
      "tokens_seen": 1849622528
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004441123370110331,
      "loss": 1.4871,
      "theoretical_loss": 3.481451546479726,
      "tokens_seen": 1849688064
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044409227683049146,
      "loss": 1.3627,
      "theoretical_loss": 3.481441188698241,
      "tokens_seen": 1849753600
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004440722166499498,
      "loss": 1.48,
      "theoretical_loss": 3.4814308313864686,
      "tokens_seen": 1849819136
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044405215646940824,
      "loss": 1.3912,
      "theoretical_loss": 3.481420474544372,
      "tokens_seen": 1849884672
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004440320962888666,
      "loss": 1.4531,
      "theoretical_loss": 3.4814101181719135,
      "tokens_seen": 1849950208
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044401203610832497,
      "loss": 1.5204,
      "theoretical_loss": 3.481399762269054,
      "tokens_seen": 1850015744
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044399197592778333,
      "loss": 1.4877,
      "theoretical_loss": 3.4813894068357563,
      "tokens_seen": 1850081280
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004439719157472417,
      "loss": 1.2629,
      "theoretical_loss": 3.481379051871982,
      "tokens_seen": 1850146816
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004439518555667001,
      "loss": 1.4742,
      "theoretical_loss": 3.4813686973776936,
      "tokens_seen": 1850212352
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044393179538615847,
      "loss": 1.3092,
      "theoretical_loss": 3.481358343352853,
      "tokens_seen": 1850277888
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044391173520561683,
      "loss": 1.4518,
      "theoretical_loss": 3.4813479897974227,
      "tokens_seen": 1850343424
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004438916750250752,
      "loss": 1.394,
      "theoretical_loss": 3.4813376367113644,
      "tokens_seen": 1850408960
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004438716148445336,
      "loss": 1.4057,
      "theoretical_loss": 3.48132728409464,
      "tokens_seen": 1850474496
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044385155466399197,
      "loss": 1.4478,
      "theoretical_loss": 3.481316931947212,
      "tokens_seen": 1850540032
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044383149448345033,
      "loss": 1.396,
      "theoretical_loss": 3.481306580269042,
      "tokens_seen": 1850605568
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004438114343029087,
      "loss": 1.5167,
      "theoretical_loss": 3.481296229060093,
      "tokens_seen": 1850671104
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004437913741223671,
      "loss": 1.5403,
      "theoretical_loss": 3.4812858783203264,
      "tokens_seen": 1850736640
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004437713139418255,
      "loss": 1.4142,
      "theoretical_loss": 3.4812755280497045,
      "tokens_seen": 1850802176
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044375125376128384,
      "loss": 1.5195,
      "theoretical_loss": 3.4812651782481896,
      "tokens_seen": 1850867712
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004437311935807422,
      "loss": 1.5058,
      "theoretical_loss": 3.4812548289157434,
      "tokens_seen": 1850933248
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044371113340020056,
      "loss": 1.5035,
      "theoretical_loss": 3.481244480052329,
      "tokens_seen": 1850998784
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044369107321965903,
      "loss": 1.3656,
      "theoretical_loss": 3.481234131657907,
      "tokens_seen": 1851064320
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004436710130391174,
      "loss": 1.4866,
      "theoretical_loss": 3.4812237837324407,
      "tokens_seen": 1851129856
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044365095285857576,
      "loss": 1.3965,
      "theoretical_loss": 3.4812134362758913,
      "tokens_seen": 1851195392
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004436308926780341,
      "loss": 1.3518,
      "theoretical_loss": 3.4812030892882224,
      "tokens_seen": 1851260928
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044361083249749254,
      "loss": 1.3071,
      "theoretical_loss": 3.4811927427693954,
      "tokens_seen": 1851326464
    },
    {
      "epoch": 0.56,
      "objective/train/advantage_avg": 0.4888497292995453,
      "objective/train/docs_used": 675199,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4120993614196777,
      "objective/train/original_loss": 1.4120994806289673,
      "objective/train/theoretical_loss": 3.4811823967193716,
      "objective/train/tokens_used": 1871852000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23948651552200317,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501022338867188,
      "objective/train/weighted_lm_loss": 1.4826780557632446,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4811823967193716,
      "tokens_seen": 1851392000
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004435907723169509,
      "loss": 1.3805,
      "theoretical_loss": 3.4811823967193716,
      "tokens_seen": 1851392000
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044357071213640926,
      "loss": 1.5146,
      "theoretical_loss": 3.4811720511381146,
      "tokens_seen": 1851457536
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004435506519558676,
      "loss": 1.3828,
      "theoretical_loss": 3.4811617060255857,
      "tokens_seen": 1851523072
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000443530591775326,
      "loss": 1.4837,
      "theoretical_loss": 3.4811513613817473,
      "tokens_seen": 1851588608
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004435105315947844,
      "loss": 1.4599,
      "theoretical_loss": 3.4811410172065615,
      "tokens_seen": 1851654144
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044349047141424276,
      "loss": 1.4869,
      "theoretical_loss": 3.481130673499991,
      "tokens_seen": 1851719680
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004434704112337011,
      "loss": 1.2904,
      "theoretical_loss": 3.481120330261997,
      "tokens_seen": 1851785216
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004434503510531595,
      "loss": 1.3326,
      "theoretical_loss": 3.481109987492543,
      "tokens_seen": 1851850752
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004434302908726179,
      "loss": 1.4077,
      "theoretical_loss": 3.4810996451915894,
      "tokens_seen": 1851916288
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044341023069207627,
      "loss": 1.389,
      "theoretical_loss": 3.4810893033591004,
      "tokens_seen": 1851981824
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044339017051153463,
      "loss": 1.4694,
      "theoretical_loss": 3.4810789619950366,
      "tokens_seen": 1852047360
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000443370110330993,
      "loss": 1.3834,
      "theoretical_loss": 3.481068621099361,
      "tokens_seen": 1852112896
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044335005015045135,
      "loss": 1.4911,
      "theoretical_loss": 3.481058280672036,
      "tokens_seen": 1852178432
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044332998996990977,
      "loss": 1.5031,
      "theoretical_loss": 3.4810479407130233,
      "tokens_seen": 1852243968
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044330992978936813,
      "loss": 1.252,
      "theoretical_loss": 3.481037601222285,
      "tokens_seen": 1852309504
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004432898696088265,
      "loss": 1.4533,
      "theoretical_loss": 3.4810272621997838,
      "tokens_seen": 1852375040
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044326980942828486,
      "loss": 1.4161,
      "theoretical_loss": 3.4810169236454818,
      "tokens_seen": 1852440576
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044324974924774327,
      "loss": 1.4538,
      "theoretical_loss": 3.4810065855593413,
      "tokens_seen": 1852506112
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044322968906720163,
      "loss": 1.4054,
      "theoretical_loss": 3.480996247941324,
      "tokens_seen": 1852571648
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044320962888666,
      "loss": 1.3867,
      "theoretical_loss": 3.480985910791393,
      "tokens_seen": 1852637184
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044318956870611836,
      "loss": 1.3323,
      "theoretical_loss": 3.48097557410951,
      "tokens_seen": 1852702720
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004431695085255767,
      "loss": 1.4835,
      "theoretical_loss": 3.4809652378956373,
      "tokens_seen": 1852768256
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044314944834503514,
      "loss": 1.4424,
      "theoretical_loss": 3.4809549021497372,
      "tokens_seen": 1852833792
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004431293881644935,
      "loss": 1.4228,
      "theoretical_loss": 3.4809445668717722,
      "tokens_seen": 1852899328
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044310932798395186,
      "loss": 1.4012,
      "theoretical_loss": 3.4809342320617045,
      "tokens_seen": 1852964864
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004430892678034102,
      "loss": 1.5008,
      "theoretical_loss": 3.480923897719496,
      "tokens_seen": 1853030400
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044306920762286864,
      "loss": 1.378,
      "theoretical_loss": 3.480913563845109,
      "tokens_seen": 1853095936
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000443049147442327,
      "loss": 1.3129,
      "theoretical_loss": 3.480903230438506,
      "tokens_seen": 1853161472
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044302908726178537,
      "loss": 1.3836,
      "theoretical_loss": 3.4808928974996496,
      "tokens_seen": 1853227008
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044300902708124373,
      "loss": 1.3344,
      "theoretical_loss": 3.4808825650285016,
      "tokens_seen": 1853292544
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004429889669007021,
      "loss": 1.4357,
      "theoretical_loss": 3.480872233025024,
      "tokens_seen": 1853358080
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004429689067201605,
      "loss": 1.3416,
      "theoretical_loss": 3.48086190148918,
      "tokens_seen": 1853423616
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044294884653961887,
      "loss": 1.5662,
      "theoretical_loss": 3.4808515704209317,
      "tokens_seen": 1853489152
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044292878635907723,
      "loss": 1.4399,
      "theoretical_loss": 3.480841239820241,
      "tokens_seen": 1853554688
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004429087261785356,
      "loss": 1.4001,
      "theoretical_loss": 3.4808309096870698,
      "tokens_seen": 1853620224
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000442888665997994,
      "loss": 1.4147,
      "theoretical_loss": 3.4808205800213816,
      "tokens_seen": 1853685760
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044286860581745237,
      "loss": 1.3574,
      "theoretical_loss": 3.4808102508231378,
      "tokens_seen": 1853751296
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044284854563691073,
      "loss": 1.4469,
      "theoretical_loss": 3.4807999220923014,
      "tokens_seen": 1853816832
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004428284854563691,
      "loss": 1.3123,
      "theoretical_loss": 3.480789593828834,
      "tokens_seen": 1853882368
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004428084252758275,
      "loss": 1.4231,
      "theoretical_loss": 3.480779266032698,
      "tokens_seen": 1853947904
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004427883650952859,
      "loss": 1.3885,
      "theoretical_loss": 3.4807689387038563,
      "tokens_seen": 1854013440
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044276830491474424,
      "loss": 1.4787,
      "theoretical_loss": 3.480758611842271,
      "tokens_seen": 1854078976
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004427482447342026,
      "loss": 1.4775,
      "theoretical_loss": 3.4807482854479037,
      "tokens_seen": 1854144512
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044272818455366096,
      "loss": 1.3905,
      "theoretical_loss": 3.4807379595207184,
      "tokens_seen": 1854210048
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004427081243731194,
      "loss": 1.4715,
      "theoretical_loss": 3.480727634060676,
      "tokens_seen": 1854275584
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044268806419257774,
      "loss": 1.3996,
      "theoretical_loss": 3.4807173090677397,
      "tokens_seen": 1854341120
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004426680040120361,
      "loss": 1.4835,
      "theoretical_loss": 3.4807069845418708,
      "tokens_seen": 1854406656
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044264794383149447,
      "loss": 1.5103,
      "theoretical_loss": 3.480696660483033,
      "tokens_seen": 1854472192
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004426278836509529,
      "loss": 1.4289,
      "theoretical_loss": 3.480686336891188,
      "tokens_seen": 1854537728
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044260782347041124,
      "loss": 1.4533,
      "theoretical_loss": 3.480676013766298,
      "tokens_seen": 1854603264
    },
    {
      "epoch": 0.56,
      "objective/train/advantage_avg": 0.4945298731327057,
      "objective/train/docs_used": 676483,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4758983850479126,
      "objective/train/original_loss": 1.475898265838623,
      "objective/train/theoretical_loss": 3.4806656911083254,
      "objective/train/tokens_used": 1875128800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24503852427005768,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506986379623413,
      "objective/train/weighted_lm_loss": 1.5508050918579102,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4806656911083254,
      "tokens_seen": 1854668800
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004425877632898696,
      "loss": 1.5948,
      "theoretical_loss": 3.4806656911083254,
      "tokens_seen": 1854668800
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044256770310932797,
      "loss": 1.3602,
      "theoretical_loss": 3.4806553689172333,
      "tokens_seen": 1854734336
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044254764292878633,
      "loss": 1.4845,
      "theoretical_loss": 3.4806450471929833,
      "tokens_seen": 1854799872
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044252758274824475,
      "loss": 1.4442,
      "theoretical_loss": 3.480634725935538,
      "tokens_seen": 1854865408
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004425075225677031,
      "loss": 1.424,
      "theoretical_loss": 3.48062440514486,
      "tokens_seen": 1854930944
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044248746238716147,
      "loss": 1.4168,
      "theoretical_loss": 3.4806140848209113,
      "tokens_seen": 1854996480
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044246740220661983,
      "loss": 1.4029,
      "theoretical_loss": 3.480603764963655,
      "tokens_seen": 1855062016
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044244734202607825,
      "loss": 1.5013,
      "theoretical_loss": 3.480593445573053,
      "tokens_seen": 1855127552
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004424272818455366,
      "loss": 1.4178,
      "theoretical_loss": 3.4805831266490674,
      "tokens_seen": 1855193088
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000442407221664995,
      "loss": 1.3648,
      "theoretical_loss": 3.480572808191661,
      "tokens_seen": 1855258624
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044238716148445334,
      "loss": 1.3769,
      "theoretical_loss": 3.4805624902007963,
      "tokens_seen": 1855324160
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004423671013039117,
      "loss": 1.3492,
      "theoretical_loss": 3.480552172676436,
      "tokens_seen": 1855389696
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004423470411233701,
      "loss": 1.4636,
      "theoretical_loss": 3.480541855618542,
      "tokens_seen": 1855455232
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004423269809428285,
      "loss": 1.469,
      "theoretical_loss": 3.4805315390270772,
      "tokens_seen": 1855520768
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044230692076228684,
      "loss": 1.3329,
      "theoretical_loss": 3.480521222902003,
      "tokens_seen": 1855586304
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004422868605817452,
      "loss": 1.3722,
      "theoretical_loss": 3.4805109072432834,
      "tokens_seen": 1855651840
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004422668004012036,
      "loss": 1.363,
      "theoretical_loss": 3.4805005920508796,
      "tokens_seen": 1855717376
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000442246740220662,
      "loss": 1.3701,
      "theoretical_loss": 3.480490277324755,
      "tokens_seen": 1855782912
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044222668004012034,
      "loss": 1.3943,
      "theoretical_loss": 3.480479963064871,
      "tokens_seen": 1855848448
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004422066198595787,
      "loss": 1.5046,
      "theoretical_loss": 3.480469649271191,
      "tokens_seen": 1855913984
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004421865596790371,
      "loss": 1.4402,
      "theoretical_loss": 3.480459335943676,
      "tokens_seen": 1855979520
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044216649949849554,
      "loss": 1.5538,
      "theoretical_loss": 3.480449023082291,
      "tokens_seen": 1856045056
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004421464393179539,
      "loss": 1.414,
      "theoretical_loss": 3.4804387106869967,
      "tokens_seen": 1856110592
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044212637913741226,
      "loss": 1.3584,
      "theoretical_loss": 3.4804283987577556,
      "tokens_seen": 1856176128
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004421063189568706,
      "loss": 1.3512,
      "theoretical_loss": 3.4804180872945305,
      "tokens_seen": 1856241664
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044208625877632904,
      "loss": 1.3055,
      "theoretical_loss": 3.480407776297284,
      "tokens_seen": 1856307200
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004420661985957874,
      "loss": 1.4005,
      "theoretical_loss": 3.4803974657659786,
      "tokens_seen": 1856372736
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044204613841524577,
      "loss": 1.3464,
      "theoretical_loss": 3.4803871557005763,
      "tokens_seen": 1856438272
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044202607823470413,
      "loss": 1.3662,
      "theoretical_loss": 3.4803768461010405,
      "tokens_seen": 1856503808
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004420060180541625,
      "loss": 1.4625,
      "theoretical_loss": 3.480366536967333,
      "tokens_seen": 1856569344
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004419859578736209,
      "loss": 1.452,
      "theoretical_loss": 3.4803562282994163,
      "tokens_seen": 1856634880
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044196589769307927,
      "loss": 1.4331,
      "theoretical_loss": 3.4803459200972533,
      "tokens_seen": 1856700416
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044194583751253763,
      "loss": 1.5279,
      "theoretical_loss": 3.4803356123608062,
      "tokens_seen": 1856765952
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000441925777331996,
      "loss": 1.3755,
      "theoretical_loss": 3.4803253050900373,
      "tokens_seen": 1856831488
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004419057171514544,
      "loss": 1.3453,
      "theoretical_loss": 3.48031499828491,
      "tokens_seen": 1856897024
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044188565697091277,
      "loss": 1.4411,
      "theoretical_loss": 3.4803046919453866,
      "tokens_seen": 1856962560
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044186559679037113,
      "loss": 1.4222,
      "theoretical_loss": 3.480294386071429,
      "tokens_seen": 1857028096
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004418455366098295,
      "loss": 1.4876,
      "theoretical_loss": 3.480284080663,
      "tokens_seen": 1857093632
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004418254764292879,
      "loss": 1.4532,
      "theoretical_loss": 3.480273775720062,
      "tokens_seen": 1857159168
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004418054162487463,
      "loss": 1.2828,
      "theoretical_loss": 3.480263471242578,
      "tokens_seen": 1857224704
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044178535606820464,
      "loss": 1.423,
      "theoretical_loss": 3.4802531672305106,
      "tokens_seen": 1857290240
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000441765295887663,
      "loss": 1.4851,
      "theoretical_loss": 3.4802428636838214,
      "tokens_seen": 1857355776
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044174523570712136,
      "loss": 1.4205,
      "theoretical_loss": 3.480232560602474,
      "tokens_seen": 1857421312
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004417251755265798,
      "loss": 1.424,
      "theoretical_loss": 3.480222257986431,
      "tokens_seen": 1857486848
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044170511534603814,
      "loss": 1.4754,
      "theoretical_loss": 3.480211955835654,
      "tokens_seen": 1857552384
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004416850551654965,
      "loss": 1.435,
      "theoretical_loss": 3.4802016541501066,
      "tokens_seen": 1857617920
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044166499498495487,
      "loss": 1.3268,
      "theoretical_loss": 3.480191352929751,
      "tokens_seen": 1857683456
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004416449348044133,
      "loss": 1.4616,
      "theoretical_loss": 3.480181052174549,
      "tokens_seen": 1857748992
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044162487462387164,
      "loss": 1.4641,
      "theoretical_loss": 3.4801707518844647,
      "tokens_seen": 1857814528
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044160481444333,
      "loss": 1.3097,
      "theoretical_loss": 3.48016045205946,
      "tokens_seen": 1857880064
    },
    {
      "epoch": 0.56,
      "objective/train/advantage_avg": 0.49308258295059204,
      "objective/train/docs_used": 677697,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2494289875030518,
      "objective/train/original_loss": 1.2494289875030518,
      "objective/train/theoretical_loss": 3.4801501526994967,
      "objective/train/tokens_used": 1878405600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24350279569625854,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505461692810059,
      "objective/train/weighted_lm_loss": 1.3122785091400146,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4801501526994967,
      "tokens_seen": 1857945600
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044158475426278837,
      "loss": 1.3879,
      "theoretical_loss": 3.4801501526994967,
      "tokens_seen": 1857945600
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044156469408224673,
      "loss": 1.3513,
      "theoretical_loss": 3.480139853804538,
      "tokens_seen": 1858011136
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044154463390170515,
      "loss": 1.3886,
      "theoretical_loss": 3.480129555374547,
      "tokens_seen": 1858076672
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004415245737211635,
      "loss": 1.4277,
      "theoretical_loss": 3.480119257409486,
      "tokens_seen": 1858142208
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044150451354062187,
      "loss": 1.4339,
      "theoretical_loss": 3.480108959909318,
      "tokens_seen": 1858207744
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044148445336008023,
      "loss": 1.4523,
      "theoretical_loss": 3.480098662874004,
      "tokens_seen": 1858273280
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044146439317953865,
      "loss": 1.4162,
      "theoretical_loss": 3.4800883663035083,
      "tokens_seen": 1858338816
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000441444332998997,
      "loss": 1.3697,
      "theoretical_loss": 3.480078070197793,
      "tokens_seen": 1858404352
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004414242728184554,
      "loss": 1.3736,
      "theoretical_loss": 3.480067774556821,
      "tokens_seen": 1858469888
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044140421263791374,
      "loss": 1.3172,
      "theoretical_loss": 3.4800574793805543,
      "tokens_seen": 1858535424
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004413841524573721,
      "loss": 1.361,
      "theoretical_loss": 3.4800471846689556,
      "tokens_seen": 1858600960
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004413640922768305,
      "loss": 1.4998,
      "theoretical_loss": 3.4800368904219887,
      "tokens_seen": 1858666496
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004413440320962889,
      "loss": 1.4603,
      "theoretical_loss": 3.4800265966396147,
      "tokens_seen": 1858732032
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044132397191574724,
      "loss": 1.4295,
      "theoretical_loss": 3.480016303321797,
      "tokens_seen": 1858797568
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004413039117352056,
      "loss": 1.4536,
      "theoretical_loss": 3.4800060104684984,
      "tokens_seen": 1858863104
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000441283851554664,
      "loss": 1.3218,
      "theoretical_loss": 3.4799957180796817,
      "tokens_seen": 1858928640
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004412637913741224,
      "loss": 1.3421,
      "theoretical_loss": 3.4799854261553085,
      "tokens_seen": 1858994176
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044124373119358074,
      "loss": 1.3409,
      "theoretical_loss": 3.4799751346953425,
      "tokens_seen": 1859059712
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004412236710130391,
      "loss": 1.3173,
      "theoretical_loss": 3.4799648436997463,
      "tokens_seen": 1859125248
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044120361083249747,
      "loss": 1.3969,
      "theoretical_loss": 3.479954553168482,
      "tokens_seen": 1859190784
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004411835506519559,
      "loss": 1.4035,
      "theoretical_loss": 3.4799442631015127,
      "tokens_seen": 1859256320
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044116349047141425,
      "loss": 1.3725,
      "theoretical_loss": 3.4799339734988015,
      "tokens_seen": 1859321856
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004411434302908726,
      "loss": 1.3882,
      "theoretical_loss": 3.47992368436031,
      "tokens_seen": 1859387392
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044112337011033097,
      "loss": 1.3872,
      "theoretical_loss": 3.479913395686002,
      "tokens_seen": 1859452928
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004411033099297894,
      "loss": 1.3037,
      "theoretical_loss": 3.479903107475839,
      "tokens_seen": 1859518464
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044108324974924775,
      "loss": 1.4362,
      "theoretical_loss": 3.479892819729785,
      "tokens_seen": 1859584000
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004410631895687061,
      "loss": 1.3524,
      "theoretical_loss": 3.4798825324478018,
      "tokens_seen": 1859649536
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004410431293881645,
      "loss": 1.3958,
      "theoretical_loss": 3.4798722456298528,
      "tokens_seen": 1859715072
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044102306920762284,
      "loss": 1.5407,
      "theoretical_loss": 3.4798619592759,
      "tokens_seen": 1859780608
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044100300902708125,
      "loss": 1.3255,
      "theoretical_loss": 3.4798516733859066,
      "tokens_seen": 1859846144
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004409829488465396,
      "loss": 1.5385,
      "theoretical_loss": 3.4798413879598353,
      "tokens_seen": 1859911680
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000440962888665998,
      "loss": 1.3602,
      "theoretical_loss": 3.4798311029976485,
      "tokens_seen": 1859977216
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044094282848545634,
      "loss": 1.3782,
      "theoretical_loss": 3.4798208184993094,
      "tokens_seen": 1860042752
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044092276830491476,
      "loss": 1.2952,
      "theoretical_loss": 3.4798105344647805,
      "tokens_seen": 1860108288
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004409027081243731,
      "loss": 1.4085,
      "theoretical_loss": 3.4798002508940242,
      "tokens_seen": 1860173824
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004408826479438315,
      "loss": 1.4382,
      "theoretical_loss": 3.4797899677870037,
      "tokens_seen": 1860239360
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044086258776328984,
      "loss": 1.4435,
      "theoretical_loss": 3.479779685143682,
      "tokens_seen": 1860304896
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004408425275827482,
      "loss": 1.3911,
      "theoretical_loss": 3.479769402964021,
      "tokens_seen": 1860370432
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004408224674022066,
      "loss": 1.4942,
      "theoretical_loss": 3.479759121247984,
      "tokens_seen": 1860435968
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000440802407221665,
      "loss": 1.3743,
      "theoretical_loss": 3.4797488399955343,
      "tokens_seen": 1860501504
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044078234704112335,
      "loss": 1.5584,
      "theoretical_loss": 3.479738559206633,
      "tokens_seen": 1860567040
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004407622868605817,
      "loss": 1.4223,
      "theoretical_loss": 3.479728278881245,
      "tokens_seen": 1860632576
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004407422266800401,
      "loss": 1.3786,
      "theoretical_loss": 3.479717999019332,
      "tokens_seen": 1860698112
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004407221664994985,
      "loss": 1.3975,
      "theoretical_loss": 3.4797077196208566,
      "tokens_seen": 1860763648
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044070210631895685,
      "loss": 1.4072,
      "theoretical_loss": 3.4796974406857815,
      "tokens_seen": 1860829184
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044068204613841527,
      "loss": 1.3739,
      "theoretical_loss": 3.47968716221407,
      "tokens_seen": 1860894720
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004406619859578737,
      "loss": 1.3686,
      "theoretical_loss": 3.4796768842056847,
      "tokens_seen": 1860960256
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044064192577733204,
      "loss": 1.4927,
      "theoretical_loss": 3.4796666066605884,
      "tokens_seen": 1861025792
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004406218655967904,
      "loss": 1.5127,
      "theoretical_loss": 3.479656329578744,
      "tokens_seen": 1861091328
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044060180541624877,
      "loss": 1.3924,
      "theoretical_loss": 3.4796460529601143,
      "tokens_seen": 1861156864
    },
    {
      "epoch": 0.56,
      "objective/train/advantage_avg": 0.4917011260986328,
      "objective/train/docs_used": 678968,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3713939189910889,
      "objective/train/original_loss": 1.3713939189910889,
      "objective/train/theoretical_loss": 3.4796357768046615,
      "objective/train/tokens_used": 1881682400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2426432967185974,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504034757614136,
      "objective/train/weighted_lm_loss": 1.44036066532135,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.4796357768046615,
      "tokens_seen": 1861222400
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044058174523570713,
      "loss": 1.3985,
      "theoretical_loss": 3.4796357768046615,
      "tokens_seen": 1861222400
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044056168505516555,
      "loss": 1.4453,
      "theoretical_loss": 3.4796255011123494,
      "tokens_seen": 1861287936
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004405416248746239,
      "loss": 1.3333,
      "theoretical_loss": 3.4796152258831405,
      "tokens_seen": 1861353472
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044052156469408227,
      "loss": 1.4282,
      "theoretical_loss": 3.479604951116997,
      "tokens_seen": 1861419008
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044050150451354063,
      "loss": 1.3746,
      "theoretical_loss": 3.4795946768138823,
      "tokens_seen": 1861484544
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044048144433299905,
      "loss": 1.3949,
      "theoretical_loss": 3.4795844029737593,
      "tokens_seen": 1861550080
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004404613841524574,
      "loss": 1.4332,
      "theoretical_loss": 3.4795741295965907,
      "tokens_seen": 1861615616
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004404413239719158,
      "loss": 1.4263,
      "theoretical_loss": 3.479563856682339,
      "tokens_seen": 1861681152
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044042126379137414,
      "loss": 1.4479,
      "theoretical_loss": 3.4795535842309677,
      "tokens_seen": 1861746688
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004404012036108325,
      "loss": 1.5118,
      "theoretical_loss": 3.4795433122424395,
      "tokens_seen": 1861812224
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004403811434302909,
      "loss": 1.5397,
      "theoretical_loss": 3.4795330407167167,
      "tokens_seen": 1861877760
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004403610832497493,
      "loss": 1.406,
      "theoretical_loss": 3.4795227696537623,
      "tokens_seen": 1861943296
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044034102306920764,
      "loss": 1.4602,
      "theoretical_loss": 3.4795124990535395,
      "tokens_seen": 1862008832
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000440320962888666,
      "loss": 1.3469,
      "theoretical_loss": 3.4795022289160116,
      "tokens_seen": 1862074368
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004403009027081244,
      "loss": 1.2699,
      "theoretical_loss": 3.4794919592411406,
      "tokens_seen": 1862139904
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004402808425275828,
      "loss": 1.4047,
      "theoretical_loss": 3.47948169002889,
      "tokens_seen": 1862205440
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044026078234704114,
      "loss": 1.4329,
      "theoretical_loss": 3.479471421279222,
      "tokens_seen": 1862270976
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004402407221664995,
      "loss": 1.4441,
      "theoretical_loss": 3.4794611529920996,
      "tokens_seen": 1862336512
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044022066198595787,
      "loss": 1.4177,
      "theoretical_loss": 3.479450885167487,
      "tokens_seen": 1862402048
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004402006018054163,
      "loss": 1.4456,
      "theoretical_loss": 3.479440617805345,
      "tokens_seen": 1862467584
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044018054162487465,
      "loss": 1.4536,
      "theoretical_loss": 3.4794303509056377,
      "tokens_seen": 1862533120
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000440160481444333,
      "loss": 1.3749,
      "theoretical_loss": 3.4794200844683285,
      "tokens_seen": 1862598656
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044014042126379137,
      "loss": 1.3729,
      "theoretical_loss": 3.4794098184933793,
      "tokens_seen": 1862664192
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004401203610832498,
      "loss": 1.4301,
      "theoretical_loss": 3.479399552980753,
      "tokens_seen": 1862729728
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044010030090270815,
      "loss": 1.4621,
      "theoretical_loss": 3.479389287930413,
      "tokens_seen": 1862795264
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004400802407221665,
      "loss": 1.4409,
      "theoretical_loss": 3.4793790233423225,
      "tokens_seen": 1862860800
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004400601805416249,
      "loss": 1.3714,
      "theoretical_loss": 3.479368759216444,
      "tokens_seen": 1862926336
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044004012036108324,
      "loss": 1.4337,
      "theoretical_loss": 3.4793584955527406,
      "tokens_seen": 1862991872
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044002006018054165,
      "loss": 1.4151,
      "theoretical_loss": 3.4793482323511746,
      "tokens_seen": 1863057408
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00044,
      "loss": 1.4487,
      "theoretical_loss": 3.4793379696117093,
      "tokens_seen": 1863122944
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004399799398194584,
      "loss": 1.4476,
      "theoretical_loss": 3.479327707334308,
      "tokens_seen": 1863188480
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00043995987963891674,
      "loss": 1.4357,
      "theoretical_loss": 3.4793174455189337,
      "tokens_seen": 1863254016
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00043993981945837516,
      "loss": 1.5529,
      "theoretical_loss": 3.479307184165549,
      "tokens_seen": 1863319552
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004399197592778335,
      "loss": 1.5103,
      "theoretical_loss": 3.4792969232741164,
      "tokens_seen": 1863385088
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004398996990972919,
      "loss": 1.288,
      "theoretical_loss": 3.4792866628446,
      "tokens_seen": 1863450624
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00043987963891675024,
      "loss": 1.4593,
      "theoretical_loss": 3.479276402876962,
      "tokens_seen": 1863516160
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004398595787362086,
      "loss": 1.3556,
      "theoretical_loss": 3.4792661433711656,
      "tokens_seen": 1863581696
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.000439839518555667,
      "loss": 1.501,
      "theoretical_loss": 3.4792558843271735,
      "tokens_seen": 1863647232
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004398194583751254,
      "loss": 1.4251,
      "theoretical_loss": 3.4792456257449484,
      "tokens_seen": 1863712768
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00043979939819458375,
      "loss": 1.442,
      "theoretical_loss": 3.4792353676244545,
      "tokens_seen": 1863778304
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004397793380140421,
      "loss": 1.3412,
      "theoretical_loss": 3.479225109965653,
      "tokens_seen": 1863843840
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004397592778335005,
      "loss": 1.4419,
      "theoretical_loss": 3.479214852768509,
      "tokens_seen": 1863909376
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004397392176529589,
      "loss": 1.4793,
      "theoretical_loss": 3.4792045960329836,
      "tokens_seen": 1863974912
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00043971915747241725,
      "loss": 1.4431,
      "theoretical_loss": 3.479194339759041,
      "tokens_seen": 1864040448
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004396990972918756,
      "loss": 1.4998,
      "theoretical_loss": 3.4791840839466435,
      "tokens_seen": 1864105984
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00043967903711133403,
      "loss": 1.4012,
      "theoretical_loss": 3.479173828595755,
      "tokens_seen": 1864171520
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004396589769307924,
      "loss": 1.4571,
      "theoretical_loss": 3.4791635737063373,
      "tokens_seen": 1864237056
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00043963891675025075,
      "loss": 1.3954,
      "theoretical_loss": 3.479153319278354,
      "tokens_seen": 1864302592
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004396188565697091,
      "loss": 1.4621,
      "theoretical_loss": 3.479143065311768,
      "tokens_seen": 1864368128
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004395987963891675,
      "loss": 1.2811,
      "theoretical_loss": 3.4791328118065428,
      "tokens_seen": 1864433664
    },
    {
      "epoch": 0.56,
      "objective/train/advantage_avg": 0.4944264590740204,
      "objective/train/docs_used": 680280,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1517951488494873,
      "objective/train/original_loss": 1.1517951488494873,
      "objective/train/theoretical_loss": 3.479122558762641,
      "objective/train/tokens_used": 1884959200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24487784504890442,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506876707077026,
      "objective/train/weighted_lm_loss": 1.2100492715835571,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.479122558762641,
      "tokens_seen": 1864499200
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0004395787362086259,
      "loss": 1.3032,
      "theoretical_loss": 3.479122558762641,
      "tokens_seen": 1864499200
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043955867602808426,
      "loss": 1.4356,
      "theoretical_loss": 3.479112306180026,
      "tokens_seen": 1864564736
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004395386158475426,
      "loss": 1.5101,
      "theoretical_loss": 3.47910205405866,
      "tokens_seen": 1864630272
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000439518555667001,
      "loss": 1.2962,
      "theoretical_loss": 3.4790918023985067,
      "tokens_seen": 1864695808
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004394984954864594,
      "loss": 1.3672,
      "theoretical_loss": 3.4790815511995294,
      "tokens_seen": 1864761344
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043947843530591776,
      "loss": 1.3554,
      "theoretical_loss": 3.4790713004616904,
      "tokens_seen": 1864826880
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004394583751253761,
      "loss": 1.3552,
      "theoretical_loss": 3.479061050184953,
      "tokens_seen": 1864892416
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004394383149448345,
      "loss": 1.4299,
      "theoretical_loss": 3.479050800369281,
      "tokens_seen": 1864957952
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043941825476429285,
      "loss": 1.4462,
      "theoretical_loss": 3.479040551014636,
      "tokens_seen": 1865023488
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043939819458375126,
      "loss": 1.4849,
      "theoretical_loss": 3.4790303021209823,
      "tokens_seen": 1865089024
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004393781344032096,
      "loss": 1.3441,
      "theoretical_loss": 3.4790200536882825,
      "tokens_seen": 1865154560
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000439358074222668,
      "loss": 1.4041,
      "theoretical_loss": 3.4790098057164998,
      "tokens_seen": 1865220096
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043933801404212635,
      "loss": 1.3914,
      "theoretical_loss": 3.4789995582055973,
      "tokens_seen": 1865285632
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043931795386158477,
      "loss": 1.4397,
      "theoretical_loss": 3.478989311155538,
      "tokens_seen": 1865351168
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043929789368104313,
      "loss": 1.4597,
      "theoretical_loss": 3.4789790645662846,
      "tokens_seen": 1865416704
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004392778335005015,
      "loss": 1.4834,
      "theoretical_loss": 3.478968818437801,
      "tokens_seen": 1865482240
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043925777331995985,
      "loss": 1.4663,
      "theoretical_loss": 3.4789585727700496,
      "tokens_seen": 1865547776
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004392377131394182,
      "loss": 1.4428,
      "theoretical_loss": 3.478948327562994,
      "tokens_seen": 1865613312
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043921765295887663,
      "loss": 1.4218,
      "theoretical_loss": 3.478938082816597,
      "tokens_seen": 1865678848
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000439197592778335,
      "loss": 1.4502,
      "theoretical_loss": 3.4789278385308218,
      "tokens_seen": 1865744384
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004391775325977934,
      "loss": 1.3524,
      "theoretical_loss": 3.4789175947056314,
      "tokens_seen": 1865809920
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043915747241725177,
      "loss": 1.4482,
      "theoretical_loss": 3.4789073513409887,
      "tokens_seen": 1865875456
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004391374122367102,
      "loss": 1.384,
      "theoretical_loss": 3.4788971084368576,
      "tokens_seen": 1865940992
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043911735205616855,
      "loss": 1.3867,
      "theoretical_loss": 3.478886865993201,
      "tokens_seen": 1866006528
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004390972918756269,
      "loss": 1.5377,
      "theoretical_loss": 3.478876624009981,
      "tokens_seen": 1866072064
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004390772316950853,
      "loss": 1.4743,
      "theoretical_loss": 3.478866382487162,
      "tokens_seen": 1866137600
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043905717151454364,
      "loss": 1.4481,
      "theoretical_loss": 3.478856141424706,
      "tokens_seen": 1866203136
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043903711133400205,
      "loss": 1.4218,
      "theoretical_loss": 3.478845900822577,
      "tokens_seen": 1866268672
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004390170511534604,
      "loss": 1.4588,
      "theoretical_loss": 3.4788356606807387,
      "tokens_seen": 1866334208
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004389969909729188,
      "loss": 1.4841,
      "theoretical_loss": 3.4788254209991525,
      "tokens_seen": 1866399744
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043897693079237714,
      "loss": 1.3953,
      "theoretical_loss": 3.478815181777783,
      "tokens_seen": 1866465280
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043895687061183556,
      "loss": 1.3911,
      "theoretical_loss": 3.478804943016593,
      "tokens_seen": 1866530816
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004389368104312939,
      "loss": 1.3205,
      "theoretical_loss": 3.478794704715545,
      "tokens_seen": 1866596352
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004389167502507523,
      "loss": 1.4108,
      "theoretical_loss": 3.4787844668746035,
      "tokens_seen": 1866661888
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043889669007021064,
      "loss": 1.3921,
      "theoretical_loss": 3.4787742294937303,
      "tokens_seen": 1866727424
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000438876629889669,
      "loss": 1.5271,
      "theoretical_loss": 3.4787639925728895,
      "tokens_seen": 1866792960
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004388565697091274,
      "loss": 1.3392,
      "theoretical_loss": 3.478753756112044,
      "tokens_seen": 1866858496
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004388365095285858,
      "loss": 1.3468,
      "theoretical_loss": 3.4787435201111565,
      "tokens_seen": 1866924032
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043881644934804415,
      "loss": 1.4751,
      "theoretical_loss": 3.4787332845701906,
      "tokens_seen": 1866989568
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004387963891675025,
      "loss": 1.366,
      "theoretical_loss": 3.47872304948911,
      "tokens_seen": 1867055104
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004387763289869609,
      "loss": 1.4606,
      "theoretical_loss": 3.478712814867877,
      "tokens_seen": 1867120640
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004387562688064193,
      "loss": 1.3653,
      "theoretical_loss": 3.4787025807064555,
      "tokens_seen": 1867186176
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043873620862587765,
      "loss": 1.3719,
      "theoretical_loss": 3.4786923470048077,
      "tokens_seen": 1867251712
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000438716148445336,
      "loss": 1.5714,
      "theoretical_loss": 3.4786821137628983,
      "tokens_seen": 1867317248
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043869608826479443,
      "loss": 1.4109,
      "theoretical_loss": 3.4786718809806896,
      "tokens_seen": 1867382784
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004386760280842528,
      "loss": 1.4265,
      "theoretical_loss": 3.4786616486581448,
      "tokens_seen": 1867448320
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043865596790371115,
      "loss": 1.3861,
      "theoretical_loss": 3.478651416795227,
      "tokens_seen": 1867513856
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004386359077231695,
      "loss": 1.3935,
      "theoretical_loss": 3.4786411853918997,
      "tokens_seen": 1867579392
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004386158475426279,
      "loss": 1.4247,
      "theoretical_loss": 3.4786309544481266,
      "tokens_seen": 1867644928
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004385957873620863,
      "loss": 1.3702,
      "theoretical_loss": 3.4786207239638696,
      "tokens_seen": 1867710464
    },
    {
      "epoch": 0.57,
      "objective/train/advantage_avg": 0.4935499131679535,
      "objective/train/docs_used": 681429,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3021905422210693,
      "objective/train/original_loss": 1.3021904230117798,
      "objective/train/theoretical_loss": 3.478610493939094,
      "objective/train/tokens_used": 1888236000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2443898618221283,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505974292755127,
      "objective/train/weighted_lm_loss": 1.3679802417755127,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.478610493939094,
      "tokens_seen": 1867776000
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043857572718154466,
      "loss": 1.3699,
      "theoretical_loss": 3.478610493939094,
      "tokens_seen": 1867776000
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000438555667001003,
      "loss": 1.3726,
      "theoretical_loss": 3.478600264373761,
      "tokens_seen": 1867841536
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004385356068204614,
      "loss": 1.477,
      "theoretical_loss": 3.4785900352678345,
      "tokens_seen": 1867907072
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004385155466399198,
      "loss": 1.3919,
      "theoretical_loss": 3.478579806621278,
      "tokens_seen": 1867972608
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043849548645937816,
      "loss": 1.4453,
      "theoretical_loss": 3.4785695784340556,
      "tokens_seen": 1868038144
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004384754262788365,
      "loss": 1.4241,
      "theoretical_loss": 3.4785593507061288,
      "tokens_seen": 1868103680
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004384553660982949,
      "loss": 1.378,
      "theoretical_loss": 3.4785491234374617,
      "tokens_seen": 1868169216
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043843530591775325,
      "loss": 1.3633,
      "theoretical_loss": 3.478538896628018,
      "tokens_seen": 1868234752
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043841524573721166,
      "loss": 1.3652,
      "theoretical_loss": 3.4785286702777602,
      "tokens_seen": 1868300288
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043839518555667,
      "loss": 1.3377,
      "theoretical_loss": 3.478518444386652,
      "tokens_seen": 1868365824
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004383751253761284,
      "loss": 1.4517,
      "theoretical_loss": 3.478508218954657,
      "tokens_seen": 1868431360
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043835506519558675,
      "loss": 1.3431,
      "theoretical_loss": 3.478497993981737,
      "tokens_seen": 1868496896
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043833500501504517,
      "loss": 1.3591,
      "theoretical_loss": 3.4784877694678573,
      "tokens_seen": 1868562432
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043831494483450353,
      "loss": 1.4268,
      "theoretical_loss": 3.47847754541298,
      "tokens_seen": 1868627968
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004382948846539619,
      "loss": 1.5802,
      "theoretical_loss": 3.4784673218170687,
      "tokens_seen": 1868693504
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043827482447342025,
      "loss": 1.3938,
      "theoretical_loss": 3.478457098680087,
      "tokens_seen": 1868759040
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004382547642928786,
      "loss": 1.366,
      "theoretical_loss": 3.478446876001997,
      "tokens_seen": 1868824576
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043823470411233703,
      "loss": 1.5466,
      "theoretical_loss": 3.4784366537827633,
      "tokens_seen": 1868890112
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004382146439317954,
      "loss": 1.5028,
      "theoretical_loss": 3.4784264320223492,
      "tokens_seen": 1868955648
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043819458375125376,
      "loss": 1.3453,
      "theoretical_loss": 3.4784162107207175,
      "tokens_seen": 1869021184
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004381745235707121,
      "loss": 1.444,
      "theoretical_loss": 3.4784059898778312,
      "tokens_seen": 1869086720
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043815446339017053,
      "loss": 1.4222,
      "theoretical_loss": 3.4783957694936545,
      "tokens_seen": 1869152256
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004381344032096289,
      "loss": 1.5185,
      "theoretical_loss": 3.47838554956815,
      "tokens_seen": 1869217792
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043811434302908726,
      "loss": 1.3959,
      "theoretical_loss": 3.478375330101281,
      "tokens_seen": 1869283328
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004380942828485456,
      "loss": 1.4233,
      "theoretical_loss": 3.4783651110930123,
      "tokens_seen": 1869348864
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000438074222668004,
      "loss": 1.4033,
      "theoretical_loss": 3.478354892543305,
      "tokens_seen": 1869414400
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004380541624874624,
      "loss": 1.4891,
      "theoretical_loss": 3.4783446744521243,
      "tokens_seen": 1869479936
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043803410230692076,
      "loss": 1.4401,
      "theoretical_loss": 3.4783344568194323,
      "tokens_seen": 1869545472
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004380140421263791,
      "loss": 1.453,
      "theoretical_loss": 3.478324239645193,
      "tokens_seen": 1869611008
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004379939819458375,
      "loss": 1.4691,
      "theoretical_loss": 3.4783140229293696,
      "tokens_seen": 1869676544
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004379739217652959,
      "loss": 1.499,
      "theoretical_loss": 3.4783038066719256,
      "tokens_seen": 1869742080
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043795386158475427,
      "loss": 1.4892,
      "theoretical_loss": 3.4782935908728243,
      "tokens_seen": 1869807616
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043793380140421263,
      "loss": 1.4717,
      "theoretical_loss": 3.478283375532029,
      "tokens_seen": 1869873152
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000437913741223671,
      "loss": 1.3958,
      "theoretical_loss": 3.478273160649503,
      "tokens_seen": 1869938688
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043789368104312935,
      "loss": 1.5567,
      "theoretical_loss": 3.47826294622521,
      "tokens_seen": 1870004224
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043787362086258777,
      "loss": 1.4008,
      "theoretical_loss": 3.4782527322591132,
      "tokens_seen": 1870069760
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043785356068204613,
      "loss": 1.4781,
      "theoretical_loss": 3.4782425187511756,
      "tokens_seen": 1870135296
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004378335005015045,
      "loss": 1.4473,
      "theoretical_loss": 3.4782323057013613,
      "tokens_seen": 1870200832
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043781344032096286,
      "loss": 1.4268,
      "theoretical_loss": 3.4782220931096335,
      "tokens_seen": 1870266368
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043779338014042127,
      "loss": 1.3812,
      "theoretical_loss": 3.4782118809759552,
      "tokens_seen": 1870331904
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043777331995987963,
      "loss": 1.4538,
      "theoretical_loss": 3.47820166930029,
      "tokens_seen": 1870397440
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000437753259779338,
      "loss": 1.4889,
      "theoretical_loss": 3.4781914580826014,
      "tokens_seen": 1870462976
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043773319959879636,
      "loss": 1.396,
      "theoretical_loss": 3.478181247322853,
      "tokens_seen": 1870528512
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004377131394182547,
      "loss": 1.3092,
      "theoretical_loss": 3.478171037021008,
      "tokens_seen": 1870594048
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043769307923771314,
      "loss": 1.4147,
      "theoretical_loss": 3.47816082717703,
      "tokens_seen": 1870659584
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043767301905717155,
      "loss": 1.3273,
      "theoretical_loss": 3.478150617790882,
      "tokens_seen": 1870725120
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004376529588766299,
      "loss": 1.3827,
      "theoretical_loss": 3.4781404088625276,
      "tokens_seen": 1870790656
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004376328986960883,
      "loss": 1.4909,
      "theoretical_loss": 3.4781302003919303,
      "tokens_seen": 1870856192
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004376128385155467,
      "loss": 1.3214,
      "theoretical_loss": 3.478119992379054,
      "tokens_seen": 1870921728
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043759277833500506,
      "loss": 1.5715,
      "theoretical_loss": 3.478109784823861,
      "tokens_seen": 1870987264
    },
    {
      "epoch": 0.57,
      "objective/train/advantage_avg": 0.49209684133529663,
      "objective/train/docs_used": 682627,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3585387468338013,
      "objective/train/original_loss": 1.3585386276245117,
      "objective/train/theoretical_loss": 3.4780995777263164,
      "objective/train/tokens_used": 1891512800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2439141720533371,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504496097564697,
      "objective/train/weighted_lm_loss": 1.4270821809768677,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4780995777263164,
      "tokens_seen": 1871052800
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004375727181544634,
      "loss": 1.3885,
      "theoretical_loss": 3.4780995777263164,
      "tokens_seen": 1871052800
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004375526579739218,
      "loss": 1.4459,
      "theoretical_loss": 3.4780893710863823,
      "tokens_seen": 1871118336
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004375325977933802,
      "loss": 1.5713,
      "theoretical_loss": 3.478079164904022,
      "tokens_seen": 1871183872
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043751253761283856,
      "loss": 1.442,
      "theoretical_loss": 3.4780689591792004,
      "tokens_seen": 1871249408
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004374924774322969,
      "loss": 1.4867,
      "theoretical_loss": 3.47805875391188,
      "tokens_seen": 1871314944
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004374724172517553,
      "loss": 1.5314,
      "theoretical_loss": 3.478048549102024,
      "tokens_seen": 1871380480
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043745235707121365,
      "loss": 1.4335,
      "theoretical_loss": 3.4780383447495966,
      "tokens_seen": 1871446016
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043743229689067206,
      "loss": 1.3928,
      "theoretical_loss": 3.478028140854561,
      "tokens_seen": 1871511552
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004374122367101304,
      "loss": 1.389,
      "theoretical_loss": 3.4780179374168805,
      "tokens_seen": 1871577088
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004373921765295888,
      "loss": 1.3308,
      "theoretical_loss": 3.4780077344365186,
      "tokens_seen": 1871642624
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043737211634904715,
      "loss": 1.4386,
      "theoretical_loss": 3.477997531913439,
      "tokens_seen": 1871708160
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043735205616850557,
      "loss": 1.4446,
      "theoretical_loss": 3.477987329847605,
      "tokens_seen": 1871773696
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043733199598796393,
      "loss": 1.4811,
      "theoretical_loss": 3.47797712823898,
      "tokens_seen": 1871839232
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004373119358074223,
      "loss": 1.5373,
      "theoretical_loss": 3.477966927087528,
      "tokens_seen": 1871904768
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043729187562688065,
      "loss": 1.4768,
      "theoretical_loss": 3.477956726393212,
      "tokens_seen": 1871970304
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000437271815446339,
      "loss": 1.3935,
      "theoretical_loss": 3.477946526155996,
      "tokens_seen": 1872035840
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043725175526579743,
      "loss": 1.5294,
      "theoretical_loss": 3.477936326375843,
      "tokens_seen": 1872101376
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004372316950852558,
      "loss": 1.3951,
      "theoretical_loss": 3.477926127052717,
      "tokens_seen": 1872166912
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043721163490471416,
      "loss": 1.4886,
      "theoretical_loss": 3.477915928186581,
      "tokens_seen": 1872232448
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004371915747241725,
      "loss": 1.4647,
      "theoretical_loss": 3.4779057297773983,
      "tokens_seen": 1872297984
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043717151454363093,
      "loss": 1.5075,
      "theoretical_loss": 3.4778955318251334,
      "tokens_seen": 1872363520
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004371514543630893,
      "loss": 1.4248,
      "theoretical_loss": 3.47788533432975,
      "tokens_seen": 1872429056
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043713139418254766,
      "loss": 1.3731,
      "theoretical_loss": 3.4778751372912105,
      "tokens_seen": 1872494592
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000437111334002006,
      "loss": 1.4269,
      "theoretical_loss": 3.4778649407094786,
      "tokens_seen": 1872560128
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004370912738214644,
      "loss": 1.4102,
      "theoretical_loss": 3.4778547445845187,
      "tokens_seen": 1872625664
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004370712136409228,
      "loss": 1.466,
      "theoretical_loss": 3.477844548916294,
      "tokens_seen": 1872691200
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043705115346038116,
      "loss": 1.3726,
      "theoretical_loss": 3.4778343537047673,
      "tokens_seen": 1872756736
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004370310932798395,
      "loss": 1.3961,
      "theoretical_loss": 3.477824158949903,
      "tokens_seen": 1872822272
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004370110330992979,
      "loss": 1.4974,
      "theoretical_loss": 3.4778139646516646,
      "tokens_seen": 1872887808
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004369909729187563,
      "loss": 1.3451,
      "theoretical_loss": 3.4778037708100156,
      "tokens_seen": 1872953344
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043697091273821467,
      "loss": 1.4586,
      "theoretical_loss": 3.4777935774249196,
      "tokens_seen": 1873018880
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043695085255767303,
      "loss": 1.4644,
      "theoretical_loss": 3.4777833844963397,
      "tokens_seen": 1873084416
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004369307923771314,
      "loss": 1.3951,
      "theoretical_loss": 3.47777319202424,
      "tokens_seen": 1873149952
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043691073219658975,
      "loss": 1.5266,
      "theoretical_loss": 3.4777630000085837,
      "tokens_seen": 1873215488
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043689067201604817,
      "loss": 1.4555,
      "theoretical_loss": 3.4777528084493348,
      "tokens_seen": 1873281024
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043687061183550653,
      "loss": 1.4507,
      "theoretical_loss": 3.4777426173464567,
      "tokens_seen": 1873346560
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004368505516549649,
      "loss": 1.3518,
      "theoretical_loss": 3.4777324266999132,
      "tokens_seen": 1873412096
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043683049147442326,
      "loss": 1.4007,
      "theoretical_loss": 3.477722236509668,
      "tokens_seen": 1873477632
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043681043129388167,
      "loss": 1.4565,
      "theoretical_loss": 3.477712046775684,
      "tokens_seen": 1873543168
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043679037111334003,
      "loss": 1.5301,
      "theoretical_loss": 3.4777018574979253,
      "tokens_seen": 1873608704
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004367703109327984,
      "loss": 1.3676,
      "theoretical_loss": 3.4776916686763557,
      "tokens_seen": 1873674240
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043675025075225676,
      "loss": 1.5049,
      "theoretical_loss": 3.477681480310938,
      "tokens_seen": 1873739776
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004367301905717151,
      "loss": 1.3805,
      "theoretical_loss": 3.477671292401637,
      "tokens_seen": 1873805312
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043671013039117354,
      "loss": 1.5346,
      "theoretical_loss": 3.4776611049484156,
      "tokens_seen": 1873870848
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004366900702106319,
      "loss": 1.3519,
      "theoretical_loss": 3.4776509179512374,
      "tokens_seen": 1873936384
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043667001003009026,
      "loss": 1.4303,
      "theoretical_loss": 3.477640731410067,
      "tokens_seen": 1874001920
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004366499498495486,
      "loss": 1.333,
      "theoretical_loss": 3.477630545324866,
      "tokens_seen": 1874067456
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043662988966900704,
      "loss": 1.4554,
      "theoretical_loss": 3.4776203596956003,
      "tokens_seen": 1874132992
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004366098294884654,
      "loss": 1.4042,
      "theoretical_loss": 3.477610174522232,
      "tokens_seen": 1874198528
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043658976930792376,
      "loss": 1.5274,
      "theoretical_loss": 3.4775999898047254,
      "tokens_seen": 1874264064
    },
    {
      "epoch": 0.57,
      "objective/train/advantage_avg": 0.4912225604057312,
      "objective/train/docs_used": 683764,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4714360237121582,
      "objective/train/original_loss": 1.471435785293579,
      "objective/train/theoretical_loss": 3.477589805543044,
      "objective/train/tokens_used": 1894789600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24204428493976593,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050352692604065,
      "objective/train/weighted_lm_loss": 1.545599341392517,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.477589805543044,
      "tokens_seen": 1874329600
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043656970912738213,
      "loss": 1.5089,
      "theoretical_loss": 3.477589805543044,
      "tokens_seen": 1874329600
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004365496489468405,
      "loss": 1.3827,
      "theoretical_loss": 3.477579621737152,
      "tokens_seen": 1874395136
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004365295887662989,
      "loss": 1.5549,
      "theoretical_loss": 3.477569438387012,
      "tokens_seen": 1874460672
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043650952858575727,
      "loss": 1.5388,
      "theoretical_loss": 3.4775592554925887,
      "tokens_seen": 1874526208
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043648946840521563,
      "loss": 1.5098,
      "theoretical_loss": 3.477549073053845,
      "tokens_seen": 1874591744
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000436469408224674,
      "loss": 1.3819,
      "theoretical_loss": 3.477538891070745,
      "tokens_seen": 1874657280
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004364493480441324,
      "loss": 1.5436,
      "theoretical_loss": 3.4775287095432525,
      "tokens_seen": 1874722816
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043642928786359077,
      "loss": 1.4929,
      "theoretical_loss": 3.4775185284713306,
      "tokens_seen": 1874788352
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043640922768304913,
      "loss": 1.409,
      "theoretical_loss": 3.477508347854944,
      "tokens_seen": 1874853888
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004363891675025075,
      "loss": 1.3954,
      "theoretical_loss": 3.4774981676940553,
      "tokens_seen": 1874919424
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004363691073219659,
      "loss": 1.3718,
      "theoretical_loss": 3.4774879879886287,
      "tokens_seen": 1874984960
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004363490471414243,
      "loss": 1.4625,
      "theoretical_loss": 3.477477808738628,
      "tokens_seen": 1875050496
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043632898696088264,
      "loss": 1.4773,
      "theoretical_loss": 3.477467629944017,
      "tokens_seen": 1875116032
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000436308926780341,
      "loss": 1.3452,
      "theoretical_loss": 3.4774574516047587,
      "tokens_seen": 1875181568
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043628886659979936,
      "loss": 1.4904,
      "theoretical_loss": 3.4774472737208173,
      "tokens_seen": 1875247104
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004362688064192578,
      "loss": 1.3179,
      "theoretical_loss": 3.477437096292157,
      "tokens_seen": 1875312640
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043624874623871614,
      "loss": 1.379,
      "theoretical_loss": 3.4774269193187406,
      "tokens_seen": 1875378176
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004362286860581745,
      "loss": 1.5267,
      "theoretical_loss": 3.477416742800533,
      "tokens_seen": 1875443712
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043620862587763286,
      "loss": 1.3837,
      "theoretical_loss": 3.4774065667374963,
      "tokens_seen": 1875509248
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004361885656970913,
      "loss": 1.362,
      "theoretical_loss": 3.477396391129596,
      "tokens_seen": 1875574784
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004361685055165497,
      "loss": 1.4721,
      "theoretical_loss": 3.4773862159767943,
      "tokens_seen": 1875640320
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043614844533600806,
      "loss": 1.4982,
      "theoretical_loss": 3.477376041279056,
      "tokens_seen": 1875705856
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004361283851554664,
      "loss": 1.4682,
      "theoretical_loss": 3.4773658670363448,
      "tokens_seen": 1875771392
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004361083249749248,
      "loss": 1.4286,
      "theoretical_loss": 3.477355693248623,
      "tokens_seen": 1875836928
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004360882647943832,
      "loss": 1.4535,
      "theoretical_loss": 3.4773455199158567,
      "tokens_seen": 1875902464
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043606820461384156,
      "loss": 1.6225,
      "theoretical_loss": 3.477335347038008,
      "tokens_seen": 1875968000
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004360481444332999,
      "loss": 1.5179,
      "theoretical_loss": 3.477325174615041,
      "tokens_seen": 1876033536
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004360280842527583,
      "loss": 1.3951,
      "theoretical_loss": 3.47731500264692,
      "tokens_seen": 1876099072
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004360080240722167,
      "loss": 1.4963,
      "theoretical_loss": 3.4773048311336083,
      "tokens_seen": 1876164608
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043598796389167507,
      "loss": 1.3607,
      "theoretical_loss": 3.47729466007507,
      "tokens_seen": 1876230144
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043596790371113343,
      "loss": 1.4665,
      "theoretical_loss": 3.477284489471268,
      "tokens_seen": 1876295680
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004359478435305918,
      "loss": 1.4345,
      "theoretical_loss": 3.477274319322167,
      "tokens_seen": 1876361216
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043592778335005015,
      "loss": 1.4469,
      "theoretical_loss": 3.4772641496277306,
      "tokens_seen": 1876426752
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043590772316950857,
      "loss": 1.4007,
      "theoretical_loss": 3.477253980387922,
      "tokens_seen": 1876492288
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043588766298896693,
      "loss": 1.3963,
      "theoretical_loss": 3.4772438116027065,
      "tokens_seen": 1876557824
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004358676028084253,
      "loss": 1.4212,
      "theoretical_loss": 3.4772336432720463,
      "tokens_seen": 1876623360
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043584754262788366,
      "loss": 1.4086,
      "theoretical_loss": 3.4772234753959057,
      "tokens_seen": 1876688896
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043582748244734207,
      "loss": 1.4385,
      "theoretical_loss": 3.4772133079742487,
      "tokens_seen": 1876754432
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043580742226680043,
      "loss": 1.412,
      "theoretical_loss": 3.477203141007039,
      "tokens_seen": 1876819968
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004357873620862588,
      "loss": 1.504,
      "theoretical_loss": 3.4771929744942414,
      "tokens_seen": 1876885504
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043576730190571716,
      "loss": 1.3988,
      "theoretical_loss": 3.477182808435818,
      "tokens_seen": 1876951040
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004357472417251755,
      "loss": 1.3663,
      "theoretical_loss": 3.477172642831733,
      "tokens_seen": 1877016576
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043572718154463394,
      "loss": 1.4551,
      "theoretical_loss": 3.4771624776819516,
      "tokens_seen": 1877082112
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004357071213640923,
      "loss": 1.3625,
      "theoretical_loss": 3.477152312986436,
      "tokens_seen": 1877147648
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043568706118355066,
      "loss": 1.4617,
      "theoretical_loss": 3.477142148745151,
      "tokens_seen": 1877213184
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000435667001003009,
      "loss": 1.5037,
      "theoretical_loss": 3.47713198495806,
      "tokens_seen": 1877278720
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043564694082246744,
      "loss": 1.3687,
      "theoretical_loss": 3.4771218216251274,
      "tokens_seen": 1877344256
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004356268806419258,
      "loss": 1.4578,
      "theoretical_loss": 3.4771116587463164,
      "tokens_seen": 1877409792
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043560682046138417,
      "loss": 1.4916,
      "theoretical_loss": 3.477101496321591,
      "tokens_seen": 1877475328
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043558676028084253,
      "loss": 1.3766,
      "theoretical_loss": 3.4770913343509156,
      "tokens_seen": 1877540864
    },
    {
      "epoch": 0.57,
      "objective/train/advantage_avg": 0.4919297695159912,
      "objective/train/docs_used": 684931,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.384324073791504,
      "objective/train/original_loss": 1.384324073791504,
      "objective/train/theoretical_loss": 3.4770811728342537,
      "objective/train/tokens_used": 1898066400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2424512654542923,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504255294799805,
      "objective/train/weighted_lm_loss": 1.454155445098877,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4770811728342537,
      "tokens_seen": 1877606400
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043558676028084253,
      "loss": 1.7065,
      "theoretical_loss": 3.4770811728342537,
      "tokens_seen": 1877606400
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004355667001003009,
      "loss": 1.3908,
      "theoretical_loss": 3.477071011771569,
      "tokens_seen": 1877671936
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004355466399197593,
      "loss": 1.3859,
      "theoretical_loss": 3.4770608511628254,
      "tokens_seen": 1877737472
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043552657973921767,
      "loss": 1.5351,
      "theoretical_loss": 3.477050691007987,
      "tokens_seen": 1877803008
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043550651955867603,
      "loss": 1.5536,
      "theoretical_loss": 3.4770405313070176,
      "tokens_seen": 1877868544
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004354864593781344,
      "loss": 1.3888,
      "theoretical_loss": 3.4770303720598807,
      "tokens_seen": 1877934080
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004354663991975928,
      "loss": 1.5566,
      "theoretical_loss": 3.477020213266541,
      "tokens_seen": 1877999616
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043544633901705117,
      "loss": 1.5541,
      "theoretical_loss": 3.4770100549269616,
      "tokens_seen": 1878065152
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043542627883650953,
      "loss": 1.4432,
      "theoretical_loss": 3.4769998970411073,
      "tokens_seen": 1878130688
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004354062186559679,
      "loss": 1.4714,
      "theoretical_loss": 3.476989739608941,
      "tokens_seen": 1878196224
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004353861584754263,
      "loss": 1.4456,
      "theoretical_loss": 3.476979582630427,
      "tokens_seen": 1878261760
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004353660982948847,
      "loss": 1.4212,
      "theoretical_loss": 3.476969426105529,
      "tokens_seen": 1878327296
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043534603811434304,
      "loss": 1.4736,
      "theoretical_loss": 3.476959270034212,
      "tokens_seen": 1878392832
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004353259779338014,
      "loss": 1.4348,
      "theoretical_loss": 3.476949114416439,
      "tokens_seen": 1878458368
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043530591775325976,
      "loss": 1.4793,
      "theoretical_loss": 3.4769389592521733,
      "tokens_seen": 1878523904
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004352858575727182,
      "loss": 1.5465,
      "theoretical_loss": 3.47692880454138,
      "tokens_seen": 1878589440
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043526579739217654,
      "loss": 1.4615,
      "theoretical_loss": 3.4769186502840226,
      "tokens_seen": 1878654976
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004352457372116349,
      "loss": 1.5018,
      "theoretical_loss": 3.476908496480065,
      "tokens_seen": 1878720512
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043522567703109326,
      "loss": 1.4764,
      "theoretical_loss": 3.476898343129471,
      "tokens_seen": 1878786048
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004352056168505517,
      "loss": 1.4082,
      "theoretical_loss": 3.476888190232205,
      "tokens_seen": 1878851584
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043518555667001004,
      "loss": 1.3649,
      "theoretical_loss": 3.4768780377882305,
      "tokens_seen": 1878917120
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004351654964894684,
      "loss": 1.3775,
      "theoretical_loss": 3.4768678857975113,
      "tokens_seen": 1878982656
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043514543630892677,
      "loss": 1.3141,
      "theoretical_loss": 3.476857734260012,
      "tokens_seen": 1879048192
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043512537612838513,
      "loss": 1.4935,
      "theoretical_loss": 3.4768475831756964,
      "tokens_seen": 1879113728
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043510531594784355,
      "loss": 1.4015,
      "theoretical_loss": 3.476837432544528,
      "tokens_seen": 1879179264
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004350852557673019,
      "loss": 1.4678,
      "theoretical_loss": 3.4768272823664708,
      "tokens_seen": 1879244800
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043506519558676027,
      "loss": 1.3687,
      "theoretical_loss": 3.4768171326414894,
      "tokens_seen": 1879310336
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043504513540621863,
      "loss": 1.4062,
      "theoretical_loss": 3.4768069833695474,
      "tokens_seen": 1879375872
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043502507522567705,
      "loss": 1.4084,
      "theoretical_loss": 3.4767968345506093,
      "tokens_seen": 1879441408
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004350050150451354,
      "loss": 1.3907,
      "theoretical_loss": 3.4767866861846377,
      "tokens_seen": 1879506944
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004349849548645938,
      "loss": 1.3997,
      "theoretical_loss": 3.4767765382715976,
      "tokens_seen": 1879572480
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043496489468405214,
      "loss": 1.4052,
      "theoretical_loss": 3.476766390811453,
      "tokens_seen": 1879638016
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004349448345035105,
      "loss": 1.4838,
      "theoretical_loss": 3.476756243804168,
      "tokens_seen": 1879703552
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004349247743229689,
      "loss": 1.4224,
      "theoretical_loss": 3.4767460972497064,
      "tokens_seen": 1879769088
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004349047141424273,
      "loss": 1.4522,
      "theoretical_loss": 3.4767359511480316,
      "tokens_seen": 1879834624
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043488465396188564,
      "loss": 1.4197,
      "theoretical_loss": 3.4767258054991084,
      "tokens_seen": 1879900160
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000434864593781344,
      "loss": 1.4591,
      "theoretical_loss": 3.476715660302901,
      "tokens_seen": 1879965696
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004348445336008024,
      "loss": 1.436,
      "theoretical_loss": 3.476705515559372,
      "tokens_seen": 1880031232
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004348244734202608,
      "loss": 1.4032,
      "theoretical_loss": 3.4766953712684874,
      "tokens_seen": 1880096768
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043480441323971914,
      "loss": 1.4612,
      "theoretical_loss": 3.47668522743021,
      "tokens_seen": 1880162304
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004347843530591775,
      "loss": 1.3688,
      "theoretical_loss": 3.476675084044504,
      "tokens_seen": 1880227840
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043476429287863587,
      "loss": 1.4341,
      "theoretical_loss": 3.4766649411113337,
      "tokens_seen": 1880293376
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004347442326980943,
      "loss": 1.3391,
      "theoretical_loss": 3.476654798630663,
      "tokens_seen": 1880358912
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043472417251755265,
      "loss": 1.3613,
      "theoretical_loss": 3.4766446566024554,
      "tokens_seen": 1880424448
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000434704112337011,
      "loss": 1.3953,
      "theoretical_loss": 3.4766345150266758,
      "tokens_seen": 1880489984
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004346840521564694,
      "loss": 1.5031,
      "theoretical_loss": 3.476624373903288,
      "tokens_seen": 1880555520
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043466399197592784,
      "loss": 1.3813,
      "theoretical_loss": 3.476614233232256,
      "tokens_seen": 1880621056
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004346439317953862,
      "loss": 1.4419,
      "theoretical_loss": 3.4766040930135436,
      "tokens_seen": 1880686592
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043462387161484457,
      "loss": 1.3887,
      "theoretical_loss": 3.476593953247115,
      "tokens_seen": 1880752128
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043460381143430293,
      "loss": 1.3289,
      "theoretical_loss": 3.476583813932934,
      "tokens_seen": 1880817664
    },
    {
      "epoch": 0.57,
      "objective/train/advantage_avg": 0.4939194619655609,
      "objective/train/docs_used": 686045,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2888964414596558,
      "objective/train/original_loss": 1.2888963222503662,
      "objective/train/theoretical_loss": 3.476573675070966,
      "objective/train/tokens_used": 1901343200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2441953718662262,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506335496902466,
      "objective/train/weighted_lm_loss": 1.3540970087051392,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.476573675070966,
      "tokens_seen": 1880883200
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004345837512537613,
      "loss": 1.4233,
      "theoretical_loss": 3.476573675070966,
      "tokens_seen": 1880883200
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004345636910732197,
      "loss": 1.5089,
      "theoretical_loss": 3.476563536661174,
      "tokens_seen": 1880948736
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043454363089267807,
      "loss": 1.4017,
      "theoretical_loss": 3.4765533987035213,
      "tokens_seen": 1881014272
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043452357071213643,
      "loss": 1.5543,
      "theoretical_loss": 3.476543261197973,
      "tokens_seen": 1881079808
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004345035105315948,
      "loss": 1.3535,
      "theoretical_loss": 3.4765331241444937,
      "tokens_seen": 1881145344
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004344834503510532,
      "loss": 1.4525,
      "theoretical_loss": 3.4765229875430466,
      "tokens_seen": 1881210880
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043446339017051157,
      "loss": 1.423,
      "theoretical_loss": 3.476512851393596,
      "tokens_seen": 1881276416
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043444332998996993,
      "loss": 1.2929,
      "theoretical_loss": 3.476502715696106,
      "tokens_seen": 1881341952
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004344232698094283,
      "loss": 1.3372,
      "theoretical_loss": 3.4764925804505404,
      "tokens_seen": 1881407488
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004344032096288867,
      "loss": 1.3151,
      "theoretical_loss": 3.476482445656864,
      "tokens_seen": 1881473024
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004343831494483451,
      "loss": 1.4166,
      "theoretical_loss": 3.4764723113150406,
      "tokens_seen": 1881538560
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043436308926780344,
      "loss": 1.4237,
      "theoretical_loss": 3.4764621774250344,
      "tokens_seen": 1881604096
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004343430290872618,
      "loss": 1.382,
      "theoretical_loss": 3.476452043986809,
      "tokens_seen": 1881669632
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043432296890672016,
      "loss": 1.3931,
      "theoretical_loss": 3.4764419110003293,
      "tokens_seen": 1881735168
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004343029087261786,
      "loss": 1.3698,
      "theoretical_loss": 3.476431778465559,
      "tokens_seen": 1881800704
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043428284854563694,
      "loss": 1.3714,
      "theoretical_loss": 3.476421646382462,
      "tokens_seen": 1881866240
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004342627883650953,
      "loss": 1.418,
      "theoretical_loss": 3.4764115147510033,
      "tokens_seen": 1881931776
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043424272818455367,
      "loss": 1.5748,
      "theoretical_loss": 3.4764013835711456,
      "tokens_seen": 1881997312
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004342226680040121,
      "loss": 1.3771,
      "theoretical_loss": 3.4763912528428547,
      "tokens_seen": 1882062848
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043420260782347044,
      "loss": 1.5418,
      "theoretical_loss": 3.4763811225660937,
      "tokens_seen": 1882128384
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004341825476429288,
      "loss": 1.3445,
      "theoretical_loss": 3.4763709927408266,
      "tokens_seen": 1882193920
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043416248746238717,
      "loss": 1.3813,
      "theoretical_loss": 3.4763608633670184,
      "tokens_seen": 1882259456
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043414242728184553,
      "loss": 1.3853,
      "theoretical_loss": 3.476350734444633,
      "tokens_seen": 1882324992
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043412236710130395,
      "loss": 1.3771,
      "theoretical_loss": 3.4763406059736344,
      "tokens_seen": 1882390528
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004341023069207623,
      "loss": 1.468,
      "theoretical_loss": 3.4763304779539865,
      "tokens_seen": 1882456064
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043408224674022067,
      "loss": 1.4965,
      "theoretical_loss": 3.4763203503856537,
      "tokens_seen": 1882521600
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043406218655967903,
      "loss": 1.4043,
      "theoretical_loss": 3.4763102232686003,
      "tokens_seen": 1882587136
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043404212637913745,
      "loss": 1.4717,
      "theoretical_loss": 3.4763000966027904,
      "tokens_seen": 1882652672
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004340220661985958,
      "loss": 1.5218,
      "theoretical_loss": 3.476289970388188,
      "tokens_seen": 1882718208
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004340020060180542,
      "loss": 1.5754,
      "theoretical_loss": 3.476279844624758,
      "tokens_seen": 1882783744
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043398194583751254,
      "loss": 1.4628,
      "theoretical_loss": 3.476269719312464,
      "tokens_seen": 1882849280
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004339618856569709,
      "loss": 1.354,
      "theoretical_loss": 3.47625959445127,
      "tokens_seen": 1882914816
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004339418254764293,
      "loss": 1.3517,
      "theoretical_loss": 3.4762494700411404,
      "tokens_seen": 1882980352
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004339217652958877,
      "loss": 1.447,
      "theoretical_loss": 3.47623934608204,
      "tokens_seen": 1883045888
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043390170511534604,
      "loss": 1.4466,
      "theoretical_loss": 3.4762292225739317,
      "tokens_seen": 1883111424
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004338816449348044,
      "loss": 1.2907,
      "theoretical_loss": 3.4762190995167814,
      "tokens_seen": 1883176960
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004338615847542628,
      "loss": 1.4407,
      "theoretical_loss": 3.476208976910552,
      "tokens_seen": 1883242496
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004338415245737212,
      "loss": 1.3077,
      "theoretical_loss": 3.476198854755208,
      "tokens_seen": 1883308032
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043382146439317954,
      "loss": 1.4687,
      "theoretical_loss": 3.476188733050714,
      "tokens_seen": 1883373568
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004338014042126379,
      "loss": 1.5632,
      "theoretical_loss": 3.476178611797034,
      "tokens_seen": 1883439104
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043378134403209627,
      "loss": 1.3729,
      "theoretical_loss": 3.476168490994132,
      "tokens_seen": 1883504640
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004337612838515547,
      "loss": 1.4104,
      "theoretical_loss": 3.476158370641973,
      "tokens_seen": 1883570176
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043374122367101305,
      "loss": 1.4118,
      "theoretical_loss": 3.47614825074052,
      "tokens_seen": 1883635712
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004337211634904714,
      "loss": 1.4484,
      "theoretical_loss": 3.476138131289739,
      "tokens_seen": 1883701248
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043370110330992977,
      "loss": 1.4509,
      "theoretical_loss": 3.4761280122895926,
      "tokens_seen": 1883766784
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004336810431293882,
      "loss": 1.4473,
      "theoretical_loss": 3.476117893740046,
      "tokens_seen": 1883832320
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043366098294884655,
      "loss": 1.4753,
      "theoretical_loss": 3.4761077756410628,
      "tokens_seen": 1883897856
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004336409227683049,
      "loss": 1.428,
      "theoretical_loss": 3.4760976579926077,
      "tokens_seen": 1883963392
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004336208625877633,
      "loss": 1.4985,
      "theoretical_loss": 3.476087540794645,
      "tokens_seen": 1884028928
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043360080240722164,
      "loss": 1.4223,
      "theoretical_loss": 3.476077424047139,
      "tokens_seen": 1884094464
    },
    {
      "epoch": 0.57,
      "objective/train/advantage_avg": 0.4946485757827759,
      "objective/train/docs_used": 687181,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3779020309448242,
      "objective/train/original_loss": 1.3779020309448242,
      "objective/train/theoretical_loss": 3.4760673077500535,
      "objective/train/tokens_used": 1904620000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24493667483329773,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507099628448486,
      "objective/train/weighted_lm_loss": 1.4477553367614746,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4760673077500535,
      "tokens_seen": 1884160000
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043358074222668005,
      "loss": 1.5151,
      "theoretical_loss": 3.4760673077500535,
      "tokens_seen": 1884160000
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004335606820461384,
      "loss": 1.3914,
      "theoretical_loss": 3.476057191903353,
      "tokens_seen": 1884225536
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004335406218655968,
      "loss": 1.5354,
      "theoretical_loss": 3.476047076507002,
      "tokens_seen": 1884291072
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043352056168505514,
      "loss": 1.4736,
      "theoretical_loss": 3.4760369615609648,
      "tokens_seen": 1884356608
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043350050150451356,
      "loss": 1.412,
      "theoretical_loss": 3.4760268470652056,
      "tokens_seen": 1884422144
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004334804413239719,
      "loss": 1.4332,
      "theoretical_loss": 3.476016733019689,
      "tokens_seen": 1884487680
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004334603811434303,
      "loss": 1.361,
      "theoretical_loss": 3.4760066194243784,
      "tokens_seen": 1884553216
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043344032096288864,
      "loss": 1.494,
      "theoretical_loss": 3.4759965062792393,
      "tokens_seen": 1884618752
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000433420260782347,
      "loss": 1.5173,
      "theoretical_loss": 3.475986393584235,
      "tokens_seen": 1884684288
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004334002006018054,
      "loss": 1.4705,
      "theoretical_loss": 3.47597628133933,
      "tokens_seen": 1884749824
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004333801404212638,
      "loss": 1.5485,
      "theoretical_loss": 3.4759661695444892,
      "tokens_seen": 1884815360
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043336008024072215,
      "loss": 1.5069,
      "theoretical_loss": 3.4759560581996762,
      "tokens_seen": 1884880896
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004333400200601805,
      "loss": 1.3702,
      "theoretical_loss": 3.475945947304856,
      "tokens_seen": 1884946432
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004333199598796389,
      "loss": 1.4284,
      "theoretical_loss": 3.4759358368599926,
      "tokens_seen": 1885011968
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004332998996990973,
      "loss": 1.5801,
      "theoretical_loss": 3.47592572686505,
      "tokens_seen": 1885077504
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043327983951855565,
      "loss": 1.5093,
      "theoretical_loss": 3.475915617319993,
      "tokens_seen": 1885143040
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000433259779338014,
      "loss": 1.4876,
      "theoretical_loss": 3.475905508224786,
      "tokens_seen": 1885208576
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043323971915747243,
      "loss": 1.3971,
      "theoretical_loss": 3.475895399579393,
      "tokens_seen": 1885274112
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004332196589769308,
      "loss": 1.4828,
      "theoretical_loss": 3.4758852913837783,
      "tokens_seen": 1885339648
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043319959879638915,
      "loss": 1.3452,
      "theoretical_loss": 3.475875183637907,
      "tokens_seen": 1885405184
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043317953861584757,
      "loss": 1.3473,
      "theoretical_loss": 3.4758650763417425,
      "tokens_seen": 1885470720
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043315947843530593,
      "loss": 1.4111,
      "theoretical_loss": 3.4758549694952494,
      "tokens_seen": 1885536256
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043313941825476435,
      "loss": 1.4575,
      "theoretical_loss": 3.475844863098393,
      "tokens_seen": 1885601792
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004331193580742227,
      "loss": 1.4094,
      "theoretical_loss": 3.4758347571511363,
      "tokens_seen": 1885667328
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043309929789368107,
      "loss": 1.3924,
      "theoretical_loss": 3.4758246516534443,
      "tokens_seen": 1885732864
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043307923771313943,
      "loss": 1.469,
      "theoretical_loss": 3.4758145466052817,
      "tokens_seen": 1885798400
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043305917753259785,
      "loss": 1.3873,
      "theoretical_loss": 3.475804442006612,
      "tokens_seen": 1885863936
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004330391173520562,
      "loss": 1.5078,
      "theoretical_loss": 3.4757943378574008,
      "tokens_seen": 1885929472
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004330190571715146,
      "loss": 1.4953,
      "theoretical_loss": 3.4757842341576115,
      "tokens_seen": 1885995008
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043299899699097294,
      "loss": 1.4629,
      "theoretical_loss": 3.475774130907209,
      "tokens_seen": 1886060544
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004329789368104313,
      "loss": 1.4875,
      "theoretical_loss": 3.4757640281061573,
      "tokens_seen": 1886126080
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004329588766298897,
      "loss": 1.4905,
      "theoretical_loss": 3.4757539257544208,
      "tokens_seen": 1886191616
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004329388164493481,
      "loss": 1.4511,
      "theoretical_loss": 3.4757438238519645,
      "tokens_seen": 1886257152
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043291875626880644,
      "loss": 1.4004,
      "theoretical_loss": 3.4757337223987523,
      "tokens_seen": 1886322688
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004328986960882648,
      "loss": 1.4018,
      "theoretical_loss": 3.4757236213947484,
      "tokens_seen": 1886388224
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004328786359077232,
      "loss": 1.4005,
      "theoretical_loss": 3.475713520839918,
      "tokens_seen": 1886453760
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004328585757271816,
      "loss": 1.3993,
      "theoretical_loss": 3.4757034207342254,
      "tokens_seen": 1886519296
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043283851554663994,
      "loss": 1.509,
      "theoretical_loss": 3.4756933210776344,
      "tokens_seen": 1886584832
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004328184553660983,
      "loss": 1.3853,
      "theoretical_loss": 3.4756832218701095,
      "tokens_seen": 1886650368
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043279839518555667,
      "loss": 1.4423,
      "theoretical_loss": 3.4756731231116156,
      "tokens_seen": 1886715904
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004327783350050151,
      "loss": 1.475,
      "theoretical_loss": 3.4756630248021168,
      "tokens_seen": 1886781440
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043275827482447345,
      "loss": 1.3473,
      "theoretical_loss": 3.475652926941578,
      "tokens_seen": 1886846976
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004327382146439318,
      "loss": 1.3907,
      "theoretical_loss": 3.475642829529963,
      "tokens_seen": 1886912512
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043271815446339017,
      "loss": 1.3843,
      "theoretical_loss": 3.4756327325672367,
      "tokens_seen": 1886978048
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004326980942828486,
      "loss": 1.4285,
      "theoretical_loss": 3.4756226360533633,
      "tokens_seen": 1887043584
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043267803410230695,
      "loss": 1.3906,
      "theoretical_loss": 3.4756125399883073,
      "tokens_seen": 1887109120
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004326579739217653,
      "loss": 1.4769,
      "theoretical_loss": 3.4756024443720337,
      "tokens_seen": 1887174656
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004326379137412237,
      "loss": 1.3242,
      "theoretical_loss": 3.475592349204506,
      "tokens_seen": 1887240192
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043261785356068204,
      "loss": 1.4067,
      "theoretical_loss": 3.4755822544856896,
      "tokens_seen": 1887305728
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043259779338014045,
      "loss": 1.5399,
      "theoretical_loss": 3.475572160215548,
      "tokens_seen": 1887371264
    },
    {
      "epoch": 0.57,
      "objective/train/advantage_avg": 0.4945085346698761,
      "objective/train/docs_used": 688313,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5170170068740845,
      "objective/train/original_loss": 1.517016887664795,
      "objective/train/theoretical_loss": 3.4755620663940467,
      "objective/train/tokens_used": 1907896800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24520368874073029,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506974458694458,
      "objective/train/weighted_lm_loss": 1.5938998460769653,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4755620663940467,
      "tokens_seen": 1887436800
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004325777331995988,
      "loss": 1.4292,
      "theoretical_loss": 3.4755620663940467,
      "tokens_seen": 1887436800
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004325576730190572,
      "loss": 1.5261,
      "theoretical_loss": 3.47555197302115,
      "tokens_seen": 1887502336
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043253761283851554,
      "loss": 1.4669,
      "theoretical_loss": 3.475541880096822,
      "tokens_seen": 1887567872
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043251755265797396,
      "loss": 1.406,
      "theoretical_loss": 3.4755317876210263,
      "tokens_seen": 1887633408
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004324974924774323,
      "loss": 1.4889,
      "theoretical_loss": 3.4755216955937294,
      "tokens_seen": 1887698944
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004324774322968907,
      "loss": 1.3977,
      "theoretical_loss": 3.4755116040148946,
      "tokens_seen": 1887764480
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043245737211634904,
      "loss": 1.3814,
      "theoretical_loss": 3.4755015128844864,
      "tokens_seen": 1887830016
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004324373119358074,
      "loss": 1.4692,
      "theoretical_loss": 3.4754914222024698,
      "tokens_seen": 1887895552
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004324172517552658,
      "loss": 1.3085,
      "theoretical_loss": 3.4754813319688087,
      "tokens_seen": 1887961088
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004323971915747242,
      "loss": 1.3709,
      "theoretical_loss": 3.4754712421834677,
      "tokens_seen": 1888026624
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043237713139418255,
      "loss": 1.3753,
      "theoretical_loss": 3.475461152846412,
      "tokens_seen": 1888092160
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004323570712136409,
      "loss": 1.4128,
      "theoretical_loss": 3.475451063957606,
      "tokens_seen": 1888157696
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004323370110330993,
      "loss": 1.4852,
      "theoretical_loss": 3.4754409755170133,
      "tokens_seen": 1888223232
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004323169508525577,
      "loss": 1.3538,
      "theoretical_loss": 3.475430887524599,
      "tokens_seen": 1888288768
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043229689067201605,
      "loss": 1.4306,
      "theoretical_loss": 3.475420799980328,
      "tokens_seen": 1888354304
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004322768304914744,
      "loss": 1.4075,
      "theoretical_loss": 3.4754107128841647,
      "tokens_seen": 1888419840
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043225677031093283,
      "loss": 1.4537,
      "theoretical_loss": 3.4754006262360733,
      "tokens_seen": 1888485376
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004322367101303912,
      "loss": 1.2544,
      "theoretical_loss": 3.4753905400360185,
      "tokens_seen": 1888550912
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043221664994984955,
      "loss": 1.3702,
      "theoretical_loss": 3.4753804542839646,
      "tokens_seen": 1888616448
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004321965897693079,
      "loss": 1.2866,
      "theoretical_loss": 3.4753703689798763,
      "tokens_seen": 1888681984
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004321765295887663,
      "loss": 1.3337,
      "theoretical_loss": 3.4753602841237186,
      "tokens_seen": 1888747520
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004321564694082247,
      "loss": 1.3989,
      "theoretical_loss": 3.4753501997154554,
      "tokens_seen": 1888813056
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043213640922768306,
      "loss": 1.3254,
      "theoretical_loss": 3.475340115755052,
      "tokens_seen": 1888878592
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004321163490471414,
      "loss": 1.4377,
      "theoretical_loss": 3.4753300322424727,
      "tokens_seen": 1888944128
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004320962888665998,
      "loss": 1.3877,
      "theoretical_loss": 3.4753199491776816,
      "tokens_seen": 1889009664
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004320762286860582,
      "loss": 1.3286,
      "theoretical_loss": 3.475309866560644,
      "tokens_seen": 1889075200
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043205616850551656,
      "loss": 1.3998,
      "theoretical_loss": 3.4752997843913236,
      "tokens_seen": 1889140736
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004320361083249749,
      "loss": 1.4527,
      "theoretical_loss": 3.475289702669686,
      "tokens_seen": 1889206272
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004320160481444333,
      "loss": 1.3827,
      "theoretical_loss": 3.475279621395695,
      "tokens_seen": 1889271808
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043199598796389165,
      "loss": 1.5137,
      "theoretical_loss": 3.4752695405693155,
      "tokens_seen": 1889337344
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043197592778335006,
      "loss": 1.4394,
      "theoretical_loss": 3.475259460190512,
      "tokens_seen": 1889402880
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004319558676028084,
      "loss": 1.3021,
      "theoretical_loss": 3.4752493802592497,
      "tokens_seen": 1889468416
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004319358074222668,
      "loss": 1.3865,
      "theoretical_loss": 3.4752393007754923,
      "tokens_seen": 1889533952
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043191574724172515,
      "loss": 1.4765,
      "theoretical_loss": 3.475229221739205,
      "tokens_seen": 1889599488
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043189568706118357,
      "loss": 1.4165,
      "theoretical_loss": 3.475219143150352,
      "tokens_seen": 1889665024
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043187562688064193,
      "loss": 1.435,
      "theoretical_loss": 3.4752090650088983,
      "tokens_seen": 1889730560
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004318555667001003,
      "loss": 1.4029,
      "theoretical_loss": 3.4751989873148084,
      "tokens_seen": 1889796096
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043183550651955865,
      "loss": 1.498,
      "theoretical_loss": 3.475188910068047,
      "tokens_seen": 1889861632
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000431815446339017,
      "loss": 1.4092,
      "theoretical_loss": 3.4751788332685782,
      "tokens_seen": 1889927168
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043179538615847543,
      "loss": 1.4878,
      "theoretical_loss": 3.475168756916368,
      "tokens_seen": 1889992704
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004317753259779338,
      "loss": 1.5218,
      "theoretical_loss": 3.4751586810113793,
      "tokens_seen": 1890058240
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043175526579739215,
      "loss": 1.2746,
      "theoretical_loss": 3.475148605553578,
      "tokens_seen": 1890123776
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004317352056168505,
      "loss": 1.3765,
      "theoretical_loss": 3.475138530542928,
      "tokens_seen": 1890189312
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043171514543630893,
      "loss": 1.448,
      "theoretical_loss": 3.4751284559793945,
      "tokens_seen": 1890254848
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004316950852557673,
      "loss": 1.4184,
      "theoretical_loss": 3.4751183818629414,
      "tokens_seen": 1890320384
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004316750250752257,
      "loss": 1.3502,
      "theoretical_loss": 3.4751083081935343,
      "tokens_seen": 1890385920
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004316549648946841,
      "loss": 1.4553,
      "theoretical_loss": 3.4750982349711377,
      "tokens_seen": 1890451456
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043163490471414244,
      "loss": 1.3263,
      "theoretical_loss": 3.475088162195716,
      "tokens_seen": 1890516992
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043161484453360085,
      "loss": 1.5692,
      "theoretical_loss": 3.4750780898672335,
      "tokens_seen": 1890582528
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004315947843530592,
      "loss": 1.3745,
      "theoretical_loss": 3.4750680179856555,
      "tokens_seen": 1890648064
    },
    {
      "epoch": 0.57,
      "objective/train/advantage_avg": 0.4934093952178955,
      "objective/train/docs_used": 689504,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3186650276184082,
      "objective/train/original_loss": 1.3186650276184082,
      "objective/train/theoretical_loss": 3.4750579465509466,
      "objective/train/tokens_used": 1911173600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24384815990924835,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050580620765686,
      "objective/train/weighted_lm_loss": 1.385423183441162,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4750579465509466,
      "tokens_seen": 1890713600
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004315747241725176,
      "loss": 1.3588,
      "theoretical_loss": 3.4750579465509466,
      "tokens_seen": 1890713600
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043155466399197594,
      "loss": 1.3438,
      "theoretical_loss": 3.475047875563071,
      "tokens_seen": 1890779136
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043153460381143436,
      "loss": 1.3197,
      "theoretical_loss": 3.475037805021994,
      "tokens_seen": 1890844672
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004315145436308927,
      "loss": 1.36,
      "theoretical_loss": 3.4750277349276795,
      "tokens_seen": 1890910208
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004314944834503511,
      "loss": 1.4822,
      "theoretical_loss": 3.4750176652800935,
      "tokens_seen": 1890975744
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043147442326980944,
      "loss": 1.3433,
      "theoretical_loss": 3.4750075960791995,
      "tokens_seen": 1891041280
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004314543630892678,
      "loss": 1.3781,
      "theoretical_loss": 3.4749975273249625,
      "tokens_seen": 1891106816
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004314343029087262,
      "loss": 1.3978,
      "theoretical_loss": 3.4749874590173473,
      "tokens_seen": 1891172352
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004314142427281846,
      "loss": 1.5054,
      "theoretical_loss": 3.474977391156319,
      "tokens_seen": 1891237888
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043139418254764295,
      "loss": 1.4476,
      "theoretical_loss": 3.474967323741842,
      "tokens_seen": 1891303424
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004313741223671013,
      "loss": 1.4053,
      "theoretical_loss": 3.4749572567738807,
      "tokens_seen": 1891368960
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004313540621865597,
      "loss": 1.3811,
      "theoretical_loss": 3.4749471902524003,
      "tokens_seen": 1891434496
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004313340020060181,
      "loss": 1.3914,
      "theoretical_loss": 3.474937124177365,
      "tokens_seen": 1891500032
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043131394182547645,
      "loss": 1.4675,
      "theoretical_loss": 3.47492705854874,
      "tokens_seen": 1891565568
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004312938816449348,
      "loss": 1.4798,
      "theoretical_loss": 3.4749169933664903,
      "tokens_seen": 1891631104
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043127382146439323,
      "loss": 1.3581,
      "theoretical_loss": 3.4749069286305803,
      "tokens_seen": 1891696640
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004312537612838516,
      "loss": 1.4019,
      "theoretical_loss": 3.474896864340974,
      "tokens_seen": 1891762176
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043123370110330995,
      "loss": 1.3767,
      "theoretical_loss": 3.4748868004976377,
      "tokens_seen": 1891827712
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004312136409227683,
      "loss": 1.4084,
      "theoretical_loss": 3.4748767371005345,
      "tokens_seen": 1891893248
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004311935807422267,
      "loss": 1.4562,
      "theoretical_loss": 3.47486667414963,
      "tokens_seen": 1891958784
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004311735205616851,
      "loss": 1.4079,
      "theoretical_loss": 3.4748566116448893,
      "tokens_seen": 1892024320
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043115346038114346,
      "loss": 1.4179,
      "theoretical_loss": 3.474846549586277,
      "tokens_seen": 1892089856
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004311334002006018,
      "loss": 1.3995,
      "theoretical_loss": 3.474836487973757,
      "tokens_seen": 1892155392
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004311133400200602,
      "loss": 1.4342,
      "theoretical_loss": 3.474826426807295,
      "tokens_seen": 1892220928
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004310932798395186,
      "loss": 1.5022,
      "theoretical_loss": 3.4748163660868556,
      "tokens_seen": 1892286464
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043107321965897696,
      "loss": 1.3824,
      "theoretical_loss": 3.474806305812403,
      "tokens_seen": 1892352000
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004310531594784353,
      "loss": 1.4806,
      "theoretical_loss": 3.474796245983903,
      "tokens_seen": 1892417536
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004310330992978937,
      "loss": 1.4399,
      "theoretical_loss": 3.47478618660132,
      "tokens_seen": 1892483072
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043101303911735205,
      "loss": 1.4881,
      "theoretical_loss": 3.4747761276646183,
      "tokens_seen": 1892548608
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043099297893681046,
      "loss": 1.4968,
      "theoretical_loss": 3.4747660691737634,
      "tokens_seen": 1892614144
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004309729187562688,
      "loss": 1.3922,
      "theoretical_loss": 3.4747560111287195,
      "tokens_seen": 1892679680
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004309528585757272,
      "loss": 1.3539,
      "theoretical_loss": 3.4747459535294514,
      "tokens_seen": 1892745216
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043093279839518555,
      "loss": 1.4434,
      "theoretical_loss": 3.4747358963759245,
      "tokens_seen": 1892810752
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043091273821464397,
      "loss": 1.459,
      "theoretical_loss": 3.474725839668103,
      "tokens_seen": 1892876288
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043089267803410233,
      "loss": 1.3094,
      "theoretical_loss": 3.4747157834059523,
      "tokens_seen": 1892941824
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004308726178535607,
      "loss": 1.3663,
      "theoretical_loss": 3.4747057275894364,
      "tokens_seen": 1893007360
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043085255767301905,
      "loss": 1.4072,
      "theoretical_loss": 3.4746956722185214,
      "tokens_seen": 1893072896
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004308324974924774,
      "loss": 1.39,
      "theoretical_loss": 3.4746856172931704,
      "tokens_seen": 1893138432
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043081243731193583,
      "loss": 1.4542,
      "theoretical_loss": 3.47467556281335,
      "tokens_seen": 1893203968
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004307923771313942,
      "loss": 1.3887,
      "theoretical_loss": 3.474665508779024,
      "tokens_seen": 1893269504
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043077231695085256,
      "loss": 1.4925,
      "theoretical_loss": 3.4746554551901574,
      "tokens_seen": 1893335040
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004307522567703109,
      "loss": 1.4692,
      "theoretical_loss": 3.4746454020467152,
      "tokens_seen": 1893400576
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043073219658976933,
      "loss": 1.4308,
      "theoretical_loss": 3.474635349348662,
      "tokens_seen": 1893466112
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004307121364092277,
      "loss": 1.4611,
      "theoretical_loss": 3.474625297095963,
      "tokens_seen": 1893531648
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043069207622868606,
      "loss": 1.3999,
      "theoretical_loss": 3.4746152452885832,
      "tokens_seen": 1893597184
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004306720160481444,
      "loss": 1.3943,
      "theoretical_loss": 3.474605193926487,
      "tokens_seen": 1893662720
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004306519558676028,
      "loss": 1.39,
      "theoretical_loss": 3.4745951430096387,
      "tokens_seen": 1893728256
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004306318956870612,
      "loss": 1.501,
      "theoretical_loss": 3.4745850925380046,
      "tokens_seen": 1893793792
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043061183550651956,
      "loss": 1.3688,
      "theoretical_loss": 3.4745750425115487,
      "tokens_seen": 1893859328
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004305917753259779,
      "loss": 1.45,
      "theoretical_loss": 3.474564992930236,
      "tokens_seen": 1893924864
    },
    {
      "epoch": 0.57,
      "objective/train/advantage_avg": 0.49572598934173584,
      "objective/train/docs_used": 690600,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4033095836639404,
      "objective/train/original_loss": 1.4033095836639404,
      "objective/train/theoretical_loss": 3.474554943794031,
      "objective/train/tokens_used": 1914450400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24590560793876648,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508227348327637,
      "objective/train/weighted_lm_loss": 1.4746100902557373,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.474554943794031,
      "tokens_seen": 1893990400
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004305717151454363,
      "loss": 1.3995,
      "theoretical_loss": 3.474554943794031,
      "tokens_seen": 1893990400
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004305516549648947,
      "loss": 1.4783,
      "theoretical_loss": 3.4745448951028997,
      "tokens_seen": 1894055936
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043053159478435306,
      "loss": 1.4373,
      "theoretical_loss": 3.474534846856806,
      "tokens_seen": 1894121472
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004305115346038114,
      "loss": 1.3378,
      "theoretical_loss": 3.4745247990557147,
      "tokens_seen": 1894187008
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004304914744232698,
      "loss": 1.5208,
      "theoretical_loss": 3.4745147516995916,
      "tokens_seen": 1894252544
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043047141424272815,
      "loss": 1.5108,
      "theoretical_loss": 3.4745047047884006,
      "tokens_seen": 1894318080
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043045135406218657,
      "loss": 1.3755,
      "theoretical_loss": 3.4744946583221075,
      "tokens_seen": 1894383616
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043043129388164493,
      "loss": 1.4517,
      "theoretical_loss": 3.4744846123006767,
      "tokens_seen": 1894449152
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004304112337011033,
      "loss": 1.3504,
      "theoretical_loss": 3.474474566724073,
      "tokens_seen": 1894514688
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043039117352056165,
      "loss": 1.4931,
      "theoretical_loss": 3.474464521592262,
      "tokens_seen": 1894580224
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043037111334002007,
      "loss": 1.4277,
      "theoretical_loss": 3.4744544769052075,
      "tokens_seen": 1894645760
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043035105315947843,
      "loss": 1.3738,
      "theoretical_loss": 3.4744444326628754,
      "tokens_seen": 1894711296
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004303309929789368,
      "loss": 1.4337,
      "theoretical_loss": 3.4744343888652303,
      "tokens_seen": 1894776832
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043031093279839516,
      "loss": 1.3818,
      "theoretical_loss": 3.474424345512237,
      "tokens_seen": 1894842368
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004302908726178535,
      "loss": 1.4066,
      "theoretical_loss": 3.4744143026038605,
      "tokens_seen": 1894907904
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043027081243731194,
      "loss": 1.3114,
      "theoretical_loss": 3.4744042601400658,
      "tokens_seen": 1894973440
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004302507522567703,
      "loss": 1.4035,
      "theoretical_loss": 3.474394218120818,
      "tokens_seen": 1895038976
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043023069207622866,
      "loss": 1.4797,
      "theoretical_loss": 3.474384176546082,
      "tokens_seen": 1895104512
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000430210631895687,
      "loss": 1.3455,
      "theoretical_loss": 3.4743741354158226,
      "tokens_seen": 1895170048
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043019057171514544,
      "loss": 1.4803,
      "theoretical_loss": 3.4743640947300047,
      "tokens_seen": 1895235584
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043017051153460386,
      "loss": 1.4119,
      "theoretical_loss": 3.474354054488593,
      "tokens_seen": 1895301120
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004301504513540622,
      "loss": 1.3522,
      "theoretical_loss": 3.4743440146915536,
      "tokens_seen": 1895366656
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004301303911735206,
      "loss": 1.4574,
      "theoretical_loss": 3.4743339753388502,
      "tokens_seen": 1895432192
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.000430110330992979,
      "loss": 1.4088,
      "theoretical_loss": 3.4743239364304483,
      "tokens_seen": 1895497728
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043009027081243736,
      "loss": 1.3782,
      "theoretical_loss": 3.4743138979663133,
      "tokens_seen": 1895563264
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004300702106318957,
      "loss": 1.45,
      "theoretical_loss": 3.474303859946409,
      "tokens_seen": 1895628800
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004300501504513541,
      "loss": 1.5218,
      "theoretical_loss": 3.4742938223707016,
      "tokens_seen": 1895694336
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043003009027081245,
      "loss": 1.5625,
      "theoretical_loss": 3.4742837852391553,
      "tokens_seen": 1895759872
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00043001003009027086,
      "loss": 1.4203,
      "theoretical_loss": 3.4742737485517354,
      "tokens_seen": 1895825408
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004299899699097292,
      "loss": 1.5133,
      "theoretical_loss": 3.4742637123084075,
      "tokens_seen": 1895890944
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004299699097291876,
      "loss": 1.3576,
      "theoretical_loss": 3.4742536765091354,
      "tokens_seen": 1895956480
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00042994984954864595,
      "loss": 1.4143,
      "theoretical_loss": 3.474243641153885,
      "tokens_seen": 1896022016
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00042992978936810437,
      "loss": 1.3649,
      "theoretical_loss": 3.4742336062426205,
      "tokens_seen": 1896087552
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00042990972918756273,
      "loss": 1.4921,
      "theoretical_loss": 3.474223571775308,
      "tokens_seen": 1896153088
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004298896690070211,
      "loss": 1.3386,
      "theoretical_loss": 3.474213537751911,
      "tokens_seen": 1896218624
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00042986960882647945,
      "loss": 1.3984,
      "theoretical_loss": 3.4742035041723964,
      "tokens_seen": 1896284160
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004298495486459378,
      "loss": 1.43,
      "theoretical_loss": 3.4741934710367284,
      "tokens_seen": 1896349696
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00042982948846539623,
      "loss": 1.4545,
      "theoretical_loss": 3.4741834383448715,
      "tokens_seen": 1896415232
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004298094282848546,
      "loss": 1.5451,
      "theoretical_loss": 3.4741734060967913,
      "tokens_seen": 1896480768
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00042978936810431296,
      "loss": 1.3919,
      "theoretical_loss": 3.474163374292452,
      "tokens_seen": 1896546304
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004297693079237713,
      "loss": 1.3576,
      "theoretical_loss": 3.47415334293182,
      "tokens_seen": 1896611840
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00042974924774322973,
      "loss": 1.4357,
      "theoretical_loss": 3.474143312014859,
      "tokens_seen": 1896677376
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004297291875626881,
      "loss": 1.3164,
      "theoretical_loss": 3.4741332815415356,
      "tokens_seen": 1896742912
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00042970912738214646,
      "loss": 1.4314,
      "theoretical_loss": 3.4741232515118137,
      "tokens_seen": 1896808448
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004296890672016048,
      "loss": 1.4751,
      "theoretical_loss": 3.4741132219256583,
      "tokens_seen": 1896873984
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004296690070210632,
      "loss": 1.3752,
      "theoretical_loss": 3.474103192783035,
      "tokens_seen": 1896939520
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004296489468405216,
      "loss": 1.3832,
      "theoretical_loss": 3.474093164083908,
      "tokens_seen": 1897005056
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00042962888665997996,
      "loss": 1.529,
      "theoretical_loss": 3.474083135828244,
      "tokens_seen": 1897070592
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004296088264794383,
      "loss": 1.4063,
      "theoretical_loss": 3.4740731080160066,
      "tokens_seen": 1897136128
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004295887662988967,
      "loss": 1.5351,
      "theoretical_loss": 3.474063080647161,
      "tokens_seen": 1897201664
    },
    {
      "epoch": 0.57,
      "objective/train/advantage_avg": 0.4932161867618561,
      "objective/train/docs_used": 691800,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4548717737197876,
      "objective/train/original_loss": 1.4548718929290771,
      "objective/train/theoretical_loss": 3.474053053721673,
      "objective/train/tokens_used": 1917727200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2437119036912918,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050560474395752,
      "objective/train/weighted_lm_loss": 1.5283620357513428,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.474053053721673,
      "tokens_seen": 1897267200
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004295687061183551,
      "loss": 1.4606,
      "theoretical_loss": 3.474053053721673,
      "tokens_seen": 1897267200
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00042954864593781347,
      "loss": 1.4966,
      "theoretical_loss": 3.4740430272395075,
      "tokens_seen": 1897332736
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.00042952858575727183,
      "loss": 1.4459,
      "theoretical_loss": 3.4740330012006293,
      "tokens_seen": 1897398272
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0004295085255767302,
      "loss": 1.3513,
      "theoretical_loss": 3.4740229756050036,
      "tokens_seen": 1897463808
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042948846539618855,
      "loss": 1.4435,
      "theoretical_loss": 3.474012950452595,
      "tokens_seen": 1897529344
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042946840521564697,
      "loss": 1.4227,
      "theoretical_loss": 3.4740029257433696,
      "tokens_seen": 1897594880
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042944834503510533,
      "loss": 1.4271,
      "theoretical_loss": 3.473992901477292,
      "tokens_seen": 1897660416
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004294282848545637,
      "loss": 1.41,
      "theoretical_loss": 3.4739828776543265,
      "tokens_seen": 1897725952
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042940822467402206,
      "loss": 1.3718,
      "theoretical_loss": 3.47397285427444,
      "tokens_seen": 1897791488
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042938816449348047,
      "loss": 1.2845,
      "theoretical_loss": 3.4739628313375963,
      "tokens_seen": 1897857024
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042936810431293883,
      "loss": 1.4995,
      "theoretical_loss": 3.4739528088437606,
      "tokens_seen": 1897922560
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004293480441323972,
      "loss": 1.2932,
      "theoretical_loss": 3.4739427867928985,
      "tokens_seen": 1897988096
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042932798395185556,
      "loss": 1.4294,
      "theoretical_loss": 3.473932765184975,
      "tokens_seen": 1898053632
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004293079237713139,
      "loss": 1.4541,
      "theoretical_loss": 3.473922744019955,
      "tokens_seen": 1898119168
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042928786359077234,
      "loss": 1.4103,
      "theoretical_loss": 3.4739127232978033,
      "tokens_seen": 1898184704
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004292678034102307,
      "loss": 1.2837,
      "theoretical_loss": 3.473902703018486,
      "tokens_seen": 1898250240
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042924774322968906,
      "loss": 1.3664,
      "theoretical_loss": 3.473892683181968,
      "tokens_seen": 1898315776
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004292276830491474,
      "loss": 1.3811,
      "theoretical_loss": 3.4738826637882134,
      "tokens_seen": 1898381312
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042920762286860584,
      "loss": 1.4058,
      "theoretical_loss": 3.473872644837189,
      "tokens_seen": 1898446848
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004291875626880642,
      "loss": 1.4057,
      "theoretical_loss": 3.4738626263288586,
      "tokens_seen": 1898512384
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042916750250752256,
      "loss": 1.5068,
      "theoretical_loss": 3.4738526082631878,
      "tokens_seen": 1898577920
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004291474423269809,
      "loss": 1.4123,
      "theoretical_loss": 3.473842590640142,
      "tokens_seen": 1898643456
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004291273821464393,
      "loss": 1.4039,
      "theoretical_loss": 3.4738325734596858,
      "tokens_seen": 1898708992
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004291073219658977,
      "loss": 1.3165,
      "theoretical_loss": 3.4738225567217853,
      "tokens_seen": 1898774528
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042908726178535607,
      "loss": 1.3251,
      "theoretical_loss": 3.4738125404264046,
      "tokens_seen": 1898840064
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042906720160481443,
      "loss": 1.3697,
      "theoretical_loss": 3.47380252457351,
      "tokens_seen": 1898905600
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004290471414242728,
      "loss": 1.3942,
      "theoretical_loss": 3.473792509163066,
      "tokens_seen": 1898971136
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004290270812437312,
      "loss": 1.2969,
      "theoretical_loss": 3.4737824941950373,
      "tokens_seen": 1899036672
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042900702106318957,
      "loss": 1.43,
      "theoretical_loss": 3.47377247966939,
      "tokens_seen": 1899102208
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042898696088264793,
      "loss": 1.3692,
      "theoretical_loss": 3.473762465586089,
      "tokens_seen": 1899167744
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004289669007021063,
      "loss": 1.5179,
      "theoretical_loss": 3.4737524519450993,
      "tokens_seen": 1899233280
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004289468405215647,
      "loss": 1.4178,
      "theoretical_loss": 3.4737424387463864,
      "tokens_seen": 1899298816
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004289267803410231,
      "loss": 1.431,
      "theoretical_loss": 3.4737324259899154,
      "tokens_seen": 1899364352
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042890672016048144,
      "loss": 1.421,
      "theoretical_loss": 3.4737224136756515,
      "tokens_seen": 1899429888
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004288866599799398,
      "loss": 1.3467,
      "theoretical_loss": 3.4737124018035597,
      "tokens_seen": 1899495424
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042886659979939816,
      "loss": 1.3768,
      "theoretical_loss": 3.4737023903736053,
      "tokens_seen": 1899560960
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004288465396188566,
      "loss": 1.3189,
      "theoretical_loss": 3.4736923793857537,
      "tokens_seen": 1899626496
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042882647943831494,
      "loss": 1.3922,
      "theoretical_loss": 3.4736823688399703,
      "tokens_seen": 1899692032
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004288064192577733,
      "loss": 1.489,
      "theoretical_loss": 3.47367235873622,
      "tokens_seen": 1899757568
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042878635907723166,
      "loss": 1.3897,
      "theoretical_loss": 3.473662349074468,
      "tokens_seen": 1899823104
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004287662988966901,
      "loss": 1.3391,
      "theoretical_loss": 3.4736523398546795,
      "tokens_seen": 1899888640
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042874623871614844,
      "loss": 1.3797,
      "theoretical_loss": 3.47364233107682,
      "tokens_seen": 1899954176
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004287261785356068,
      "loss": 1.4804,
      "theoretical_loss": 3.4736323227408548,
      "tokens_seen": 1900019712
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042870611835506517,
      "loss": 1.394,
      "theoretical_loss": 3.473622314846749,
      "tokens_seen": 1900085248
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042868605817452353,
      "loss": 1.4617,
      "theoretical_loss": 3.4736123073944674,
      "tokens_seen": 1900150784
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000428665997993982,
      "loss": 1.4285,
      "theoretical_loss": 3.473602300383976,
      "tokens_seen": 1900216320
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042864593781344036,
      "loss": 1.3867,
      "theoretical_loss": 3.4735922938152397,
      "tokens_seen": 1900281856
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004286258776328987,
      "loss": 1.4094,
      "theoretical_loss": 3.473582287688224,
      "tokens_seen": 1900347392
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004286058174523571,
      "loss": 1.4667,
      "theoretical_loss": 3.473572282002894,
      "tokens_seen": 1900412928
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004285857572718155,
      "loss": 1.3478,
      "theoretical_loss": 3.473562276759215,
      "tokens_seen": 1900478464
    },
    {
      "epoch": 0.58,
      "objective/train/advantage_avg": 0.49316853284835815,
      "objective/train/docs_used": 693007,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.283091425895691,
      "objective/train/original_loss": 1.2830915451049805,
      "objective/train/theoretical_loss": 3.4735522719571517,
      "objective/train/tokens_used": 1921004000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24380354583263397,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505561828613281,
      "objective/train/weighted_lm_loss": 1.3479235172271729,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4735522719571517,
      "tokens_seen": 1900544000
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042856569709127387,
      "loss": 1.378,
      "theoretical_loss": 3.4735522719571517,
      "tokens_seen": 1900544000
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042854563691073223,
      "loss": 1.3463,
      "theoretical_loss": 3.4735422675966703,
      "tokens_seen": 1900609536
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004285255767301906,
      "loss": 1.3249,
      "theoretical_loss": 3.4735322636777357,
      "tokens_seen": 1900675072
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042850551654964895,
      "loss": 1.4219,
      "theoretical_loss": 3.4735222602003133,
      "tokens_seen": 1900740608
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042848545636910737,
      "loss": 1.438,
      "theoretical_loss": 3.473512257164368,
      "tokens_seen": 1900806144
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042846539618856573,
      "loss": 1.4666,
      "theoretical_loss": 3.4735022545698655,
      "tokens_seen": 1900871680
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004284453360080241,
      "loss": 1.4116,
      "theoretical_loss": 3.4734922524167713,
      "tokens_seen": 1900937216
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042842527582748246,
      "loss": 1.3683,
      "theoretical_loss": 3.47348225070505,
      "tokens_seen": 1901002752
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042840521564694087,
      "loss": 1.3571,
      "theoretical_loss": 3.4734722494346673,
      "tokens_seen": 1901068288
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042838515546639923,
      "loss": 1.4213,
      "theoretical_loss": 3.4734622486055886,
      "tokens_seen": 1901133824
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004283650952858576,
      "loss": 1.3981,
      "theoretical_loss": 3.473452248217779,
      "tokens_seen": 1901199360
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042834503510531596,
      "loss": 1.4318,
      "theoretical_loss": 3.473442248271204,
      "tokens_seen": 1901264896
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004283249749247743,
      "loss": 1.3993,
      "theoretical_loss": 3.473432248765829,
      "tokens_seen": 1901330432
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042830491474423274,
      "loss": 1.4171,
      "theoretical_loss": 3.473422249701619,
      "tokens_seen": 1901395968
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004282848545636911,
      "loss": 1.4621,
      "theoretical_loss": 3.47341225107854,
      "tokens_seen": 1901461504
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042826479438314946,
      "loss": 1.4512,
      "theoretical_loss": 3.4734022528965562,
      "tokens_seen": 1901527040
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004282447342026078,
      "loss": 1.3951,
      "theoretical_loss": 3.473392255155634,
      "tokens_seen": 1901592576
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042822467402206624,
      "loss": 1.3522,
      "theoretical_loss": 3.4733822578557385,
      "tokens_seen": 1901658112
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004282046138415246,
      "loss": 1.4647,
      "theoretical_loss": 3.4733722609968343,
      "tokens_seen": 1901723648
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042818455366098296,
      "loss": 1.399,
      "theoretical_loss": 3.4733622645788875,
      "tokens_seen": 1901789184
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004281644934804413,
      "loss": 1.4865,
      "theoretical_loss": 3.4733522686018636,
      "tokens_seen": 1901854720
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004281444332998997,
      "loss": 1.4418,
      "theoretical_loss": 3.473342273065728,
      "tokens_seen": 1901920256
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004281243731193581,
      "loss": 1.4286,
      "theoretical_loss": 3.473332277970445,
      "tokens_seen": 1901985792
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042810431293881647,
      "loss": 1.4763,
      "theoretical_loss": 3.473322283315981,
      "tokens_seen": 1902051328
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042808425275827483,
      "loss": 1.4029,
      "theoretical_loss": 3.4733122891023007,
      "tokens_seen": 1902116864
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004280641925777332,
      "loss": 1.4653,
      "theoretical_loss": 3.47330229532937,
      "tokens_seen": 1902182400
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004280441323971916,
      "loss": 1.318,
      "theoretical_loss": 3.473292301997154,
      "tokens_seen": 1902247936
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042802407221664997,
      "loss": 1.4843,
      "theoretical_loss": 3.4732823091056186,
      "tokens_seen": 1902313472
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042800401203610833,
      "loss": 1.3632,
      "theoretical_loss": 3.473272316654729,
      "tokens_seen": 1902379008
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004279839518555667,
      "loss": 1.5324,
      "theoretical_loss": 3.4732623246444496,
      "tokens_seen": 1902444544
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004279638916750251,
      "loss": 1.4874,
      "theoretical_loss": 3.473252333074747,
      "tokens_seen": 1902510080
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004279438314944835,
      "loss": 1.4481,
      "theoretical_loss": 3.473242341945586,
      "tokens_seen": 1902575616
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042792377131394184,
      "loss": 1.4185,
      "theoretical_loss": 3.4732323512569323,
      "tokens_seen": 1902641152
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004279037111334002,
      "loss": 1.4028,
      "theoretical_loss": 3.4732223610087507,
      "tokens_seen": 1902706688
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042788365095285856,
      "loss": 1.3682,
      "theoretical_loss": 3.473212371201008,
      "tokens_seen": 1902772224
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000427863590772317,
      "loss": 1.402,
      "theoretical_loss": 3.473202381833668,
      "tokens_seen": 1902837760
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042784353059177534,
      "loss": 1.4297,
      "theoretical_loss": 3.473192392906697,
      "tokens_seen": 1902903296
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004278234704112337,
      "loss": 1.5206,
      "theoretical_loss": 3.47318240442006,
      "tokens_seen": 1902968832
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042780341023069206,
      "loss": 1.4115,
      "theoretical_loss": 3.473172416373723,
      "tokens_seen": 1903034368
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004277833500501505,
      "loss": 1.3025,
      "theoretical_loss": 3.473162428767651,
      "tokens_seen": 1903099904
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042776328986960884,
      "loss": 1.4649,
      "theoretical_loss": 3.473152441601809,
      "tokens_seen": 1903165440
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004277432296890672,
      "loss": 1.3666,
      "theoretical_loss": 3.4731424548761636,
      "tokens_seen": 1903230976
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042772316950852557,
      "loss": 1.571,
      "theoretical_loss": 3.4731324685906797,
      "tokens_seen": 1903296512
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042770310932798393,
      "loss": 1.4915,
      "theoretical_loss": 3.473122482745322,
      "tokens_seen": 1903362048
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042768304914744235,
      "loss": 1.5141,
      "theoretical_loss": 3.473112497340057,
      "tokens_seen": 1903427584
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004276629889669007,
      "loss": 1.3328,
      "theoretical_loss": 3.47310251237485,
      "tokens_seen": 1903493120
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042764292878635907,
      "loss": 1.5147,
      "theoretical_loss": 3.4730925278496656,
      "tokens_seen": 1903558656
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042762286860581743,
      "loss": 1.4297,
      "theoretical_loss": 3.47308254376447,
      "tokens_seen": 1903624192
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042760280842527585,
      "loss": 1.3644,
      "theoretical_loss": 3.473072560119229,
      "tokens_seen": 1903689728
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004275827482447342,
      "loss": 1.4597,
      "theoretical_loss": 3.4730625769139074,
      "tokens_seen": 1903755264
    },
    {
      "epoch": 0.58,
      "objective/train/advantage_avg": 0.4940413534641266,
      "objective/train/docs_used": 694210,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.338824987411499,
      "objective/train/original_loss": 1.3388248682022095,
      "objective/train/theoretical_loss": 3.4730525941484705,
      "objective/train/tokens_used": 1924280800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24445657432079315,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506469011306763,
      "objective/train/weighted_lm_loss": 1.4065638780593872,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4730525941484705,
      "tokens_seen": 1903820800
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004275626880641926,
      "loss": 1.4223,
      "theoretical_loss": 3.4730525941484705,
      "tokens_seen": 1903820800
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042754262788365094,
      "loss": 1.4739,
      "theoretical_loss": 3.4730426118228843,
      "tokens_seen": 1903886336
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004275225677031093,
      "loss": 1.421,
      "theoretical_loss": 3.473032629937114,
      "tokens_seen": 1903951872
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004275025075225677,
      "loss": 1.3012,
      "theoretical_loss": 3.473022648491126,
      "tokens_seen": 1904017408
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004274824473420261,
      "loss": 1.4374,
      "theoretical_loss": 3.473012667484884,
      "tokens_seen": 1904082944
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042746238716148444,
      "loss": 1.408,
      "theoretical_loss": 3.4730026869183552,
      "tokens_seen": 1904148480
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004274423269809428,
      "loss": 1.4809,
      "theoretical_loss": 3.472992706791504,
      "tokens_seen": 1904214016
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004274222668004012,
      "loss": 1.4419,
      "theoretical_loss": 3.472982727104296,
      "tokens_seen": 1904279552
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004274022066198596,
      "loss": 1.4732,
      "theoretical_loss": 3.4729727478566974,
      "tokens_seen": 1904345088
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042738214643931794,
      "loss": 1.3422,
      "theoretical_loss": 3.4729627690486735,
      "tokens_seen": 1904410624
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004273620862587763,
      "loss": 1.4281,
      "theoretical_loss": 3.472952790680189,
      "tokens_seen": 1904476160
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042734202607823467,
      "loss": 1.3668,
      "theoretical_loss": 3.47294281275121,
      "tokens_seen": 1904541696
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004273219658976931,
      "loss": 1.4421,
      "theoretical_loss": 3.4729328352617026,
      "tokens_seen": 1904607232
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042730190571715145,
      "loss": 1.5182,
      "theoretical_loss": 3.472922858211632,
      "tokens_seen": 1904672768
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004272818455366098,
      "loss": 1.359,
      "theoretical_loss": 3.472912881600963,
      "tokens_seen": 1904738304
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042726178535606817,
      "loss": 1.3996,
      "theoretical_loss": 3.4729029054296614,
      "tokens_seen": 1904803840
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004272417251755266,
      "loss": 1.4959,
      "theoretical_loss": 3.4728929296976934,
      "tokens_seen": 1904869376
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042722166499498495,
      "loss": 1.4757,
      "theoretical_loss": 3.4728829544050237,
      "tokens_seen": 1904934912
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004272016048144433,
      "loss": 1.3176,
      "theoretical_loss": 3.4728729795516182,
      "tokens_seen": 1905000448
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004271815446339017,
      "loss": 1.4733,
      "theoretical_loss": 3.4728630051374427,
      "tokens_seen": 1905065984
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004271614844533601,
      "loss": 1.3922,
      "theoretical_loss": 3.4728530311624626,
      "tokens_seen": 1905131520
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004271414242728185,
      "loss": 1.4833,
      "theoretical_loss": 3.4728430576266436,
      "tokens_seen": 1905197056
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042712136409227687,
      "loss": 1.4002,
      "theoretical_loss": 3.4728330845299507,
      "tokens_seen": 1905262592
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042710130391173523,
      "loss": 1.3819,
      "theoretical_loss": 3.4728231118723496,
      "tokens_seen": 1905328128
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004270812437311936,
      "loss": 1.4116,
      "theoretical_loss": 3.4728131396538067,
      "tokens_seen": 1905393664
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000427061183550652,
      "loss": 1.4801,
      "theoretical_loss": 3.472803167874287,
      "tokens_seen": 1905459200
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042704112337011037,
      "loss": 1.4531,
      "theoretical_loss": 3.472793196533755,
      "tokens_seen": 1905524736
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042702106318956873,
      "loss": 1.479,
      "theoretical_loss": 3.472783225632178,
      "tokens_seen": 1905590272
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004270010030090271,
      "loss": 1.3845,
      "theoretical_loss": 3.4727732551695207,
      "tokens_seen": 1905655808
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004269809428284855,
      "loss": 1.4231,
      "theoretical_loss": 3.472763285145749,
      "tokens_seen": 1905721344
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004269608826479439,
      "loss": 1.5266,
      "theoretical_loss": 3.472753315560828,
      "tokens_seen": 1905786880
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042694082246740224,
      "loss": 1.4221,
      "theoretical_loss": 3.472743346414724,
      "tokens_seen": 1905852416
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004269207622868606,
      "loss": 1.4118,
      "theoretical_loss": 3.472733377707402,
      "tokens_seen": 1905917952
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042690070210631896,
      "loss": 1.3432,
      "theoretical_loss": 3.472723409438828,
      "tokens_seen": 1905983488
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004268806419257774,
      "loss": 1.364,
      "theoretical_loss": 3.4727134416089678,
      "tokens_seen": 1906049024
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042686058174523574,
      "loss": 1.5245,
      "theoretical_loss": 3.472703474217786,
      "tokens_seen": 1906114560
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004268405215646941,
      "loss": 1.4615,
      "theoretical_loss": 3.4726935072652494,
      "tokens_seen": 1906180096
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042682046138415246,
      "loss": 1.4588,
      "theoretical_loss": 3.472683540751323,
      "tokens_seen": 1906245632
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004268004012036109,
      "loss": 1.4538,
      "theoretical_loss": 3.472673574675972,
      "tokens_seen": 1906311168
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042678034102306924,
      "loss": 1.4418,
      "theoretical_loss": 3.472663609039163,
      "tokens_seen": 1906376704
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004267602808425276,
      "loss": 1.3316,
      "theoretical_loss": 3.472653643840861,
      "tokens_seen": 1906442240
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042674022066198597,
      "loss": 1.3187,
      "theoretical_loss": 3.4726436790810316,
      "tokens_seen": 1906507776
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042672016048144433,
      "loss": 1.4028,
      "theoretical_loss": 3.472633714759641,
      "tokens_seen": 1906573312
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042670010030090275,
      "loss": 1.5348,
      "theoretical_loss": 3.4726237508766538,
      "tokens_seen": 1906638848
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004266800401203611,
      "loss": 1.3457,
      "theoretical_loss": 3.472613787432037,
      "tokens_seen": 1906704384
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042665997993981947,
      "loss": 1.3603,
      "theoretical_loss": 3.472603824425755,
      "tokens_seen": 1906769920
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042663991975927783,
      "loss": 1.4916,
      "theoretical_loss": 3.4725938618577743,
      "tokens_seen": 1906835456
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042661985957873625,
      "loss": 1.4947,
      "theoretical_loss": 3.4725838997280603,
      "tokens_seen": 1906900992
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004265997993981946,
      "loss": 1.4058,
      "theoretical_loss": 3.4725739380365788,
      "tokens_seen": 1906966528
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000426579739217653,
      "loss": 1.4222,
      "theoretical_loss": 3.4725639767832948,
      "tokens_seen": 1907032064
    },
    {
      "epoch": 0.58,
      "objective/train/advantage_avg": 0.49523991346359253,
      "objective/train/docs_used": 695522,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.316782832145691,
      "objective/train/original_loss": 1.3167827129364014,
      "objective/train/theoretical_loss": 3.4725540159681745,
      "objective/train/tokens_used": 1927557600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24565571546554565,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050773024559021,
      "objective/train/weighted_lm_loss": 1.3836686611175537,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4725540159681745,
      "tokens_seen": 1907097600
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042655967903711134,
      "loss": 1.3812,
      "theoretical_loss": 3.4725540159681745,
      "tokens_seen": 1907097600
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004265396188565697,
      "loss": 1.3675,
      "theoretical_loss": 3.4725440555911837,
      "tokens_seen": 1907163136
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004265195586760281,
      "loss": 1.3198,
      "theoretical_loss": 3.4725340956522874,
      "tokens_seen": 1907228672
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004264994984954865,
      "loss": 1.4733,
      "theoretical_loss": 3.4725241361514527,
      "tokens_seen": 1907294208
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042647943831494484,
      "loss": 1.3658,
      "theoretical_loss": 3.4725141770886436,
      "tokens_seen": 1907359744
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004264593781344032,
      "loss": 1.3921,
      "theoretical_loss": 3.472504218463827,
      "tokens_seen": 1907425280
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004264393179538616,
      "loss": 1.3273,
      "theoretical_loss": 3.472494260276968,
      "tokens_seen": 1907490816
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042641925777332,
      "loss": 1.4431,
      "theoretical_loss": 3.4724843025280325,
      "tokens_seen": 1907556352
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042639919759277834,
      "loss": 1.3423,
      "theoretical_loss": 3.4724743452169857,
      "tokens_seen": 1907621888
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004263791374122367,
      "loss": 1.4318,
      "theoretical_loss": 3.472464388343794,
      "tokens_seen": 1907687424
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042635907723169507,
      "loss": 1.427,
      "theoretical_loss": 3.4724544319084227,
      "tokens_seen": 1907752960
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004263390170511535,
      "loss": 1.3113,
      "theoretical_loss": 3.4724444759108373,
      "tokens_seen": 1907818496
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042631895687061185,
      "loss": 1.443,
      "theoretical_loss": 3.4724345203510047,
      "tokens_seen": 1907884032
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004262988966900702,
      "loss": 1.493,
      "theoretical_loss": 3.472424565228889,
      "tokens_seen": 1907949568
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042627883650952857,
      "loss": 1.3185,
      "theoretical_loss": 3.4724146105444573,
      "tokens_seen": 1908015104
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000426258776328987,
      "loss": 1.4224,
      "theoretical_loss": 3.472404656297674,
      "tokens_seen": 1908080640
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042623871614844535,
      "loss": 1.3899,
      "theoretical_loss": 3.472394702488506,
      "tokens_seen": 1908146176
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004262186559679037,
      "loss": 1.4098,
      "theoretical_loss": 3.4723847491169186,
      "tokens_seen": 1908211712
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004261985957873621,
      "loss": 1.3537,
      "theoretical_loss": 3.4723747961828773,
      "tokens_seen": 1908277248
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042617853560682044,
      "loss": 1.5019,
      "theoretical_loss": 3.472364843686348,
      "tokens_seen": 1908342784
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042615847542627885,
      "loss": 1.3118,
      "theoretical_loss": 3.4723548916272966,
      "tokens_seen": 1908408320
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004261384152457372,
      "loss": 1.4324,
      "theoretical_loss": 3.472344940005689,
      "tokens_seen": 1908473856
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004261183550651956,
      "loss": 1.4111,
      "theoretical_loss": 3.4723349888214905,
      "tokens_seen": 1908539392
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042609829488465394,
      "loss": 1.3739,
      "theoretical_loss": 3.4723250380746666,
      "tokens_seen": 1908604928
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042607823470411236,
      "loss": 1.5025,
      "theoretical_loss": 3.4723150877651836,
      "tokens_seen": 1908670464
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004260581745235707,
      "loss": 1.3066,
      "theoretical_loss": 3.4723051378930077,
      "tokens_seen": 1908736000
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004260381143430291,
      "loss": 1.3993,
      "theoretical_loss": 3.4722951884581037,
      "tokens_seen": 1908801536
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042601805416248744,
      "loss": 1.5235,
      "theoretical_loss": 3.472285239460438,
      "tokens_seen": 1908867072
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004259979939819458,
      "loss": 1.4065,
      "theoretical_loss": 3.472275290899976,
      "tokens_seen": 1908932608
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004259779338014042,
      "loss": 1.4566,
      "theoretical_loss": 3.4722653427766836,
      "tokens_seen": 1908998144
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004259578736208626,
      "loss": 1.4198,
      "theoretical_loss": 3.472255395090527,
      "tokens_seen": 1909063680
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042593781344032095,
      "loss": 1.3505,
      "theoretical_loss": 3.4722454478414715,
      "tokens_seen": 1909129216
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004259177532597793,
      "loss": 1.4697,
      "theoretical_loss": 3.472235501029483,
      "tokens_seen": 1909194752
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004258976930792377,
      "loss": 1.3651,
      "theoretical_loss": 3.472225554654527,
      "tokens_seen": 1909260288
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004258776328986961,
      "loss": 1.469,
      "theoretical_loss": 3.47221560871657,
      "tokens_seen": 1909325824
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042585757271815445,
      "loss": 1.464,
      "theoretical_loss": 3.4722056632155773,
      "tokens_seen": 1909391360
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004258375125376128,
      "loss": 1.4844,
      "theoretical_loss": 3.4721957181515144,
      "tokens_seen": 1909456896
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004258174523570712,
      "loss": 1.5369,
      "theoretical_loss": 3.472185773524348,
      "tokens_seen": 1909522432
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004257973921765296,
      "loss": 1.3858,
      "theoretical_loss": 3.4721758293340432,
      "tokens_seen": 1909587968
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042577733199598795,
      "loss": 1.4174,
      "theoretical_loss": 3.472165885580566,
      "tokens_seen": 1909653504
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004257572718154463,
      "loss": 1.4091,
      "theoretical_loss": 3.472155942263883,
      "tokens_seen": 1909719040
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004257372116349047,
      "loss": 1.4862,
      "theoretical_loss": 3.472145999383958,
      "tokens_seen": 1909784576
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004257171514543631,
      "loss": 1.3451,
      "theoretical_loss": 3.472136056940759,
      "tokens_seen": 1909850112
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042569709127382145,
      "loss": 1.4765,
      "theoretical_loss": 3.472126114934251,
      "tokens_seen": 1909915648
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004256770310932798,
      "loss": 1.4927,
      "theoretical_loss": 3.4721161733643995,
      "tokens_seen": 1909981184
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042565697091273823,
      "loss": 1.3976,
      "theoretical_loss": 3.4721062322311704,
      "tokens_seen": 1910046720
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042563691073219665,
      "loss": 1.329,
      "theoretical_loss": 3.4720962915345304,
      "tokens_seen": 1910112256
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000425616850551655,
      "loss": 1.3729,
      "theoretical_loss": 3.472086351274444,
      "tokens_seen": 1910177792
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004255967903711134,
      "loss": 1.4771,
      "theoretical_loss": 3.472076411450878,
      "tokens_seen": 1910243328
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042557673019057174,
      "loss": 1.399,
      "theoretical_loss": 3.4720664720637986,
      "tokens_seen": 1910308864
    },
    {
      "epoch": 0.58,
      "objective/train/advantage_avg": 0.4915177822113037,
      "objective/train/docs_used": 696705,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2636632919311523,
      "objective/train/original_loss": 1.2636632919311523,
      "objective/train/theoretical_loss": 3.4720565331131708,
      "objective/train/tokens_used": 1930834400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24238885939121246,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503840446472168,
      "objective/train/weighted_lm_loss": 1.327191948890686,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4720565331131708,
      "tokens_seen": 1910374400
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004255566700100301,
      "loss": 1.3571,
      "theoretical_loss": 3.4720565331131708,
      "tokens_seen": 1910374400
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004255366098294885,
      "loss": 1.4216,
      "theoretical_loss": 3.4720465945989605,
      "tokens_seen": 1910439936
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004255165496489469,
      "loss": 1.4373,
      "theoretical_loss": 3.472036656521134,
      "tokens_seen": 1910505472
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042549648946840524,
      "loss": 1.4506,
      "theoretical_loss": 3.4720267188796567,
      "tokens_seen": 1910571008
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004254764292878636,
      "loss": 1.3563,
      "theoretical_loss": 3.4720167816744953,
      "tokens_seen": 1910636544
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000425456369107322,
      "loss": 1.2864,
      "theoretical_loss": 3.472006844905615,
      "tokens_seen": 1910702080
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004254363089267804,
      "loss": 1.3859,
      "theoretical_loss": 3.4719969085729816,
      "tokens_seen": 1910767616
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042541624874623874,
      "loss": 1.4816,
      "theoretical_loss": 3.471986972676561,
      "tokens_seen": 1910833152
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004253961885656971,
      "loss": 1.5515,
      "theoretical_loss": 3.4719770372163197,
      "tokens_seen": 1910898688
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042537612838515547,
      "loss": 1.4268,
      "theoretical_loss": 3.471967102192223,
      "tokens_seen": 1910964224
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004253560682046139,
      "loss": 1.4606,
      "theoretical_loss": 3.4719571676042373,
      "tokens_seen": 1911029760
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042533600802407225,
      "loss": 1.5323,
      "theoretical_loss": 3.471947233452328,
      "tokens_seen": 1911095296
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004253159478435306,
      "loss": 1.4687,
      "theoretical_loss": 3.471937299736461,
      "tokens_seen": 1911160832
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042529588766298897,
      "loss": 1.4044,
      "theoretical_loss": 3.471927366456603,
      "tokens_seen": 1911226368
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004252758274824474,
      "loss": 1.4708,
      "theoretical_loss": 3.4719174336127185,
      "tokens_seen": 1911291904
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042525576730190575,
      "loss": 1.4627,
      "theoretical_loss": 3.471907501204775,
      "tokens_seen": 1911357440
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004252357071213641,
      "loss": 1.4236,
      "theoretical_loss": 3.471897569232737,
      "tokens_seen": 1911422976
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004252156469408225,
      "loss": 1.4666,
      "theoretical_loss": 3.4718876376965717,
      "tokens_seen": 1911488512
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042519558676028084,
      "loss": 1.3773,
      "theoretical_loss": 3.471877706596244,
      "tokens_seen": 1911554048
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042517552657973925,
      "loss": 1.4599,
      "theoretical_loss": 3.4718677759317202,
      "tokens_seen": 1911619584
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004251554663991976,
      "loss": 1.2891,
      "theoretical_loss": 3.4718578457029663,
      "tokens_seen": 1911685120
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000425135406218656,
      "loss": 1.4079,
      "theoretical_loss": 3.4718479159099487,
      "tokens_seen": 1911750656
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042511534603811434,
      "loss": 1.3764,
      "theoretical_loss": 3.4718379865526323,
      "tokens_seen": 1911816192
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042509528585757276,
      "loss": 1.3904,
      "theoretical_loss": 3.471828057630984,
      "tokens_seen": 1911881728
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004250752256770311,
      "loss": 1.4109,
      "theoretical_loss": 3.4718181291449692,
      "tokens_seen": 1911947264
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004250551654964895,
      "loss": 1.4056,
      "theoretical_loss": 3.471808201094554,
      "tokens_seen": 1912012800
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042503510531594784,
      "loss": 1.6166,
      "theoretical_loss": 3.4717982734797044,
      "tokens_seen": 1912078336
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004250150451354062,
      "loss": 1.3416,
      "theoretical_loss": 3.4717883463003867,
      "tokens_seen": 1912143872
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004249949849548646,
      "loss": 1.3533,
      "theoretical_loss": 3.471778419556566,
      "tokens_seen": 1912209408
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000424974924774323,
      "loss": 1.5119,
      "theoretical_loss": 3.471768493248209,
      "tokens_seen": 1912274944
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042495486459378135,
      "loss": 1.5352,
      "theoretical_loss": 3.4717585673752813,
      "tokens_seen": 1912340480
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004249348044132397,
      "loss": 1.3675,
      "theoretical_loss": 3.471748641937749,
      "tokens_seen": 1912406016
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004249147442326981,
      "loss": 1.5666,
      "theoretical_loss": 3.4717387169355787,
      "tokens_seen": 1912471552
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004248946840521565,
      "loss": 1.3823,
      "theoretical_loss": 3.4717287923687348,
      "tokens_seen": 1912537088
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042487462387161485,
      "loss": 1.4247,
      "theoretical_loss": 3.4717188682371853,
      "tokens_seen": 1912602624
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004248545636910732,
      "loss": 1.298,
      "theoretical_loss": 3.471708944540895,
      "tokens_seen": 1912668160
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004248345035105316,
      "loss": 1.4758,
      "theoretical_loss": 3.4716990212798295,
      "tokens_seen": 1912733696
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042481444332999,
      "loss": 1.4387,
      "theoretical_loss": 3.4716890984539557,
      "tokens_seen": 1912799232
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042479438314944835,
      "loss": 1.3989,
      "theoretical_loss": 3.471679176063239,
      "tokens_seen": 1912864768
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004247743229689067,
      "loss": 1.3997,
      "theoretical_loss": 3.4716692541076464,
      "tokens_seen": 1912930304
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004247542627883651,
      "loss": 1.3466,
      "theoretical_loss": 3.4716593325871425,
      "tokens_seen": 1912995840
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004247342026078235,
      "loss": 1.4853,
      "theoretical_loss": 3.4716494115016943,
      "tokens_seen": 1913061376
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042471414242728186,
      "loss": 1.4098,
      "theoretical_loss": 3.4716394908512678,
      "tokens_seen": 1913126912
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004246940822467402,
      "loss": 1.425,
      "theoretical_loss": 3.4716295706358284,
      "tokens_seen": 1913192448
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004246740220661986,
      "loss": 1.2167,
      "theoretical_loss": 3.4716196508553425,
      "tokens_seen": 1913257984
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000424653961885657,
      "loss": 1.3475,
      "theoretical_loss": 3.4716097315097763,
      "tokens_seen": 1913323520
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042463390170511536,
      "loss": 1.4807,
      "theoretical_loss": 3.471599812599095,
      "tokens_seen": 1913389056
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004246138415245737,
      "loss": 1.4491,
      "theoretical_loss": 3.471589894123266,
      "tokens_seen": 1913454592
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004245937813440321,
      "loss": 1.5587,
      "theoretical_loss": 3.4715799760822548,
      "tokens_seen": 1913520128
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042457372116349044,
      "loss": 1.4179,
      "theoretical_loss": 3.471570058476027,
      "tokens_seen": 1913585664
    },
    {
      "epoch": 0.58,
      "objective/train/advantage_avg": 0.49120426177978516,
      "objective/train/docs_used": 697951,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4675827026367188,
      "objective/train/original_loss": 1.4675827026367188,
      "objective/train/theoretical_loss": 3.4715601413045487,
      "objective/train/tokens_used": 1934111200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24178384244441986,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050349473953247,
      "objective/train/weighted_lm_loss": 1.541298270225525,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.4715601413045487,
      "tokens_seen": 1913651200
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042455366098294886,
      "loss": 1.4664,
      "theoretical_loss": 3.4715601413045487,
      "tokens_seen": 1913651200
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004245336008024072,
      "loss": 1.3563,
      "theoretical_loss": 3.4715502245677867,
      "tokens_seen": 1913716736
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004245135406218656,
      "loss": 1.4123,
      "theoretical_loss": 3.4715403082657064,
      "tokens_seen": 1913782272
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042449348044132395,
      "loss": 1.4518,
      "theoretical_loss": 3.471530392398274,
      "tokens_seen": 1913847808
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042447342026078236,
      "loss": 1.4484,
      "theoretical_loss": 3.4715204769654555,
      "tokens_seen": 1913913344
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004244533600802407,
      "loss": 1.4258,
      "theoretical_loss": 3.471510561967217,
      "tokens_seen": 1913978880
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004244332998996991,
      "loss": 1.4178,
      "theoretical_loss": 3.4715006474035244,
      "tokens_seen": 1914044416
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042441323971915745,
      "loss": 1.4168,
      "theoretical_loss": 3.4714907332743445,
      "tokens_seen": 1914109952
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004243931795386158,
      "loss": 1.477,
      "theoretical_loss": 3.471480819579643,
      "tokens_seen": 1914175488
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042437311935807423,
      "loss": 1.339,
      "theoretical_loss": 3.4714709063193854,
      "tokens_seen": 1914241024
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004243530591775326,
      "loss": 1.3067,
      "theoretical_loss": 3.4714609934935385,
      "tokens_seen": 1914306560
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042433299899699095,
      "loss": 1.4963,
      "theoretical_loss": 3.4714510811020682,
      "tokens_seen": 1914372096
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004243129388164493,
      "loss": 1.5163,
      "theoretical_loss": 3.4714411691449403,
      "tokens_seen": 1914437632
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042429287863590773,
      "loss": 1.5299,
      "theoretical_loss": 3.471431257622121,
      "tokens_seen": 1914503168
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004242728184553661,
      "loss": 1.3763,
      "theoretical_loss": 3.4714213465335773,
      "tokens_seen": 1914568704
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042425275827482446,
      "loss": 1.4033,
      "theoretical_loss": 3.471411435879274,
      "tokens_seen": 1914634240
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004242326980942828,
      "loss": 1.2783,
      "theoretical_loss": 3.4714015256591777,
      "tokens_seen": 1914699776
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004242126379137412,
      "loss": 1.4917,
      "theoretical_loss": 3.4713916158732547,
      "tokens_seen": 1914765312
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004241925777331996,
      "loss": 1.4634,
      "theoretical_loss": 3.4713817065214707,
      "tokens_seen": 1914830848
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000424172517552658,
      "loss": 1.4688,
      "theoretical_loss": 3.4713717976037928,
      "tokens_seen": 1914896384
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004241524573721164,
      "loss": 1.4957,
      "theoretical_loss": 3.4713618891201863,
      "tokens_seen": 1914961920
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042413239719157474,
      "loss": 1.4359,
      "theoretical_loss": 3.471351981070617,
      "tokens_seen": 1915027456
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042411233701103316,
      "loss": 1.373,
      "theoretical_loss": 3.4713420734550517,
      "tokens_seen": 1915092992
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004240922768304915,
      "loss": 1.4127,
      "theoretical_loss": 3.4713321662734566,
      "tokens_seen": 1915158528
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004240722166499499,
      "loss": 1.5087,
      "theoretical_loss": 3.471322259525798,
      "tokens_seen": 1915224064
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042405215646940824,
      "loss": 1.482,
      "theoretical_loss": 3.471312353212041,
      "tokens_seen": 1915289600
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004240320962888666,
      "loss": 1.4276,
      "theoretical_loss": 3.4713024473321523,
      "tokens_seen": 1915355136
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000424012036108325,
      "loss": 1.4269,
      "theoretical_loss": 3.4712925418860987,
      "tokens_seen": 1915420672
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004239919759277834,
      "loss": 1.4198,
      "theoretical_loss": 3.4712826368738456,
      "tokens_seen": 1915486208
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042397191574724175,
      "loss": 1.3776,
      "theoretical_loss": 3.471272732295359,
      "tokens_seen": 1915551744
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004239518555667001,
      "loss": 1.4431,
      "theoretical_loss": 3.4712628281506057,
      "tokens_seen": 1915617280
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004239317953861585,
      "loss": 1.3818,
      "theoretical_loss": 3.4712529244395514,
      "tokens_seen": 1915682816
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004239117352056169,
      "loss": 1.4656,
      "theoretical_loss": 3.471243021162163,
      "tokens_seen": 1915748352
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042389167502507525,
      "loss": 1.4408,
      "theoretical_loss": 3.4712331183184055,
      "tokens_seen": 1915813888
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004238716148445336,
      "loss": 1.4088,
      "theoretical_loss": 3.471223215908246,
      "tokens_seen": 1915879424
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000423851554663992,
      "loss": 1.3899,
      "theoretical_loss": 3.471213313931651,
      "tokens_seen": 1915944960
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004238314944834504,
      "loss": 1.3305,
      "theoretical_loss": 3.4712034123885855,
      "tokens_seen": 1916010496
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042381143430290875,
      "loss": 1.2591,
      "theoretical_loss": 3.4711935112790164,
      "tokens_seen": 1916076032
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004237913741223671,
      "loss": 1.4532,
      "theoretical_loss": 3.47118361060291,
      "tokens_seen": 1916141568
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004237713139418255,
      "loss": 1.409,
      "theoretical_loss": 3.4711737103602323,
      "tokens_seen": 1916207104
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004237512537612839,
      "loss": 1.3512,
      "theoretical_loss": 3.471163810550949,
      "tokens_seen": 1916272640
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042373119358074226,
      "loss": 1.4926,
      "theoretical_loss": 3.471153911175027,
      "tokens_seen": 1916338176
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004237111334002006,
      "loss": 1.4279,
      "theoretical_loss": 3.4711440122324326,
      "tokens_seen": 1916403712
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000423691073219659,
      "loss": 1.3795,
      "theoretical_loss": 3.4711341137231315,
      "tokens_seen": 1916469248
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004236710130391174,
      "loss": 1.4696,
      "theoretical_loss": 3.47112421564709,
      "tokens_seen": 1916534784
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042365095285857576,
      "loss": 1.2752,
      "theoretical_loss": 3.4711143180042745,
      "tokens_seen": 1916600320
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004236308926780341,
      "loss": 1.3777,
      "theoretical_loss": 3.471104420794651,
      "tokens_seen": 1916665856
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004236108324974925,
      "loss": 1.4337,
      "theoretical_loss": 3.4710945240181865,
      "tokens_seen": 1916731392
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042359077231695085,
      "loss": 1.4927,
      "theoretical_loss": 3.4710846276748466,
      "tokens_seen": 1916796928
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042357071213640926,
      "loss": 1.3319,
      "theoretical_loss": 3.471074731764597,
      "tokens_seen": 1916862464
    },
    {
      "epoch": 0.58,
      "objective/train/advantage_avg": 0.49333640933036804,
      "objective/train/docs_used": 699035,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2895686626434326,
      "objective/train/original_loss": 1.2895686626434326,
      "objective/train/theoretical_loss": 3.471064836287405,
      "objective/train/tokens_used": 1937388000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24371019005775452,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505725145339966,
      "objective/train/weighted_lm_loss": 1.3547755479812622,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.471064836287405,
      "tokens_seen": 1916928000
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004235506519558676,
      "loss": 1.4438,
      "theoretical_loss": 3.471064836287405,
      "tokens_seen": 1916928000
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000423530591775326,
      "loss": 1.4504,
      "theoretical_loss": 3.4710549412432363,
      "tokens_seen": 1916993536
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042351053159478435,
      "loss": 1.4189,
      "theoretical_loss": 3.4710450466320575,
      "tokens_seen": 1917059072
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042349047141424276,
      "loss": 1.519,
      "theoretical_loss": 3.471035152453834,
      "tokens_seen": 1917124608
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042347041123370113,
      "loss": 1.3534,
      "theoretical_loss": 3.471025258708533,
      "tokens_seen": 1917190144
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004234503510531595,
      "loss": 1.3235,
      "theoretical_loss": 3.47101536539612,
      "tokens_seen": 1917255680
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042343029087261785,
      "loss": 1.4089,
      "theoretical_loss": 3.471005472516562,
      "tokens_seen": 1917321216
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004234102306920762,
      "loss": 1.422,
      "theoretical_loss": 3.470995580069825,
      "tokens_seen": 1917386752
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042339017051153463,
      "loss": 1.4628,
      "theoretical_loss": 3.470985688055875,
      "tokens_seen": 1917452288
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000423370110330993,
      "loss": 1.4247,
      "theoretical_loss": 3.4709757964746784,
      "tokens_seen": 1917517824
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042335005015045135,
      "loss": 1.385,
      "theoretical_loss": 3.4709659053262016,
      "tokens_seen": 1917583360
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004233299899699097,
      "loss": 1.3974,
      "theoretical_loss": 3.470956014610411,
      "tokens_seen": 1917648896
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042330992978936813,
      "loss": 1.5407,
      "theoretical_loss": 3.4709461243272726,
      "tokens_seen": 1917714432
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004232898696088265,
      "loss": 1.4337,
      "theoretical_loss": 3.4709362344767527,
      "tokens_seen": 1917779968
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042326980942828486,
      "loss": 1.5307,
      "theoretical_loss": 3.470926345058818,
      "tokens_seen": 1917845504
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004232497492477432,
      "loss": 1.4549,
      "theoretical_loss": 3.470916456073434,
      "tokens_seen": 1917911040
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004232296890672016,
      "loss": 1.4107,
      "theoretical_loss": 3.4709065675205677,
      "tokens_seen": 1917976576
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042320962888666,
      "loss": 1.3601,
      "theoretical_loss": 3.4708966794001856,
      "tokens_seen": 1918042112
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042318956870611836,
      "loss": 1.3113,
      "theoretical_loss": 3.4708867917122532,
      "tokens_seen": 1918107648
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004231695085255767,
      "loss": 1.5084,
      "theoretical_loss": 3.4708769044567376,
      "tokens_seen": 1918173184
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004231494483450351,
      "loss": 1.3959,
      "theoretical_loss": 3.470867017633604,
      "tokens_seen": 1918238720
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004231293881644935,
      "loss": 1.324,
      "theoretical_loss": 3.47085713124282,
      "tokens_seen": 1918304256
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042310932798395186,
      "loss": 1.5313,
      "theoretical_loss": 3.470847245284351,
      "tokens_seen": 1918369792
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004230892678034102,
      "loss": 1.3635,
      "theoretical_loss": 3.470837359758164,
      "tokens_seen": 1918435328
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004230692076228686,
      "loss": 1.4401,
      "theoretical_loss": 3.470827474664225,
      "tokens_seen": 1918500864
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042304914744232695,
      "loss": 1.348,
      "theoretical_loss": 3.4708175900025005,
      "tokens_seen": 1918566400
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042302908726178537,
      "loss": 1.3823,
      "theoretical_loss": 3.4708077057729567,
      "tokens_seen": 1918631936
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042300902708124373,
      "loss": 1.4714,
      "theoretical_loss": 3.4707978219755597,
      "tokens_seen": 1918697472
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004229889669007021,
      "loss": 1.4766,
      "theoretical_loss": 3.4707879386102762,
      "tokens_seen": 1918763008
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042296890672016045,
      "loss": 1.3831,
      "theoretical_loss": 3.470778055677073,
      "tokens_seen": 1918828544
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042294884653961887,
      "loss": 1.3729,
      "theoretical_loss": 3.4707681731759155,
      "tokens_seen": 1918894080
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042292878635907723,
      "loss": 1.4214,
      "theoretical_loss": 3.4707582911067703,
      "tokens_seen": 1918959616
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004229087261785356,
      "loss": 1.4335,
      "theoretical_loss": 3.470748409469604,
      "tokens_seen": 1919025152
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042288866599799396,
      "loss": 1.4138,
      "theoretical_loss": 3.4707385282643832,
      "tokens_seen": 1919090688
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004228686058174523,
      "loss": 1.352,
      "theoretical_loss": 3.4707286474910735,
      "tokens_seen": 1919156224
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042284854563691074,
      "loss": 1.3398,
      "theoretical_loss": 3.470718767149642,
      "tokens_seen": 1919221760
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004228284854563691,
      "loss": 1.372,
      "theoretical_loss": 3.4707088872400544,
      "tokens_seen": 1919287296
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042280842527582746,
      "loss": 1.4037,
      "theoretical_loss": 3.4706990077622777,
      "tokens_seen": 1919352832
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004227883650952858,
      "loss": 1.3822,
      "theoretical_loss": 3.4706891287162787,
      "tokens_seen": 1919418368
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042276830491474424,
      "loss": 1.3759,
      "theoretical_loss": 3.4706792501020223,
      "tokens_seen": 1919483904
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004227482447342026,
      "loss": 1.5474,
      "theoretical_loss": 3.4706693719194766,
      "tokens_seen": 1919549440
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042272818455366096,
      "loss": 1.3759,
      "theoretical_loss": 3.4706594941686064,
      "tokens_seen": 1919614976
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004227081243731193,
      "loss": 1.4302,
      "theoretical_loss": 3.470649616849379,
      "tokens_seen": 1919680512
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004226880641925777,
      "loss": 1.3024,
      "theoretical_loss": 3.4706397399617606,
      "tokens_seen": 1919746048
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042266800401203616,
      "loss": 1.4292,
      "theoretical_loss": 3.4706298635057182,
      "tokens_seen": 1919811584
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004226479438314945,
      "loss": 1.4443,
      "theoretical_loss": 3.470619987481217,
      "tokens_seen": 1919877120
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004226278836509529,
      "loss": 1.4754,
      "theoretical_loss": 3.4706101118882247,
      "tokens_seen": 1919942656
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042260782347041125,
      "loss": 1.4254,
      "theoretical_loss": 3.470600236726707,
      "tokens_seen": 1920008192
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042258776328986966,
      "loss": 1.4354,
      "theoretical_loss": 3.4705903619966296,
      "tokens_seen": 1920073728
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000422567703109328,
      "loss": 1.4485,
      "theoretical_loss": 3.4705804876979602,
      "tokens_seen": 1920139264
    },
    {
      "epoch": 0.58,
      "objective/train/advantage_avg": 0.49608397483825684,
      "objective/train/docs_used": 700307,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4548554420471191,
      "objective/train/original_loss": 1.4548554420471191,
      "objective/train/theoretical_loss": 3.4705706138306653,
      "objective/train/tokens_used": 1940664800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2463647723197937,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508610010147095,
      "objective/train/weighted_lm_loss": 1.528975009918213,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.4705706138306653,
      "tokens_seen": 1920204800
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004225476429287864,
      "loss": 1.5317,
      "theoretical_loss": 3.4705706138306653,
      "tokens_seen": 1920204800
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042252758274824475,
      "loss": 1.4188,
      "theoretical_loss": 3.4705607403947107,
      "tokens_seen": 1920270336
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042250752256770317,
      "loss": 1.4227,
      "theoretical_loss": 3.4705508673900622,
      "tokens_seen": 1920335872
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042248746238716153,
      "loss": 1.4267,
      "theoretical_loss": 3.470540994816688,
      "tokens_seen": 1920401408
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004224674022066199,
      "loss": 1.4246,
      "theoretical_loss": 3.4705311226745525,
      "tokens_seen": 1920466944
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042244734202607825,
      "loss": 1.4445,
      "theoretical_loss": 3.470521250963624,
      "tokens_seen": 1920532480
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004224272818455366,
      "loss": 1.3352,
      "theoretical_loss": 3.470511379683868,
      "tokens_seen": 1920598016
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042240722166499503,
      "loss": 1.3258,
      "theoretical_loss": 3.470501508835251,
      "tokens_seen": 1920663552
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004223871614844534,
      "loss": 1.516,
      "theoretical_loss": 3.470491638417739,
      "tokens_seen": 1920729088
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042236710130391176,
      "loss": 1.327,
      "theoretical_loss": 3.4704817684313,
      "tokens_seen": 1920794624
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004223470411233701,
      "loss": 1.3227,
      "theoretical_loss": 3.470471898875899,
      "tokens_seen": 1920860160
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042232698094282853,
      "loss": 1.4738,
      "theoretical_loss": 3.470462029751503,
      "tokens_seen": 1920925696
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004223069207622869,
      "loss": 1.3961,
      "theoretical_loss": 3.470452161058078,
      "tokens_seen": 1920991232
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042228686058174526,
      "loss": 1.4635,
      "theoretical_loss": 3.4704422927955916,
      "tokens_seen": 1921056768
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004222668004012036,
      "loss": 1.2602,
      "theoretical_loss": 3.4704324249640086,
      "tokens_seen": 1921122304
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000422246740220662,
      "loss": 1.5033,
      "theoretical_loss": 3.4704225575632974,
      "tokens_seen": 1921187840
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004222266800401204,
      "loss": 1.5195,
      "theoretical_loss": 3.470412690593423,
      "tokens_seen": 1921253376
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042220661985957876,
      "loss": 1.3375,
      "theoretical_loss": 3.4704028240543527,
      "tokens_seen": 1921318912
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004221865596790371,
      "loss": 1.3774,
      "theoretical_loss": 3.470392957946053,
      "tokens_seen": 1921384448
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004221664994984955,
      "loss": 1.3552,
      "theoretical_loss": 3.4703830922684897,
      "tokens_seen": 1921449984
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004221464393179539,
      "loss": 1.4443,
      "theoretical_loss": 3.4703732270216303,
      "tokens_seen": 1921515520
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042212637913741226,
      "loss": 1.5174,
      "theoretical_loss": 3.4703633622054406,
      "tokens_seen": 1921581056
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004221063189568706,
      "loss": 1.2861,
      "theoretical_loss": 3.470353497819887,
      "tokens_seen": 1921646592
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000422086258776329,
      "loss": 1.4304,
      "theoretical_loss": 3.4703436338649363,
      "tokens_seen": 1921712128
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042206619859578735,
      "loss": 1.3216,
      "theoretical_loss": 3.470333770340555,
      "tokens_seen": 1921777664
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042204613841524577,
      "loss": 1.4052,
      "theoretical_loss": 3.47032390724671,
      "tokens_seen": 1921843200
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042202607823470413,
      "loss": 1.411,
      "theoretical_loss": 3.4703140445833673,
      "tokens_seen": 1921908736
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004220060180541625,
      "loss": 1.4203,
      "theoretical_loss": 3.4703041823504934,
      "tokens_seen": 1921974272
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042198595787362085,
      "loss": 1.3977,
      "theoretical_loss": 3.4702943205480548,
      "tokens_seen": 1922039808
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042196589769307927,
      "loss": 1.4659,
      "theoretical_loss": 3.4702844591760185,
      "tokens_seen": 1922105344
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042194583751253763,
      "loss": 1.4698,
      "theoretical_loss": 3.4702745982343512,
      "tokens_seen": 1922170880
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000421925777331996,
      "loss": 1.29,
      "theoretical_loss": 3.470264737723019,
      "tokens_seen": 1922236416
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042190571715145436,
      "loss": 1.3899,
      "theoretical_loss": 3.470254877641988,
      "tokens_seen": 1922301952
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004218856569709127,
      "loss": 1.4389,
      "theoretical_loss": 3.470245017991225,
      "tokens_seen": 1922367488
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042186559679037114,
      "loss": 1.4468,
      "theoretical_loss": 3.4702351587706977,
      "tokens_seen": 1922433024
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004218455366098295,
      "loss": 1.4163,
      "theoretical_loss": 3.4702252999803713,
      "tokens_seen": 1922498560
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042182547642928786,
      "loss": 1.3815,
      "theoretical_loss": 3.470215441620213,
      "tokens_seen": 1922564096
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004218054162487462,
      "loss": 1.4501,
      "theoretical_loss": 3.4702055836901886,
      "tokens_seen": 1922629632
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042178535606820464,
      "loss": 1.5005,
      "theoretical_loss": 3.470195726190266,
      "tokens_seen": 1922695168
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000421765295887663,
      "loss": 1.4971,
      "theoretical_loss": 3.470185869120411,
      "tokens_seen": 1922760704
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042174523570712136,
      "loss": 1.4655,
      "theoretical_loss": 3.47017601248059,
      "tokens_seen": 1922826240
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004217251755265797,
      "loss": 1.4828,
      "theoretical_loss": 3.4701661562707695,
      "tokens_seen": 1922891776
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004217051153460381,
      "loss": 1.3273,
      "theoretical_loss": 3.470156300490917,
      "tokens_seen": 1922957312
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004216850551654965,
      "loss": 1.4219,
      "theoretical_loss": 3.4701464451409985,
      "tokens_seen": 1923022848
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042166499498495487,
      "loss": 1.3857,
      "theoretical_loss": 3.4701365902209798,
      "tokens_seen": 1923088384
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042164493480441323,
      "loss": 1.3922,
      "theoretical_loss": 3.4701267357308287,
      "tokens_seen": 1923153920
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004216248746238716,
      "loss": 1.4635,
      "theoretical_loss": 3.4701168816705117,
      "tokens_seen": 1923219456
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042160481444333,
      "loss": 1.4296,
      "theoretical_loss": 3.470107028039995,
      "tokens_seen": 1923284992
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042158475426278837,
      "loss": 1.5832,
      "theoretical_loss": 3.4700971748392453,
      "tokens_seen": 1923350528
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042156469408224673,
      "loss": 1.4333,
      "theoretical_loss": 3.4700873220682285,
      "tokens_seen": 1923416064
    },
    {
      "epoch": 0.58,
      "objective/train/advantage_avg": 0.49544912576675415,
      "objective/train/docs_used": 701572,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3334323167800903,
      "objective/train/original_loss": 1.3334321975708008,
      "objective/train/theoretical_loss": 3.4700774697269128,
      "objective/train/tokens_used": 1943941600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24565473198890686,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507937669754028,
      "objective/train/weighted_lm_loss": 1.4011366367340088,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4700774697269128,
      "tokens_seen": 1923481600
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004215446339017051,
      "loss": 1.4117,
      "theoretical_loss": 3.4700774697269128,
      "tokens_seen": 1923481600
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004215245737211635,
      "loss": 1.3324,
      "theoretical_loss": 3.4700676178152636,
      "tokens_seen": 1923547136
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004215045135406219,
      "loss": 1.4414,
      "theoretical_loss": 3.470057766333248,
      "tokens_seen": 1923612672
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042148445336008024,
      "loss": 1.4241,
      "theoretical_loss": 3.4700479152808326,
      "tokens_seen": 1923678208
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004214643931795386,
      "loss": 1.3876,
      "theoretical_loss": 3.4700380646579836,
      "tokens_seen": 1923743744
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042144433299899696,
      "loss": 1.4274,
      "theoretical_loss": 3.470028214464668,
      "tokens_seen": 1923809280
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004214242728184554,
      "loss": 1.6239,
      "theoretical_loss": 3.4700183647008522,
      "tokens_seen": 1923874816
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042140421263791374,
      "loss": 1.3366,
      "theoretical_loss": 3.4700085153665032,
      "tokens_seen": 1923940352
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004213841524573721,
      "loss": 1.4267,
      "theoretical_loss": 3.469998666461588,
      "tokens_seen": 1924005888
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042136409227683046,
      "loss": 1.4595,
      "theoretical_loss": 3.4699888179860725,
      "tokens_seen": 1924071424
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004213440320962889,
      "loss": 1.3769,
      "theoretical_loss": 3.4699789699399233,
      "tokens_seen": 1924136960
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042132397191574724,
      "loss": 1.4398,
      "theoretical_loss": 3.4699691223231075,
      "tokens_seen": 1924202496
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004213039117352056,
      "loss": 1.3267,
      "theoretical_loss": 3.469959275135592,
      "tokens_seen": 1924268032
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042128385155466397,
      "loss": 1.2961,
      "theoretical_loss": 3.469949428377342,
      "tokens_seen": 1924333568
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042126379137412233,
      "loss": 1.3908,
      "theoretical_loss": 3.469939582048326,
      "tokens_seen": 1924399104
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042124373119358075,
      "loss": 1.3672,
      "theoretical_loss": 3.4699297361485097,
      "tokens_seen": 1924464640
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004212236710130391,
      "loss": 1.3621,
      "theoretical_loss": 3.4699198906778603,
      "tokens_seen": 1924530176
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042120361083249747,
      "loss": 1.3861,
      "theoretical_loss": 3.469910045636344,
      "tokens_seen": 1924595712
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042118355065195583,
      "loss": 1.4873,
      "theoretical_loss": 3.469900201023928,
      "tokens_seen": 1924661248
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004211634904714143,
      "loss": 1.216,
      "theoretical_loss": 3.4698903568405783,
      "tokens_seen": 1924726784
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042114343029087266,
      "loss": 1.4207,
      "theoretical_loss": 3.469880513086262,
      "tokens_seen": 1924792320
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042112337011033103,
      "loss": 1.3825,
      "theoretical_loss": 3.4698706697609456,
      "tokens_seen": 1924857856
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004211033099297894,
      "loss": 1.3821,
      "theoretical_loss": 3.469860826864596,
      "tokens_seen": 1924923392
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042108324974924775,
      "loss": 1.4138,
      "theoretical_loss": 3.46985098439718,
      "tokens_seen": 1924988928
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042106318956870617,
      "loss": 1.4852,
      "theoretical_loss": 3.469841142358664,
      "tokens_seen": 1925054464
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042104312938816453,
      "loss": 1.3801,
      "theoretical_loss": 3.469831300749015,
      "tokens_seen": 1925120000
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004210230692076229,
      "loss": 1.4421,
      "theoretical_loss": 3.4698214595681995,
      "tokens_seen": 1925185536
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042100300902708125,
      "loss": 1.3938,
      "theoretical_loss": 3.469811618816184,
      "tokens_seen": 1925251072
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042098294884653967,
      "loss": 1.3519,
      "theoretical_loss": 3.469801778492936,
      "tokens_seen": 1925316608
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042096288866599803,
      "loss": 1.3983,
      "theoretical_loss": 3.469791938598421,
      "tokens_seen": 1925382144
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004209428284854564,
      "loss": 1.4874,
      "theoretical_loss": 3.4697820991326074,
      "tokens_seen": 1925447680
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042092276830491476,
      "loss": 1.3385,
      "theoretical_loss": 3.46977226009546,
      "tokens_seen": 1925513216
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004209027081243731,
      "loss": 1.3673,
      "theoretical_loss": 3.4697624214869474,
      "tokens_seen": 1925578752
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042088264794383154,
      "loss": 1.367,
      "theoretical_loss": 3.469752583307035,
      "tokens_seen": 1925644288
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004208625877632899,
      "loss": 1.3745,
      "theoretical_loss": 3.46974274555569,
      "tokens_seen": 1925709824
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042084252758274826,
      "loss": 1.4176,
      "theoretical_loss": 3.469732908232879,
      "tokens_seen": 1925775360
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004208224674022066,
      "loss": 1.4398,
      "theoretical_loss": 3.4697230713385694,
      "tokens_seen": 1925840896
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042080240722166504,
      "loss": 1.4835,
      "theoretical_loss": 3.469713234872727,
      "tokens_seen": 1925906432
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004207823470411234,
      "loss": 1.4607,
      "theoretical_loss": 3.4697033988353194,
      "tokens_seen": 1925971968
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042076228686058176,
      "loss": 1.4203,
      "theoretical_loss": 3.4696935632263126,
      "tokens_seen": 1926037504
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004207422266800401,
      "loss": 1.342,
      "theoretical_loss": 3.469683728045674,
      "tokens_seen": 1926103040
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004207221664994985,
      "loss": 1.3981,
      "theoretical_loss": 3.46967389329337,
      "tokens_seen": 1926168576
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004207021063189569,
      "loss": 1.3476,
      "theoretical_loss": 3.4696640589693675,
      "tokens_seen": 1926234112
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042068204613841527,
      "loss": 1.4264,
      "theoretical_loss": 3.469654225073633,
      "tokens_seen": 1926299648
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042066198595787363,
      "loss": 1.3943,
      "theoretical_loss": 3.469644391606134,
      "tokens_seen": 1926365184
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000420641925777332,
      "loss": 1.3849,
      "theoretical_loss": 3.4696345585668364,
      "tokens_seen": 1926430720
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004206218655967904,
      "loss": 1.4138,
      "theoretical_loss": 3.4696247259557076,
      "tokens_seen": 1926496256
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042060180541624877,
      "loss": 1.5058,
      "theoretical_loss": 3.4696148937727145,
      "tokens_seen": 1926561792
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042058174523570713,
      "loss": 1.4074,
      "theoretical_loss": 3.4696050620178234,
      "tokens_seen": 1926627328
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004205616850551655,
      "loss": 1.4678,
      "theoretical_loss": 3.469595230691001,
      "tokens_seen": 1926692864
    },
    {
      "epoch": 0.58,
      "objective/train/advantage_avg": 0.49047499895095825,
      "objective/train/docs_used": 702709,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.55850350856781,
      "objective/train/original_loss": 1.5585033893585205,
      "objective/train/theoretical_loss": 3.469585399792215,
      "objective/train/tokens_used": 1947218400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2415967434644699,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502758026123047,
      "objective/train/weighted_lm_loss": 1.6368056535720825,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.469585399792215,
      "tokens_seen": 1926758400
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004205416248746239,
      "loss": 1.4634,
      "theoretical_loss": 3.469585399792215,
      "tokens_seen": 1926758400
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004205215646940823,
      "loss": 1.4098,
      "theoretical_loss": 3.4695755693214307,
      "tokens_seen": 1926823936
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042050150451354064,
      "loss": 1.4036,
      "theoretical_loss": 3.4695657392786163,
      "tokens_seen": 1926889472
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000420481444332999,
      "loss": 1.2789,
      "theoretical_loss": 3.469555909663738,
      "tokens_seen": 1926955008
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042046138415245736,
      "loss": 1.4303,
      "theoretical_loss": 3.4695460804767633,
      "tokens_seen": 1927020544
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004204413239719158,
      "loss": 1.4668,
      "theoretical_loss": 3.469536251717658,
      "tokens_seen": 1927086080
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042042126379137414,
      "loss": 1.3097,
      "theoretical_loss": 3.469526423386389,
      "tokens_seen": 1927151616
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004204012036108325,
      "loss": 1.497,
      "theoretical_loss": 3.4695165954829243,
      "tokens_seen": 1927217152
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042038114343029086,
      "loss": 1.4724,
      "theoretical_loss": 3.4695067680072293,
      "tokens_seen": 1927282688
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004203610832497493,
      "loss": 1.497,
      "theoretical_loss": 3.4694969409592717,
      "tokens_seen": 1927348224
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042034102306920764,
      "loss": 1.4562,
      "theoretical_loss": 3.4694871143390182,
      "tokens_seen": 1927413760
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000420320962888666,
      "loss": 1.4742,
      "theoretical_loss": 3.469477288146436,
      "tokens_seen": 1927479296
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042030090270812437,
      "loss": 1.4508,
      "theoretical_loss": 3.469467462381491,
      "tokens_seen": 1927544832
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042028084252758273,
      "loss": 1.3142,
      "theoretical_loss": 3.4694576370441506,
      "tokens_seen": 1927610368
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042026078234704115,
      "loss": 1.4416,
      "theoretical_loss": 3.4694478121343817,
      "tokens_seen": 1927675904
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004202407221664995,
      "loss": 1.3816,
      "theoretical_loss": 3.4694379876521513,
      "tokens_seen": 1927741440
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042022066198595787,
      "loss": 1.4999,
      "theoretical_loss": 3.469428163597426,
      "tokens_seen": 1927806976
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042020060180541623,
      "loss": 1.4001,
      "theoretical_loss": 3.469418339970172,
      "tokens_seen": 1927872512
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042018054162487465,
      "loss": 1.4413,
      "theoretical_loss": 3.4694085167703577,
      "tokens_seen": 1927938048
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000420160481444333,
      "loss": 1.4044,
      "theoretical_loss": 3.469398693997949,
      "tokens_seen": 1928003584
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004201404212637914,
      "loss": 1.3392,
      "theoretical_loss": 3.4693888716529124,
      "tokens_seen": 1928069120
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042012036108324974,
      "loss": 1.4023,
      "theoretical_loss": 3.469379049735216,
      "tokens_seen": 1928134656
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004201003009027081,
      "loss": 1.5465,
      "theoretical_loss": 3.469369228244825,
      "tokens_seen": 1928200192
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004200802407221665,
      "loss": 1.4235,
      "theoretical_loss": 3.469359407181708,
      "tokens_seen": 1928265728
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004200601805416249,
      "loss": 1.4611,
      "theoretical_loss": 3.4693495865458313,
      "tokens_seen": 1928331264
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042004012036108324,
      "loss": 1.4147,
      "theoretical_loss": 3.4693397663371615,
      "tokens_seen": 1928396800
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004200200601805416,
      "loss": 1.3464,
      "theoretical_loss": 3.4693299465556655,
      "tokens_seen": 1928462336
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00042,
      "loss": 1.4893,
      "theoretical_loss": 3.4693201272013106,
      "tokens_seen": 1928527872
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004199799398194584,
      "loss": 1.3873,
      "theoretical_loss": 3.4693103082740633,
      "tokens_seen": 1928593408
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041995987963891674,
      "loss": 1.547,
      "theoretical_loss": 3.4693004897738904,
      "tokens_seen": 1928658944
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004199398194583751,
      "loss": 1.4923,
      "theoretical_loss": 3.4692906717007594,
      "tokens_seen": 1928724480
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041991975927783347,
      "loss": 1.495,
      "theoretical_loss": 3.469280854054637,
      "tokens_seen": 1928790016
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004198996990972919,
      "loss": 1.4089,
      "theoretical_loss": 3.4692710368354898,
      "tokens_seen": 1928855552
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041987963891675025,
      "loss": 1.501,
      "theoretical_loss": 3.4692612200432853,
      "tokens_seen": 1928921088
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004198595787362086,
      "loss": 1.4119,
      "theoretical_loss": 3.4692514036779896,
      "tokens_seen": 1928986624
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041983951855566697,
      "loss": 1.4191,
      "theoretical_loss": 3.46924158773957,
      "tokens_seen": 1929052160
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004198194583751254,
      "loss": 1.3346,
      "theoretical_loss": 3.469231772227994,
      "tokens_seen": 1929117696
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041979939819458375,
      "loss": 1.3208,
      "theoretical_loss": 3.4692219571432283,
      "tokens_seen": 1929183232
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004197793380140421,
      "loss": 1.3347,
      "theoretical_loss": 3.4692121424852393,
      "tokens_seen": 1929248768
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041975927783350047,
      "loss": 1.4244,
      "theoretical_loss": 3.469202328253994,
      "tokens_seen": 1929314304
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041973921765295883,
      "loss": 1.4326,
      "theoretical_loss": 3.4691925144494604,
      "tokens_seen": 1929379840
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041971915747241725,
      "loss": 1.5762,
      "theoretical_loss": 3.469182701071604,
      "tokens_seen": 1929445376
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004196990972918756,
      "loss": 1.5049,
      "theoretical_loss": 3.4691728881203927,
      "tokens_seen": 1929510912
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.000419679037111334,
      "loss": 1.3679,
      "theoretical_loss": 3.4691630755957927,
      "tokens_seen": 1929576448
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004196589769307924,
      "loss": 1.488,
      "theoretical_loss": 3.4691532634977724,
      "tokens_seen": 1929641984
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004196389167502508,
      "loss": 1.4084,
      "theoretical_loss": 3.469143451826297,
      "tokens_seen": 1929707520
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041961885656970917,
      "loss": 1.4561,
      "theoretical_loss": 3.469133640581335,
      "tokens_seen": 1929773056
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041959879638916753,
      "loss": 1.5052,
      "theoretical_loss": 3.469123829762852,
      "tokens_seen": 1929838592
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004195787362086259,
      "loss": 1.3491,
      "theoretical_loss": 3.469114019370816,
      "tokens_seen": 1929904128
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004195586760280843,
      "loss": 1.3479,
      "theoretical_loss": 3.4691042094051934,
      "tokens_seen": 1929969664
    },
    {
      "epoch": 0.58,
      "objective/train/advantage_avg": 0.4927586615085602,
      "objective/train/docs_used": 703903,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.386762022972107,
      "objective/train/original_loss": 1.3867621421813965,
      "objective/train/theoretical_loss": 3.4690943998659516,
      "objective/train/tokens_used": 1950495200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24326469004154205,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505125522613525,
      "objective/train/weighted_lm_loss": 1.4566758871078491,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4690943998659516,
      "tokens_seen": 1930035200
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004195386158475427,
      "loss": 1.4286,
      "theoretical_loss": 3.4690943998659516,
      "tokens_seen": 1930035200
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041951855566700104,
      "loss": 1.3906,
      "theoretical_loss": 3.4690845907530576,
      "tokens_seen": 1930100736
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004194984954864594,
      "loss": 1.5118,
      "theoretical_loss": 3.4690747820664782,
      "tokens_seen": 1930166272
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041947843530591776,
      "loss": 1.4682,
      "theoretical_loss": 3.46906497380618,
      "tokens_seen": 1930231808
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004194583751253762,
      "loss": 1.2619,
      "theoretical_loss": 3.4690551659721307,
      "tokens_seen": 1930297344
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041943831494483454,
      "loss": 1.4955,
      "theoretical_loss": 3.469045358564297,
      "tokens_seen": 1930362880
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0004194182547642929,
      "loss": 1.4005,
      "theoretical_loss": 3.469035551582646,
      "tokens_seen": 1930428416
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00041939819458375126,
      "loss": 1.4188,
      "theoretical_loss": 3.4690257450271442,
      "tokens_seen": 1930493952
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004193781344032097,
      "loss": 1.4737,
      "theoretical_loss": 3.4690159388977593,
      "tokens_seen": 1930559488
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041935807422266804,
      "loss": 1.3788,
      "theoretical_loss": 3.4690061331944584,
      "tokens_seen": 1930625024
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004193380140421264,
      "loss": 1.3751,
      "theoretical_loss": 3.468996327917208,
      "tokens_seen": 1930690560
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041931795386158477,
      "loss": 1.4781,
      "theoretical_loss": 3.4689865230659755,
      "tokens_seen": 1930756096
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041929789368104313,
      "loss": 1.4408,
      "theoretical_loss": 3.4689767186407274,
      "tokens_seen": 1930821632
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041927783350050155,
      "loss": 1.3744,
      "theoretical_loss": 3.4689669146414315,
      "tokens_seen": 1930887168
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004192577733199599,
      "loss": 1.468,
      "theoretical_loss": 3.468957111068054,
      "tokens_seen": 1930952704
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041923771313941827,
      "loss": 1.432,
      "theoretical_loss": 3.468947307920563,
      "tokens_seen": 1931018240
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041921765295887663,
      "loss": 1.3018,
      "theoretical_loss": 3.468937505198924,
      "tokens_seen": 1931083776
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041919759277833505,
      "loss": 1.4616,
      "theoretical_loss": 3.4689277029031054,
      "tokens_seen": 1931149312
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004191775325977934,
      "loss": 1.3599,
      "theoretical_loss": 3.468917901033074,
      "tokens_seen": 1931214848
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004191574724172518,
      "loss": 1.4295,
      "theoretical_loss": 3.4689080995887966,
      "tokens_seen": 1931280384
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041913741223671014,
      "loss": 1.4572,
      "theoretical_loss": 3.4688982985702403,
      "tokens_seen": 1931345920
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004191173520561685,
      "loss": 1.4583,
      "theoretical_loss": 3.468888497977372,
      "tokens_seen": 1931411456
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004190972918756269,
      "loss": 1.4229,
      "theoretical_loss": 3.468878697810159,
      "tokens_seen": 1931476992
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004190772316950853,
      "loss": 1.4538,
      "theoretical_loss": 3.468868898068569,
      "tokens_seen": 1931542528
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041905717151454364,
      "loss": 1.4545,
      "theoretical_loss": 3.468859098752568,
      "tokens_seen": 1931608064
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000419037111334002,
      "loss": 1.4645,
      "theoretical_loss": 3.468849299862123,
      "tokens_seen": 1931673600
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004190170511534604,
      "loss": 1.4356,
      "theoretical_loss": 3.468839501397202,
      "tokens_seen": 1931739136
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004189969909729188,
      "loss": 1.5123,
      "theoretical_loss": 3.4688297033577715,
      "tokens_seen": 1931804672
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041897693079237714,
      "loss": 1.3838,
      "theoretical_loss": 3.468819905743799,
      "tokens_seen": 1931870208
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004189568706118355,
      "loss": 1.4746,
      "theoretical_loss": 3.4688101085552514,
      "tokens_seen": 1931935744
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041893681043129387,
      "loss": 1.443,
      "theoretical_loss": 3.4688003117920956,
      "tokens_seen": 1932001280
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004189167502507523,
      "loss": 1.4427,
      "theoretical_loss": 3.468790515454299,
      "tokens_seen": 1932066816
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041889669007021065,
      "loss": 1.4386,
      "theoretical_loss": 3.4687807195418285,
      "tokens_seen": 1932132352
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000418876629889669,
      "loss": 1.3446,
      "theoretical_loss": 3.4687709240546507,
      "tokens_seen": 1932197888
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041885656970912737,
      "loss": 1.3529,
      "theoretical_loss": 3.4687611289927336,
      "tokens_seen": 1932263424
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004188365095285858,
      "loss": 1.3429,
      "theoretical_loss": 3.4687513343560443,
      "tokens_seen": 1932328960
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041881644934804415,
      "loss": 1.4134,
      "theoretical_loss": 3.468741540144549,
      "tokens_seen": 1932394496
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004187963891675025,
      "loss": 1.4475,
      "theoretical_loss": 3.468731746358216,
      "tokens_seen": 1932460032
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004187763289869609,
      "loss": 1.4613,
      "theoretical_loss": 3.468721952997012,
      "tokens_seen": 1932525568
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041875626880641924,
      "loss": 1.3835,
      "theoretical_loss": 3.4687121600609037,
      "tokens_seen": 1932591104
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041873620862587765,
      "loss": 1.3563,
      "theoretical_loss": 3.4687023675498585,
      "tokens_seen": 1932656640
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000418716148445336,
      "loss": 1.4552,
      "theoretical_loss": 3.4686925754638436,
      "tokens_seen": 1932722176
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004186960882647944,
      "loss": 1.4825,
      "theoretical_loss": 3.4686827838028256,
      "tokens_seen": 1932787712
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041867602808425274,
      "loss": 1.4656,
      "theoretical_loss": 3.4686729925667725,
      "tokens_seen": 1932853248
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041865596790371115,
      "loss": 1.4734,
      "theoretical_loss": 3.4686632017556516,
      "tokens_seen": 1932918784
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004186359077231695,
      "loss": 1.574,
      "theoretical_loss": 3.468653411369429,
      "tokens_seen": 1932984320
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004186158475426279,
      "loss": 1.2482,
      "theoretical_loss": 3.4686436214080727,
      "tokens_seen": 1933049856
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041859578736208624,
      "loss": 1.2984,
      "theoretical_loss": 3.468633831871549,
      "tokens_seen": 1933115392
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004185757271815446,
      "loss": 1.3233,
      "theoretical_loss": 3.468624042759826,
      "tokens_seen": 1933180928
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000418555667001003,
      "loss": 1.4654,
      "theoretical_loss": 3.4686142540728704,
      "tokens_seen": 1933246464
    },
    {
      "epoch": 0.59,
      "objective/train/advantage_avg": 0.491926908493042,
      "objective/train/docs_used": 705162,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3340891599655151,
      "objective/train/original_loss": 1.3340892791748047,
      "objective/train/theoretical_loss": 3.4686044658106496,
      "objective/train/tokens_used": 1953772000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24255746603012085,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504257678985596,
      "objective/train/weighted_lm_loss": 1.4013338088989258,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4686044658106496,
      "tokens_seen": 1933312000
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004185356068204614,
      "loss": 1.4149,
      "theoretical_loss": 3.4686044658106496,
      "tokens_seen": 1933312000
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041851554663991974,
      "loss": 1.2179,
      "theoretical_loss": 3.46859467797313,
      "tokens_seen": 1933377536
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004184954864593781,
      "loss": 1.3531,
      "theoretical_loss": 3.46858489056028,
      "tokens_seen": 1933443072
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004184754262788365,
      "loss": 1.4316,
      "theoretical_loss": 3.468575103572066,
      "tokens_seen": 1933508608
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004184553660982949,
      "loss": 1.5412,
      "theoretical_loss": 3.468565317008456,
      "tokens_seen": 1933574144
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041843530591775325,
      "loss": 1.4494,
      "theoretical_loss": 3.4685555308694154,
      "tokens_seen": 1933639680
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004184152457372116,
      "loss": 1.609,
      "theoretical_loss": 3.4685457451549135,
      "tokens_seen": 1933705216
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041839518555666997,
      "loss": 1.4731,
      "theoretical_loss": 3.468535959864916,
      "tokens_seen": 1933770752
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004183751253761284,
      "loss": 1.4711,
      "theoretical_loss": 3.4685261749993908,
      "tokens_seen": 1933836288
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041835506519558675,
      "loss": 1.3255,
      "theoretical_loss": 3.468516390558305,
      "tokens_seen": 1933901824
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004183350050150451,
      "loss": 1.4496,
      "theoretical_loss": 3.468506606541626,
      "tokens_seen": 1933967360
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004183149448345035,
      "loss": 1.4003,
      "theoretical_loss": 3.4684968229493203,
      "tokens_seen": 1934032896
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004182948846539619,
      "loss": 1.2854,
      "theoretical_loss": 3.4684870397813556,
      "tokens_seen": 1934098432
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041827482447342025,
      "loss": 1.452,
      "theoretical_loss": 3.4684772570376996,
      "tokens_seen": 1934163968
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004182547642928786,
      "loss": 1.4071,
      "theoretical_loss": 3.4684674747183184,
      "tokens_seen": 1934229504
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000418234704112337,
      "loss": 1.4633,
      "theoretical_loss": 3.46845769282318,
      "tokens_seen": 1934295040
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004182146439317954,
      "loss": 1.4851,
      "theoretical_loss": 3.4684479113522517,
      "tokens_seen": 1934360576
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041819458375125376,
      "loss": 1.3611,
      "theoretical_loss": 3.4684381303055005,
      "tokens_seen": 1934426112
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004181745235707121,
      "loss": 1.46,
      "theoretical_loss": 3.4684283496828936,
      "tokens_seen": 1934491648
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041815446339017054,
      "loss": 1.4239,
      "theoretical_loss": 3.468418569484398,
      "tokens_seen": 1934557184
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004181344032096289,
      "loss": 1.4588,
      "theoretical_loss": 3.4684087897099816,
      "tokens_seen": 1934622720
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004181143430290873,
      "loss": 1.3931,
      "theoretical_loss": 3.4683990103596107,
      "tokens_seen": 1934688256
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004180942828485457,
      "loss": 1.4009,
      "theoretical_loss": 3.468389231433254,
      "tokens_seen": 1934753792
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041807422266800404,
      "loss": 1.3947,
      "theoretical_loss": 3.4683794529308773,
      "tokens_seen": 1934819328
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004180541624874624,
      "loss": 1.4818,
      "theoretical_loss": 3.4683696748524486,
      "tokens_seen": 1934884864
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004180341023069208,
      "loss": 1.4777,
      "theoretical_loss": 3.4683598971979346,
      "tokens_seen": 1934950400
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004180140421263792,
      "loss": 1.4579,
      "theoretical_loss": 3.4683501199673037,
      "tokens_seen": 1935015936
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041799398194583754,
      "loss": 1.4598,
      "theoretical_loss": 3.4683403431605218,
      "tokens_seen": 1935081472
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004179739217652959,
      "loss": 1.4832,
      "theoretical_loss": 3.468330566777557,
      "tokens_seen": 1935147008
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041795386158475427,
      "loss": 1.3494,
      "theoretical_loss": 3.4683207908183764,
      "tokens_seen": 1935212544
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004179338014042127,
      "loss": 1.3788,
      "theoretical_loss": 3.468311015282947,
      "tokens_seen": 1935278080
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041791374122367105,
      "loss": 1.4692,
      "theoretical_loss": 3.4683012401712365,
      "tokens_seen": 1935343616
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004178936810431294,
      "loss": 1.4332,
      "theoretical_loss": 3.4682914654832118,
      "tokens_seen": 1935409152
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041787362086258777,
      "loss": 1.3672,
      "theoretical_loss": 3.4682816912188406,
      "tokens_seen": 1935474688
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004178535606820462,
      "loss": 1.4138,
      "theoretical_loss": 3.4682719173780905,
      "tokens_seen": 1935540224
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041783350050150455,
      "loss": 1.2928,
      "theoretical_loss": 3.468262143960928,
      "tokens_seen": 1935605760
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004178134403209629,
      "loss": 1.3266,
      "theoretical_loss": 3.46825237096732,
      "tokens_seen": 1935671296
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004177933801404213,
      "loss": 1.3693,
      "theoretical_loss": 3.468242598397235,
      "tokens_seen": 1935736832
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041777331995987964,
      "loss": 1.3814,
      "theoretical_loss": 3.46823282625064,
      "tokens_seen": 1935802368
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041775325977933805,
      "loss": 1.4694,
      "theoretical_loss": 3.468223054527502,
      "tokens_seen": 1935867904
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004177331995987964,
      "loss": 1.3413,
      "theoretical_loss": 3.4682132832277883,
      "tokens_seen": 1935933440
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004177131394182548,
      "loss": 1.4282,
      "theoretical_loss": 3.468203512351467,
      "tokens_seen": 1935998976
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041769307923771314,
      "loss": 1.3697,
      "theoretical_loss": 3.468193741898504,
      "tokens_seen": 1936064512
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041767301905717156,
      "loss": 1.4269,
      "theoretical_loss": 3.468183971868868,
      "tokens_seen": 1936130048
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004176529588766299,
      "loss": 1.4894,
      "theoretical_loss": 3.4681742022625253,
      "tokens_seen": 1936195584
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004176328986960883,
      "loss": 1.5301,
      "theoretical_loss": 3.4681644330794437,
      "tokens_seen": 1936261120
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041761283851554664,
      "loss": 1.4571,
      "theoretical_loss": 3.4681546643195906,
      "tokens_seen": 1936326656
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000417592778335005,
      "loss": 1.6328,
      "theoretical_loss": 3.4681448959829337,
      "tokens_seen": 1936392192
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004175727181544634,
      "loss": 1.4787,
      "theoretical_loss": 3.468135128069439,
      "tokens_seen": 1936457728
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004175526579739218,
      "loss": 1.5119,
      "theoretical_loss": 3.4681253605790756,
      "tokens_seen": 1936523264
    },
    {
      "epoch": 0.59,
      "objective/train/advantage_avg": 0.4913966655731201,
      "objective/train/docs_used": 706032,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2860032320022583,
      "objective/train/original_loss": 1.2860031127929688,
      "objective/train/theoretical_loss": 3.4681155935118095,
      "objective/train/tokens_used": 1957048800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24176520109176636,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503686666488647,
      "objective/train/weighted_lm_loss": 1.350783109664917,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 3.4681155935118095,
      "tokens_seen": 1936588800
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041753259779338015,
      "loss": 1.4001,
      "theoretical_loss": 3.4681155935118095,
      "tokens_seen": 1936588800
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004175125376128385,
      "loss": 1.3903,
      "theoretical_loss": 3.4681058268676086,
      "tokens_seen": 1936654336
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004174924774322969,
      "loss": 1.4153,
      "theoretical_loss": 3.4680960606464404,
      "tokens_seen": 1936719872
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004174724172517553,
      "loss": 1.4531,
      "theoretical_loss": 3.4680862948482716,
      "tokens_seen": 1936785408
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041745235707121365,
      "loss": 1.4726,
      "theoretical_loss": 3.4680765294730707,
      "tokens_seen": 1936850944
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000417432296890672,
      "loss": 1.4267,
      "theoretical_loss": 3.4680667645208043,
      "tokens_seen": 1936916480
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041741223671013037,
      "loss": 1.388,
      "theoretical_loss": 3.4680569999914397,
      "tokens_seen": 1936982016
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004173921765295888,
      "loss": 1.3867,
      "theoretical_loss": 3.468047235884945,
      "tokens_seen": 1937047552
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041737211634904715,
      "loss": 1.3831,
      "theoretical_loss": 3.468037472201286,
      "tokens_seen": 1937113088
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004173520561685055,
      "loss": 1.4001,
      "theoretical_loss": 3.468027708940432,
      "tokens_seen": 1937178624
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004173319959879639,
      "loss": 1.4594,
      "theoretical_loss": 3.468017946102349,
      "tokens_seen": 1937244160
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004173119358074223,
      "loss": 1.4221,
      "theoretical_loss": 3.468008183687005,
      "tokens_seen": 1937309696
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041729187562688065,
      "loss": 1.4699,
      "theoretical_loss": 3.467998421694367,
      "tokens_seen": 1937375232
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000417271815446339,
      "loss": 1.4159,
      "theoretical_loss": 3.467988660124403,
      "tokens_seen": 1937440768
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004172517552657974,
      "loss": 1.4401,
      "theoretical_loss": 3.467978898977081,
      "tokens_seen": 1937506304
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004172316950852558,
      "loss": 1.4878,
      "theoretical_loss": 3.467969138252366,
      "tokens_seen": 1937571840
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041721163490471416,
      "loss": 1.4322,
      "theoretical_loss": 3.467959377950228,
      "tokens_seen": 1937637376
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004171915747241725,
      "loss": 1.4686,
      "theoretical_loss": 3.467949618070633,
      "tokens_seen": 1937702912
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004171715145436309,
      "loss": 1.4192,
      "theoretical_loss": 3.4679398586135486,
      "tokens_seen": 1937768448
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041715145436308924,
      "loss": 1.407,
      "theoretical_loss": 3.467930099578943,
      "tokens_seen": 1937833984
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041713139418254766,
      "loss": 1.3358,
      "theoretical_loss": 3.4679203409667823,
      "tokens_seen": 1937899520
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000417111334002006,
      "loss": 1.4435,
      "theoretical_loss": 3.4679105827770345,
      "tokens_seen": 1937965056
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004170912738214644,
      "loss": 1.4279,
      "theoretical_loss": 3.467900825009668,
      "tokens_seen": 1938030592
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041707121364092275,
      "loss": 1.3976,
      "theoretical_loss": 3.467891067664649,
      "tokens_seen": 1938096128
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041705115346038116,
      "loss": 1.5413,
      "theoretical_loss": 3.467881310741945,
      "tokens_seen": 1938161664
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004170310932798395,
      "loss": 1.3903,
      "theoretical_loss": 3.467871554241524,
      "tokens_seen": 1938227200
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004170110330992979,
      "loss": 1.4482,
      "theoretical_loss": 3.4678617981633533,
      "tokens_seen": 1938292736
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041699097291875625,
      "loss": 1.4313,
      "theoretical_loss": 3.4678520425074,
      "tokens_seen": 1938358272
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004169709127382146,
      "loss": 1.433,
      "theoretical_loss": 3.467842287273632,
      "tokens_seen": 1938423808
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041695085255767303,
      "loss": 1.3913,
      "theoretical_loss": 3.4678325324620167,
      "tokens_seen": 1938489344
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004169307923771314,
      "loss": 1.3984,
      "theoretical_loss": 3.467822778072521,
      "tokens_seen": 1938554880
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041691073219658975,
      "loss": 1.438,
      "theoretical_loss": 3.4678130241051135,
      "tokens_seen": 1938620416
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004168906720160481,
      "loss": 1.3642,
      "theoretical_loss": 3.4678032705597603,
      "tokens_seen": 1938685952
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041687061183550653,
      "loss": 1.3859,
      "theoretical_loss": 3.46779351743643,
      "tokens_seen": 1938751488
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004168505516549649,
      "loss": 1.4449,
      "theoretical_loss": 3.4677837647350898,
      "tokens_seen": 1938817024
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041683049147442326,
      "loss": 1.4686,
      "theoretical_loss": 3.4677740124557066,
      "tokens_seen": 1938882560
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004168104312938816,
      "loss": 1.3234,
      "theoretical_loss": 3.4677642605982486,
      "tokens_seen": 1938948096
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041679037111334,
      "loss": 1.4131,
      "theoretical_loss": 3.467754509162683,
      "tokens_seen": 1939013632
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004167703109327984,
      "loss": 1.4358,
      "theoretical_loss": 3.467744758148976,
      "tokens_seen": 1939079168
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041675025075225676,
      "loss": 1.3384,
      "theoretical_loss": 3.467735007557098,
      "tokens_seen": 1939144704
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004167301905717151,
      "loss": 1.4278,
      "theoretical_loss": 3.467725257387014,
      "tokens_seen": 1939210240
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004167101303911735,
      "loss": 1.4368,
      "theoretical_loss": 3.467715507638693,
      "tokens_seen": 1939275776
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004166900702106319,
      "loss": 1.4753,
      "theoretical_loss": 3.4677057583121007,
      "tokens_seen": 1939341312
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041667001003009026,
      "loss": 1.4335,
      "theoretical_loss": 3.4676960094072067,
      "tokens_seen": 1939406848
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004166499498495487,
      "loss": 1.4696,
      "theoretical_loss": 3.467686260923977,
      "tokens_seen": 1939472384
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041662988966900704,
      "loss": 1.4309,
      "theoretical_loss": 3.46767651286238,
      "tokens_seen": 1939537920
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004166098294884654,
      "loss": 1.3781,
      "theoretical_loss": 3.467666765222383,
      "tokens_seen": 1939603456
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004165897693079238,
      "loss": 1.4327,
      "theoretical_loss": 3.467657018003954,
      "tokens_seen": 1939668992
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004165697091273822,
      "loss": 1.3674,
      "theoretical_loss": 3.467647271207059,
      "tokens_seen": 1939734528
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041654964894684055,
      "loss": 1.4225,
      "theoretical_loss": 3.4676375248316664,
      "tokens_seen": 1939800064
    },
    {
      "epoch": 0.59,
      "objective/train/advantage_avg": 0.4904329180717468,
      "objective/train/docs_used": 707259,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.387086033821106,
      "objective/train/original_loss": 1.3870857954025269,
      "objective/train/theoretical_loss": 3.4676277788777443,
      "objective/train/tokens_used": 1960325600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2412327527999878,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050269603729248,
      "objective/train/weighted_lm_loss": 1.4567883014678955,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.4676277788777443,
      "tokens_seen": 1939865600
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004165295887662989,
      "loss": 1.4447,
      "theoretical_loss": 3.4676277788777443,
      "tokens_seen": 1939865600
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004165095285857573,
      "loss": 1.3827,
      "theoretical_loss": 3.4676180333452598,
      "tokens_seen": 1939931136
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004164894684052157,
      "loss": 1.4077,
      "theoretical_loss": 3.46760828823418,
      "tokens_seen": 1939996672
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041646940822467405,
      "loss": 1.4773,
      "theoretical_loss": 3.467598543544473,
      "tokens_seen": 1940062208
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004164493480441324,
      "loss": 1.3943,
      "theoretical_loss": 3.467588799276106,
      "tokens_seen": 1940127744
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004164292878635908,
      "loss": 1.3467,
      "theoretical_loss": 3.467579055429047,
      "tokens_seen": 1940193280
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004164092276830492,
      "loss": 1.443,
      "theoretical_loss": 3.4675693120032633,
      "tokens_seen": 1940258816
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041638916750250755,
      "loss": 1.3965,
      "theoretical_loss": 3.4675595689987224,
      "tokens_seen": 1940324352
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004163691073219659,
      "loss": 1.4362,
      "theoretical_loss": 3.4675498264153912,
      "tokens_seen": 1940389888
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004163490471414243,
      "loss": 1.4704,
      "theoretical_loss": 3.4675400842532387,
      "tokens_seen": 1940455424
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004163289869608827,
      "loss": 1.4823,
      "theoretical_loss": 3.467530342512232,
      "tokens_seen": 1940520960
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041630892678034105,
      "loss": 1.3937,
      "theoretical_loss": 3.4675206011923376,
      "tokens_seen": 1940586496
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004162888665997994,
      "loss": 1.4073,
      "theoretical_loss": 3.4675108602935243,
      "tokens_seen": 1940652032
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004162688064192578,
      "loss": 1.4104,
      "theoretical_loss": 3.4675011198157595,
      "tokens_seen": 1940717568
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004162487462387162,
      "loss": 1.393,
      "theoretical_loss": 3.46749137975901,
      "tokens_seen": 1940783104
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041622868605817456,
      "loss": 1.3,
      "theoretical_loss": 3.4674816401232444,
      "tokens_seen": 1940848640
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004162086258776329,
      "loss": 1.3807,
      "theoretical_loss": 3.4674719009084294,
      "tokens_seen": 1940914176
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004161885656970913,
      "loss": 1.4019,
      "theoretical_loss": 3.4674621621145336,
      "tokens_seen": 1940979712
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041616850551654964,
      "loss": 1.4536,
      "theoretical_loss": 3.4674524237415234,
      "tokens_seen": 1941045248
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041614844533600806,
      "loss": 1.4379,
      "theoretical_loss": 3.467442685789367,
      "tokens_seen": 1941110784
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004161283851554664,
      "loss": 1.4382,
      "theoretical_loss": 3.4674329482580326,
      "tokens_seen": 1941176320
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004161083249749248,
      "loss": 1.3856,
      "theoretical_loss": 3.467423211147487,
      "tokens_seen": 1941241856
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041608826479438315,
      "loss": 1.3531,
      "theoretical_loss": 3.4674134744576977,
      "tokens_seen": 1941307392
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041606820461384156,
      "loss": 1.4532,
      "theoretical_loss": 3.467403738188633,
      "tokens_seen": 1941372928
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004160481444332999,
      "loss": 1.32,
      "theoretical_loss": 3.46739400234026,
      "tokens_seen": 1941438464
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004160280842527583,
      "loss": 1.5269,
      "theoretical_loss": 3.4673842669125463,
      "tokens_seen": 1941504000
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041600802407221665,
      "loss": 1.4503,
      "theoretical_loss": 3.46737453190546,
      "tokens_seen": 1941569536
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000415987963891675,
      "loss": 1.4778,
      "theoretical_loss": 3.4673647973189685,
      "tokens_seen": 1941635072
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041596790371113343,
      "loss": 1.4034,
      "theoretical_loss": 3.467355063153039,
      "tokens_seen": 1941700608
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004159478435305918,
      "loss": 1.2859,
      "theoretical_loss": 3.4673453294076397,
      "tokens_seen": 1941766144
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041592778335005015,
      "loss": 1.2914,
      "theoretical_loss": 3.467335596082738,
      "tokens_seen": 1941831680
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004159077231695085,
      "loss": 1.3737,
      "theoretical_loss": 3.467325863178302,
      "tokens_seen": 1941897216
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041588766298896693,
      "loss": 1.4176,
      "theoretical_loss": 3.467316130694299,
      "tokens_seen": 1941962752
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004158676028084253,
      "loss": 1.4135,
      "theoretical_loss": 3.4673063986306962,
      "tokens_seen": 1942028288
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041584754262788366,
      "loss": 1.3626,
      "theoretical_loss": 3.467296666987462,
      "tokens_seen": 1942093824
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000415827482447342,
      "loss": 1.3682,
      "theoretical_loss": 3.4672869357645633,
      "tokens_seen": 1942159360
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004158074222668004,
      "loss": 1.3449,
      "theoretical_loss": 3.467277204961968,
      "tokens_seen": 1942224896
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004157873620862588,
      "loss": 1.4418,
      "theoretical_loss": 3.4672674745796446,
      "tokens_seen": 1942290432
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041576730190571716,
      "loss": 1.364,
      "theoretical_loss": 3.4672577446175596,
      "tokens_seen": 1942355968
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004157472417251755,
      "loss": 1.3605,
      "theoretical_loss": 3.467248015075681,
      "tokens_seen": 1942421504
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004157271815446339,
      "loss": 1.4031,
      "theoretical_loss": 3.467238285953977,
      "tokens_seen": 1942487040
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004157071213640923,
      "loss": 1.5021,
      "theoretical_loss": 3.4672285572524153,
      "tokens_seen": 1942552576
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041568706118355066,
      "loss": 1.4279,
      "theoretical_loss": 3.4672188289709625,
      "tokens_seen": 1942618112
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000415667001003009,
      "loss": 1.4938,
      "theoretical_loss": 3.467209101109587,
      "tokens_seen": 1942683648
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004156469408224674,
      "loss": 1.3492,
      "theoretical_loss": 3.467199373668257,
      "tokens_seen": 1942749184
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041562688064192575,
      "loss": 1.4418,
      "theoretical_loss": 3.46718964664694,
      "tokens_seen": 1942814720
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041560682046138417,
      "loss": 1.4195,
      "theoretical_loss": 3.467179920045602,
      "tokens_seen": 1942880256
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041558676028084253,
      "loss": 1.3449,
      "theoretical_loss": 3.4671701938642134,
      "tokens_seen": 1942945792
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004155667001003009,
      "loss": 1.3446,
      "theoretical_loss": 3.4671604681027404,
      "tokens_seen": 1943011328
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041554663991975925,
      "loss": 1.5057,
      "theoretical_loss": 3.4671507427611505,
      "tokens_seen": 1943076864
    },
    {
      "epoch": 0.59,
      "objective/train/advantage_avg": 0.4921308755874634,
      "objective/train/docs_used": 708379,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3154819011688232,
      "objective/train/original_loss": 1.3154819011688232,
      "objective/train/theoretical_loss": 3.4671410178394115,
      "objective/train/tokens_used": 1963602400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24292311072349548,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050447940826416,
      "objective/train/weighted_lm_loss": 1.3815838098526,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4671410178394115,
      "tokens_seen": 1943142400
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041552657973921767,
      "loss": 1.3327,
      "theoretical_loss": 3.4671410178394115,
      "tokens_seen": 1943142400
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041550651955867603,
      "loss": 1.5045,
      "theoretical_loss": 3.4671312933374923,
      "tokens_seen": 1943207936
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004154864593781344,
      "loss": 1.4947,
      "theoretical_loss": 3.467121569255359,
      "tokens_seen": 1943273472
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041546639919759276,
      "loss": 1.4612,
      "theoretical_loss": 3.4671118455929806,
      "tokens_seen": 1943339008
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004154463390170511,
      "loss": 1.3689,
      "theoretical_loss": 3.4671021223503242,
      "tokens_seen": 1943404544
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041542627883650954,
      "loss": 1.3964,
      "theoretical_loss": 3.4670923995273575,
      "tokens_seen": 1943470080
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004154062186559679,
      "loss": 1.3922,
      "theoretical_loss": 3.4670826771240484,
      "tokens_seen": 1943535616
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041538615847542626,
      "loss": 1.4363,
      "theoretical_loss": 3.4670729551403645,
      "tokens_seen": 1943601152
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004153660982948846,
      "loss": 1.454,
      "theoretical_loss": 3.467063233576274,
      "tokens_seen": 1943666688
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041534603811434304,
      "loss": 1.4495,
      "theoretical_loss": 3.4670535124317436,
      "tokens_seen": 1943732224
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004153259779338014,
      "loss": 1.5504,
      "theoretical_loss": 3.4670437917067423,
      "tokens_seen": 1943797760
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041530591775325976,
      "loss": 1.3946,
      "theoretical_loss": 3.467034071401237,
      "tokens_seen": 1943863296
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004152858575727181,
      "loss": 1.3525,
      "theoretical_loss": 3.467024351515196,
      "tokens_seen": 1943928832
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004152657973921765,
      "loss": 1.4342,
      "theoretical_loss": 3.4670146320485866,
      "tokens_seen": 1943994368
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004152457372116349,
      "loss": 1.5079,
      "theoretical_loss": 3.4670049130013765,
      "tokens_seen": 1944059904
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041522567703109327,
      "loss": 1.3785,
      "theoretical_loss": 3.466995194373534,
      "tokens_seen": 1944125440
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041520561685055163,
      "loss": 1.3465,
      "theoretical_loss": 3.4669854761650267,
      "tokens_seen": 1944190976
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041518555667001,
      "loss": 1.1851,
      "theoretical_loss": 3.4669757583758223,
      "tokens_seen": 1944256512
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004151654964894684,
      "loss": 1.5317,
      "theoretical_loss": 3.466966041005888,
      "tokens_seen": 1944322048
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004151454363089268,
      "loss": 1.3846,
      "theoretical_loss": 3.466956324055192,
      "tokens_seen": 1944387584
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004151253761283852,
      "loss": 1.3908,
      "theoretical_loss": 3.466946607523703,
      "tokens_seen": 1944453120
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041510531594784355,
      "loss": 1.5588,
      "theoretical_loss": 3.466936891411388,
      "tokens_seen": 1944518656
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041508525576730196,
      "loss": 1.424,
      "theoretical_loss": 3.4669271757182143,
      "tokens_seen": 1944584192
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004150651955867603,
      "loss": 1.4477,
      "theoretical_loss": 3.46691746044415,
      "tokens_seen": 1944649728
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004150451354062187,
      "loss": 1.4001,
      "theoretical_loss": 3.4669077455891637,
      "tokens_seen": 1944715264
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041502507522567705,
      "loss": 1.4495,
      "theoretical_loss": 3.466898031153222,
      "tokens_seen": 1944780800
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004150050150451354,
      "loss": 1.4049,
      "theoretical_loss": 3.466888317136293,
      "tokens_seen": 1944846336
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041498495486459383,
      "loss": 1.4699,
      "theoretical_loss": 3.466878603538346,
      "tokens_seen": 1944911872
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004149648946840522,
      "loss": 1.3981,
      "theoretical_loss": 3.466868890359346,
      "tokens_seen": 1944977408
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041494483450351055,
      "loss": 1.3697,
      "theoretical_loss": 3.4668591775992637,
      "tokens_seen": 1945042944
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004149247743229689,
      "loss": 1.3824,
      "theoretical_loss": 3.466849465258065,
      "tokens_seen": 1945108480
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041490471414242733,
      "loss": 1.4507,
      "theoretical_loss": 3.4668397533357185,
      "tokens_seen": 1945174016
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004148846539618857,
      "loss": 1.467,
      "theoretical_loss": 3.466830041832192,
      "tokens_seen": 1945239552
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041486459378134406,
      "loss": 1.2642,
      "theoretical_loss": 3.4668203307474528,
      "tokens_seen": 1945305088
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004148445336008024,
      "loss": 1.4485,
      "theoretical_loss": 3.4668106200814695,
      "tokens_seen": 1945370624
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004148244734202608,
      "loss": 1.4335,
      "theoretical_loss": 3.4668009098342094,
      "tokens_seen": 1945436160
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004148044132397192,
      "loss": 1.392,
      "theoretical_loss": 3.4667912000056402,
      "tokens_seen": 1945501696
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041478435305917756,
      "loss": 1.4039,
      "theoretical_loss": 3.466781490595731,
      "tokens_seen": 1945567232
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004147642928786359,
      "loss": 1.3602,
      "theoretical_loss": 3.466771781604448,
      "tokens_seen": 1945632768
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004147442326980943,
      "loss": 1.2672,
      "theoretical_loss": 3.4667620730317594,
      "tokens_seen": 1945698304
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004147241725175527,
      "loss": 1.4768,
      "theoretical_loss": 3.4667523648776344,
      "tokens_seen": 1945763840
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041470411233701106,
      "loss": 1.4252,
      "theoretical_loss": 3.4667426571420394,
      "tokens_seen": 1945829376
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004146840521564694,
      "loss": 1.3296,
      "theoretical_loss": 3.4667329498249426,
      "tokens_seen": 1945894912
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004146639919759278,
      "loss": 1.4598,
      "theoretical_loss": 3.466723242926312,
      "tokens_seen": 1945960448
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041464393179538615,
      "loss": 1.4318,
      "theoretical_loss": 3.4667135364461155,
      "tokens_seen": 1946025984
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041462387161484457,
      "loss": 1.3821,
      "theoretical_loss": 3.466703830384321,
      "tokens_seen": 1946091520
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041460381143430293,
      "loss": 1.3231,
      "theoretical_loss": 3.466694124740896,
      "tokens_seen": 1946157056
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004145837512537613,
      "loss": 1.3716,
      "theoretical_loss": 3.466684419515809,
      "tokens_seen": 1946222592
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041456369107321965,
      "loss": 1.4886,
      "theoretical_loss": 3.4666747147090273,
      "tokens_seen": 1946288128
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041454363089267807,
      "loss": 1.415,
      "theoretical_loss": 3.4666650103205194,
      "tokens_seen": 1946353664
    },
    {
      "epoch": 0.59,
      "objective/train/advantage_avg": 0.4936843812465668,
      "objective/train/docs_used": 709628,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3543987274169922,
      "objective/train/original_loss": 1.3543986082077026,
      "objective/train/theoretical_loss": 3.466655306350253,
      "objective/train/tokens_used": 1966879200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24424119293689728,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506101846694946,
      "objective/train/weighted_lm_loss": 1.4228193759918213,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.466655306350253,
      "tokens_seen": 1946419200
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041452357071213643,
      "loss": 1.3733,
      "theoretical_loss": 3.466655306350253,
      "tokens_seen": 1946419200
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004145035105315948,
      "loss": 1.3869,
      "theoretical_loss": 3.466645602798195,
      "tokens_seen": 1946484736
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041448345035105316,
      "loss": 1.5025,
      "theoretical_loss": 3.4666358996643147,
      "tokens_seen": 1946550272
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004144633901705115,
      "loss": 1.3269,
      "theoretical_loss": 3.4666261969485794,
      "tokens_seen": 1946615808
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041444332998996994,
      "loss": 1.4222,
      "theoretical_loss": 3.4666164946509572,
      "tokens_seen": 1946681344
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004144232698094283,
      "loss": 1.2897,
      "theoretical_loss": 3.4666067927714157,
      "tokens_seen": 1946746880
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041440320962888666,
      "loss": 1.4941,
      "theoretical_loss": 3.466597091309923,
      "tokens_seen": 1946812416
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000414383149448345,
      "loss": 1.4996,
      "theoretical_loss": 3.466587390266447,
      "tokens_seen": 1946877952
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041436308926780344,
      "loss": 1.4658,
      "theoretical_loss": 3.466577689640955,
      "tokens_seen": 1946943488
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004143430290872618,
      "loss": 1.5227,
      "theoretical_loss": 3.4665679894334165,
      "tokens_seen": 1947009024
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041432296890672016,
      "loss": 1.442,
      "theoretical_loss": 3.466558289643798,
      "tokens_seen": 1947074560
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004143029087261785,
      "loss": 1.3317,
      "theoretical_loss": 3.466548590272067,
      "tokens_seen": 1947140096
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004142828485456369,
      "loss": 1.4446,
      "theoretical_loss": 3.4665388913181934,
      "tokens_seen": 1947205632
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004142627883650953,
      "loss": 1.3546,
      "theoretical_loss": 3.4665291927821436,
      "tokens_seen": 1947271168
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041424272818455367,
      "loss": 1.5075,
      "theoretical_loss": 3.466519494663886,
      "tokens_seen": 1947336704
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041422266800401203,
      "loss": 1.3531,
      "theoretical_loss": 3.466509796963389,
      "tokens_seen": 1947402240
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004142026078234704,
      "loss": 1.3157,
      "theoretical_loss": 3.4665000996806192,
      "tokens_seen": 1947467776
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004141825476429288,
      "loss": 1.3955,
      "theoretical_loss": 3.466490402815546,
      "tokens_seen": 1947533312
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041416248746238717,
      "loss": 1.4536,
      "theoretical_loss": 3.466480706368137,
      "tokens_seen": 1947598848
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041414242728184553,
      "loss": 1.4422,
      "theoretical_loss": 3.4664710103383594,
      "tokens_seen": 1947664384
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004141223671013039,
      "loss": 1.3914,
      "theoretical_loss": 3.466461314726182,
      "tokens_seen": 1947729920
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004141023069207623,
      "loss": 1.4016,
      "theoretical_loss": 3.466451619531572,
      "tokens_seen": 1947795456
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004140822467402207,
      "loss": 1.3619,
      "theoretical_loss": 3.466441924754498,
      "tokens_seen": 1947860992
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041406218655967904,
      "loss": 1.439,
      "theoretical_loss": 3.466432230394928,
      "tokens_seen": 1947926528
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004140421263791374,
      "loss": 1.3635,
      "theoretical_loss": 3.4664225364528294,
      "tokens_seen": 1947992064
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041402206619859576,
      "loss": 1.3693,
      "theoretical_loss": 3.466412842928171,
      "tokens_seen": 1948057600
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004140020060180542,
      "loss": 1.4629,
      "theoretical_loss": 3.46640314982092,
      "tokens_seen": 1948123136
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041398194583751254,
      "loss": 1.4369,
      "theoretical_loss": 3.466393457131045,
      "tokens_seen": 1948188672
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004139618856569709,
      "loss": 1.4996,
      "theoretical_loss": 3.466383764858513,
      "tokens_seen": 1948254208
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041394182547642926,
      "loss": 1.4168,
      "theoretical_loss": 3.4663740730032933,
      "tokens_seen": 1948319744
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004139217652958877,
      "loss": 1.3874,
      "theoretical_loss": 3.466364381565353,
      "tokens_seen": 1948385280
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041390170511534604,
      "loss": 1.4545,
      "theoretical_loss": 3.4663546905446605,
      "tokens_seen": 1948450816
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004138816449348044,
      "loss": 1.4502,
      "theoretical_loss": 3.466344999941184,
      "tokens_seen": 1948516352
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041386158475426277,
      "loss": 1.3825,
      "theoretical_loss": 3.466335309754891,
      "tokens_seen": 1948581888
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041384152457372113,
      "loss": 1.3056,
      "theoretical_loss": 3.4663256199857493,
      "tokens_seen": 1948647424
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041382146439317954,
      "loss": 1.5453,
      "theoretical_loss": 3.4663159306337277,
      "tokens_seen": 1948712960
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004138014042126379,
      "loss": 1.4392,
      "theoretical_loss": 3.4663062416987938,
      "tokens_seen": 1948778496
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041378134403209627,
      "loss": 1.3002,
      "theoretical_loss": 3.4662965531809156,
      "tokens_seen": 1948844032
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041376128385155463,
      "loss": 1.4675,
      "theoretical_loss": 3.466286865080061,
      "tokens_seen": 1948909568
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041374122367101305,
      "loss": 1.512,
      "theoretical_loss": 3.4662771773961985,
      "tokens_seen": 1948975104
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004137211634904714,
      "loss": 1.4876,
      "theoretical_loss": 3.4662674901292956,
      "tokens_seen": 1949040640
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041370110330992977,
      "loss": 1.3363,
      "theoretical_loss": 3.466257803279321,
      "tokens_seen": 1949106176
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041368104312938813,
      "loss": 1.3115,
      "theoretical_loss": 3.466248116846242,
      "tokens_seen": 1949171712
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041366098294884655,
      "loss": 1.3595,
      "theoretical_loss": 3.4662384308300265,
      "tokens_seen": 1949237248
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041364092276830497,
      "loss": 1.2792,
      "theoretical_loss": 3.4662287452306435,
      "tokens_seen": 1949302784
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041362086258776333,
      "loss": 1.5343,
      "theoretical_loss": 3.4662190600480605,
      "tokens_seen": 1949368320
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004136008024072217,
      "loss": 1.4384,
      "theoretical_loss": 3.4662093752822454,
      "tokens_seen": 1949433856
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041358074222668005,
      "loss": 1.3957,
      "theoretical_loss": 3.4661996909331667,
      "tokens_seen": 1949499392
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041356068204613847,
      "loss": 1.2758,
      "theoretical_loss": 3.466190007000792,
      "tokens_seen": 1949564928
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041354062186559683,
      "loss": 1.554,
      "theoretical_loss": 3.46618032348509,
      "tokens_seen": 1949630464
    },
    {
      "epoch": 0.59,
      "objective/train/advantage_avg": 0.49210435152053833,
      "objective/train/docs_used": 710798,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3685261011123657,
      "objective/train/original_loss": 1.3685259819030762,
      "objective/train/theoretical_loss": 3.4661706403860277,
      "objective/train/tokens_used": 1970156000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24257782101631165,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504435300827026,
      "objective/train/weighted_lm_loss": 1.43767249584198,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4661706403860277,
      "tokens_seen": 1949696000
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004135205616850552,
      "loss": 1.4228,
      "theoretical_loss": 3.4661706403860277,
      "tokens_seen": 1949696000
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041350050150451356,
      "loss": 1.4562,
      "theoretical_loss": 3.4661609577035746,
      "tokens_seen": 1949761536
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004134804413239719,
      "loss": 1.4531,
      "theoretical_loss": 3.466151275437697,
      "tokens_seen": 1949827072
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041346038114343034,
      "loss": 1.527,
      "theoretical_loss": 3.4661415935883646,
      "tokens_seen": 1949892608
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004134403209628887,
      "loss": 1.3594,
      "theoretical_loss": 3.4661319121555447,
      "tokens_seen": 1949958144
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041342026078234706,
      "loss": 1.3832,
      "theoretical_loss": 3.4661222311392055,
      "tokens_seen": 1950023680
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004134002006018054,
      "loss": 1.2076,
      "theoretical_loss": 3.4661125505393153,
      "tokens_seen": 1950089216
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041338014042126384,
      "loss": 1.4686,
      "theoretical_loss": 3.466102870355842,
      "tokens_seen": 1950154752
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004133600802407222,
      "loss": 1.3774,
      "theoretical_loss": 3.4660931905887535,
      "tokens_seen": 1950220288
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041334002006018056,
      "loss": 1.4291,
      "theoretical_loss": 3.466083511238018,
      "tokens_seen": 1950285824
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004133199598796389,
      "loss": 1.4329,
      "theoretical_loss": 3.4660738323036036,
      "tokens_seen": 1950351360
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004132998996990973,
      "loss": 1.5275,
      "theoretical_loss": 3.466064153785479,
      "tokens_seen": 1950416896
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004132798395185557,
      "loss": 1.4905,
      "theoretical_loss": 3.4660544756836114,
      "tokens_seen": 1950482432
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041325977933801407,
      "loss": 1.4099,
      "theoretical_loss": 3.4660447979979696,
      "tokens_seen": 1950547968
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041323971915747243,
      "loss": 1.4102,
      "theoretical_loss": 3.466035120728521,
      "tokens_seen": 1950613504
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004132196589769308,
      "loss": 1.479,
      "theoretical_loss": 3.4660254438752345,
      "tokens_seen": 1950679040
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004131995987963892,
      "loss": 1.3387,
      "theoretical_loss": 3.466015767438078,
      "tokens_seen": 1950744576
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041317953861584757,
      "loss": 1.3519,
      "theoretical_loss": 3.4660060914170194,
      "tokens_seen": 1950810112
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041315947843530593,
      "loss": 1.3681,
      "theoretical_loss": 3.465996415812027,
      "tokens_seen": 1950875648
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004131394182547643,
      "loss": 1.386,
      "theoretical_loss": 3.4659867406230687,
      "tokens_seen": 1950941184
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004131193580742227,
      "loss": 1.384,
      "theoretical_loss": 3.4659770658501126,
      "tokens_seen": 1951006720
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004130992978936811,
      "loss": 1.5167,
      "theoretical_loss": 3.4659673914931277,
      "tokens_seen": 1951072256
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041307923771313944,
      "loss": 1.5048,
      "theoretical_loss": 3.4659577175520813,
      "tokens_seen": 1951137792
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004130591775325978,
      "loss": 1.4441,
      "theoretical_loss": 3.4659480440269412,
      "tokens_seen": 1951203328
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041303911735205616,
      "loss": 1.4532,
      "theoretical_loss": 3.4659383709176765,
      "tokens_seen": 1951268864
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004130190571715146,
      "loss": 1.2902,
      "theoretical_loss": 3.4659286982242548,
      "tokens_seen": 1951334400
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041299899699097294,
      "loss": 1.3443,
      "theoretical_loss": 3.4659190259466444,
      "tokens_seen": 1951399936
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004129789368104313,
      "loss": 1.3518,
      "theoretical_loss": 3.465909354084814,
      "tokens_seen": 1951465472
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041295887662988966,
      "loss": 1.3952,
      "theoretical_loss": 3.4658996826387303,
      "tokens_seen": 1951531008
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004129388164493481,
      "loss": 1.3935,
      "theoretical_loss": 3.465890011608363,
      "tokens_seen": 1951596544
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041291875626880644,
      "loss": 1.4425,
      "theoretical_loss": 3.46588034099368,
      "tokens_seen": 1951662080
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004128986960882648,
      "loss": 1.4642,
      "theoretical_loss": 3.465870670794648,
      "tokens_seen": 1951727616
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041287863590772317,
      "loss": 1.5196,
      "theoretical_loss": 3.465861001011237,
      "tokens_seen": 1951793152
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041285857572718153,
      "loss": 1.3642,
      "theoretical_loss": 3.4658513316434147,
      "tokens_seen": 1951858688
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041283851554663995,
      "loss": 1.3647,
      "theoretical_loss": 3.465841662691149,
      "tokens_seen": 1951924224
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004128184553660983,
      "loss": 1.3687,
      "theoretical_loss": 3.465831994154408,
      "tokens_seen": 1951989760
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041279839518555667,
      "loss": 1.4385,
      "theoretical_loss": 3.4658223260331598,
      "tokens_seen": 1952055296
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041277833500501503,
      "loss": 1.3982,
      "theoretical_loss": 3.465812658327373,
      "tokens_seen": 1952120832
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041275827482447345,
      "loss": 1.382,
      "theoretical_loss": 3.465802991037016,
      "tokens_seen": 1952186368
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004127382146439318,
      "loss": 1.379,
      "theoretical_loss": 3.465793324162057,
      "tokens_seen": 1952251904
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041271815446339017,
      "loss": 1.4357,
      "theoretical_loss": 3.465783657702463,
      "tokens_seen": 1952317440
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041269809428284854,
      "loss": 1.5383,
      "theoretical_loss": 3.4657739916582035,
      "tokens_seen": 1952382976
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004126780341023069,
      "loss": 1.4681,
      "theoretical_loss": 3.4657643260292463,
      "tokens_seen": 1952448512
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004126579739217653,
      "loss": 1.4141,
      "theoretical_loss": 3.4657546608155596,
      "tokens_seen": 1952514048
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004126379137412237,
      "loss": 1.4536,
      "theoretical_loss": 3.465744996017112,
      "tokens_seen": 1952579584
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041261785356068204,
      "loss": 1.426,
      "theoretical_loss": 3.465735331633871,
      "tokens_seen": 1952645120
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004125977933801404,
      "loss": 1.4024,
      "theoretical_loss": 3.4657256676658053,
      "tokens_seen": 1952710656
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004125777331995988,
      "loss": 1.3665,
      "theoretical_loss": 3.465716004112883,
      "tokens_seen": 1952776192
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004125576730190572,
      "loss": 1.408,
      "theoretical_loss": 3.465706340975072,
      "tokens_seen": 1952841728
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041253761283851554,
      "loss": 1.5883,
      "theoretical_loss": 3.465696678252341,
      "tokens_seen": 1952907264
    },
    {
      "epoch": 0.59,
      "objective/train/advantage_avg": 0.49196383357048035,
      "objective/train/docs_used": 712021,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3089810609817505,
      "objective/train/original_loss": 1.30898118019104,
      "objective/train/theoretical_loss": 3.4656870159446584,
      "objective/train/tokens_used": 1973432800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24232245981693268,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504282712936401,
      "objective/train/weighted_lm_loss": 1.3749518394470215,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4656870159446584,
      "tokens_seen": 1952972800
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004125175526579739,
      "loss": 1.3564,
      "theoretical_loss": 3.4656870159446584,
      "tokens_seen": 1952972800
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041249749247743227,
      "loss": 1.4463,
      "theoretical_loss": 3.465677354051992,
      "tokens_seen": 1953038336
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004124774322968907,
      "loss": 1.3943,
      "theoretical_loss": 3.4656676925743106,
      "tokens_seen": 1953103872
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041245737211634904,
      "loss": 1.3787,
      "theoretical_loss": 3.4656580315115817,
      "tokens_seen": 1953169408
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004124373119358074,
      "loss": 1.3878,
      "theoretical_loss": 3.465648370863774,
      "tokens_seen": 1953234944
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041241725175526577,
      "loss": 1.3396,
      "theoretical_loss": 3.4656387106308557,
      "tokens_seen": 1953300480
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004123971915747242,
      "loss": 1.4385,
      "theoretical_loss": 3.465629050812795,
      "tokens_seen": 1953366016
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041237713139418255,
      "loss": 1.3208,
      "theoretical_loss": 3.4656193914095605,
      "tokens_seen": 1953431552
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004123570712136409,
      "loss": 1.3975,
      "theoretical_loss": 3.46560973242112,
      "tokens_seen": 1953497088
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041233701103309927,
      "loss": 1.4725,
      "theoretical_loss": 3.465600073847442,
      "tokens_seen": 1953562624
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041231695085255763,
      "loss": 1.4634,
      "theoretical_loss": 3.465590415688495,
      "tokens_seen": 1953628160
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041229689067201605,
      "loss": 1.3895,
      "theoretical_loss": 3.465580757944247,
      "tokens_seen": 1953693696
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004122768304914744,
      "loss": 1.4366,
      "theoretical_loss": 3.4655711006146657,
      "tokens_seen": 1953759232
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004122567703109328,
      "loss": 1.5132,
      "theoretical_loss": 3.4655614436997206,
      "tokens_seen": 1953824768
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041223671013039114,
      "loss": 1.3278,
      "theoretical_loss": 3.465551787199379,
      "tokens_seen": 1953890304
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041221664994984955,
      "loss": 1.4665,
      "theoretical_loss": 3.46554213111361,
      "tokens_seen": 1953955840
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004121965897693079,
      "loss": 1.5065,
      "theoretical_loss": 3.465532475442381,
      "tokens_seen": 1954021376
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004121765295887663,
      "loss": 1.3483,
      "theoretical_loss": 3.4655228201856616,
      "tokens_seen": 1954086912
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004121564694082247,
      "loss": 1.458,
      "theoretical_loss": 3.4655131653434186,
      "tokens_seen": 1954152448
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004121364092276831,
      "loss": 1.5268,
      "theoretical_loss": 3.465503510915621,
      "tokens_seen": 1954217984
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004121163490471415,
      "loss": 1.3606,
      "theoretical_loss": 3.4654938569022375,
      "tokens_seen": 1954283520
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041209628886659984,
      "loss": 1.3341,
      "theoretical_loss": 3.4654842033032356,
      "tokens_seen": 1954349056
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004120762286860582,
      "loss": 1.4367,
      "theoretical_loss": 3.465474550118584,
      "tokens_seen": 1954414592
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041205616850551656,
      "loss": 1.4248,
      "theoretical_loss": 3.4654648973482516,
      "tokens_seen": 1954480128
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000412036108324975,
      "loss": 1.4823,
      "theoretical_loss": 3.4654552449922056,
      "tokens_seen": 1954545664
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041201604814443334,
      "loss": 1.3798,
      "theoretical_loss": 3.465445593050415,
      "tokens_seen": 1954611200
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004119959879638917,
      "loss": 1.4491,
      "theoretical_loss": 3.4654359415228484,
      "tokens_seen": 1954676736
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041197592778335006,
      "loss": 1.5433,
      "theoretical_loss": 3.4654262904094737,
      "tokens_seen": 1954742272
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004119558676028085,
      "loss": 1.4056,
      "theoretical_loss": 3.4654166397102593,
      "tokens_seen": 1954807808
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041193580742226684,
      "loss": 1.3762,
      "theoretical_loss": 3.4654069894251736,
      "tokens_seen": 1954873344
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004119157472417252,
      "loss": 1.5037,
      "theoretical_loss": 3.4653973395541846,
      "tokens_seen": 1954938880
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041189568706118357,
      "loss": 1.4463,
      "theoretical_loss": 3.4653876900972613,
      "tokens_seen": 1955004416
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041187562688064193,
      "loss": 1.4218,
      "theoretical_loss": 3.4653780410543717,
      "tokens_seen": 1955069952
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041185556670010035,
      "loss": 1.416,
      "theoretical_loss": 3.465368392425484,
      "tokens_seen": 1955135488
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004118355065195587,
      "loss": 1.4709,
      "theoretical_loss": 3.465358744210567,
      "tokens_seen": 1955201024
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041181544633901707,
      "loss": 1.4151,
      "theoretical_loss": 3.4653490964095885,
      "tokens_seen": 1955266560
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041179538615847543,
      "loss": 1.3909,
      "theoretical_loss": 3.465339449022517,
      "tokens_seen": 1955332096
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041177532597793385,
      "loss": 1.507,
      "theoretical_loss": 3.4653298020493217,
      "tokens_seen": 1955397632
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004117552657973922,
      "loss": 1.4518,
      "theoretical_loss": 3.46532015548997,
      "tokens_seen": 1955463168
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004117352056168506,
      "loss": 1.4539,
      "theoretical_loss": 3.4653105093444303,
      "tokens_seen": 1955528704
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041171514543630894,
      "loss": 1.4045,
      "theoretical_loss": 3.4653008636126716,
      "tokens_seen": 1955594240
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004116950852557673,
      "loss": 1.5288,
      "theoretical_loss": 3.465291218294662,
      "tokens_seen": 1955659776
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004116750250752257,
      "loss": 1.3309,
      "theoretical_loss": 3.46528157339037,
      "tokens_seen": 1955725312
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004116549648946841,
      "loss": 1.3941,
      "theoretical_loss": 3.465271928899763,
      "tokens_seen": 1955790848
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041163490471414244,
      "loss": 1.4132,
      "theoretical_loss": 3.4652622848228107,
      "tokens_seen": 1955856384
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004116148445336008,
      "loss": 1.4397,
      "theoretical_loss": 3.4652526411594815,
      "tokens_seen": 1955921920
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004115947843530592,
      "loss": 1.4134,
      "theoretical_loss": 3.465242997909743,
      "tokens_seen": 1955987456
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004115747241725176,
      "loss": 1.4227,
      "theoretical_loss": 3.465233355073564,
      "tokens_seen": 1956052992
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041155466399197594,
      "loss": 1.4292,
      "theoretical_loss": 3.4652237126509124,
      "tokens_seen": 1956118528
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004115346038114343,
      "loss": 1.4524,
      "theoretical_loss": 3.4652140706417573,
      "tokens_seen": 1956184064
    },
    {
      "epoch": 0.59,
      "objective/train/advantage_avg": 0.49314144253730774,
      "objective/train/docs_used": 713197,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.297523856163025,
      "objective/train/original_loss": 1.297523856163025,
      "objective/train/theoretical_loss": 3.465204429046067,
      "objective/train/tokens_used": 1976709600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24374866485595703,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050553321838379,
      "objective/train/weighted_lm_loss": 1.363060712814331,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.465204429046067,
      "tokens_seen": 1956249600
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041151454363089267,
      "loss": 1.3804,
      "theoretical_loss": 3.465204429046067,
      "tokens_seen": 1956249600
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004114944834503511,
      "loss": 1.4481,
      "theoretical_loss": 3.46519478786381,
      "tokens_seen": 1956315136
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041147442326980944,
      "loss": 1.469,
      "theoretical_loss": 3.465185147094954,
      "tokens_seen": 1956380672
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004114543630892678,
      "loss": 1.4662,
      "theoretical_loss": 3.4651755067394685,
      "tokens_seen": 1956446208
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041143430290872617,
      "loss": 1.2973,
      "theoretical_loss": 3.465165866797321,
      "tokens_seen": 1956511744
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004114142427281846,
      "loss": 1.5631,
      "theoretical_loss": 3.4651562272684804,
      "tokens_seen": 1956577280
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041139418254764295,
      "loss": 1.4375,
      "theoretical_loss": 3.465146588152915,
      "tokens_seen": 1956642816
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004113741223671013,
      "loss": 1.3597,
      "theoretical_loss": 3.4651369494505935,
      "tokens_seen": 1956708352
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041135406218655967,
      "loss": 1.4131,
      "theoretical_loss": 3.465127311161484,
      "tokens_seen": 1956773888
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041133400200601803,
      "loss": 1.5106,
      "theoretical_loss": 3.4651176732855546,
      "tokens_seen": 1956839424
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041131394182547645,
      "loss": 1.3731,
      "theoretical_loss": 3.465108035822775,
      "tokens_seen": 1956904960
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004112938816449348,
      "loss": 1.4634,
      "theoretical_loss": 3.4650983987731125,
      "tokens_seen": 1956970496
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004112738214643932,
      "loss": 1.4656,
      "theoretical_loss": 3.465088762136536,
      "tokens_seen": 1957036032
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041125376128385154,
      "loss": 1.5066,
      "theoretical_loss": 3.465079125913014,
      "tokens_seen": 1957101568
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041123370110330995,
      "loss": 1.3885,
      "theoretical_loss": 3.4650694901025147,
      "tokens_seen": 1957167104
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004112136409227683,
      "loss": 1.4023,
      "theoretical_loss": 3.4650598547050064,
      "tokens_seen": 1957232640
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004111935807422267,
      "loss": 1.4603,
      "theoretical_loss": 3.4650502197204585,
      "tokens_seen": 1957298176
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041117352056168504,
      "loss": 1.3327,
      "theoretical_loss": 3.465040585148839,
      "tokens_seen": 1957363712
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004111534603811434,
      "loss": 1.3547,
      "theoretical_loss": 3.4650309509901156,
      "tokens_seen": 1957429248
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004111334002006018,
      "loss": 1.5937,
      "theoretical_loss": 3.4650213172442577,
      "tokens_seen": 1957494784
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004111133400200602,
      "loss": 1.4764,
      "theoretical_loss": 3.465011683911234,
      "tokens_seen": 1957560320
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041109327983951854,
      "loss": 1.4804,
      "theoretical_loss": 3.4650020509910116,
      "tokens_seen": 1957625856
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004110732196589769,
      "loss": 1.4317,
      "theoretical_loss": 3.4649924184835603,
      "tokens_seen": 1957691392
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004110531594784353,
      "loss": 1.5579,
      "theoretical_loss": 3.4649827863888483,
      "tokens_seen": 1957756928
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004110330992978937,
      "loss": 1.5206,
      "theoretical_loss": 3.4649731547068443,
      "tokens_seen": 1957822464
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041101303911735205,
      "loss": 1.409,
      "theoretical_loss": 3.4649635234375165,
      "tokens_seen": 1957888000
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004109929789368104,
      "loss": 1.3623,
      "theoretical_loss": 3.4649538925808328,
      "tokens_seen": 1957953536
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041097291875626877,
      "loss": 1.486,
      "theoretical_loss": 3.4649442621367625,
      "tokens_seen": 1958019072
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004109528585757272,
      "loss": 1.386,
      "theoretical_loss": 3.4649346321052743,
      "tokens_seen": 1958084608
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041093279839518555,
      "loss": 1.3808,
      "theoretical_loss": 3.4649250024863356,
      "tokens_seen": 1958150144
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004109127382146439,
      "loss": 1.4627,
      "theoretical_loss": 3.4649153732799167,
      "tokens_seen": 1958215680
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004108926780341023,
      "loss": 1.404,
      "theoretical_loss": 3.4649057444859843,
      "tokens_seen": 1958281216
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004108726178535607,
      "loss": 1.3993,
      "theoretical_loss": 3.4648961161045078,
      "tokens_seen": 1958346752
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041085255767301905,
      "loss": 1.3514,
      "theoretical_loss": 3.4648864881354564,
      "tokens_seen": 1958412288
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004108324974924774,
      "loss": 1.43,
      "theoretical_loss": 3.464876860578797,
      "tokens_seen": 1958477824
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004108124373119358,
      "loss": 1.41,
      "theoretical_loss": 3.464867233434499,
      "tokens_seen": 1958543360
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004107923771313942,
      "loss": 1.4835,
      "theoretical_loss": 3.4648576067025316,
      "tokens_seen": 1958608896
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041077231695085256,
      "loss": 1.4125,
      "theoretical_loss": 3.4648479803828627,
      "tokens_seen": 1958674432
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004107522567703109,
      "loss": 1.5272,
      "theoretical_loss": 3.46483835447546,
      "tokens_seen": 1958739968
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004107321965897693,
      "loss": 1.3302,
      "theoretical_loss": 3.4648287289802937,
      "tokens_seen": 1958805504
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041071213640922764,
      "loss": 1.4213,
      "theoretical_loss": 3.464819103897331,
      "tokens_seen": 1958871040
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041069207622868606,
      "loss": 1.4118,
      "theoretical_loss": 3.464809479226542,
      "tokens_seen": 1958936576
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004106720160481444,
      "loss": 1.4565,
      "theoretical_loss": 3.464799854967893,
      "tokens_seen": 1959002112
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041065195586760284,
      "loss": 1.4281,
      "theoretical_loss": 3.4647902311213548,
      "tokens_seen": 1959067648
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004106318956870612,
      "loss": 1.4656,
      "theoretical_loss": 3.4647806076868943,
      "tokens_seen": 1959133184
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004106118355065196,
      "loss": 1.4399,
      "theoretical_loss": 3.464770984664481,
      "tokens_seen": 1959198720
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000410591775325978,
      "loss": 1.4434,
      "theoretical_loss": 3.464761362054084,
      "tokens_seen": 1959264256
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041057171514543634,
      "loss": 1.3692,
      "theoretical_loss": 3.4647517398556698,
      "tokens_seen": 1959329792
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004105516549648947,
      "loss": 1.4408,
      "theoretical_loss": 3.4647421180692093,
      "tokens_seen": 1959395328
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041053159478435307,
      "loss": 1.351,
      "theoretical_loss": 3.46473249669467,
      "tokens_seen": 1959460864
    },
    {
      "epoch": 0.59,
      "objective/train/advantage_avg": 0.48486465215682983,
      "objective/train/docs_used": 714425,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4289131164550781,
      "objective/train/original_loss": 1.4289131164550781,
      "objective/train/theoretical_loss": 3.4647228757320203,
      "objective/train/tokens_used": 1979986400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2372017353773117,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.049692153930664,
      "objective/train/weighted_lm_loss": 1.5002632141113281,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4647228757320203,
      "tokens_seen": 1959526400
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004105115346038115,
      "loss": 1.4777,
      "theoretical_loss": 3.4647228757320203,
      "tokens_seen": 1959526400
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041049147442326985,
      "loss": 1.4599,
      "theoretical_loss": 3.4647132551812287,
      "tokens_seen": 1959591936
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004104714142427282,
      "loss": 1.4051,
      "theoretical_loss": 3.464703635042265,
      "tokens_seen": 1959657472
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041045135406218657,
      "loss": 1.4745,
      "theoretical_loss": 3.4646940153150965,
      "tokens_seen": 1959723008
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.000410431293881645,
      "loss": 1.4452,
      "theoretical_loss": 3.4646843959996927,
      "tokens_seen": 1959788544
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041041123370110335,
      "loss": 1.4131,
      "theoretical_loss": 3.4646747770960213,
      "tokens_seen": 1959854080
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004103911735205617,
      "loss": 1.3567,
      "theoretical_loss": 3.4646651586040518,
      "tokens_seen": 1959919616
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041037111334002007,
      "loss": 1.3353,
      "theoretical_loss": 3.4646555405237516,
      "tokens_seen": 1959985152
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041035105315947844,
      "loss": 1.504,
      "theoretical_loss": 3.464645922855091,
      "tokens_seen": 1960050688
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041033099297893685,
      "loss": 1.3554,
      "theoretical_loss": 3.464636305598037,
      "tokens_seen": 1960116224
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004103109327983952,
      "loss": 1.4428,
      "theoretical_loss": 3.46462668875256,
      "tokens_seen": 1960181760
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004102908726178536,
      "loss": 1.5029,
      "theoretical_loss": 3.464617072318627,
      "tokens_seen": 1960247296
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041027081243731194,
      "loss": 1.5119,
      "theoretical_loss": 3.464607456296207,
      "tokens_seen": 1960312832
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041025075225677035,
      "loss": 1.4344,
      "theoretical_loss": 3.4645978406852693,
      "tokens_seen": 1960378368
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004102306920762287,
      "loss": 1.4625,
      "theoretical_loss": 3.464588225485782,
      "tokens_seen": 1960443904
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004102106318956871,
      "loss": 1.5277,
      "theoretical_loss": 3.4645786106977137,
      "tokens_seen": 1960509440
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041019057171514544,
      "loss": 1.4594,
      "theoretical_loss": 3.464568996321033,
      "tokens_seen": 1960574976
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004101705115346038,
      "loss": 1.3425,
      "theoretical_loss": 3.4645593823557093,
      "tokens_seen": 1960640512
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004101504513540622,
      "loss": 1.5963,
      "theoretical_loss": 3.46454976880171,
      "tokens_seen": 1960706048
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004101303911735206,
      "loss": 1.4275,
      "theoretical_loss": 3.464540155659005,
      "tokens_seen": 1960771584
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041011033099297894,
      "loss": 1.4703,
      "theoretical_loss": 3.4645305429275624,
      "tokens_seen": 1960837120
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004100902708124373,
      "loss": 1.42,
      "theoretical_loss": 3.4645209306073506,
      "tokens_seen": 1960902656
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004100702106318957,
      "loss": 1.5077,
      "theoretical_loss": 3.4645113186983383,
      "tokens_seen": 1960968192
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004100501504513541,
      "loss": 1.3775,
      "theoretical_loss": 3.464501707200495,
      "tokens_seen": 1961033728
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00041003009027081245,
      "loss": 1.4528,
      "theoretical_loss": 3.464492096113788,
      "tokens_seen": 1961099264
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004100100300902708,
      "loss": 1.4163,
      "theoretical_loss": 3.464482485438187,
      "tokens_seen": 1961164800
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040998996990972917,
      "loss": 1.5449,
      "theoretical_loss": 3.4644728751736604,
      "tokens_seen": 1961230336
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004099699097291876,
      "loss": 1.4244,
      "theoretical_loss": 3.464463265320177,
      "tokens_seen": 1961295872
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040994984954864595,
      "loss": 1.3816,
      "theoretical_loss": 3.4644536558777057,
      "tokens_seen": 1961361408
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004099297893681043,
      "loss": 1.4105,
      "theoretical_loss": 3.4644440468462143,
      "tokens_seen": 1961426944
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004099097291875627,
      "loss": 1.4077,
      "theoretical_loss": 3.4644344382256724,
      "tokens_seen": 1961492480
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004098896690070211,
      "loss": 1.5133,
      "theoretical_loss": 3.464424830016048,
      "tokens_seen": 1961558016
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040986960882647945,
      "loss": 1.4823,
      "theoretical_loss": 3.4644152222173106,
      "tokens_seen": 1961623552
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004098495486459378,
      "loss": 1.498,
      "theoretical_loss": 3.4644056148294284,
      "tokens_seen": 1961689088
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004098294884653962,
      "loss": 1.3887,
      "theoretical_loss": 3.4643960078523697,
      "tokens_seen": 1961754624
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004098094282848546,
      "loss": 1.37,
      "theoretical_loss": 3.464386401286104,
      "tokens_seen": 1961820160
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040978936810431296,
      "loss": 1.4787,
      "theoretical_loss": 3.4643767951305997,
      "tokens_seen": 1961885696
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004097693079237713,
      "loss": 1.4302,
      "theoretical_loss": 3.4643671893858254,
      "tokens_seen": 1961951232
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004097492477432297,
      "loss": 1.3375,
      "theoretical_loss": 3.4643575840517498,
      "tokens_seen": 1962016768
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040972918756268804,
      "loss": 1.3618,
      "theoretical_loss": 3.464347979128342,
      "tokens_seen": 1962082304
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040970912738214646,
      "loss": 1.4056,
      "theoretical_loss": 3.4643383746155703,
      "tokens_seen": 1962147840
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004096890672016048,
      "loss": 1.3463,
      "theoretical_loss": 3.464328770513404,
      "tokens_seen": 1962213376
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004096690070210632,
      "loss": 1.4216,
      "theoretical_loss": 3.464319166821811,
      "tokens_seen": 1962278912
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040964894684052155,
      "loss": 1.4071,
      "theoretical_loss": 3.464309563540761,
      "tokens_seen": 1962344448
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040962888665997996,
      "loss": 1.4542,
      "theoretical_loss": 3.464299960670221,
      "tokens_seen": 1962409984
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004096088264794383,
      "loss": 1.4658,
      "theoretical_loss": 3.464290358210162,
      "tokens_seen": 1962475520
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004095887662988967,
      "loss": 1.4548,
      "theoretical_loss": 3.4642807561605515,
      "tokens_seen": 1962541056
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040956870611835505,
      "loss": 1.3988,
      "theoretical_loss": 3.4642711545213585,
      "tokens_seen": 1962606592
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004095486459378134,
      "loss": 1.4311,
      "theoretical_loss": 3.464261553292551,
      "tokens_seen": 1962672128
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040952858575727183,
      "loss": 1.4364,
      "theoretical_loss": 3.4642519524740996,
      "tokens_seen": 1962737664
    },
    {
      "epoch": 0.59,
      "objective/train/advantage_avg": 0.49501994252204895,
      "objective/train/docs_used": 715564,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2202401161193848,
      "objective/train/original_loss": 1.2202401161193848,
      "objective/train/theoretical_loss": 3.464242352065971,
      "objective/train/tokens_used": 1983263200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24527207016944885,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507489442825317,
      "objective/train/weighted_lm_loss": 1.2820630073547363,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.464242352065971,
      "tokens_seen": 1962803200
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004095085255767302,
      "loss": 1.4138,
      "theoretical_loss": 3.464242352065971,
      "tokens_seen": 1962803200
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040948846539618855,
      "loss": 1.497,
      "theoretical_loss": 3.4642327520681353,
      "tokens_seen": 1962868736
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004094684052156469,
      "loss": 1.4293,
      "theoretical_loss": 3.4642231524805607,
      "tokens_seen": 1962934272
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040944834503510533,
      "loss": 1.4189,
      "theoretical_loss": 3.464213553303216,
      "tokens_seen": 1962999808
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004094282848545637,
      "loss": 1.3653,
      "theoretical_loss": 3.4642039545360706,
      "tokens_seen": 1963065344
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040940822467402206,
      "loss": 1.5382,
      "theoretical_loss": 3.4641943561790924,
      "tokens_seen": 1963130880
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004093881644934804,
      "loss": 1.5261,
      "theoretical_loss": 3.4641847582322507,
      "tokens_seen": 1963196416
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004093681043129388,
      "loss": 1.5112,
      "theoretical_loss": 3.464175160695514,
      "tokens_seen": 1963261952
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004093480441323972,
      "loss": 1.4629,
      "theoretical_loss": 3.464165563568851,
      "tokens_seen": 1963327488
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00040932798395185556,
      "loss": 1.466,
      "theoretical_loss": 3.464155966852231,
      "tokens_seen": 1963393024
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0004093079237713139,
      "loss": 1.4817,
      "theoretical_loss": 3.4641463705456226,
      "tokens_seen": 1963458560
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004092878635907723,
      "loss": 1.4583,
      "theoretical_loss": 3.4641367746489946,
      "tokens_seen": 1963524096
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004092678034102307,
      "loss": 1.449,
      "theoretical_loss": 3.4641271791623156,
      "tokens_seen": 1963589632
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040924774322968906,
      "loss": 1.5003,
      "theoretical_loss": 3.464117584085554,
      "tokens_seen": 1963655168
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004092276830491474,
      "loss": 1.4346,
      "theoretical_loss": 3.4641079894186797,
      "tokens_seen": 1963720704
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004092076228686058,
      "loss": 1.3635,
      "theoretical_loss": 3.4640983951616606,
      "tokens_seen": 1963786240
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040918756268806415,
      "loss": 1.3708,
      "theoretical_loss": 3.464088801314466,
      "tokens_seen": 1963851776
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040916750250752257,
      "loss": 1.4283,
      "theoretical_loss": 3.464079207877065,
      "tokens_seen": 1963917312
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000409147442326981,
      "loss": 1.4292,
      "theoretical_loss": 3.4640696148494254,
      "tokens_seen": 1963982848
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040912738214643934,
      "loss": 1.4314,
      "theoretical_loss": 3.464060022231517,
      "tokens_seen": 1964048384
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004091073219658977,
      "loss": 1.4232,
      "theoretical_loss": 3.464050430023308,
      "tokens_seen": 1964113920
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004090872617853561,
      "loss": 1.4437,
      "theoretical_loss": 3.4640408382247676,
      "tokens_seen": 1964179456
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004090672016048145,
      "loss": 1.3521,
      "theoretical_loss": 3.464031246835864,
      "tokens_seen": 1964244992
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040904714142427285,
      "loss": 1.4135,
      "theoretical_loss": 3.464021655856567,
      "tokens_seen": 1964310528
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004090270812437312,
      "loss": 1.4272,
      "theoretical_loss": 3.4640120652868456,
      "tokens_seen": 1964376064
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040900702106318957,
      "loss": 1.4588,
      "theoretical_loss": 3.464002475126667,
      "tokens_seen": 1964441600
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000408986960882648,
      "loss": 1.385,
      "theoretical_loss": 3.4639928853760016,
      "tokens_seen": 1964507136
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040896690070210635,
      "loss": 1.4788,
      "theoretical_loss": 3.463983296034818,
      "tokens_seen": 1964572672
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004089468405215647,
      "loss": 1.3462,
      "theoretical_loss": 3.4639737071030843,
      "tokens_seen": 1964638208
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004089267803410231,
      "loss": 1.4054,
      "theoretical_loss": 3.46396411858077,
      "tokens_seen": 1964703744
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004089067201604815,
      "loss": 1.4934,
      "theoretical_loss": 3.463954530467844,
      "tokens_seen": 1964769280
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040888665997993985,
      "loss": 1.3623,
      "theoretical_loss": 3.463944942764275,
      "tokens_seen": 1964834816
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004088665997993982,
      "loss": 1.3923,
      "theoretical_loss": 3.4639353554700314,
      "tokens_seen": 1964900352
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004088465396188566,
      "loss": 1.3527,
      "theoretical_loss": 3.463925768585083,
      "tokens_seen": 1964965888
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000408826479438315,
      "loss": 1.4587,
      "theoretical_loss": 3.463916182109398,
      "tokens_seen": 1965031424
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040880641925777336,
      "loss": 1.4043,
      "theoretical_loss": 3.4639065960429454,
      "tokens_seen": 1965096960
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004087863590772317,
      "loss": 1.3936,
      "theoretical_loss": 3.4638970103856943,
      "tokens_seen": 1965162496
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004087662988966901,
      "loss": 1.485,
      "theoretical_loss": 3.4638874251376137,
      "tokens_seen": 1965228032
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040874623871614844,
      "loss": 1.4141,
      "theoretical_loss": 3.463877840298672,
      "tokens_seen": 1965293568
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040872617853560686,
      "loss": 1.4877,
      "theoretical_loss": 3.4638682558688383,
      "tokens_seen": 1965359104
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004087061183550652,
      "loss": 1.286,
      "theoretical_loss": 3.4638586718480817,
      "tokens_seen": 1965424640
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004086860581745236,
      "loss": 1.4046,
      "theoretical_loss": 3.463849088236371,
      "tokens_seen": 1965490176
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040866599799398195,
      "loss": 1.4775,
      "theoretical_loss": 3.4638395050336745,
      "tokens_seen": 1965555712
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040864593781344036,
      "loss": 1.4834,
      "theoretical_loss": 3.4638299222399622,
      "tokens_seen": 1965621248
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004086258776328987,
      "loss": 1.3564,
      "theoretical_loss": 3.4638203398552028,
      "tokens_seen": 1965686784
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004086058174523571,
      "loss": 1.4178,
      "theoretical_loss": 3.463810757879364,
      "tokens_seen": 1965752320
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040858575727181545,
      "loss": 1.5231,
      "theoretical_loss": 3.4638011763124164,
      "tokens_seen": 1965817856
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004085656970912738,
      "loss": 1.4188,
      "theoretical_loss": 3.463791595154327,
      "tokens_seen": 1965883392
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040854563691073223,
      "loss": 1.42,
      "theoretical_loss": 3.4637820144050666,
      "tokens_seen": 1965948928
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004085255767301906,
      "loss": 1.2976,
      "theoretical_loss": 3.4637724340646034,
      "tokens_seen": 1966014464
    },
    {
      "debugging/Compilability": 1.0,
      "debugging/distinct-1-grams": 0.7394139103740143,
      "debugging/entropy-1-grams": 5.1329642330953185,
      "debugging/length": 486.55555555555554,
      "debugging/num_segments": 9,
      "debugging/raw_token_scores_avg": 0.005027866922318935,
      "debugging/raw_token_scores_std": 0.022307263687253,
      "debugging/score": 0.00915515707107717,
      "debugging/score_std": 0.01661134296838666,
      "epoch": 0.6,
      "objective/train/advantage_avg": 0.49497345089912415,
      "objective/train/docs_used": 716760,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4728285074234009,
      "objective/train/original_loss": 1.4728285074234009,
      "objective/train/theoretical_loss": 3.4637628541329066,
      "objective/train/tokens_used": 1986540000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24549663066864014,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507454872131348,
      "objective/train/weighted_lm_loss": 1.5474436283111572,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.7788007855415344,
      "theoretical_loss": 3.4637628541329066,
      "tokens_seen": 1966080000
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040850551654964895,
      "loss": 1.5253,
      "theoretical_loss": 3.4637628541329066,
      "tokens_seen": 1966080000
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004084854563691073,
      "loss": 1.4174,
      "theoretical_loss": 3.463753274609944,
      "tokens_seen": 1966145536
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040846539618856573,
      "loss": 1.3519,
      "theoretical_loss": 3.463743695495686,
      "tokens_seen": 1966211072
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004084453360080241,
      "loss": 1.3913,
      "theoretical_loss": 3.463734116790101,
      "tokens_seen": 1966276608
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040842527582748246,
      "loss": 1.3612,
      "theoretical_loss": 3.4637245384931576,
      "tokens_seen": 1966342144
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004084052156469408,
      "loss": 1.4988,
      "theoretical_loss": 3.4637149606048245,
      "tokens_seen": 1966407680
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004083851554663992,
      "loss": 1.3846,
      "theoretical_loss": 3.463705383125072,
      "tokens_seen": 1966473216
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004083650952858576,
      "loss": 1.4392,
      "theoretical_loss": 3.4636958060538676,
      "tokens_seen": 1966538752
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040834503510531596,
      "loss": 1.4483,
      "theoretical_loss": 3.4636862293911816,
      "tokens_seen": 1966604288
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004083249749247743,
      "loss": 1.4009,
      "theoretical_loss": 3.4636766531369814,
      "tokens_seen": 1966669824
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004083049147442327,
      "loss": 1.4303,
      "theoretical_loss": 3.4636670772912375,
      "tokens_seen": 1966735360
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004082848545636911,
      "loss": 1.4072,
      "theoretical_loss": 3.463657501853918,
      "tokens_seen": 1966800896
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040826479438314946,
      "loss": 1.3913,
      "theoretical_loss": 3.463647926824992,
      "tokens_seen": 1966866432
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004082447342026078,
      "loss": 1.443,
      "theoretical_loss": 3.463638352204428,
      "tokens_seen": 1966931968
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004082246740220662,
      "loss": 1.4226,
      "theoretical_loss": 3.463628777992196,
      "tokens_seen": 1966997504
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040820461384152455,
      "loss": 1.5621,
      "theoretical_loss": 3.463619204188264,
      "tokens_seen": 1967063040
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040818455366098297,
      "loss": 1.309,
      "theoretical_loss": 3.4636096307926016,
      "tokens_seen": 1967128576
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040816449348044133,
      "loss": 1.3237,
      "theoretical_loss": 3.463600057805178,
      "tokens_seen": 1967194112
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004081444332998997,
      "loss": 1.5359,
      "theoretical_loss": 3.4635904852259616,
      "tokens_seen": 1967259648
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040812437311935805,
      "loss": 1.4224,
      "theoretical_loss": 3.4635809130549218,
      "tokens_seen": 1967325184
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040810431293881647,
      "loss": 1.4653,
      "theoretical_loss": 3.4635713412920275,
      "tokens_seen": 1967390720
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040808425275827483,
      "loss": 1.4941,
      "theoretical_loss": 3.463561769937247,
      "tokens_seen": 1967456256
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004080641925777332,
      "loss": 1.5181,
      "theoretical_loss": 3.4635521989905507,
      "tokens_seen": 1967521792
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040804413239719156,
      "loss": 1.4481,
      "theoretical_loss": 3.463542628451907,
      "tokens_seen": 1967587328
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004080240722166499,
      "loss": 1.3705,
      "theoretical_loss": 3.4635330583212838,
      "tokens_seen": 1967652864
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040800401203610834,
      "loss": 1.4337,
      "theoretical_loss": 3.4635234885986517,
      "tokens_seen": 1967718400
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004079839518555667,
      "loss": 1.583,
      "theoretical_loss": 3.463513919283979,
      "tokens_seen": 1967783936
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040796389167502506,
      "loss": 1.3972,
      "theoretical_loss": 3.4635043503772347,
      "tokens_seen": 1967849472
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004079438314944834,
      "loss": 1.339,
      "theoretical_loss": 3.463494781878388,
      "tokens_seen": 1967915008
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040792377131394184,
      "loss": 1.357,
      "theoretical_loss": 3.463485213787408,
      "tokens_seen": 1967980544
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004079037111334002,
      "loss": 1.5609,
      "theoretical_loss": 3.4634756461042633,
      "tokens_seen": 1968046080
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040788365095285856,
      "loss": 1.4753,
      "theoretical_loss": 3.4634660788289238,
      "tokens_seen": 1968111616
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004078635907723169,
      "loss": 1.5395,
      "theoretical_loss": 3.4634565119613576,
      "tokens_seen": 1968177152
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004078435305917753,
      "loss": 1.4128,
      "theoretical_loss": 3.463446945501534,
      "tokens_seen": 1968242688
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004078234704112337,
      "loss": 1.4423,
      "theoretical_loss": 3.4634373794494224,
      "tokens_seen": 1968308224
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040780341023069207,
      "loss": 1.4583,
      "theoretical_loss": 3.4634278138049917,
      "tokens_seen": 1968373760
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040778335005015043,
      "loss": 1.472,
      "theoretical_loss": 3.463418248568211,
      "tokens_seen": 1968439296
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004077632898696088,
      "loss": 1.4046,
      "theoretical_loss": 3.4634086837390488,
      "tokens_seen": 1968504832
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004077432296890672,
      "loss": 1.3685,
      "theoretical_loss": 3.463399119317475,
      "tokens_seen": 1968570368
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040772316950852557,
      "loss": 1.3835,
      "theoretical_loss": 3.4633895553034577,
      "tokens_seen": 1968635904
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040770310932798393,
      "loss": 1.427,
      "theoretical_loss": 3.463379991696967,
      "tokens_seen": 1968701440
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004076830491474423,
      "loss": 1.4431,
      "theoretical_loss": 3.463370428497971,
      "tokens_seen": 1968766976
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004076629889669007,
      "loss": 1.3941,
      "theoretical_loss": 3.4633608657064396,
      "tokens_seen": 1968832512
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004076429287863591,
      "loss": 1.4504,
      "theoretical_loss": 3.4633513033223418,
      "tokens_seen": 1968898048
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004076228686058175,
      "loss": 1.362,
      "theoretical_loss": 3.463341741345646,
      "tokens_seen": 1968963584
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040760280842527585,
      "loss": 1.3839,
      "theoretical_loss": 3.463332179776322,
      "tokens_seen": 1969029120
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004075827482447342,
      "loss": 1.3549,
      "theoretical_loss": 3.463322618614338,
      "tokens_seen": 1969094656
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040756268806419263,
      "loss": 1.3454,
      "theoretical_loss": 3.463313057859665,
      "tokens_seen": 1969160192
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000407542627883651,
      "loss": 1.4043,
      "theoretical_loss": 3.4633034975122694,
      "tokens_seen": 1969225728
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040752256770310935,
      "loss": 1.5057,
      "theoretical_loss": 3.4632939375721223,
      "tokens_seen": 1969291264
    },
    {
      "epoch": 0.6,
      "objective/train/advantage_avg": 0.49629923701286316,
      "objective/train/docs_used": 718017,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4240174293518066,
      "objective/train/original_loss": 1.4240174293518066,
      "objective/train/theoretical_loss": 3.4632843780391918,
      "objective/train/tokens_used": 1989816800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24651269614696503,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508832931518555,
      "objective/train/weighted_lm_loss": 1.4963409900665283,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.4632843780391918,
      "tokens_seen": 1969356800
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004075025075225677,
      "loss": 1.4286,
      "theoretical_loss": 3.4632843780391918,
      "tokens_seen": 1969356800
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040748244734202613,
      "loss": 1.4168,
      "theoretical_loss": 3.463274818913448,
      "tokens_seen": 1969422336
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004074623871614845,
      "loss": 1.4078,
      "theoretical_loss": 3.4632652601948593,
      "tokens_seen": 1969487872
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040744232698094286,
      "loss": 1.5736,
      "theoretical_loss": 3.4632557018833943,
      "tokens_seen": 1969553408
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004074222668004012,
      "loss": 1.4548,
      "theoretical_loss": 3.463246143979023,
      "tokens_seen": 1969618944
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004074022066198596,
      "loss": 1.3283,
      "theoretical_loss": 3.463236586481714,
      "tokens_seen": 1969684480
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000407382146439318,
      "loss": 1.3244,
      "theoretical_loss": 3.463227029391437,
      "tokens_seen": 1969750016
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040736208625877636,
      "loss": 1.4088,
      "theoretical_loss": 3.463217472708161,
      "tokens_seen": 1969815552
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004073420260782347,
      "loss": 1.5282,
      "theoretical_loss": 3.4632079164318545,
      "tokens_seen": 1969881088
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004073219658976931,
      "loss": 1.3963,
      "theoretical_loss": 3.4631983605624868,
      "tokens_seen": 1969946624
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004073019057171515,
      "loss": 1.4304,
      "theoretical_loss": 3.4631888051000272,
      "tokens_seen": 1970012160
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040728184553660986,
      "loss": 1.2918,
      "theoretical_loss": 3.4631792500444454,
      "tokens_seen": 1970077696
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004072617853560682,
      "loss": 1.5151,
      "theoretical_loss": 3.4631696953957096,
      "tokens_seen": 1970143232
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004072417251755266,
      "loss": 1.2829,
      "theoretical_loss": 3.4631601411537902,
      "tokens_seen": 1970208768
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040722166499498495,
      "loss": 1.4793,
      "theoretical_loss": 3.4631505873186548,
      "tokens_seen": 1970274304
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040720160481444337,
      "loss": 1.5214,
      "theoretical_loss": 3.463141033890273,
      "tokens_seen": 1970339840
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040718154463390173,
      "loss": 1.3832,
      "theoretical_loss": 3.463131480868615,
      "tokens_seen": 1970405376
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004071614844533601,
      "loss": 1.3941,
      "theoretical_loss": 3.463121928253649,
      "tokens_seen": 1970470912
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040714142427281845,
      "loss": 1.2594,
      "theoretical_loss": 3.463112376045344,
      "tokens_seen": 1970536448
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040712136409227687,
      "loss": 1.4069,
      "theoretical_loss": 3.46310282424367,
      "tokens_seen": 1970601984
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040710130391173523,
      "loss": 1.3762,
      "theoretical_loss": 3.4630932728485955,
      "tokens_seen": 1970667520
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004070812437311936,
      "loss": 1.4366,
      "theoretical_loss": 3.4630837218600896,
      "tokens_seen": 1970733056
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040706118355065196,
      "loss": 1.4226,
      "theoretical_loss": 3.463074171278122,
      "tokens_seen": 1970798592
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004070411233701103,
      "loss": 1.3626,
      "theoretical_loss": 3.4630646211026614,
      "tokens_seen": 1970864128
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040702106318956874,
      "loss": 1.3851,
      "theoretical_loss": 3.4630550713336774,
      "tokens_seen": 1970929664
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004070010030090271,
      "loss": 1.3779,
      "theoretical_loss": 3.463045521971139,
      "tokens_seen": 1970995200
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040698094282848546,
      "loss": 1.5004,
      "theoretical_loss": 3.4630359730150153,
      "tokens_seen": 1971060736
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004069608826479438,
      "loss": 1.4689,
      "theoretical_loss": 3.4630264244652755,
      "tokens_seen": 1971126272
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040694082246740224,
      "loss": 1.3777,
      "theoretical_loss": 3.4630168763218885,
      "tokens_seen": 1971191808
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004069207622868606,
      "loss": 1.5411,
      "theoretical_loss": 3.4630073285848244,
      "tokens_seen": 1971257344
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040690070210631896,
      "loss": 1.3712,
      "theoretical_loss": 3.4629977812540518,
      "tokens_seen": 1971322880
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004068806419257773,
      "loss": 1.4871,
      "theoretical_loss": 3.46298823432954,
      "tokens_seen": 1971388416
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004068605817452357,
      "loss": 1.5517,
      "theoretical_loss": 3.4629786878112574,
      "tokens_seen": 1971453952
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004068405215646941,
      "loss": 1.4396,
      "theoretical_loss": 3.462969141699175,
      "tokens_seen": 1971519488
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040682046138415247,
      "loss": 1.4194,
      "theoretical_loss": 3.46295959599326,
      "tokens_seen": 1971585024
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040680040120361083,
      "loss": 1.3953,
      "theoretical_loss": 3.4629500506934834,
      "tokens_seen": 1971650560
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004067803410230692,
      "loss": 1.4816,
      "theoretical_loss": 3.4629405057998133,
      "tokens_seen": 1971716096
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004067602808425276,
      "loss": 1.3907,
      "theoretical_loss": 3.4629309613122192,
      "tokens_seen": 1971781632
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040674022066198597,
      "loss": 1.4869,
      "theoretical_loss": 3.4629214172306706,
      "tokens_seen": 1971847168
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040672016048144433,
      "loss": 1.4545,
      "theoretical_loss": 3.4629118735551363,
      "tokens_seen": 1971912704
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004067001003009027,
      "loss": 1.5684,
      "theoretical_loss": 3.4629023302855857,
      "tokens_seen": 1971978240
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004066800401203611,
      "loss": 1.5132,
      "theoretical_loss": 3.462892787421988,
      "tokens_seen": 1972043776
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040665997993981947,
      "loss": 1.3028,
      "theoretical_loss": 3.462883244964313,
      "tokens_seen": 1972109312
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040663991975927783,
      "loss": 1.3473,
      "theoretical_loss": 3.462873702912529,
      "tokens_seen": 1972174848
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004066198595787362,
      "loss": 1.3886,
      "theoretical_loss": 3.462864161266606,
      "tokens_seen": 1972240384
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040659979939819456,
      "loss": 1.448,
      "theoretical_loss": 3.462854620026513,
      "tokens_seen": 1972305920
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000406579739217653,
      "loss": 1.3028,
      "theoretical_loss": 3.4628450791922187,
      "tokens_seen": 1972371456
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040655967903711134,
      "loss": 1.4017,
      "theoretical_loss": 3.4628355387636933,
      "tokens_seen": 1972436992
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004065396188565697,
      "loss": 1.3812,
      "theoretical_loss": 3.4628259987409056,
      "tokens_seen": 1972502528
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040651955867602806,
      "loss": 1.3159,
      "theoretical_loss": 3.4628164591238253,
      "tokens_seen": 1972568064
    },
    {
      "epoch": 0.6,
      "objective/train/advantage_avg": 0.4938172399997711,
      "objective/train/docs_used": 719209,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2467628717422485,
      "objective/train/original_loss": 1.246762990951538,
      "objective/train/theoretical_loss": 3.4628069199124205,
      "objective/train/tokens_used": 1993093600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2441314309835434,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506229400634766,
      "objective/train/weighted_lm_loss": 1.309868335723877,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4628069199124205,
      "tokens_seen": 1972633600
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004064994984954865,
      "loss": 1.3739,
      "theoretical_loss": 3.4628069199124205,
      "tokens_seen": 1972633600
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040647943831494484,
      "loss": 1.2847,
      "theoretical_loss": 3.462797381106662,
      "tokens_seen": 1972699136
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004064593781344032,
      "loss": 1.4595,
      "theoretical_loss": 3.462787842706518,
      "tokens_seen": 1972764672
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040643931795386157,
      "loss": 1.3835,
      "theoretical_loss": 3.462778304711958,
      "tokens_seen": 1972830208
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040641925777331993,
      "loss": 1.4742,
      "theoretical_loss": 3.4627687671229515,
      "tokens_seen": 1972895744
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040639919759277834,
      "loss": 1.331,
      "theoretical_loss": 3.4627592299394676,
      "tokens_seen": 1972961280
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004063791374122367,
      "loss": 1.4772,
      "theoretical_loss": 3.4627496931614754,
      "tokens_seen": 1973026816
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040635907723169507,
      "loss": 1.4576,
      "theoretical_loss": 3.4627401567889446,
      "tokens_seen": 1973092352
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040633901705115343,
      "loss": 1.435,
      "theoretical_loss": 3.462730620821845,
      "tokens_seen": 1973157888
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040631895687061185,
      "loss": 1.4508,
      "theoretical_loss": 3.4627210852601444,
      "tokens_seen": 1973223424
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004062988966900702,
      "loss": 1.4492,
      "theoretical_loss": 3.4627115501038133,
      "tokens_seen": 1973288960
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040627883650952857,
      "loss": 1.433,
      "theoretical_loss": 3.4627020153528205,
      "tokens_seen": 1973354496
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040625877632898693,
      "loss": 1.4142,
      "theoretical_loss": 3.462692481007136,
      "tokens_seen": 1973420032
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004062387161484453,
      "loss": 1.4273,
      "theoretical_loss": 3.462682947066728,
      "tokens_seen": 1973485568
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004062186559679037,
      "loss": 1.4018,
      "theoretical_loss": 3.4626734135315664,
      "tokens_seen": 1973551104
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004061985957873621,
      "loss": 1.3199,
      "theoretical_loss": 3.4626638804016205,
      "tokens_seen": 1973616640
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040617853560682044,
      "loss": 1.4682,
      "theoretical_loss": 3.46265434767686,
      "tokens_seen": 1973682176
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004061584754262788,
      "loss": 1.4326,
      "theoretical_loss": 3.4626448153572533,
      "tokens_seen": 1973747712
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040613841524573727,
      "loss": 1.36,
      "theoretical_loss": 3.4626352834427707,
      "tokens_seen": 1973813248
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040611835506519563,
      "loss": 1.4885,
      "theoretical_loss": 3.462625751933381,
      "tokens_seen": 1973878784
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000406098294884654,
      "loss": 1.3596,
      "theoretical_loss": 3.4626162208290534,
      "tokens_seen": 1973944320
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040607823470411236,
      "loss": 1.4434,
      "theoretical_loss": 3.462606690129758,
      "tokens_seen": 1974009856
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004060581745235707,
      "loss": 1.3791,
      "theoretical_loss": 3.4625971598354632,
      "tokens_seen": 1974075392
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040603811434302914,
      "loss": 1.4246,
      "theoretical_loss": 3.462587629946139,
      "tokens_seen": 1974140928
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004060180541624875,
      "loss": 1.4764,
      "theoretical_loss": 3.462578100461754,
      "tokens_seen": 1974206464
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040599799398194586,
      "loss": 1.5049,
      "theoretical_loss": 3.4625685713822785,
      "tokens_seen": 1974272000
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004059779338014042,
      "loss": 1.4331,
      "theoretical_loss": 3.4625590427076816,
      "tokens_seen": 1974337536
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040595787362086264,
      "loss": 1.4153,
      "theoretical_loss": 3.4625495144379324,
      "tokens_seen": 1974403072
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000405937813440321,
      "loss": 1.3637,
      "theoretical_loss": 3.4625399865730007,
      "tokens_seen": 1974468608
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040591775325977936,
      "loss": 1.3863,
      "theoretical_loss": 3.4625304591128545,
      "tokens_seen": 1974534144
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004058976930792377,
      "loss": 1.3612,
      "theoretical_loss": 3.462520932057465,
      "tokens_seen": 1974599680
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004058776328986961,
      "loss": 1.414,
      "theoretical_loss": 3.4625114054068007,
      "tokens_seen": 1974665216
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004058575727181545,
      "loss": 1.3933,
      "theoretical_loss": 3.462501879160831,
      "tokens_seen": 1974730752
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040583751253761287,
      "loss": 1.479,
      "theoretical_loss": 3.462492353319525,
      "tokens_seen": 1974796288
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040581745235707123,
      "loss": 1.4626,
      "theoretical_loss": 3.4624828278828526,
      "tokens_seen": 1974861824
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004057973921765296,
      "loss": 1.4346,
      "theoretical_loss": 3.4624733028507824,
      "tokens_seen": 1974927360
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000405777331995988,
      "loss": 1.4558,
      "theoretical_loss": 3.462463778223285,
      "tokens_seen": 1974992896
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040575727181544637,
      "loss": 1.4422,
      "theoretical_loss": 3.4624542540003294,
      "tokens_seen": 1975058432
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040573721163490473,
      "loss": 1.3766,
      "theoretical_loss": 3.462444730181884,
      "tokens_seen": 1975123968
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004057171514543631,
      "loss": 1.3487,
      "theoretical_loss": 3.4624352067679194,
      "tokens_seen": 1975189504
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040569709127382146,
      "loss": 1.4777,
      "theoretical_loss": 3.462425683758404,
      "tokens_seen": 1975255040
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004056770310932799,
      "loss": 1.5249,
      "theoretical_loss": 3.4624161611533086,
      "tokens_seen": 1975320576
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040565697091273824,
      "loss": 1.4772,
      "theoretical_loss": 3.4624066389526016,
      "tokens_seen": 1975386112
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004056369107321966,
      "loss": 1.4489,
      "theoretical_loss": 3.462397117156252,
      "tokens_seen": 1975451648
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040561685055165496,
      "loss": 1.5382,
      "theoretical_loss": 3.4623875957642305,
      "tokens_seen": 1975517184
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004055967903711134,
      "loss": 1.3988,
      "theoretical_loss": 3.462378074776505,
      "tokens_seen": 1975582720
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040557673019057174,
      "loss": 1.4112,
      "theoretical_loss": 3.4623685541930467,
      "tokens_seen": 1975648256
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004055566700100301,
      "loss": 1.4352,
      "theoretical_loss": 3.462359034013823,
      "tokens_seen": 1975713792
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040553660982948846,
      "loss": 1.5392,
      "theoretical_loss": 3.462349514238805,
      "tokens_seen": 1975779328
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004055165496489469,
      "loss": 1.4623,
      "theoretical_loss": 3.4623399948679614,
      "tokens_seen": 1975844864
    },
    {
      "epoch": 0.6,
      "objective/train/advantage_avg": 0.4915984272956848,
      "objective/train/docs_used": 720450,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5267773866653442,
      "objective/train/original_loss": 1.5267775058746338,
      "objective/train/theoretical_loss": 3.462330475901262,
      "objective/train/tokens_used": 1996370400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2424638271331787,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050392508506775,
      "objective/train/weighted_lm_loss": 1.6036399602890015,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.462330475901262,
      "tokens_seen": 1975910400
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040549648946840524,
      "loss": 1.572,
      "theoretical_loss": 3.462330475901262,
      "tokens_seen": 1975910400
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004054764292878636,
      "loss": 1.4637,
      "theoretical_loss": 3.462320957338676,
      "tokens_seen": 1975975936
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040545636910732197,
      "loss": 1.3565,
      "theoretical_loss": 3.462311439180173,
      "tokens_seen": 1976041472
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040543630892678033,
      "loss": 1.4189,
      "theoretical_loss": 3.4623019214257216,
      "tokens_seen": 1976107008
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040541624874623874,
      "loss": 1.41,
      "theoretical_loss": 3.462292404075292,
      "tokens_seen": 1976172544
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004053961885656971,
      "loss": 1.3285,
      "theoretical_loss": 3.462282887128854,
      "tokens_seen": 1976238080
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040537612838515547,
      "loss": 1.4471,
      "theoretical_loss": 3.4622733705863764,
      "tokens_seen": 1976303616
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040535606820461383,
      "loss": 1.2833,
      "theoretical_loss": 3.462263854447829,
      "tokens_seen": 1976369152
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040533600802407225,
      "loss": 1.4049,
      "theoretical_loss": 3.462254338713181,
      "tokens_seen": 1976434688
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004053159478435306,
      "loss": 1.3219,
      "theoretical_loss": 3.462244823382402,
      "tokens_seen": 1976500224
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040529588766298897,
      "loss": 1.4183,
      "theoretical_loss": 3.462235308455462,
      "tokens_seen": 1976565760
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040527582748244733,
      "loss": 1.519,
      "theoretical_loss": 3.462225793932329,
      "tokens_seen": 1976631296
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004052557673019057,
      "loss": 1.4565,
      "theoretical_loss": 3.4622162798129743,
      "tokens_seen": 1976696832
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004052357071213641,
      "loss": 1.424,
      "theoretical_loss": 3.462206766097366,
      "tokens_seen": 1976762368
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004052156469408225,
      "loss": 1.3035,
      "theoretical_loss": 3.4621972527854745,
      "tokens_seen": 1976827904
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040519558676028084,
      "loss": 1.4769,
      "theoretical_loss": 3.462187739877269,
      "tokens_seen": 1976893440
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004051755265797392,
      "loss": 1.4471,
      "theoretical_loss": 3.462178227372718,
      "tokens_seen": 1976958976
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004051554663991976,
      "loss": 1.4731,
      "theoretical_loss": 3.4621687152717926,
      "tokens_seen": 1977024512
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000405135406218656,
      "loss": 1.4477,
      "theoretical_loss": 3.462159203574461,
      "tokens_seen": 1977090048
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040511534603811434,
      "loss": 1.3845,
      "theoretical_loss": 3.4621496922806942,
      "tokens_seen": 1977155584
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004050952858575727,
      "loss": 1.4315,
      "theoretical_loss": 3.4621401813904598,
      "tokens_seen": 1977221120
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040507522567703107,
      "loss": 1.3628,
      "theoretical_loss": 3.4621306709037287,
      "tokens_seen": 1977286656
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004050551654964895,
      "loss": 1.5145,
      "theoretical_loss": 3.46212116082047,
      "tokens_seen": 1977352192
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040503510531594784,
      "loss": 1.4096,
      "theoretical_loss": 3.4621116511406527,
      "tokens_seen": 1977417728
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004050150451354062,
      "loss": 1.3906,
      "theoretical_loss": 3.4621021418642473,
      "tokens_seen": 1977483264
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040499498495486457,
      "loss": 1.4631,
      "theoretical_loss": 3.4620926329912227,
      "tokens_seen": 1977548800
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000404974924774323,
      "loss": 1.5372,
      "theoretical_loss": 3.4620831245215484,
      "tokens_seen": 1977614336
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040495486459378135,
      "loss": 1.4619,
      "theoretical_loss": 3.4620736164551946,
      "tokens_seen": 1977679872
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004049348044132397,
      "loss": 1.327,
      "theoretical_loss": 3.46206410879213,
      "tokens_seen": 1977745408
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040491474423269807,
      "loss": 1.4757,
      "theoretical_loss": 3.4620546015323237,
      "tokens_seen": 1977810944
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040489468405215643,
      "loss": 1.4487,
      "theoretical_loss": 3.462045094675747,
      "tokens_seen": 1977876480
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040487462387161485,
      "loss": 1.3926,
      "theoretical_loss": 3.462035588222368,
      "tokens_seen": 1977942016
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004048545636910732,
      "loss": 1.4542,
      "theoretical_loss": 3.462026082172156,
      "tokens_seen": 1978007552
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004048345035105316,
      "loss": 1.387,
      "theoretical_loss": 3.462016576525082,
      "tokens_seen": 1978073088
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040481444332998994,
      "loss": 1.4269,
      "theoretical_loss": 3.462007071281114,
      "tokens_seen": 1978138624
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040479438314944835,
      "loss": 1.4233,
      "theoretical_loss": 3.461997566440223,
      "tokens_seen": 1978204160
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004047743229689067,
      "loss": 1.4313,
      "theoretical_loss": 3.4619880620023777,
      "tokens_seen": 1978269696
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004047542627883651,
      "loss": 1.2853,
      "theoretical_loss": 3.461978557967548,
      "tokens_seen": 1978335232
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040473420260782344,
      "loss": 1.3887,
      "theoretical_loss": 3.461969054335703,
      "tokens_seen": 1978400768
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004047141424272818,
      "loss": 1.4051,
      "theoretical_loss": 3.461959551106812,
      "tokens_seen": 1978466304
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004046940822467402,
      "loss": 1.4916,
      "theoretical_loss": 3.4619500482808454,
      "tokens_seen": 1978531840
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004046740220661986,
      "loss": 1.3343,
      "theoretical_loss": 3.4619405458577726,
      "tokens_seen": 1978597376
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000404653961885657,
      "loss": 1.4764,
      "theoretical_loss": 3.461931043837563,
      "tokens_seen": 1978662912
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040463390170511536,
      "loss": 1.3268,
      "theoretical_loss": 3.4619215422201863,
      "tokens_seen": 1978728448
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004046138415245738,
      "loss": 1.462,
      "theoretical_loss": 3.4619120410056117,
      "tokens_seen": 1978793984
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040459378134403214,
      "loss": 1.4506,
      "theoretical_loss": 3.461902540193809,
      "tokens_seen": 1978859520
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004045737211634905,
      "loss": 1.402,
      "theoretical_loss": 3.4618930397847487,
      "tokens_seen": 1978925056
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040455366098294886,
      "loss": 1.2424,
      "theoretical_loss": 3.4618835397783982,
      "tokens_seen": 1978990592
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004045336008024073,
      "loss": 1.3898,
      "theoretical_loss": 3.4618740401747297,
      "tokens_seen": 1979056128
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040451354062186564,
      "loss": 1.513,
      "theoretical_loss": 3.461864540973711,
      "tokens_seen": 1979121664
    },
    {
      "epoch": 0.6,
      "objective/train/advantage_avg": 0.4959181547164917,
      "objective/train/docs_used": 721537,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3963663578033447,
      "objective/train/original_loss": 1.3963663578033447,
      "objective/train/theoretical_loss": 3.461855042175312,
      "objective/train/tokens_used": 1999647200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24620689451694489,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050843596458435,
      "objective/train/weighted_lm_loss": 1.4673304557800293,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.461855042175312,
      "tokens_seen": 1979187200
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000404493480441324,
      "loss": 1.473,
      "theoretical_loss": 3.461855042175312,
      "tokens_seen": 1979187200
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040447342026078237,
      "loss": 1.446,
      "theoretical_loss": 3.461845543779503,
      "tokens_seen": 1979252736
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040445336008024073,
      "loss": 1.4506,
      "theoretical_loss": 3.461836045786253,
      "tokens_seen": 1979318272
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040443329989969915,
      "loss": 1.4025,
      "theoretical_loss": 3.461826548195532,
      "tokens_seen": 1979383808
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004044132397191575,
      "loss": 1.4211,
      "theoretical_loss": 3.4618170510073085,
      "tokens_seen": 1979449344
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040439317953861587,
      "loss": 1.451,
      "theoretical_loss": 3.461807554221554,
      "tokens_seen": 1979514880
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040437311935807423,
      "loss": 1.4468,
      "theoretical_loss": 3.4617980578382364,
      "tokens_seen": 1979580416
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040435305917753265,
      "loss": 1.4815,
      "theoretical_loss": 3.4617885618573268,
      "tokens_seen": 1979645952
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000404332998996991,
      "loss": 1.3999,
      "theoretical_loss": 3.4617790662787935,
      "tokens_seen": 1979711488
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040431293881644937,
      "loss": 1.4679,
      "theoretical_loss": 3.461769571102607,
      "tokens_seen": 1979777024
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040429287863590773,
      "loss": 1.5785,
      "theoretical_loss": 3.461760076328736,
      "tokens_seen": 1979842560
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004042728184553661,
      "loss": 1.4254,
      "theoretical_loss": 3.4617505819571512,
      "tokens_seen": 1979908096
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004042527582748245,
      "loss": 1.5383,
      "theoretical_loss": 3.4617410879878223,
      "tokens_seen": 1979973632
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004042326980942829,
      "loss": 1.4221,
      "theoretical_loss": 3.4617315944207174,
      "tokens_seen": 1980039168
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040421263791374124,
      "loss": 1.4818,
      "theoretical_loss": 3.461722101255808,
      "tokens_seen": 1980104704
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004041925777331996,
      "loss": 1.4543,
      "theoretical_loss": 3.461712608493063,
      "tokens_seen": 1980170240
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000404172517552658,
      "loss": 1.3767,
      "theoretical_loss": 3.461703116132452,
      "tokens_seen": 1980235776
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004041524573721164,
      "loss": 1.4305,
      "theoretical_loss": 3.461693624173944,
      "tokens_seen": 1980301312
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040413239719157474,
      "loss": 1.4489,
      "theoretical_loss": 3.46168413261751,
      "tokens_seen": 1980366848
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004041123370110331,
      "loss": 1.396,
      "theoretical_loss": 3.4616746414631185,
      "tokens_seen": 1980432384
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040409227683049147,
      "loss": 1.4942,
      "theoretical_loss": 3.46166515071074,
      "tokens_seen": 1980497920
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004040722166499499,
      "loss": 1.5091,
      "theoretical_loss": 3.4616556603603437,
      "tokens_seen": 1980563456
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040405215646940824,
      "loss": 1.4511,
      "theoretical_loss": 3.4616461704118997,
      "tokens_seen": 1980628992
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004040320962888666,
      "loss": 1.3039,
      "theoretical_loss": 3.461636680865377,
      "tokens_seen": 1980694528
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040401203610832497,
      "loss": 1.4004,
      "theoretical_loss": 3.461627191720745,
      "tokens_seen": 1980760064
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004039919759277834,
      "loss": 1.409,
      "theoretical_loss": 3.461617702977975,
      "tokens_seen": 1980825600
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040397191574724175,
      "loss": 1.393,
      "theoretical_loss": 3.4616082146370353,
      "tokens_seen": 1980891136
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004039518555667001,
      "loss": 1.5191,
      "theoretical_loss": 3.4615987266978965,
      "tokens_seen": 1980956672
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040393179538615847,
      "loss": 1.4381,
      "theoretical_loss": 3.461589239160528,
      "tokens_seen": 1981022208
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040391173520561683,
      "loss": 1.4993,
      "theoretical_loss": 3.4615797520248983,
      "tokens_seen": 1981087744
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040389167502507525,
      "loss": 1.3679,
      "theoretical_loss": 3.4615702652909786,
      "tokens_seen": 1981153280
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004038716148445336,
      "loss": 1.2853,
      "theoretical_loss": 3.461560778958738,
      "tokens_seen": 1981218816
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000403851554663992,
      "loss": 1.5066,
      "theoretical_loss": 3.4615512930281467,
      "tokens_seen": 1981284352
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040383149448345034,
      "loss": 1.3474,
      "theoretical_loss": 3.4615418074991737,
      "tokens_seen": 1981349888
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040381143430290875,
      "loss": 1.4782,
      "theoretical_loss": 3.461532322371789,
      "tokens_seen": 1981415424
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004037913741223671,
      "loss": 1.3772,
      "theoretical_loss": 3.4615228376459624,
      "tokens_seen": 1981480960
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004037713139418255,
      "loss": 1.394,
      "theoretical_loss": 3.4615133533216635,
      "tokens_seen": 1981546496
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040375125376128384,
      "loss": 1.4241,
      "theoretical_loss": 3.4615038693988622,
      "tokens_seen": 1981612032
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004037311935807422,
      "loss": 1.5368,
      "theoretical_loss": 3.4614943858775282,
      "tokens_seen": 1981677568
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004037111334002006,
      "loss": 1.3127,
      "theoretical_loss": 3.461484902757631,
      "tokens_seen": 1981743104
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000403691073219659,
      "loss": 1.3771,
      "theoretical_loss": 3.4614754200391404,
      "tokens_seen": 1981808640
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040367101303911734,
      "loss": 1.3345,
      "theoretical_loss": 3.4614659377220263,
      "tokens_seen": 1981874176
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004036509528585757,
      "loss": 1.3336,
      "theoretical_loss": 3.4614564558062586,
      "tokens_seen": 1981939712
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004036308926780341,
      "loss": 1.4397,
      "theoretical_loss": 3.4614469742918064,
      "tokens_seen": 1982005248
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004036108324974925,
      "loss": 1.3435,
      "theoretical_loss": 3.4614374931786402,
      "tokens_seen": 1982070784
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040359077231695085,
      "loss": 1.3374,
      "theoretical_loss": 3.461428012466729,
      "tokens_seen": 1982136320
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004035707121364092,
      "loss": 1.4178,
      "theoretical_loss": 3.461418532156043,
      "tokens_seen": 1982201856
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040355065195586757,
      "loss": 1.3646,
      "theoretical_loss": 3.461409052246552,
      "tokens_seen": 1982267392
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000403530591775326,
      "loss": 1.4384,
      "theoretical_loss": 3.461399572738226,
      "tokens_seen": 1982332928
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040351053159478435,
      "loss": 1.4673,
      "theoretical_loss": 3.4613900936310342,
      "tokens_seen": 1982398464
    },
    {
      "epoch": 0.6,
      "objective/train/advantage_avg": 0.49234938621520996,
      "objective/train/docs_used": 722739,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3133035898208618,
      "objective/train/original_loss": 1.3133034706115723,
      "objective/train/theoretical_loss": 3.4613806149249466,
      "objective/train/tokens_used": 2002924000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24285843968391418,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504695177078247,
      "objective/train/weighted_lm_loss": 1.3795260190963745,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 3.4613806149249466,
      "tokens_seen": 1982464000
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004034904714142427,
      "loss": 1.4002,
      "theoretical_loss": 3.4613806149249466,
      "tokens_seen": 1982464000
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004034704112337011,
      "loss": 1.441,
      "theoretical_loss": 3.461371136619933,
      "tokens_seen": 1982529536
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004034503510531595,
      "loss": 1.3956,
      "theoretical_loss": 3.461361658715963,
      "tokens_seen": 1982595072
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040343029087261785,
      "loss": 1.5442,
      "theoretical_loss": 3.461352181213006,
      "tokens_seen": 1982660608
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004034102306920762,
      "loss": 1.4451,
      "theoretical_loss": 3.4613427041110327,
      "tokens_seen": 1982726144
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004033901705115346,
      "loss": 1.3769,
      "theoretical_loss": 3.461333227410013,
      "tokens_seen": 1982791680
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000403370110330993,
      "loss": 1.4062,
      "theoretical_loss": 3.4613237511099157,
      "tokens_seen": 1982857216
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040335005015045136,
      "loss": 1.396,
      "theoretical_loss": 3.461314275210711,
      "tokens_seen": 1982922752
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004033299899699097,
      "loss": 1.3159,
      "theoretical_loss": 3.4613047997123685,
      "tokens_seen": 1982988288
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004033099297893681,
      "loss": 1.4648,
      "theoretical_loss": 3.4612953246148583,
      "tokens_seen": 1983053824
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040328986960882644,
      "loss": 1.4969,
      "theoretical_loss": 3.4612858499181502,
      "tokens_seen": 1983119360
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040326980942828486,
      "loss": 1.5053,
      "theoretical_loss": 3.461276375622214,
      "tokens_seen": 1983184896
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004032497492477432,
      "loss": 1.4642,
      "theoretical_loss": 3.46126690172702,
      "tokens_seen": 1983250432
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004032296890672016,
      "loss": 1.3372,
      "theoretical_loss": 3.4612574282325363,
      "tokens_seen": 1983315968
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040320962888665995,
      "loss": 1.4995,
      "theoretical_loss": 3.4612479551387345,
      "tokens_seen": 1983381504
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040318956870611836,
      "loss": 1.4645,
      "theoretical_loss": 3.4612384824455837,
      "tokens_seen": 1983447040
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004031695085255767,
      "loss": 1.4514,
      "theoretical_loss": 3.461229010153054,
      "tokens_seen": 1983512576
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040314944834503514,
      "loss": 1.3869,
      "theoretical_loss": 3.4612195382611146,
      "tokens_seen": 1983578112
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004031293881644935,
      "loss": 1.4377,
      "theoretical_loss": 3.461210066769736,
      "tokens_seen": 1983643648
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040310932798395187,
      "loss": 1.4516,
      "theoretical_loss": 3.4612005956788874,
      "tokens_seen": 1983709184
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004030892678034103,
      "loss": 1.4407,
      "theoretical_loss": 3.4611911249885394,
      "tokens_seen": 1983774720
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040306920762286864,
      "loss": 1.3808,
      "theoretical_loss": 3.4611816546986613,
      "tokens_seen": 1983840256
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000403049147442327,
      "loss": 1.4465,
      "theoretical_loss": 3.4611721848092225,
      "tokens_seen": 1983905792
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040302908726178537,
      "loss": 1.3607,
      "theoretical_loss": 3.461162715320194,
      "tokens_seen": 1983971328
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004030090270812438,
      "loss": 1.4261,
      "theoretical_loss": 3.461153246231545,
      "tokens_seen": 1984036864
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040298896690070215,
      "loss": 1.4954,
      "theoretical_loss": 3.4611437775432456,
      "tokens_seen": 1984102400
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004029689067201605,
      "loss": 1.4581,
      "theoretical_loss": 3.461134309255265,
      "tokens_seen": 1984167936
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040294884653961887,
      "loss": 1.3473,
      "theoretical_loss": 3.4611248413675737,
      "tokens_seen": 1984233472
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040292878635907723,
      "loss": 1.3835,
      "theoretical_loss": 3.461115373880141,
      "tokens_seen": 1984299008
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040290872617853565,
      "loss": 1.4576,
      "theoretical_loss": 3.4611059067929375,
      "tokens_seen": 1984364544
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000402888665997994,
      "loss": 1.4463,
      "theoretical_loss": 3.4610964401059325,
      "tokens_seen": 1984430080
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004028686058174524,
      "loss": 1.507,
      "theoretical_loss": 3.461086973819096,
      "tokens_seen": 1984495616
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040284854563691074,
      "loss": 1.4515,
      "theoretical_loss": 3.461077507932398,
      "tokens_seen": 1984561152
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040282848545636915,
      "loss": 1.4754,
      "theoretical_loss": 3.4610680424458082,
      "tokens_seen": 1984626688
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004028084252758275,
      "loss": 1.3188,
      "theoretical_loss": 3.461058577359297,
      "tokens_seen": 1984692224
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004027883650952859,
      "loss": 1.4721,
      "theoretical_loss": 3.461049112672833,
      "tokens_seen": 1984757760
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040276830491474424,
      "loss": 1.4847,
      "theoretical_loss": 3.4610396483863877,
      "tokens_seen": 1984823296
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004027482447342026,
      "loss": 1.3775,
      "theoretical_loss": 3.461030184499929,
      "tokens_seen": 1984888832
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000402728184553661,
      "loss": 1.3342,
      "theoretical_loss": 3.4610207210134294,
      "tokens_seen": 1984954368
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004027081243731194,
      "loss": 1.4029,
      "theoretical_loss": 3.461011257926857,
      "tokens_seen": 1985019904
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040268806419257774,
      "loss": 1.3975,
      "theoretical_loss": 3.4610017952401817,
      "tokens_seen": 1985085440
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004026680040120361,
      "loss": 1.4153,
      "theoretical_loss": 3.4609923329533743,
      "tokens_seen": 1985150976
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004026479438314945,
      "loss": 1.6006,
      "theoretical_loss": 3.4609828710664035,
      "tokens_seen": 1985216512
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004026278836509529,
      "loss": 1.3417,
      "theoretical_loss": 3.4609734095792404,
      "tokens_seen": 1985282048
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040260782347041125,
      "loss": 1.4269,
      "theoretical_loss": 3.4609639484918544,
      "tokens_seen": 1985347584
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004025877632898696,
      "loss": 1.3181,
      "theoretical_loss": 3.460954487804215,
      "tokens_seen": 1985413120
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040256770310932797,
      "loss": 1.4966,
      "theoretical_loss": 3.460945027516293,
      "tokens_seen": 1985478656
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004025476429287864,
      "loss": 1.4059,
      "theoretical_loss": 3.4609355676280575,
      "tokens_seen": 1985544192
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040252758274824475,
      "loss": 1.4369,
      "theoretical_loss": 3.4609261081394793,
      "tokens_seen": 1985609728
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004025075225677031,
      "loss": 1.3598,
      "theoretical_loss": 3.460916649050527,
      "tokens_seen": 1985675264
    },
    {
      "epoch": 0.6,
      "objective/train/advantage_avg": 0.490124374628067,
      "objective/train/docs_used": 723995,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3428082466125488,
      "objective/train/original_loss": 1.342808485031128,
      "objective/train/theoretical_loss": 3.460907190361172,
      "objective/train/tokens_used": 2006200800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2411290556192398,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502382516860962,
      "objective/train/weighted_lm_loss": 1.4103271961212158,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.460907190361172,
      "tokens_seen": 1985740800
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004024874623871615,
      "loss": 1.4159,
      "theoretical_loss": 3.460907190361172,
      "tokens_seen": 1985740800
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004024674022066199,
      "loss": 1.3883,
      "theoretical_loss": 3.460897732071383,
      "tokens_seen": 1985806336
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040244734202607825,
      "loss": 1.4642,
      "theoretical_loss": 3.460888274181131,
      "tokens_seen": 1985871872
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004024272818455366,
      "loss": 1.4074,
      "theoretical_loss": 3.4608788166903857,
      "tokens_seen": 1985937408
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000402407221664995,
      "loss": 1.3897,
      "theoretical_loss": 3.460869359599116,
      "tokens_seen": 1986002944
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004023871614844534,
      "loss": 1.4346,
      "theoretical_loss": 3.460859902907293,
      "tokens_seen": 1986068480
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040236710130391176,
      "loss": 1.4132,
      "theoretical_loss": 3.460850446614886,
      "tokens_seen": 1986134016
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004023470411233701,
      "loss": 1.427,
      "theoretical_loss": 3.4608409907218656,
      "tokens_seen": 1986199552
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004023269809428285,
      "loss": 1.3836,
      "theoretical_loss": 3.460831535228201,
      "tokens_seen": 1986265088
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040230692076228684,
      "loss": 1.5515,
      "theoretical_loss": 3.4608220801338625,
      "tokens_seen": 1986330624
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040228686058174526,
      "loss": 1.38,
      "theoretical_loss": 3.4608126254388205,
      "tokens_seen": 1986396160
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004022668004012036,
      "loss": 1.4739,
      "theoretical_loss": 3.4608031711430445,
      "tokens_seen": 1986461696
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000402246740220662,
      "loss": 1.3602,
      "theoretical_loss": 3.4607937172465046,
      "tokens_seen": 1986527232
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040222668004012035,
      "loss": 1.3919,
      "theoretical_loss": 3.4607842637491704,
      "tokens_seen": 1986592768
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040220661985957876,
      "loss": 1.3905,
      "theoretical_loss": 3.460774810651012,
      "tokens_seen": 1986658304
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004021865596790371,
      "loss": 1.4327,
      "theoretical_loss": 3.460765357952,
      "tokens_seen": 1986723840
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004021664994984955,
      "loss": 1.3661,
      "theoretical_loss": 3.4607559056521033,
      "tokens_seen": 1986789376
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040214643931795385,
      "loss": 1.4694,
      "theoretical_loss": 3.460746453751293,
      "tokens_seen": 1986854912
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004021263791374122,
      "loss": 1.3857,
      "theoretical_loss": 3.4607370022495387,
      "tokens_seen": 1986920448
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040210631895687063,
      "loss": 1.4454,
      "theoretical_loss": 3.4607275511468103,
      "tokens_seen": 1986985984
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000402086258776329,
      "loss": 1.4246,
      "theoretical_loss": 3.4607181004430774,
      "tokens_seen": 1987051520
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040206619859578735,
      "loss": 1.4181,
      "theoretical_loss": 3.4607086501383106,
      "tokens_seen": 1987117056
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004020461384152457,
      "loss": 1.4834,
      "theoretical_loss": 3.4606992002324795,
      "tokens_seen": 1987182592
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040202607823470413,
      "loss": 1.3094,
      "theoretical_loss": 3.4606897507255545,
      "tokens_seen": 1987248128
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004020060180541625,
      "loss": 1.4547,
      "theoretical_loss": 3.460680301617505,
      "tokens_seen": 1987313664
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040198595787362086,
      "loss": 1.4408,
      "theoretical_loss": 3.4606708529083017,
      "tokens_seen": 1987379200
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004019658976930792,
      "loss": 1.4744,
      "theoretical_loss": 3.4606614045979147,
      "tokens_seen": 1987444736
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004019458375125376,
      "loss": 1.436,
      "theoretical_loss": 3.460651956686313,
      "tokens_seen": 1987510272
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000401925777331996,
      "loss": 1.3674,
      "theoretical_loss": 3.460642509173468,
      "tokens_seen": 1987575808
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040190571715145436,
      "loss": 1.5682,
      "theoretical_loss": 3.4606330620593484,
      "tokens_seen": 1987641344
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004018856569709127,
      "loss": 1.5127,
      "theoretical_loss": 3.4606236153439247,
      "tokens_seen": 1987706880
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004018655967903711,
      "loss": 1.4619,
      "theoretical_loss": 3.460614169027167,
      "tokens_seen": 1987772416
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004018455366098295,
      "loss": 1.4084,
      "theoretical_loss": 3.4606047231090455,
      "tokens_seen": 1987837952
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040182547642928786,
      "loss": 1.3362,
      "theoretical_loss": 3.4605952775895306,
      "tokens_seen": 1987903488
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004018054162487462,
      "loss": 1.4946,
      "theoretical_loss": 3.460585832468591,
      "tokens_seen": 1987969024
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004017853560682046,
      "loss": 1.2914,
      "theoretical_loss": 3.4605763877461984,
      "tokens_seen": 1988034560
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040176529588766295,
      "loss": 1.3256,
      "theoretical_loss": 3.4605669434223216,
      "tokens_seen": 1988100096
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040174523570712137,
      "loss": 1.4978,
      "theoretical_loss": 3.4605574994969306,
      "tokens_seen": 1988165632
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040172517552657973,
      "loss": 1.3119,
      "theoretical_loss": 3.460548055969997,
      "tokens_seen": 1988231168
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004017051153460381,
      "loss": 1.431,
      "theoretical_loss": 3.460538612841489,
      "tokens_seen": 1988296704
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040168505516549645,
      "loss": 1.3341,
      "theoretical_loss": 3.4605291701113776,
      "tokens_seen": 1988362240
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040166499498495487,
      "loss": 1.387,
      "theoretical_loss": 3.4605197277796327,
      "tokens_seen": 1988427776
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004016449348044133,
      "loss": 1.4319,
      "theoretical_loss": 3.460510285846224,
      "tokens_seen": 1988493312
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040162487462387165,
      "loss": 1.4474,
      "theoretical_loss": 3.4605008443111225,
      "tokens_seen": 1988558848
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040160481444333,
      "loss": 1.2219,
      "theoretical_loss": 3.4604914031742977,
      "tokens_seen": 1988624384
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040158475426278837,
      "loss": 1.322,
      "theoretical_loss": 3.4604819624357193,
      "tokens_seen": 1988689920
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004015646940822468,
      "loss": 1.4469,
      "theoretical_loss": 3.460472522095358,
      "tokens_seen": 1988755456
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040154463390170515,
      "loss": 1.4666,
      "theoretical_loss": 3.4604630821531837,
      "tokens_seen": 1988820992
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004015245737211635,
      "loss": 1.3607,
      "theoretical_loss": 3.460453642609166,
      "tokens_seen": 1988886528
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004015045135406219,
      "loss": 1.3881,
      "theoretical_loss": 3.460444203463276,
      "tokens_seen": 1988952064
    },
    {
      "epoch": 0.6,
      "objective/train/advantage_avg": 0.49550849199295044,
      "objective/train/docs_used": 725238,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4878947734832764,
      "objective/train/original_loss": 1.4878950119018555,
      "objective/train/theoretical_loss": 3.4604347647154827,
      "objective/train/tokens_used": 2009477600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24568907916545868,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507999658584595,
      "objective/train/weighted_lm_loss": 1.5633479356765747,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0216598510742188,
      "theoretical_loss": 3.4604347647154827,
      "tokens_seen": 1989017600
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004014844533600803,
      "loss": 1.4984,
      "theoretical_loss": 3.4604347647154827,
      "tokens_seen": 1989017600
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040146439317953865,
      "loss": 1.4386,
      "theoretical_loss": 3.460425326365757,
      "tokens_seen": 1989083136
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000401444332998997,
      "loss": 1.453,
      "theoretical_loss": 3.4604158884140683,
      "tokens_seen": 1989148672
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004014242728184554,
      "loss": 1.4247,
      "theoretical_loss": 3.4604064508603876,
      "tokens_seen": 1989214208
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004014042126379138,
      "loss": 1.48,
      "theoretical_loss": 3.460397013704684,
      "tokens_seen": 1989279744
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040138415245737216,
      "loss": 1.328,
      "theoretical_loss": 3.4603875769469283,
      "tokens_seen": 1989345280
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004013640922768305,
      "loss": 1.4331,
      "theoretical_loss": 3.460378140587091,
      "tokens_seen": 1989410816
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004013440320962889,
      "loss": 1.4028,
      "theoretical_loss": 3.4603687046251412,
      "tokens_seen": 1989476352
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040132397191574724,
      "loss": 1.496,
      "theoretical_loss": 3.4603592690610494,
      "tokens_seen": 1989541888
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040130391173520566,
      "loss": 1.4381,
      "theoretical_loss": 3.4603498338947856,
      "tokens_seen": 1989607424
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000401283851554664,
      "loss": 1.4037,
      "theoretical_loss": 3.46034039912632,
      "tokens_seen": 1989672960
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004012637913741224,
      "loss": 1.3727,
      "theoretical_loss": 3.4603309647556233,
      "tokens_seen": 1989738496
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040124373119358075,
      "loss": 1.3996,
      "theoretical_loss": 3.460321530782665,
      "tokens_seen": 1989804032
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040122367101303916,
      "loss": 1.3706,
      "theoretical_loss": 3.460312097207415,
      "tokens_seen": 1989869568
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004012036108324975,
      "loss": 1.3305,
      "theoretical_loss": 3.460302664029844,
      "tokens_seen": 1989935104
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004011835506519559,
      "loss": 1.4504,
      "theoretical_loss": 3.4602932312499224,
      "tokens_seen": 1990000640
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040116349047141425,
      "loss": 1.5756,
      "theoretical_loss": 3.4602837988676196,
      "tokens_seen": 1990066176
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004011434302908726,
      "loss": 1.3955,
      "theoretical_loss": 3.4602743668829055,
      "tokens_seen": 1990131712
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040112337011033103,
      "loss": 1.4576,
      "theoretical_loss": 3.4602649352957515,
      "tokens_seen": 1990197248
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004011033099297894,
      "loss": 1.4039,
      "theoretical_loss": 3.4602555041061267,
      "tokens_seen": 1990262784
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040108324974924775,
      "loss": 1.4646,
      "theoretical_loss": 3.4602460733140017,
      "tokens_seen": 1990328320
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004010631895687061,
      "loss": 1.3214,
      "theoretical_loss": 3.4602366429193463,
      "tokens_seen": 1990393856
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040104312938816453,
      "loss": 1.4139,
      "theoretical_loss": 3.460227212922131,
      "tokens_seen": 1990459392
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004010230692076229,
      "loss": 1.3862,
      "theoretical_loss": 3.4602177833223267,
      "tokens_seen": 1990524928
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040100300902708126,
      "loss": 1.4742,
      "theoretical_loss": 3.460208354119902,
      "tokens_seen": 1990590464
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004009829488465396,
      "loss": 1.4415,
      "theoretical_loss": 3.4601989253148275,
      "tokens_seen": 1990656000
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000400962888665998,
      "loss": 1.4496,
      "theoretical_loss": 3.4601894969070743,
      "tokens_seen": 1990721536
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004009428284854564,
      "loss": 1.4539,
      "theoretical_loss": 3.4601800688966122,
      "tokens_seen": 1990787072
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040092276830491476,
      "loss": 1.3145,
      "theoretical_loss": 3.4601706412834106,
      "tokens_seen": 1990852608
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004009027081243731,
      "loss": 1.3753,
      "theoretical_loss": 3.4601612140674405,
      "tokens_seen": 1990918144
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004008826479438315,
      "loss": 1.416,
      "theoretical_loss": 3.460151787248672,
      "tokens_seen": 1990983680
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004008625877632899,
      "loss": 1.4082,
      "theoretical_loss": 3.4601423608270743,
      "tokens_seen": 1991049216
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040084252758274826,
      "loss": 1.4103,
      "theoretical_loss": 3.460132934802619,
      "tokens_seen": 1991114752
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004008224674022066,
      "loss": 1.3977,
      "theoretical_loss": 3.4601235091752756,
      "tokens_seen": 1991180288
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000400802407221665,
      "loss": 1.4305,
      "theoretical_loss": 3.460114083945015,
      "tokens_seen": 1991245824
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040078234704112335,
      "loss": 1.361,
      "theoretical_loss": 3.4601046591118063,
      "tokens_seen": 1991311360
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040076228686058177,
      "loss": 1.4862,
      "theoretical_loss": 3.46009523467562,
      "tokens_seen": 1991376896
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040074222668004013,
      "loss": 1.4471,
      "theoretical_loss": 3.4600858106364267,
      "tokens_seen": 1991442432
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004007221664994985,
      "loss": 1.424,
      "theoretical_loss": 3.4600763869941966,
      "tokens_seen": 1991507968
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040070210631895685,
      "loss": 1.45,
      "theoretical_loss": 3.4600669637488997,
      "tokens_seen": 1991573504
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040068204613841527,
      "loss": 1.4581,
      "theoretical_loss": 3.4600575409005065,
      "tokens_seen": 1991639040
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040066198595787363,
      "loss": 1.3784,
      "theoretical_loss": 3.4600481184489866,
      "tokens_seen": 1991704576
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000400641925777332,
      "loss": 1.4242,
      "theoretical_loss": 3.460038696394311,
      "tokens_seen": 1991770112
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040062186559679036,
      "loss": 1.5224,
      "theoretical_loss": 3.460029274736449,
      "tokens_seen": 1991835648
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004006018054162487,
      "loss": 1.3989,
      "theoretical_loss": 3.460019853475372,
      "tokens_seen": 1991901184
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040058174523570713,
      "loss": 1.5453,
      "theoretical_loss": 3.4600104326110492,
      "tokens_seen": 1991966720
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004005616850551655,
      "loss": 1.3933,
      "theoretical_loss": 3.4600010121434517,
      "tokens_seen": 1992032256
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040054162487462386,
      "loss": 1.4796,
      "theoretical_loss": 3.459991592072549,
      "tokens_seen": 1992097792
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004005215646940822,
      "loss": 1.3545,
      "theoretical_loss": 3.4599821723983117,
      "tokens_seen": 1992163328
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040050150451354064,
      "loss": 1.435,
      "theoretical_loss": 3.4599727531207103,
      "tokens_seen": 1992228864
    },
    {
      "epoch": 0.6,
      "objective/train/advantage_avg": 0.49169719219207764,
      "objective/train/docs_used": 726441,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3937511444091797,
      "objective/train/original_loss": 1.3937511444091797,
      "objective/train/theoretical_loss": 3.4599633342397142,
      "objective/train/tokens_used": 2012754400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24229390919208527,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504013299942017,
      "objective/train/weighted_lm_loss": 1.4639962911605835,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4599633342397142,
      "tokens_seen": 1992294400
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000400481444332999,
      "loss": 1.3865,
      "theoretical_loss": 3.4599633342397142,
      "tokens_seen": 1992294400
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040046138415245736,
      "loss": 1.3908,
      "theoretical_loss": 3.4599539157552943,
      "tokens_seen": 1992359936
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004004413239719157,
      "loss": 1.4553,
      "theoretical_loss": 3.459944497667421,
      "tokens_seen": 1992425472
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004004212637913741,
      "loss": 1.4289,
      "theoretical_loss": 3.4599350799760646,
      "tokens_seen": 1992491008
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004004012036108325,
      "loss": 1.3793,
      "theoretical_loss": 3.4599256626811945,
      "tokens_seen": 1992556544
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040038114343029087,
      "loss": 1.3963,
      "theoretical_loss": 3.4599162457827823,
      "tokens_seen": 1992622080
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040036108324974923,
      "loss": 1.4943,
      "theoretical_loss": 3.459906829280797,
      "tokens_seen": 1992687616
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004003410230692076,
      "loss": 1.3791,
      "theoretical_loss": 3.4598974131752094,
      "tokens_seen": 1992753152
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000400320962888666,
      "loss": 1.3934,
      "theoretical_loss": 3.45988799746599,
      "tokens_seen": 1992818688
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040030090270812437,
      "loss": 1.3409,
      "theoretical_loss": 3.459878582153109,
      "tokens_seen": 1992884224
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040028084252758273,
      "loss": 1.3613,
      "theoretical_loss": 3.459869167236536,
      "tokens_seen": 1992949760
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004002607823470411,
      "loss": 1.4537,
      "theoretical_loss": 3.4598597527162425,
      "tokens_seen": 1993015296
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040024072216649946,
      "loss": 1.4527,
      "theoretical_loss": 3.4598503385921977,
      "tokens_seen": 1993080832
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040022066198595787,
      "loss": 1.3257,
      "theoretical_loss": 3.4598409248643724,
      "tokens_seen": 1993146368
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040020060180541623,
      "loss": 1.4118,
      "theoretical_loss": 3.4598315115327374,
      "tokens_seen": 1993211904
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004001805416248746,
      "loss": 1.4553,
      "theoretical_loss": 3.459822098597262,
      "tokens_seen": 1993277440
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040016048144433296,
      "loss": 1.4167,
      "theoretical_loss": 3.4598126860579166,
      "tokens_seen": 1993342976
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040014042126379143,
      "loss": 1.303,
      "theoretical_loss": 3.459803273914672,
      "tokens_seen": 1993408512
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004001203610832498,
      "loss": 1.452,
      "theoretical_loss": 3.459793862167498,
      "tokens_seen": 1993474048
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040010030090270815,
      "loss": 1.3573,
      "theoretical_loss": 3.4597844508163655,
      "tokens_seen": 1993539584
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004000802407221665,
      "loss": 1.451,
      "theoretical_loss": 3.4597750398612455,
      "tokens_seen": 1993605120
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040006018054162493,
      "loss": 1.374,
      "theoretical_loss": 3.4597656293021064,
      "tokens_seen": 1993670656
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004000401203610833,
      "loss": 1.5495,
      "theoretical_loss": 3.45975621913892,
      "tokens_seen": 1993736192
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00040002006018054166,
      "loss": 1.455,
      "theoretical_loss": 3.4597468093716555,
      "tokens_seen": 1993801728
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0004,
      "loss": 1.4535,
      "theoretical_loss": 3.459737400000284,
      "tokens_seen": 1993867264
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003999799398194584,
      "loss": 1.4881,
      "theoretical_loss": 3.459727991024776,
      "tokens_seen": 1993932800
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003999598796389168,
      "loss": 1.4135,
      "theoretical_loss": 3.4597185824451016,
      "tokens_seen": 1993998336
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039993981945837516,
      "loss": 1.3749,
      "theoretical_loss": 3.459709174261231,
      "tokens_seen": 1994063872
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003999197592778335,
      "loss": 1.3239,
      "theoretical_loss": 3.4596997664731344,
      "tokens_seen": 1994129408
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003998996990972919,
      "loss": 1.4915,
      "theoretical_loss": 3.459690359080783,
      "tokens_seen": 1994194944
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003998796389167503,
      "loss": 1.3831,
      "theoretical_loss": 3.459680952084146,
      "tokens_seen": 1994260480
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039985957873620866,
      "loss": 1.408,
      "theoretical_loss": 3.459671545483194,
      "tokens_seen": 1994326016
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000399839518555667,
      "loss": 1.391,
      "theoretical_loss": 3.4596621392778983,
      "tokens_seen": 1994391552
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003998194583751254,
      "loss": 1.4232,
      "theoretical_loss": 3.4596527334682277,
      "tokens_seen": 1994457088
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039979939819458375,
      "loss": 1.3297,
      "theoretical_loss": 3.4596433280541543,
      "tokens_seen": 1994522624
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039977933801404217,
      "loss": 1.4436,
      "theoretical_loss": 3.459633923035647,
      "tokens_seen": 1994588160
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039975927783350053,
      "loss": 1.4761,
      "theoretical_loss": 3.459624518412677,
      "tokens_seen": 1994653696
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003997392176529589,
      "loss": 1.3546,
      "theoretical_loss": 3.4596151141852145,
      "tokens_seen": 1994719232
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039971915747241725,
      "loss": 1.3848,
      "theoretical_loss": 3.45960571035323,
      "tokens_seen": 1994784768
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039969909729187567,
      "loss": 1.4149,
      "theoretical_loss": 3.4595963069166933,
      "tokens_seen": 1994850304
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039967903711133403,
      "loss": 1.3846,
      "theoretical_loss": 3.459586903875575,
      "tokens_seen": 1994915840
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003996589769307924,
      "loss": 1.4726,
      "theoretical_loss": 3.459577501229846,
      "tokens_seen": 1994981376
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039963891675025076,
      "loss": 1.4166,
      "theoretical_loss": 3.459568098979476,
      "tokens_seen": 1995046912
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003996188565697091,
      "loss": 1.4486,
      "theoretical_loss": 3.4595586971244363,
      "tokens_seen": 1995112448
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039959879638916754,
      "loss": 1.406,
      "theoretical_loss": 3.4595492956646963,
      "tokens_seen": 1995177984
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003995787362086259,
      "loss": 1.4034,
      "theoretical_loss": 3.459539894600227,
      "tokens_seen": 1995243520
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039955867602808426,
      "loss": 1.4201,
      "theoretical_loss": 3.4595304939309983,
      "tokens_seen": 1995309056
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003995386158475426,
      "loss": 1.4004,
      "theoretical_loss": 3.4595210936569813,
      "tokens_seen": 1995374592
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039951855566700104,
      "loss": 1.3731,
      "theoretical_loss": 3.459511693778146,
      "tokens_seen": 1995440128
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003994984954864594,
      "loss": 1.4973,
      "theoretical_loss": 3.459502294294462,
      "tokens_seen": 1995505664
    },
    {
      "epoch": 0.6,
      "objective/train/advantage_avg": 0.4928278923034668,
      "objective/train/docs_used": 727673,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2011752128601074,
      "objective/train/original_loss": 1.2011752128601074,
      "objective/train/theoretical_loss": 3.4594928952059014,
      "objective/train/tokens_used": 2016031200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24332687258720398,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505197048187256,
      "objective/train/weighted_lm_loss": 1.2618193626403809,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4594928952059014,
      "tokens_seen": 1995571200
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039947843530591776,
      "loss": 1.3229,
      "theoretical_loss": 3.4594928952059014,
      "tokens_seen": 1995571200
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003994583751253761,
      "loss": 1.4466,
      "theoretical_loss": 3.459483496512434,
      "tokens_seen": 1995636736
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003994383149448345,
      "loss": 1.3737,
      "theoretical_loss": 3.4594740982140295,
      "tokens_seen": 1995702272
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003994182547642929,
      "loss": 1.4881,
      "theoretical_loss": 3.459464700310659,
      "tokens_seen": 1995767808
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039939819458375127,
      "loss": 1.3975,
      "theoretical_loss": 3.4594553028022927,
      "tokens_seen": 1995833344
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039937813440320963,
      "loss": 1.4014,
      "theoretical_loss": 3.4594459056889004,
      "tokens_seen": 1995898880
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000399358074222668,
      "loss": 1.5099,
      "theoretical_loss": 3.459436508970454,
      "tokens_seen": 1995964416
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003993380140421264,
      "loss": 1.45,
      "theoretical_loss": 3.4594271126469227,
      "tokens_seen": 1996029952
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039931795386158477,
      "loss": 1.3996,
      "theoretical_loss": 3.459417716718278,
      "tokens_seen": 1996095488
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039929789368104313,
      "loss": 1.5023,
      "theoretical_loss": 3.4594083211844886,
      "tokens_seen": 1996161024
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0003992778335005015,
      "loss": 1.3589,
      "theoretical_loss": 3.4593989260455267,
      "tokens_seen": 1996226560
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039925777331995986,
      "loss": 1.371,
      "theoretical_loss": 3.459389531301362,
      "tokens_seen": 1996292096
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039923771313941827,
      "loss": 1.4619,
      "theoretical_loss": 3.459380136951965,
      "tokens_seen": 1996357632
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.00039921765295887663,
      "loss": 1.3883,
      "theoretical_loss": 3.459370742997306,
      "tokens_seen": 1996423168
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.000399197592778335,
      "loss": 1.4069,
      "theoretical_loss": 3.459361349437356,
      "tokens_seen": 1996488704
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039917753259779336,
      "loss": 1.4077,
      "theoretical_loss": 3.459351956272085,
      "tokens_seen": 1996554240
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003991574724172518,
      "loss": 1.437,
      "theoretical_loss": 3.4593425635014636,
      "tokens_seen": 1996619776
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039913741223671014,
      "loss": 1.4793,
      "theoretical_loss": 3.4593331711254622,
      "tokens_seen": 1996685312
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003991173520561685,
      "loss": 1.4458,
      "theoretical_loss": 3.459323779144051,
      "tokens_seen": 1996750848
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039909729187562686,
      "loss": 1.446,
      "theoretical_loss": 3.459314387557201,
      "tokens_seen": 1996816384
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003990772316950853,
      "loss": 1.4045,
      "theoretical_loss": 3.4593049963648825,
      "tokens_seen": 1996881920
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039905717151454364,
      "loss": 1.312,
      "theoretical_loss": 3.459295605567066,
      "tokens_seen": 1996947456
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000399037111334002,
      "loss": 1.4664,
      "theoretical_loss": 3.4592862151637216,
      "tokens_seen": 1997012992
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039901705115346037,
      "loss": 1.4935,
      "theoretical_loss": 3.45927682515482,
      "tokens_seen": 1997078528
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039899699097291873,
      "loss": 1.3514,
      "theoretical_loss": 3.4592674355403323,
      "tokens_seen": 1997144064
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039897693079237714,
      "loss": 1.4328,
      "theoretical_loss": 3.4592580463202283,
      "tokens_seen": 1997209600
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003989568706118355,
      "loss": 1.3043,
      "theoretical_loss": 3.4592486574944785,
      "tokens_seen": 1997275136
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039893681043129387,
      "loss": 1.5111,
      "theoretical_loss": 3.459239269063054,
      "tokens_seen": 1997340672
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039891675025075223,
      "loss": 1.3552,
      "theoretical_loss": 3.4592298810259243,
      "tokens_seen": 1997406208
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039889669007021065,
      "loss": 1.3661,
      "theoretical_loss": 3.459220493383061,
      "tokens_seen": 1997471744
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000398876629889669,
      "loss": 1.5117,
      "theoretical_loss": 3.4592111061344335,
      "tokens_seen": 1997537280
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039885656970912737,
      "loss": 1.6087,
      "theoretical_loss": 3.459201719280013,
      "tokens_seen": 1997602816
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039883650952858573,
      "loss": 1.4388,
      "theoretical_loss": 3.45919233281977,
      "tokens_seen": 1997668352
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003988164493480441,
      "loss": 1.4068,
      "theoretical_loss": 3.459182946753675,
      "tokens_seen": 1997733888
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003987963891675025,
      "loss": 1.3677,
      "theoretical_loss": 3.4591735610816983,
      "tokens_seen": 1997799424
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003987763289869609,
      "loss": 1.4246,
      "theoretical_loss": 3.459164175803811,
      "tokens_seen": 1997864960
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039875626880641924,
      "loss": 1.5502,
      "theoretical_loss": 3.4591547909199827,
      "tokens_seen": 1997930496
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003987362086258776,
      "loss": 1.3925,
      "theoretical_loss": 3.4591454064301845,
      "tokens_seen": 1997996032
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000398716148445336,
      "loss": 1.3377,
      "theoretical_loss": 3.459136022334387,
      "tokens_seen": 1998061568
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003986960882647944,
      "loss": 1.5505,
      "theoretical_loss": 3.4591266386325605,
      "tokens_seen": 1998127104
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039867602808425274,
      "loss": 1.4699,
      "theoretical_loss": 3.4591172553246756,
      "tokens_seen": 1998192640
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003986559679037111,
      "loss": 1.4745,
      "theoretical_loss": 3.459107872410703,
      "tokens_seen": 1998258176
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003986359077231695,
      "loss": 1.4182,
      "theoretical_loss": 3.4590984898906134,
      "tokens_seen": 1998323712
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039861584754262794,
      "loss": 1.2969,
      "theoretical_loss": 3.4590891077643766,
      "tokens_seen": 1998389248
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003985957873620863,
      "loss": 1.3608,
      "theoretical_loss": 3.4590797260319635,
      "tokens_seen": 1998454784
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039857572718154466,
      "loss": 1.4209,
      "theoretical_loss": 3.459070344693345,
      "tokens_seen": 1998520320
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000398555667001003,
      "loss": 1.5053,
      "theoretical_loss": 3.4590609637484913,
      "tokens_seen": 1998585856
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039853560682046144,
      "loss": 1.3794,
      "theoretical_loss": 3.4590515831973736,
      "tokens_seen": 1998651392
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003985155466399198,
      "loss": 1.4782,
      "theoretical_loss": 3.459042203039961,
      "tokens_seen": 1998716928
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039849548645937816,
      "loss": 1.3979,
      "theoretical_loss": 3.4590328232762255,
      "tokens_seen": 1998782464
    },
    {
      "epoch": 0.61,
      "objective/train/advantage_avg": 0.49043774604797363,
      "objective/train/docs_used": 728818,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.390101432800293,
      "objective/train/original_loss": 1.390101432800293,
      "objective/train/theoretical_loss": 3.459023443906138,
      "objective/train/tokens_used": 2019308000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2409725785255432,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502686500549316,
      "objective/train/weighted_lm_loss": 1.4598644971847534,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.459023443906138,
      "tokens_seen": 1998848000
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003984754262788365,
      "loss": 1.4187,
      "theoretical_loss": 3.459023443906138,
      "tokens_seen": 1998848000
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003984553660982949,
      "loss": 1.3616,
      "theoretical_loss": 3.459014064929667,
      "tokens_seen": 1998913536
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003984353059177533,
      "loss": 1.4525,
      "theoretical_loss": 3.459004686346785,
      "tokens_seen": 1998979072
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039841524573721167,
      "loss": 1.3138,
      "theoretical_loss": 3.4589953081574616,
      "tokens_seen": 1999044608
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039839518555667003,
      "loss": 1.4755,
      "theoretical_loss": 3.458985930361668,
      "tokens_seen": 1999110144
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003983751253761284,
      "loss": 1.4358,
      "theoretical_loss": 3.458976552959374,
      "tokens_seen": 1999175680
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003983550651955868,
      "loss": 1.4683,
      "theoretical_loss": 3.4589671759505514,
      "tokens_seen": 1999241216
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039833500501504517,
      "loss": 1.3109,
      "theoretical_loss": 3.4589577993351694,
      "tokens_seen": 1999306752
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039831494483450353,
      "loss": 1.4158,
      "theoretical_loss": 3.4589484231132,
      "tokens_seen": 1999372288
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003982948846539619,
      "loss": 1.3574,
      "theoretical_loss": 3.4589390472846127,
      "tokens_seen": 1999437824
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039827482447342026,
      "loss": 1.4738,
      "theoretical_loss": 3.4589296718493783,
      "tokens_seen": 1999503360
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039825476429287867,
      "loss": 1.369,
      "theoretical_loss": 3.4589202968074675,
      "tokens_seen": 1999568896
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039823470411233703,
      "loss": 1.3529,
      "theoretical_loss": 3.4589109221588514,
      "tokens_seen": 1999634432
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003982146439317954,
      "loss": 1.3427,
      "theoretical_loss": 3.4589015479035,
      "tokens_seen": 1999699968
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039819458375125376,
      "loss": 1.5433,
      "theoretical_loss": 3.4588921740413845,
      "tokens_seen": 1999765504
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003981745235707122,
      "loss": 1.6186,
      "theoretical_loss": 3.458882800572474,
      "tokens_seen": 1999831040
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039815446339017054,
      "loss": 1.4834,
      "theoretical_loss": 3.4588734274967416,
      "tokens_seen": 1999896576
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003981344032096289,
      "loss": 1.3848,
      "theoretical_loss": 3.4588640548141556,
      "tokens_seen": 1999962112
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039811434302908726,
      "loss": 1.4801,
      "theoretical_loss": 3.458854682524688,
      "tokens_seen": 2000027648
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003980942828485457,
      "loss": 1.4721,
      "theoretical_loss": 3.458845310628309,
      "tokens_seen": 2000093184
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039807422266800404,
      "loss": 1.4564,
      "theoretical_loss": 3.4588359391249894,
      "tokens_seen": 2000158720
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003980541624874624,
      "loss": 1.3894,
      "theoretical_loss": 3.4588265680146995,
      "tokens_seen": 2000224256
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039803410230692077,
      "loss": 1.4697,
      "theoretical_loss": 3.45881719729741,
      "tokens_seen": 2000289792
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039801404212637913,
      "loss": 1.4627,
      "theoretical_loss": 3.458807826973092,
      "tokens_seen": 2000355328
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039799398194583754,
      "loss": 1.3149,
      "theoretical_loss": 3.458798457041716,
      "tokens_seen": 2000420864
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003979739217652959,
      "loss": 1.3964,
      "theoretical_loss": 3.458789087503252,
      "tokens_seen": 2000486400
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039795386158475427,
      "loss": 1.2901,
      "theoretical_loss": 3.458779718357672,
      "tokens_seen": 2000551936
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039793380140421263,
      "loss": 1.4535,
      "theoretical_loss": 3.4587703496049444,
      "tokens_seen": 2000617472
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039791374122367105,
      "loss": 1.5163,
      "theoretical_loss": 3.4587609812450424,
      "tokens_seen": 2000683008
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003978936810431294,
      "loss": 1.4633,
      "theoretical_loss": 3.458751613277935,
      "tokens_seen": 2000748544
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039787362086258777,
      "loss": 1.3297,
      "theoretical_loss": 3.4587422457035935,
      "tokens_seen": 2000814080
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039785356068204613,
      "loss": 1.4964,
      "theoretical_loss": 3.4587328785219884,
      "tokens_seen": 2000879616
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003978335005015045,
      "loss": 1.4537,
      "theoretical_loss": 3.4587235117330906,
      "tokens_seen": 2000945152
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003978134403209629,
      "loss": 1.4172,
      "theoretical_loss": 3.45871414533687,
      "tokens_seen": 2001010688
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003977933801404213,
      "loss": 1.3692,
      "theoretical_loss": 3.458704779333299,
      "tokens_seen": 2001076224
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039777331995987964,
      "loss": 1.3362,
      "theoretical_loss": 3.4586954137223467,
      "tokens_seen": 2001141760
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000397753259779338,
      "loss": 1.3696,
      "theoretical_loss": 3.4586860485039836,
      "tokens_seen": 2001207296
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003977331995987964,
      "loss": 1.337,
      "theoretical_loss": 3.4586766836781813,
      "tokens_seen": 2001272832
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003977131394182548,
      "loss": 1.3946,
      "theoretical_loss": 3.4586673192449107,
      "tokens_seen": 2001338368
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039769307923771314,
      "loss": 1.4309,
      "theoretical_loss": 3.4586579552041417,
      "tokens_seen": 2001403904
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003976730190571715,
      "loss": 1.5066,
      "theoretical_loss": 3.4586485915558454,
      "tokens_seen": 2001469440
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039765295887662987,
      "loss": 1.4699,
      "theoretical_loss": 3.458639228299992,
      "tokens_seen": 2001534976
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003976328986960883,
      "loss": 1.4115,
      "theoretical_loss": 3.4586298654365533,
      "tokens_seen": 2001600512
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039761283851554664,
      "loss": 1.4703,
      "theoretical_loss": 3.458620502965499,
      "tokens_seen": 2001666048
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000397592778335005,
      "loss": 1.4339,
      "theoretical_loss": 3.4586111408868,
      "tokens_seen": 2001731584
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039757271815446337,
      "loss": 1.4108,
      "theoretical_loss": 3.458601779200427,
      "tokens_seen": 2001797120
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003975526579739218,
      "loss": 1.3987,
      "theoretical_loss": 3.458592417906351,
      "tokens_seen": 2001862656
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039753259779338015,
      "loss": 1.489,
      "theoretical_loss": 3.458583057004543,
      "tokens_seen": 2001928192
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003975125376128385,
      "loss": 1.4067,
      "theoretical_loss": 3.4585736964949727,
      "tokens_seen": 2001993728
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039749247743229687,
      "loss": 1.3229,
      "theoretical_loss": 3.458564336377612,
      "tokens_seen": 2002059264
    },
    {
      "epoch": 0.61,
      "objective/train/advantage_avg": 0.4955725073814392,
      "objective/train/docs_used": 729965,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.29952871799469,
      "objective/train/original_loss": 1.2995284795761108,
      "objective/train/theoretical_loss": 3.4585549766524304,
      "objective/train/tokens_used": 2022584800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24589885771274567,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508073568344116,
      "objective/train/weighted_lm_loss": 1.3655732870101929,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4585549766524304,
      "tokens_seen": 2002124800
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039747241725175523,
      "loss": 1.4035,
      "theoretical_loss": 3.4585549766524304,
      "tokens_seen": 2002124800
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039745235707121365,
      "loss": 1.3314,
      "theoretical_loss": 3.4585456173194,
      "tokens_seen": 2002190336
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000397432296890672,
      "loss": 1.4807,
      "theoretical_loss": 3.45853625837849,
      "tokens_seen": 2002255872
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003974122367101304,
      "loss": 1.3892,
      "theoretical_loss": 3.4585268998296725,
      "tokens_seen": 2002321408
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039739217652958874,
      "loss": 1.4204,
      "theoretical_loss": 3.4585175416729177,
      "tokens_seen": 2002386944
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039737211634904715,
      "loss": 1.3231,
      "theoretical_loss": 3.4585081839081964,
      "tokens_seen": 2002452480
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003973520561685055,
      "loss": 1.3534,
      "theoretical_loss": 3.458498826535479,
      "tokens_seen": 2002518016
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003973319959879639,
      "loss": 1.447,
      "theoretical_loss": 3.4584894695547366,
      "tokens_seen": 2002583552
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039731193580742224,
      "loss": 1.4953,
      "theoretical_loss": 3.45848011296594,
      "tokens_seen": 2002649088
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003972918756268806,
      "loss": 1.4031,
      "theoretical_loss": 3.4584707567690596,
      "tokens_seen": 2002714624
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000397271815446339,
      "loss": 1.4482,
      "theoretical_loss": 3.4584614009640666,
      "tokens_seen": 2002780160
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003972517552657974,
      "loss": 1.4082,
      "theoretical_loss": 3.4584520455509313,
      "tokens_seen": 2002845696
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039723169508525574,
      "loss": 1.4164,
      "theoretical_loss": 3.4584426905296253,
      "tokens_seen": 2002911232
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003972116349047141,
      "loss": 1.3907,
      "theoretical_loss": 3.4584333359001187,
      "tokens_seen": 2002976768
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003971915747241725,
      "loss": 1.3219,
      "theoretical_loss": 3.4584239816623823,
      "tokens_seen": 2003042304
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003971715145436309,
      "loss": 1.3898,
      "theoretical_loss": 3.4584146278163868,
      "tokens_seen": 2003107840
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039715145436308925,
      "loss": 1.3622,
      "theoretical_loss": 3.4584052743621028,
      "tokens_seen": 2003173376
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039713139418254766,
      "loss": 1.4846,
      "theoretical_loss": 3.458395921299502,
      "tokens_seen": 2003238912
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003971113340020061,
      "loss": 1.4795,
      "theoretical_loss": 3.4583865686285544,
      "tokens_seen": 2003304448
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039709127382146444,
      "loss": 1.3054,
      "theoretical_loss": 3.458377216349231,
      "tokens_seen": 2003369984
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003970712136409228,
      "loss": 1.5477,
      "theoretical_loss": 3.4583678644615024,
      "tokens_seen": 2003435520
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039705115346038117,
      "loss": 1.5342,
      "theoretical_loss": 3.45835851296534,
      "tokens_seen": 2003501056
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039703109327983953,
      "loss": 1.4531,
      "theoretical_loss": 3.4583491618607134,
      "tokens_seen": 2003566592
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039701103309929794,
      "loss": 1.3789,
      "theoretical_loss": 3.458339811147595,
      "tokens_seen": 2003632128
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003969909729187563,
      "loss": 1.4489,
      "theoretical_loss": 3.4583304608259544,
      "tokens_seen": 2003697664
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039697091273821467,
      "loss": 1.4241,
      "theoretical_loss": 3.458321110895763,
      "tokens_seen": 2003763200
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039695085255767303,
      "loss": 1.4571,
      "theoretical_loss": 3.458311761356991,
      "tokens_seen": 2003828736
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039693079237713145,
      "loss": 1.4547,
      "theoretical_loss": 3.45830241220961,
      "tokens_seen": 2003894272
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003969107321965898,
      "loss": 1.4034,
      "theoretical_loss": 3.45829306345359,
      "tokens_seen": 2003959808
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039689067201604817,
      "loss": 1.4601,
      "theoretical_loss": 3.4582837150889025,
      "tokens_seen": 2004025344
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039687061183550653,
      "loss": 1.4532,
      "theoretical_loss": 3.4582743671155183,
      "tokens_seen": 2004090880
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003968505516549649,
      "loss": 1.5111,
      "theoretical_loss": 3.4582650195334077,
      "tokens_seen": 2004156416
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003968304914744233,
      "loss": 1.4149,
      "theoretical_loss": 3.458255672342542,
      "tokens_seen": 2004221952
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003968104312938817,
      "loss": 1.4934,
      "theoretical_loss": 3.4582463255428917,
      "tokens_seen": 2004287488
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039679037111334004,
      "loss": 1.3303,
      "theoretical_loss": 3.458236979134428,
      "tokens_seen": 2004353024
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003967703109327984,
      "loss": 1.4439,
      "theoretical_loss": 3.4582276331171213,
      "tokens_seen": 2004418560
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003967502507522568,
      "loss": 1.4022,
      "theoretical_loss": 3.4582182874909426,
      "tokens_seen": 2004484096
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003967301905717152,
      "loss": 1.3848,
      "theoretical_loss": 3.4582089422558626,
      "tokens_seen": 2004549632
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039671013039117354,
      "loss": 1.3375,
      "theoretical_loss": 3.458199597411853,
      "tokens_seen": 2004615168
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003966900702106319,
      "loss": 1.458,
      "theoretical_loss": 3.4581902529588833,
      "tokens_seen": 2004680704
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039667001003009027,
      "loss": 1.4188,
      "theoretical_loss": 3.4581809088969253,
      "tokens_seen": 2004746240
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003966499498495487,
      "loss": 1.5052,
      "theoretical_loss": 3.4581715652259497,
      "tokens_seen": 2004811776
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039662988966900704,
      "loss": 1.3443,
      "theoretical_loss": 3.4581622219459276,
      "tokens_seen": 2004877312
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003966098294884654,
      "loss": 1.3147,
      "theoretical_loss": 3.458152879056829,
      "tokens_seen": 2004942848
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039658976930792377,
      "loss": 1.3504,
      "theoretical_loss": 3.4581435365586257,
      "tokens_seen": 2005008384
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003965697091273822,
      "loss": 1.3259,
      "theoretical_loss": 3.4581341944512882,
      "tokens_seen": 2005073920
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039654964894684055,
      "loss": 1.4543,
      "theoretical_loss": 3.4581248527347874,
      "tokens_seen": 2005139456
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003965295887662989,
      "loss": 1.3533,
      "theoretical_loss": 3.458115511409094,
      "tokens_seen": 2005204992
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039650952858575727,
      "loss": 1.4382,
      "theoretical_loss": 3.4581061704741787,
      "tokens_seen": 2005270528
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039648946840521563,
      "loss": 1.3953,
      "theoretical_loss": 3.458096829930013,
      "tokens_seen": 2005336064
    },
    {
      "epoch": 0.61,
      "objective/train/advantage_avg": 0.4937439262866974,
      "objective/train/docs_used": 731237,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3300145864486694,
      "objective/train/original_loss": 1.330014705657959,
      "objective/train/theoretical_loss": 3.458087489776567,
      "objective/train/tokens_used": 2025861600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24411408603191376,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506154298782349,
      "objective/train/weighted_lm_loss": 1.3972079753875732,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.458087489776567,
      "tokens_seen": 2005401600
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039646940822467405,
      "loss": 1.4191,
      "theoretical_loss": 3.458087489776567,
      "tokens_seen": 2005401600
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003964493480441324,
      "loss": 1.4618,
      "theoretical_loss": 3.4580781500138125,
      "tokens_seen": 2005467136
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003964292878635908,
      "loss": 1.4398,
      "theoretical_loss": 3.45806881064172,
      "tokens_seen": 2005532672
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039640922768304914,
      "loss": 1.3018,
      "theoretical_loss": 3.4580594716602597,
      "tokens_seen": 2005598208
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039638916750250755,
      "loss": 1.423,
      "theoretical_loss": 3.458050133069404,
      "tokens_seen": 2005663744
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003963691073219659,
      "loss": 1.4063,
      "theoretical_loss": 3.458040794869123,
      "tokens_seen": 2005729280
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003963490471414243,
      "loss": 1.4553,
      "theoretical_loss": 3.458031457059387,
      "tokens_seen": 2005794816
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039632898696088264,
      "loss": 1.4237,
      "theoretical_loss": 3.4580221196401673,
      "tokens_seen": 2005860352
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000396308926780341,
      "loss": 1.3688,
      "theoretical_loss": 3.4580127826114353,
      "tokens_seen": 2005925888
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003962888665997994,
      "loss": 1.4358,
      "theoretical_loss": 3.4580034459731612,
      "tokens_seen": 2005991424
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003962688064192578,
      "loss": 1.364,
      "theoretical_loss": 3.457994109725317,
      "tokens_seen": 2006056960
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039624874623871614,
      "loss": 1.3997,
      "theoretical_loss": 3.4579847738678726,
      "tokens_seen": 2006122496
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003962286860581745,
      "loss": 1.3838,
      "theoretical_loss": 3.457975438400799,
      "tokens_seen": 2006188032
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003962086258776329,
      "loss": 1.3197,
      "theoretical_loss": 3.4579661033240674,
      "tokens_seen": 2006253568
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003961885656970913,
      "loss": 1.3617,
      "theoretical_loss": 3.457956768637649,
      "tokens_seen": 2006319104
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039616850551654965,
      "loss": 1.4137,
      "theoretical_loss": 3.457947434341514,
      "tokens_seen": 2006384640
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000396148445336008,
      "loss": 1.332,
      "theoretical_loss": 3.4579381004356344,
      "tokens_seen": 2006450176
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039612838515546637,
      "loss": 1.4804,
      "theoretical_loss": 3.45792876691998,
      "tokens_seen": 2006515712
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003961083249749248,
      "loss": 1.5271,
      "theoretical_loss": 3.457919433794522,
      "tokens_seen": 2006581248
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039608826479438315,
      "loss": 1.3094,
      "theoretical_loss": 3.457910101059232,
      "tokens_seen": 2006646784
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003960682046138415,
      "loss": 1.3648,
      "theoretical_loss": 3.4579007687140804,
      "tokens_seen": 2006712320
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003960481444332999,
      "loss": 1.3876,
      "theoretical_loss": 3.4578914367590388,
      "tokens_seen": 2006777856
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003960280842527583,
      "loss": 1.5431,
      "theoretical_loss": 3.4578821051940767,
      "tokens_seen": 2006843392
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039600802407221665,
      "loss": 1.5127,
      "theoretical_loss": 3.457872774019166,
      "tokens_seen": 2006908928
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000395987963891675,
      "loss": 1.3736,
      "theoretical_loss": 3.4578634432342783,
      "tokens_seen": 2006974464
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003959679037111334,
      "loss": 1.4907,
      "theoretical_loss": 3.457854112839384,
      "tokens_seen": 2007040000
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003959478435305918,
      "loss": 1.3711,
      "theoretical_loss": 3.4578447828344534,
      "tokens_seen": 2007105536
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039592778335005016,
      "loss": 1.4517,
      "theoretical_loss": 3.4578354532194586,
      "tokens_seen": 2007171072
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003959077231695085,
      "loss": 1.5145,
      "theoretical_loss": 3.4578261239943693,
      "tokens_seen": 2007236608
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003958876629889669,
      "loss": 1.4272,
      "theoretical_loss": 3.4578167951591574,
      "tokens_seen": 2007302144
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039586760280842524,
      "loss": 1.4953,
      "theoretical_loss": 3.457807466713794,
      "tokens_seen": 2007367680
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039584754262788366,
      "loss": 1.3977,
      "theoretical_loss": 3.4577981386582497,
      "tokens_seen": 2007433216
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000395827482447342,
      "loss": 1.5222,
      "theoretical_loss": 3.4577888109924952,
      "tokens_seen": 2007498752
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003958074222668004,
      "loss": 1.3435,
      "theoretical_loss": 3.4577794837165023,
      "tokens_seen": 2007564288
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039578736208625875,
      "loss": 1.535,
      "theoretical_loss": 3.4577701568302412,
      "tokens_seen": 2007629824
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039576730190571716,
      "loss": 1.3711,
      "theoretical_loss": 3.457760830333683,
      "tokens_seen": 2007695360
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003957472417251755,
      "loss": 1.4571,
      "theoretical_loss": 3.4577515042267994,
      "tokens_seen": 2007760896
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003957271815446339,
      "loss": 1.3253,
      "theoretical_loss": 3.4577421785095606,
      "tokens_seen": 2007826432
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039570712136409225,
      "loss": 1.3904,
      "theoretical_loss": 3.4577328531819385,
      "tokens_seen": 2007891968
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003956870611835506,
      "loss": 1.4138,
      "theoretical_loss": 3.4577235282439025,
      "tokens_seen": 2007957504
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039566700100300903,
      "loss": 1.306,
      "theoretical_loss": 3.457714203695425,
      "tokens_seen": 2008023040
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003956469408224674,
      "loss": 1.3498,
      "theoretical_loss": 3.457704879536477,
      "tokens_seen": 2008088576
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003956268806419258,
      "loss": 1.3921,
      "theoretical_loss": 3.457695555767029,
      "tokens_seen": 2008154112
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039560682046138417,
      "loss": 1.421,
      "theoretical_loss": 3.457686232387052,
      "tokens_seen": 2008219648
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003955867602808426,
      "loss": 1.3837,
      "theoretical_loss": 3.4576769093965174,
      "tokens_seen": 2008285184
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039556670010030095,
      "loss": 1.4161,
      "theoretical_loss": 3.457667586795396,
      "tokens_seen": 2008350720
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003955466399197593,
      "loss": 1.4595,
      "theoretical_loss": 3.4576582645836584,
      "tokens_seen": 2008416256
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039552657973921767,
      "loss": 1.3417,
      "theoretical_loss": 3.457648942761277,
      "tokens_seen": 2008481792
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039550651955867603,
      "loss": 1.4411,
      "theoretical_loss": 3.4576396213282212,
      "tokens_seen": 2008547328
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039548645937813445,
      "loss": 1.4045,
      "theoretical_loss": 3.457630300284463,
      "tokens_seen": 2008612864
    },
    {
      "epoch": 0.61,
      "objective/train/advantage_avg": 0.4922792613506317,
      "objective/train/docs_used": 732466,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.360693335533142,
      "objective/train/original_loss": 1.360693335533142,
      "objective/train/theoretical_loss": 3.457620979629973,
      "objective/train/tokens_used": 2029138400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2426467090845108,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504614114761353,
      "objective/train/weighted_lm_loss": 1.4292813539505005,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.457620979629973,
      "tokens_seen": 2008678400
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003954663991975928,
      "loss": 1.4084,
      "theoretical_loss": 3.457620979629973,
      "tokens_seen": 2008678400
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003954463390170512,
      "loss": 1.5053,
      "theoretical_loss": 3.4576116593647224,
      "tokens_seen": 2008743936
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039542627883650954,
      "loss": 1.502,
      "theoretical_loss": 3.457602339488682,
      "tokens_seen": 2008809472
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039540621865596795,
      "loss": 1.3648,
      "theoretical_loss": 3.457593020001824,
      "tokens_seen": 2008875008
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003953861584754263,
      "loss": 1.4235,
      "theoretical_loss": 3.457583700904118,
      "tokens_seen": 2008940544
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003953660982948847,
      "loss": 1.471,
      "theoretical_loss": 3.457574382195536,
      "tokens_seen": 2009006080
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039534603811434304,
      "loss": 1.3985,
      "theoretical_loss": 3.4575650638760482,
      "tokens_seen": 2009071616
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003953259779338014,
      "loss": 1.4411,
      "theoretical_loss": 3.4575557459456263,
      "tokens_seen": 2009137152
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003953059177532598,
      "loss": 1.3838,
      "theoretical_loss": 3.457546428404241,
      "tokens_seen": 2009202688
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003952858575727182,
      "loss": 1.4155,
      "theoretical_loss": 3.457537111251864,
      "tokens_seen": 2009268224
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039526579739217654,
      "loss": 1.4716,
      "theoretical_loss": 3.457527794488466,
      "tokens_seen": 2009333760
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003952457372116349,
      "loss": 1.3939,
      "theoretical_loss": 3.457518478114018,
      "tokens_seen": 2009399296
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003952256770310933,
      "loss": 1.4214,
      "theoretical_loss": 3.457509162128491,
      "tokens_seen": 2009464832
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003952056168505517,
      "loss": 1.5203,
      "theoretical_loss": 3.4574998465318565,
      "tokens_seen": 2009530368
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039518555667001005,
      "loss": 1.352,
      "theoretical_loss": 3.457490531324085,
      "tokens_seen": 2009595904
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003951654964894684,
      "loss": 1.363,
      "theoretical_loss": 3.4574812165051476,
      "tokens_seen": 2009661440
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039514543630892677,
      "loss": 1.4844,
      "theoretical_loss": 3.4574719020750164,
      "tokens_seen": 2009726976
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003951253761283852,
      "loss": 1.3577,
      "theoretical_loss": 3.457462588033661,
      "tokens_seen": 2009792512
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039510531594784355,
      "loss": 1.2888,
      "theoretical_loss": 3.4574532743810535,
      "tokens_seen": 2009858048
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003950852557673019,
      "loss": 1.3738,
      "theoretical_loss": 3.457443961117165,
      "tokens_seen": 2009923584
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003950651955867603,
      "loss": 1.4649,
      "theoretical_loss": 3.457434648241966,
      "tokens_seen": 2009989120
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003950451354062187,
      "loss": 1.3353,
      "theoretical_loss": 3.457425335755428,
      "tokens_seen": 2010054656
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039502507522567705,
      "loss": 1.4121,
      "theoretical_loss": 3.4574160236575224,
      "tokens_seen": 2010120192
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003950050150451354,
      "loss": 1.3864,
      "theoretical_loss": 3.4574067119482197,
      "tokens_seen": 2010185728
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003949849548645938,
      "loss": 1.3981,
      "theoretical_loss": 3.4573974006274915,
      "tokens_seen": 2010251264
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003949648946840522,
      "loss": 1.5045,
      "theoretical_loss": 3.4573880896953084,
      "tokens_seen": 2010316800
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039494483450351056,
      "loss": 1.4079,
      "theoretical_loss": 3.457378779151642,
      "tokens_seen": 2010382336
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003949247743229689,
      "loss": 1.3432,
      "theoretical_loss": 3.457369468996463,
      "tokens_seen": 2010447872
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003949047141424273,
      "loss": 1.4297,
      "theoretical_loss": 3.457360159229743,
      "tokens_seen": 2010513408
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039488465396188564,
      "loss": 1.3231,
      "theoretical_loss": 3.457350849851453,
      "tokens_seen": 2010578944
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039486459378134406,
      "loss": 1.4986,
      "theoretical_loss": 3.457341540861564,
      "tokens_seen": 2010644480
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003948445336008024,
      "loss": 1.3307,
      "theoretical_loss": 3.4573322322600473,
      "tokens_seen": 2010710016
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003948244734202608,
      "loss": 1.4603,
      "theoretical_loss": 3.4573229240468737,
      "tokens_seen": 2010775552
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039480441323971915,
      "loss": 1.2984,
      "theoretical_loss": 3.457313616222015,
      "tokens_seen": 2010841088
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039478435305917756,
      "loss": 1.4519,
      "theoretical_loss": 3.4573043087854414,
      "tokens_seen": 2010906624
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003947642928786359,
      "loss": 1.3566,
      "theoretical_loss": 3.4572950017371245,
      "tokens_seen": 2010972160
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003947442326980943,
      "loss": 1.4212,
      "theoretical_loss": 3.4572856950770356,
      "tokens_seen": 2011037696
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039472417251755265,
      "loss": 1.4788,
      "theoretical_loss": 3.4572763888051457,
      "tokens_seen": 2011103232
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000394704112337011,
      "loss": 1.4503,
      "theoretical_loss": 3.4572670829214265,
      "tokens_seen": 2011168768
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039468405215646943,
      "loss": 1.4246,
      "theoretical_loss": 3.4572577774258484,
      "tokens_seen": 2011234304
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003946639919759278,
      "loss": 1.4372,
      "theoretical_loss": 3.4572484723183825,
      "tokens_seen": 2011299840
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039464393179538615,
      "loss": 1.3151,
      "theoretical_loss": 3.457239167599001,
      "tokens_seen": 2011365376
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003946238716148445,
      "loss": 1.4789,
      "theoretical_loss": 3.457229863267674,
      "tokens_seen": 2011430912
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039460381143430293,
      "loss": 1.4861,
      "theoretical_loss": 3.457220559324373,
      "tokens_seen": 2011496448
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003945837512537613,
      "loss": 1.3598,
      "theoretical_loss": 3.457211255769069,
      "tokens_seen": 2011561984
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039456369107321966,
      "loss": 1.3814,
      "theoretical_loss": 3.457201952601734,
      "tokens_seen": 2011627520
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000394543630892678,
      "loss": 1.379,
      "theoretical_loss": 3.457192649822338,
      "tokens_seen": 2011693056
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003945235707121364,
      "loss": 1.4324,
      "theoretical_loss": 3.4571833474308535,
      "tokens_seen": 2011758592
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003945035105315948,
      "loss": 1.4143,
      "theoretical_loss": 3.4571740454272506,
      "tokens_seen": 2011824128
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039448345035105316,
      "loss": 1.4269,
      "theoretical_loss": 3.457164743811501,
      "tokens_seen": 2011889664
    },
    {
      "epoch": 0.61,
      "objective/train/advantage_avg": 0.492768257856369,
      "objective/train/docs_used": 733698,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4772740602493286,
      "objective/train/original_loss": 1.4772741794586182,
      "objective/train/theoretical_loss": 3.4571554425835753,
      "objective/train/tokens_used": 2032415200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24320749938488007,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505132675170898,
      "objective/train/weighted_lm_loss": 1.5518702268600464,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4571554425835753,
      "tokens_seen": 2011955200
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003944633901705115,
      "loss": 1.5215,
      "theoretical_loss": 3.4571554425835753,
      "tokens_seen": 2011955200
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003944433299899699,
      "loss": 1.4184,
      "theoretical_loss": 3.4571461417434453,
      "tokens_seen": 2012020736
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003944232698094283,
      "loss": 1.4962,
      "theoretical_loss": 3.4571368412910823,
      "tokens_seen": 2012086272
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039440320962888666,
      "loss": 1.2997,
      "theoretical_loss": 3.4571275412264573,
      "tokens_seen": 2012151808
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000394383149448345,
      "loss": 1.4925,
      "theoretical_loss": 3.457118241549541,
      "tokens_seen": 2012217344
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003943630892678034,
      "loss": 1.4672,
      "theoretical_loss": 3.4571089422603056,
      "tokens_seen": 2012282880
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039434302908726175,
      "loss": 1.3912,
      "theoretical_loss": 3.4570996433587218,
      "tokens_seen": 2012348416
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039432296890672017,
      "loss": 1.4738,
      "theoretical_loss": 3.4570903448447607,
      "tokens_seen": 2012413952
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039430290872617853,
      "loss": 1.468,
      "theoretical_loss": 3.4570810467183932,
      "tokens_seen": 2012479488
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003942828485456369,
      "loss": 1.3675,
      "theoretical_loss": 3.4570717489795912,
      "tokens_seen": 2012545024
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039426278836509525,
      "loss": 1.4219,
      "theoretical_loss": 3.457062451628326,
      "tokens_seen": 2012610560
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039424272818455367,
      "loss": 1.4764,
      "theoretical_loss": 3.4570531546645684,
      "tokens_seen": 2012676096
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039422266800401203,
      "loss": 1.5133,
      "theoretical_loss": 3.45704385808829,
      "tokens_seen": 2012741632
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003942026078234704,
      "loss": 1.3754,
      "theoretical_loss": 3.457034561899461,
      "tokens_seen": 2012807168
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039418254764292876,
      "loss": 1.4696,
      "theoretical_loss": 3.4570252660980536,
      "tokens_seen": 2012872704
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003941624874623871,
      "loss": 1.5434,
      "theoretical_loss": 3.457015970684039,
      "tokens_seen": 2012938240
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003941424272818456,
      "loss": 1.4534,
      "theoretical_loss": 3.4570066756573885,
      "tokens_seen": 2013003776
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039412236710130395,
      "loss": 1.3514,
      "theoretical_loss": 3.4569973810180725,
      "tokens_seen": 2013069312
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003941023069207623,
      "loss": 1.4428,
      "theoretical_loss": 3.4569880867660636,
      "tokens_seen": 2013134848
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003940822467402207,
      "loss": 1.4802,
      "theoretical_loss": 3.4569787929013325,
      "tokens_seen": 2013200384
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003940621865596791,
      "loss": 1.324,
      "theoretical_loss": 3.45696949942385,
      "tokens_seen": 2013265920
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039404212637913745,
      "loss": 1.4077,
      "theoretical_loss": 3.456960206333587,
      "tokens_seen": 2013331456
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003940220661985958,
      "loss": 1.4983,
      "theoretical_loss": 3.456950913630516,
      "tokens_seen": 2013396992
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003940020060180542,
      "loss": 1.3905,
      "theoretical_loss": 3.4569416213146082,
      "tokens_seen": 2013462528
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003939819458375126,
      "loss": 1.3826,
      "theoretical_loss": 3.4569323293858334,
      "tokens_seen": 2013528064
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039396188565697096,
      "loss": 1.3919,
      "theoretical_loss": 3.4569230378441644,
      "tokens_seen": 2013593600
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003939418254764293,
      "loss": 1.4774,
      "theoretical_loss": 3.4569137466895716,
      "tokens_seen": 2013659136
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003939217652958877,
      "loss": 1.3544,
      "theoretical_loss": 3.456904455922027,
      "tokens_seen": 2013724672
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039390170511534604,
      "loss": 1.4352,
      "theoretical_loss": 3.4568951655415017,
      "tokens_seen": 2013790208
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039388164493480446,
      "loss": 1.3423,
      "theoretical_loss": 3.456885875547966,
      "tokens_seen": 2013855744
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003938615847542628,
      "loss": 1.3733,
      "theoretical_loss": 3.456876585941392,
      "tokens_seen": 2013921280
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003938415245737212,
      "loss": 1.3943,
      "theoretical_loss": 3.456867296721751,
      "tokens_seen": 2013986816
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039382146439317955,
      "loss": 1.4077,
      "theoretical_loss": 3.4568580078890143,
      "tokens_seen": 2014052352
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039380140421263796,
      "loss": 1.3758,
      "theoretical_loss": 3.4568487194431534,
      "tokens_seen": 2014117888
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003937813440320963,
      "loss": 1.4157,
      "theoretical_loss": 3.456839431384139,
      "tokens_seen": 2014183424
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003937612838515547,
      "loss": 1.5131,
      "theoretical_loss": 3.456830143711943,
      "tokens_seen": 2014248960
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039374122367101305,
      "loss": 1.4516,
      "theoretical_loss": 3.4568208564265364,
      "tokens_seen": 2014314496
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003937211634904714,
      "loss": 1.429,
      "theoretical_loss": 3.4568115695278903,
      "tokens_seen": 2014380032
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039370110330992983,
      "loss": 1.4588,
      "theoretical_loss": 3.456802283015976,
      "tokens_seen": 2014445568
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003936810431293882,
      "loss": 1.3018,
      "theoretical_loss": 3.4567929968907656,
      "tokens_seen": 2014511104
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039366098294884655,
      "loss": 1.4255,
      "theoretical_loss": 3.4567837111522293,
      "tokens_seen": 2014576640
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003936409227683049,
      "loss": 1.4157,
      "theoretical_loss": 3.4567744258003392,
      "tokens_seen": 2014642176
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039362086258776333,
      "loss": 1.4635,
      "theoretical_loss": 3.456765140835067,
      "tokens_seen": 2014707712
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003936008024072217,
      "loss": 1.3287,
      "theoretical_loss": 3.456755856256383,
      "tokens_seen": 2014773248
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039358074222668006,
      "loss": 1.4047,
      "theoretical_loss": 3.456746572064259,
      "tokens_seen": 2014838784
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003935606820461384,
      "loss": 1.3065,
      "theoretical_loss": 3.456737288258666,
      "tokens_seen": 2014904320
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003935406218655968,
      "loss": 1.4203,
      "theoretical_loss": 3.456728004839576,
      "tokens_seen": 2014969856
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003935205616850552,
      "loss": 1.4102,
      "theoretical_loss": 3.45671872180696,
      "tokens_seen": 2015035392
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039350050150451356,
      "loss": 1.4738,
      "theoretical_loss": 3.456709439160789,
      "tokens_seen": 2015100928
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003934804413239719,
      "loss": 1.4113,
      "theoretical_loss": 3.456700156901035,
      "tokens_seen": 2015166464
    },
    {
      "epoch": 0.61,
      "objective/train/advantage_avg": 0.4926469326019287,
      "objective/train/docs_used": 734924,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4056129455566406,
      "objective/train/original_loss": 1.405612826347351,
      "objective/train/theoretical_loss": 3.456690875027669,
      "objective/train/tokens_used": 2035692000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24303746223449707,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505002737045288,
      "objective/train/weighted_lm_loss": 1.476578950881958,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.456690875027669,
      "tokens_seen": 2015232000
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003934603811434303,
      "loss": 1.4399,
      "theoretical_loss": 3.456690875027669,
      "tokens_seen": 2015232000
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003934403209628887,
      "loss": 1.3149,
      "theoretical_loss": 3.4566815935406625,
      "tokens_seen": 2015297536
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039342026078234706,
      "loss": 1.4563,
      "theoretical_loss": 3.456672312439986,
      "tokens_seen": 2015363072
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003934002006018054,
      "loss": 1.4581,
      "theoretical_loss": 3.456663031725612,
      "tokens_seen": 2015428608
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003933801404212638,
      "loss": 1.3999,
      "theoretical_loss": 3.4566537513975115,
      "tokens_seen": 2015494144
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039336008024072215,
      "loss": 1.3202,
      "theoretical_loss": 3.456644471455656,
      "tokens_seen": 2015559680
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039334002006018057,
      "loss": 1.4532,
      "theoretical_loss": 3.4566351919000167,
      "tokens_seen": 2015625216
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039331995987963893,
      "loss": 1.325,
      "theoretical_loss": 3.4566259127305647,
      "tokens_seen": 2015690752
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003932998996990973,
      "loss": 1.3272,
      "theoretical_loss": 3.456616633947272,
      "tokens_seen": 2015756288
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039327983951855565,
      "loss": 1.4346,
      "theoretical_loss": 3.456607355550109,
      "tokens_seen": 2015821824
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039325977933801407,
      "loss": 1.2946,
      "theoretical_loss": 3.4565980775390477,
      "tokens_seen": 2015887360
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039323971915747243,
      "loss": 1.4557,
      "theoretical_loss": 3.45658879991406,
      "tokens_seen": 2015952896
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003932196589769308,
      "loss": 1.4056,
      "theoretical_loss": 3.456579522675116,
      "tokens_seen": 2016018432
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039319959879638916,
      "loss": 1.3866,
      "theoretical_loss": 3.4565702458221885,
      "tokens_seen": 2016083968
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003931795386158475,
      "loss": 1.4698,
      "theoretical_loss": 3.456560969355248,
      "tokens_seen": 2016149504
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039315947843530593,
      "loss": 1.4175,
      "theoretical_loss": 3.4565516932742657,
      "tokens_seen": 2016215040
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003931394182547643,
      "loss": 1.4659,
      "theoretical_loss": 3.4565424175792137,
      "tokens_seen": 2016280576
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039311935807422266,
      "loss": 1.3382,
      "theoretical_loss": 3.4565331422700636,
      "tokens_seen": 2016346112
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000393099297893681,
      "loss": 1.3592,
      "theoretical_loss": 3.456523867346786,
      "tokens_seen": 2016411648
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039307923771313944,
      "loss": 1.3694,
      "theoretical_loss": 3.456514592809352,
      "tokens_seen": 2016477184
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003930591775325978,
      "loss": 1.3498,
      "theoretical_loss": 3.4565053186577344,
      "tokens_seen": 2016542720
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039303911735205616,
      "loss": 1.3654,
      "theoretical_loss": 3.4564960448919035,
      "tokens_seen": 2016608256
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003930190571715145,
      "loss": 1.3334,
      "theoretical_loss": 3.4564867715118313,
      "tokens_seen": 2016673792
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003929989969909729,
      "loss": 1.4667,
      "theoretical_loss": 3.4564774985174886,
      "tokens_seen": 2016739328
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003929789368104313,
      "loss": 1.4363,
      "theoretical_loss": 3.4564682259088473,
      "tokens_seen": 2016804864
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039295887662988967,
      "loss": 1.4034,
      "theoretical_loss": 3.456458953685879,
      "tokens_seen": 2016870400
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039293881644934803,
      "loss": 1.3747,
      "theoretical_loss": 3.4564496818485546,
      "tokens_seen": 2016935936
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003929187562688064,
      "loss": 1.4258,
      "theoretical_loss": 3.456440410396846,
      "tokens_seen": 2017001472
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003928986960882648,
      "loss": 1.3807,
      "theoretical_loss": 3.456431139330724,
      "tokens_seen": 2017067008
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039287863590772317,
      "loss": 1.5189,
      "theoretical_loss": 3.456421868650161,
      "tokens_seen": 2017132544
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039285857572718153,
      "loss": 1.4164,
      "theoretical_loss": 3.4564125983551275,
      "tokens_seen": 2017198080
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003928385155466399,
      "loss": 1.5125,
      "theoretical_loss": 3.456403328445595,
      "tokens_seen": 2017263616
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039281845536609825,
      "loss": 1.5368,
      "theoretical_loss": 3.456394058921536,
      "tokens_seen": 2017329152
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039279839518555667,
      "loss": 1.3639,
      "theoretical_loss": 3.4563847897829207,
      "tokens_seen": 2017394688
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039277833500501503,
      "loss": 1.3603,
      "theoretical_loss": 3.4563755210297216,
      "tokens_seen": 2017460224
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003927582748244734,
      "loss": 1.3748,
      "theoretical_loss": 3.456366252661909,
      "tokens_seen": 2017525760
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039273821464393176,
      "loss": 1.443,
      "theoretical_loss": 3.4563569846794557,
      "tokens_seen": 2017591296
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003927181544633902,
      "loss": 1.5111,
      "theoretical_loss": 3.456347717082332,
      "tokens_seen": 2017656832
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039269809428284854,
      "loss": 1.3591,
      "theoretical_loss": 3.4563384498705094,
      "tokens_seen": 2017722368
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003926780341023069,
      "loss": 1.3753,
      "theoretical_loss": 3.4563291830439606,
      "tokens_seen": 2017787904
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039265797392176526,
      "loss": 1.416,
      "theoretical_loss": 3.4563199166026557,
      "tokens_seen": 2017853440
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039263791374122373,
      "loss": 1.4318,
      "theoretical_loss": 3.456310650546567,
      "tokens_seen": 2017918976
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003926178535606821,
      "loss": 1.4117,
      "theoretical_loss": 3.456301384875666,
      "tokens_seen": 2017984512
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039259779338014046,
      "loss": 1.3232,
      "theoretical_loss": 3.456292119589923,
      "tokens_seen": 2018050048
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003925777331995988,
      "loss": 1.354,
      "theoretical_loss": 3.456282854689311,
      "tokens_seen": 2018115584
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003925576730190572,
      "loss": 1.4316,
      "theoretical_loss": 3.456273590173801,
      "tokens_seen": 2018181120
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003925376128385156,
      "loss": 1.4117,
      "theoretical_loss": 3.456264326043364,
      "tokens_seen": 2018246656
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039251755265797396,
      "loss": 1.3434,
      "theoretical_loss": 3.4562550622979717,
      "tokens_seen": 2018312192
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003924974924774323,
      "loss": 1.4444,
      "theoretical_loss": 3.456245798937596,
      "tokens_seen": 2018377728
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003924774322968907,
      "loss": 1.4934,
      "theoretical_loss": 3.456236535962208,
      "tokens_seen": 2018443264
    },
    {
      "epoch": 0.61,
      "objective/train/advantage_avg": 0.4932577311992645,
      "objective/train/docs_used": 736110,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2976014614105225,
      "objective/train/original_loss": 1.2976012229919434,
      "objective/train/theoretical_loss": 3.4562272733717796,
      "objective/train/tokens_used": 2038968800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24378927052021027,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505650043487549,
      "objective/train/weighted_lm_loss": 1.3633230924606323,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4562272733717796,
      "tokens_seen": 2018508800
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003924573721163491,
      "loss": 1.4088,
      "theoretical_loss": 3.4562272733717796,
      "tokens_seen": 2018508800
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039243731193580746,
      "loss": 1.4467,
      "theoretical_loss": 3.4562180111662815,
      "tokens_seen": 2018574336
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003924172517552658,
      "loss": 1.4447,
      "theoretical_loss": 3.456208749345686,
      "tokens_seen": 2018639872
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003923971915747242,
      "loss": 1.3937,
      "theoretical_loss": 3.4561994879099647,
      "tokens_seen": 2018705408
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039237713139418255,
      "loss": 1.3508,
      "theoretical_loss": 3.4561902268590883,
      "tokens_seen": 2018770944
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039235707121364097,
      "loss": 1.3632,
      "theoretical_loss": 3.4561809661930285,
      "tokens_seen": 2018836480
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039233701103309933,
      "loss": 1.3815,
      "theoretical_loss": 3.456171705911758,
      "tokens_seen": 2018902016
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003923169508525577,
      "loss": 1.4599,
      "theoretical_loss": 3.4561624460152465,
      "tokens_seen": 2018967552
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039229689067201605,
      "loss": 1.3758,
      "theoretical_loss": 3.4561531865034665,
      "tokens_seen": 2019033088
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039227683049147447,
      "loss": 1.316,
      "theoretical_loss": 3.45614392737639,
      "tokens_seen": 2019098624
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039225677031093283,
      "loss": 1.2815,
      "theoretical_loss": 3.4561346686339878,
      "tokens_seen": 2019164160
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003922367101303912,
      "loss": 1.3834,
      "theoretical_loss": 3.4561254102762318,
      "tokens_seen": 2019229696
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039221664994984956,
      "loss": 1.4324,
      "theoretical_loss": 3.4561161523030925,
      "tokens_seen": 2019295232
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003921965897693079,
      "loss": 1.3424,
      "theoretical_loss": 3.4561068947145435,
      "tokens_seen": 2019360768
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039217652958876633,
      "loss": 1.3912,
      "theoretical_loss": 3.4560976375105543,
      "tokens_seen": 2019426304
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003921564694082247,
      "loss": 1.4405,
      "theoretical_loss": 3.4560883806910976,
      "tokens_seen": 2019491840
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039213640922768306,
      "loss": 1.4029,
      "theoretical_loss": 3.456079124256145,
      "tokens_seen": 2019557376
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003921163490471414,
      "loss": 1.3533,
      "theoretical_loss": 3.456069868205667,
      "tokens_seen": 2019622912
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039209628886659984,
      "loss": 1.5274,
      "theoretical_loss": 3.4560606125396363,
      "tokens_seen": 2019688448
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003920762286860582,
      "loss": 1.4919,
      "theoretical_loss": 3.4560513572580236,
      "tokens_seen": 2019753984
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039205616850551656,
      "loss": 1.482,
      "theoretical_loss": 3.4560421023608012,
      "tokens_seen": 2019819520
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003920361083249749,
      "loss": 1.3703,
      "theoretical_loss": 3.4560328478479407,
      "tokens_seen": 2019885056
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003920160481444333,
      "loss": 1.5662,
      "theoretical_loss": 3.4560235937194124,
      "tokens_seen": 2019950592
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003919959879638917,
      "loss": 1.4147,
      "theoretical_loss": 3.4560143399751895,
      "tokens_seen": 2020016128
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039197592778335007,
      "loss": 1.4988,
      "theoretical_loss": 3.4560050866152423,
      "tokens_seen": 2020081664
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039195586760280843,
      "loss": 1.3508,
      "theoretical_loss": 3.4559958336395438,
      "tokens_seen": 2020147200
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003919358074222668,
      "loss": 1.3871,
      "theoretical_loss": 3.455986581048064,
      "tokens_seen": 2020212736
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003919157472417252,
      "loss": 1.4053,
      "theoretical_loss": 3.455977328840775,
      "tokens_seen": 2020278272
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039189568706118357,
      "loss": 1.3756,
      "theoretical_loss": 3.455968077017649,
      "tokens_seen": 2020343808
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039187562688064193,
      "loss": 1.3621,
      "theoretical_loss": 3.455958825578657,
      "tokens_seen": 2020409344
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003918555667001003,
      "loss": 1.4479,
      "theoretical_loss": 3.4559495745237707,
      "tokens_seen": 2020474880
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039183550651955866,
      "loss": 1.4232,
      "theoretical_loss": 3.4559403238529622,
      "tokens_seen": 2020540416
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039181544633901707,
      "loss": 1.3619,
      "theoretical_loss": 3.455931073566202,
      "tokens_seen": 2020605952
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039179538615847543,
      "loss": 1.4162,
      "theoretical_loss": 3.455921823663463,
      "tokens_seen": 2020671488
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003917753259779338,
      "loss": 1.3742,
      "theoretical_loss": 3.4559125741447154,
      "tokens_seen": 2020737024
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039175526579739216,
      "loss": 1.385,
      "theoretical_loss": 3.455903325009932,
      "tokens_seen": 2020802560
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003917352056168506,
      "loss": 1.4357,
      "theoretical_loss": 3.4558940762590837,
      "tokens_seen": 2020868096
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039171514543630894,
      "loss": 1.523,
      "theoretical_loss": 3.4558848278921426,
      "tokens_seen": 2020933632
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003916950852557673,
      "loss": 1.5114,
      "theoretical_loss": 3.4558755799090797,
      "tokens_seen": 2020999168
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039167502507522566,
      "loss": 1.415,
      "theoretical_loss": 3.4558663323098675,
      "tokens_seen": 2021064704
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003916549648946841,
      "loss": 1.345,
      "theoretical_loss": 3.455857085094477,
      "tokens_seen": 2021130240
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039163490471414244,
      "loss": 1.3354,
      "theoretical_loss": 3.45584783826288,
      "tokens_seen": 2021195776
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003916148445336008,
      "loss": 1.3979,
      "theoretical_loss": 3.455838591815048,
      "tokens_seen": 2021261312
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039159478435305916,
      "loss": 1.4126,
      "theoretical_loss": 3.455829345750953,
      "tokens_seen": 2021326848
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003915747241725175,
      "loss": 1.3297,
      "theoretical_loss": 3.4558201000705653,
      "tokens_seen": 2021392384
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039155466399197594,
      "loss": 1.3572,
      "theoretical_loss": 3.4558108547738584,
      "tokens_seen": 2021457920
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003915346038114343,
      "loss": 1.5,
      "theoretical_loss": 3.4558016098608033,
      "tokens_seen": 2021523456
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039151454363089267,
      "loss": 1.4588,
      "theoretical_loss": 3.455792365331371,
      "tokens_seen": 2021588992
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039149448345035103,
      "loss": 1.4933,
      "theoretical_loss": 3.455783121185534,
      "tokens_seen": 2021654528
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039147442326980945,
      "loss": 1.3876,
      "theoretical_loss": 3.4557738774232636,
      "tokens_seen": 2021720064
    },
    {
      "epoch": 0.61,
      "objective/train/advantage_avg": 0.4938117563724518,
      "objective/train/docs_used": 737310,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3226114511489868,
      "objective/train/original_loss": 1.3226115703582764,
      "objective/train/theoretical_loss": 3.455764634044531,
      "objective/train/tokens_used": 2042245600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2442186176776886,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050622820854187,
      "objective/train/weighted_lm_loss": 1.3894543647766113,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.455764634044531,
      "tokens_seen": 2021785600
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003914543630892678,
      "loss": 1.4308,
      "theoretical_loss": 3.455764634044531,
      "tokens_seen": 2021785600
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039143430290872617,
      "loss": 1.3608,
      "theoretical_loss": 3.4557553910493084,
      "tokens_seen": 2021851136
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039141424272818453,
      "loss": 1.2168,
      "theoretical_loss": 3.4557461484375676,
      "tokens_seen": 2021916672
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003913941825476429,
      "loss": 1.2809,
      "theoretical_loss": 3.455736906209279,
      "tokens_seen": 2021982208
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003913741223671013,
      "loss": 1.4093,
      "theoretical_loss": 3.4557276643644164,
      "tokens_seen": 2022047744
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003913540621865597,
      "loss": 1.5632,
      "theoretical_loss": 3.45571842290295,
      "tokens_seen": 2022113280
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039133400200601804,
      "loss": 1.493,
      "theoretical_loss": 3.4557091818248518,
      "tokens_seen": 2022178816
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003913139418254764,
      "loss": 1.3598,
      "theoretical_loss": 3.455699941130093,
      "tokens_seen": 2022244352
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003912938816449348,
      "loss": 1.4692,
      "theoretical_loss": 3.4556907008186464,
      "tokens_seen": 2022309888
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003912738214643932,
      "loss": 1.3565,
      "theoretical_loss": 3.4556814608904824,
      "tokens_seen": 2022375424
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039125376128385154,
      "loss": 1.4592,
      "theoretical_loss": 3.4556722213455737,
      "tokens_seen": 2022440960
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003912337011033099,
      "loss": 1.3425,
      "theoretical_loss": 3.455662982183892,
      "tokens_seen": 2022506496
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039121364092276826,
      "loss": 1.4492,
      "theoretical_loss": 3.4556537434054073,
      "tokens_seen": 2022572032
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003911935807422267,
      "loss": 1.2982,
      "theoretical_loss": 3.4556445050100937,
      "tokens_seen": 2022637568
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039117352056168504,
      "loss": 1.4923,
      "theoretical_loss": 3.455635266997921,
      "tokens_seen": 2022703104
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003911534603811434,
      "loss": 1.2983,
      "theoretical_loss": 3.455626029368862,
      "tokens_seen": 2022768640
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003911334002006018,
      "loss": 1.4023,
      "theoretical_loss": 3.455616792122888,
      "tokens_seen": 2022834176
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039111334002006024,
      "loss": 1.3921,
      "theoretical_loss": 3.455607555259971,
      "tokens_seen": 2022899712
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003910932798395186,
      "loss": 1.4192,
      "theoretical_loss": 3.4555983187800825,
      "tokens_seen": 2022965248
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039107321965897696,
      "loss": 1.5231,
      "theoretical_loss": 3.4555890826831934,
      "tokens_seen": 2023030784
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003910531594784353,
      "loss": 1.2966,
      "theoretical_loss": 3.455579846969277,
      "tokens_seen": 2023096320
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003910330992978937,
      "loss": 1.351,
      "theoretical_loss": 3.455570611638304,
      "tokens_seen": 2023161856
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003910130391173521,
      "loss": 1.4726,
      "theoretical_loss": 3.455561376690246,
      "tokens_seen": 2023227392
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039099297893681047,
      "loss": 1.3664,
      "theoretical_loss": 3.455552142125075,
      "tokens_seen": 2023292928
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039097291875626883,
      "loss": 1.3845,
      "theoretical_loss": 3.455542907942763,
      "tokens_seen": 2023358464
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003909528585757272,
      "loss": 1.4797,
      "theoretical_loss": 3.4555336741432816,
      "tokens_seen": 2023424000
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003909327983951856,
      "loss": 1.3857,
      "theoretical_loss": 3.4555244407266024,
      "tokens_seen": 2023489536
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039091273821464397,
      "loss": 1.4434,
      "theoretical_loss": 3.455515207692697,
      "tokens_seen": 2023555072
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039089267803410233,
      "loss": 1.4386,
      "theoretical_loss": 3.4555059750415373,
      "tokens_seen": 2023620608
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003908726178535607,
      "loss": 1.4595,
      "theoretical_loss": 3.455496742773095,
      "tokens_seen": 2023686144
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039085255767301906,
      "loss": 1.3618,
      "theoretical_loss": 3.455487510887342,
      "tokens_seen": 2023751680
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039083249749247747,
      "loss": 1.3693,
      "theoretical_loss": 3.45547827938425,
      "tokens_seen": 2023817216
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039081243731193583,
      "loss": 1.4597,
      "theoretical_loss": 3.45546904826379,
      "tokens_seen": 2023882752
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003907923771313942,
      "loss": 1.4671,
      "theoretical_loss": 3.4554598175259352,
      "tokens_seen": 2023948288
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039077231695085256,
      "loss": 1.4068,
      "theoretical_loss": 3.455450587170656,
      "tokens_seen": 2024013824
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000390752256770311,
      "loss": 1.4109,
      "theoretical_loss": 3.455441357197925,
      "tokens_seen": 2024079360
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039073219658976934,
      "loss": 1.4729,
      "theoretical_loss": 3.455432127607714,
      "tokens_seen": 2024144896
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003907121364092277,
      "loss": 1.5376,
      "theoretical_loss": 3.455422898399994,
      "tokens_seen": 2024210432
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039069207622868606,
      "loss": 1.4105,
      "theoretical_loss": 3.455413669574737,
      "tokens_seen": 2024275968
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003906720160481445,
      "loss": 1.4363,
      "theoretical_loss": 3.4554044411319156,
      "tokens_seen": 2024341504
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039065195586760284,
      "loss": 1.4304,
      "theoretical_loss": 3.4553952130715,
      "tokens_seen": 2024407040
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003906318956870612,
      "loss": 1.3729,
      "theoretical_loss": 3.4553859853934643,
      "tokens_seen": 2024472576
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039061183550651957,
      "loss": 1.4075,
      "theoretical_loss": 3.4553767580977777,
      "tokens_seen": 2024538112
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039059177532597793,
      "loss": 1.4145,
      "theoretical_loss": 3.4553675311844136,
      "tokens_seen": 2024603648
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039057171514543634,
      "loss": 1.3797,
      "theoretical_loss": 3.455358304653344,
      "tokens_seen": 2024669184
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003905516549648947,
      "loss": 1.3625,
      "theoretical_loss": 3.455349078504539,
      "tokens_seen": 2024734720
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039053159478435307,
      "loss": 1.4945,
      "theoretical_loss": 3.4553398527379717,
      "tokens_seen": 2024800256
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039051153460381143,
      "loss": 1.494,
      "theoretical_loss": 3.455330627353614,
      "tokens_seen": 2024865792
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039049147442326985,
      "loss": 1.372,
      "theoretical_loss": 3.455321402351437,
      "tokens_seen": 2024931328
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003904714142427282,
      "loss": 1.3557,
      "theoretical_loss": 3.455312177731413,
      "tokens_seen": 2024996864
    },
    {
      "epoch": 0.61,
      "objective/train/advantage_avg": 0.49007049202919006,
      "objective/train/docs_used": 738608,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2743072509765625,
      "objective/train/original_loss": 1.2743072509765625,
      "objective/train/theoretical_loss": 3.4553029534935136,
      "objective/train/tokens_used": 2045522400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24120016396045685,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050233244895935,
      "objective/train/weighted_lm_loss": 1.3384039402008057,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4553029534935136,
      "tokens_seen": 2025062400
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039045135406218657,
      "loss": 1.345,
      "theoretical_loss": 3.4553029534935136,
      "tokens_seen": 2025062400
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039043129388164493,
      "loss": 1.4788,
      "theoretical_loss": 3.455293729637711,
      "tokens_seen": 2025127936
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003904112337011033,
      "loss": 1.3085,
      "theoretical_loss": 3.4552845061639763,
      "tokens_seen": 2025193472
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003903911735205617,
      "loss": 1.3898,
      "theoretical_loss": 3.4552752830722815,
      "tokens_seen": 2025259008
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003903711133400201,
      "loss": 1.3765,
      "theoretical_loss": 3.4552660603625984,
      "tokens_seen": 2025324544
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039035105315947844,
      "loss": 1.3705,
      "theoretical_loss": 3.4552568380348996,
      "tokens_seen": 2025390080
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003903309929789368,
      "loss": 1.3228,
      "theoretical_loss": 3.455247616089156,
      "tokens_seen": 2025455616
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003903109327983952,
      "loss": 1.3811,
      "theoretical_loss": 3.45523839452534,
      "tokens_seen": 2025521152
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003902908726178536,
      "loss": 1.4719,
      "theoretical_loss": 3.455229173343423,
      "tokens_seen": 2025586688
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039027081243731194,
      "loss": 1.3953,
      "theoretical_loss": 3.4552199525433767,
      "tokens_seen": 2025652224
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003902507522567703,
      "loss": 1.4232,
      "theoretical_loss": 3.4552107321251735,
      "tokens_seen": 2025717760
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039023069207622866,
      "loss": 1.403,
      "theoretical_loss": 3.4552015120887845,
      "tokens_seen": 2025783296
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003902106318956871,
      "loss": 1.4158,
      "theoretical_loss": 3.455192292434183,
      "tokens_seen": 2025848832
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039019057171514544,
      "loss": 1.3987,
      "theoretical_loss": 3.455183073161339,
      "tokens_seen": 2025914368
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003901705115346038,
      "loss": 1.4085,
      "theoretical_loss": 3.4551738542702255,
      "tokens_seen": 2025979904
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039015045135406217,
      "loss": 1.3606,
      "theoretical_loss": 3.4551646357608137,
      "tokens_seen": 2026045440
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003901303911735206,
      "loss": 1.2915,
      "theoretical_loss": 3.455155417633076,
      "tokens_seen": 2026110976
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039011033099297895,
      "loss": 1.3124,
      "theoretical_loss": 3.455146199886984,
      "tokens_seen": 2026176512
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003900902708124373,
      "loss": 1.3109,
      "theoretical_loss": 3.455136982522509,
      "tokens_seen": 2026242048
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039007021063189567,
      "loss": 1.4065,
      "theoretical_loss": 3.4551277655396246,
      "tokens_seen": 2026307584
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039005015045135403,
      "loss": 1.3968,
      "theoretical_loss": 3.4551185489383007,
      "tokens_seen": 2026373120
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00039003009027081245,
      "loss": 1.3653,
      "theoretical_loss": 3.4551093327185103,
      "tokens_seen": 2026438656
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003900100300902708,
      "loss": 1.3854,
      "theoretical_loss": 3.455100116880225,
      "tokens_seen": 2026504192
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003899899699097292,
      "loss": 1.3702,
      "theoretical_loss": 3.455090901423416,
      "tokens_seen": 2026569728
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038996990972918754,
      "loss": 1.4745,
      "theoretical_loss": 3.4550816863480565,
      "tokens_seen": 2026635264
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038994984954864595,
      "loss": 1.3235,
      "theoretical_loss": 3.4550724716541175,
      "tokens_seen": 2026700800
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003899297893681043,
      "loss": 1.3877,
      "theoretical_loss": 3.455063257341571,
      "tokens_seen": 2026766336
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003899097291875627,
      "loss": 1.3886,
      "theoretical_loss": 3.4550540434103887,
      "tokens_seen": 2026831872
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038988966900702104,
      "loss": 1.5277,
      "theoretical_loss": 3.455044829860543,
      "tokens_seen": 2026897408
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003898696088264794,
      "loss": 1.4021,
      "theoretical_loss": 3.455035616692005,
      "tokens_seen": 2026962944
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003898495486459378,
      "loss": 1.3776,
      "theoretical_loss": 3.455026403904747,
      "tokens_seen": 2027028480
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003898294884653962,
      "loss": 1.5052,
      "theoretical_loss": 3.455017191498742,
      "tokens_seen": 2027094016
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038980942828485454,
      "loss": 1.4491,
      "theoretical_loss": 3.45500797947396,
      "tokens_seen": 2027159552
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003897893681043129,
      "loss": 1.3569,
      "theoretical_loss": 3.454998767830374,
      "tokens_seen": 2027225088
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003897693079237713,
      "loss": 1.3742,
      "theoretical_loss": 3.4549895565679556,
      "tokens_seen": 2027290624
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003897492477432297,
      "loss": 1.5427,
      "theoretical_loss": 3.4549803456866774,
      "tokens_seen": 2027356160
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038972918756268805,
      "loss": 1.3656,
      "theoretical_loss": 3.45497113518651,
      "tokens_seen": 2027421696
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003897091273821464,
      "loss": 1.4186,
      "theoretical_loss": 3.454961925067426,
      "tokens_seen": 2027487232
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038968906720160477,
      "loss": 1.4372,
      "theoretical_loss": 3.4549527153293975,
      "tokens_seen": 2027552768
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003896690070210632,
      "loss": 1.4457,
      "theoretical_loss": 3.4549435059723965,
      "tokens_seen": 2027618304
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038964894684052155,
      "loss": 1.514,
      "theoretical_loss": 3.4549342969963943,
      "tokens_seen": 2027683840
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038962888665997997,
      "loss": 1.3262,
      "theoretical_loss": 3.4549250884013634,
      "tokens_seen": 2027749376
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038960882647943833,
      "loss": 1.4484,
      "theoretical_loss": 3.4549158801872757,
      "tokens_seen": 2027814912
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038958876629889674,
      "loss": 1.4735,
      "theoretical_loss": 3.4549066723541024,
      "tokens_seen": 2027880448
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003895687061183551,
      "loss": 1.3898,
      "theoretical_loss": 3.4548974649018165,
      "tokens_seen": 2027945984
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038954864593781347,
      "loss": 1.5454,
      "theoretical_loss": 3.454888257830389,
      "tokens_seen": 2028011520
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038952858575727183,
      "loss": 1.3874,
      "theoretical_loss": 3.4548790511397924,
      "tokens_seen": 2028077056
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038950852557673025,
      "loss": 1.3961,
      "theoretical_loss": 3.4548698448299984,
      "tokens_seen": 2028142592
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003894884653961886,
      "loss": 1.3403,
      "theoretical_loss": 3.4548606389009793,
      "tokens_seen": 2028208128
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038946840521564697,
      "loss": 1.2751,
      "theoretical_loss": 3.4548514333527067,
      "tokens_seen": 2028273664
    },
    {
      "epoch": 0.61,
      "objective/train/advantage_avg": 0.496250718832016,
      "objective/train/docs_used": 739750,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3699121475219727,
      "objective/train/original_loss": 1.3699121475219727,
      "objective/train/theoretical_loss": 3.4548422281851527,
      "objective/train/tokens_used": 2048799200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2464478760957718,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508781671524048,
      "objective/train/weighted_lm_loss": 1.4396265745162964,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4548422281851527,
      "tokens_seen": 2028339200
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038944834503510533,
      "loss": 1.3783,
      "theoretical_loss": 3.4548422281851527,
      "tokens_seen": 2028339200
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003894282848545637,
      "loss": 1.3481,
      "theoretical_loss": 3.454833023398289,
      "tokens_seen": 2028404736
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003894082246740221,
      "loss": 1.4285,
      "theoretical_loss": 3.4548238189920877,
      "tokens_seen": 2028470272
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003893881644934805,
      "loss": 1.4571,
      "theoretical_loss": 3.454814614966521,
      "tokens_seen": 2028535808
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038936810431293884,
      "loss": 1.3629,
      "theoretical_loss": 3.4548054113215607,
      "tokens_seen": 2028601344
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003893480441323972,
      "loss": 1.2484,
      "theoretical_loss": 3.4547962080571786,
      "tokens_seen": 2028666880
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003893279839518556,
      "loss": 1.4111,
      "theoretical_loss": 3.4547870051733467,
      "tokens_seen": 2028732416
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000389307923771314,
      "loss": 1.4577,
      "theoretical_loss": 3.4547778026700375,
      "tokens_seen": 2028797952
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038928786359077234,
      "loss": 1.4342,
      "theoretical_loss": 3.4547686005472222,
      "tokens_seen": 2028863488
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003892678034102307,
      "loss": 1.3942,
      "theoretical_loss": 3.4547593988048737,
      "tokens_seen": 2028929024
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038924774322968906,
      "loss": 1.319,
      "theoretical_loss": 3.4547501974429626,
      "tokens_seen": 2028994560
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003892276830491475,
      "loss": 1.3742,
      "theoretical_loss": 3.454740996461462,
      "tokens_seen": 2029060096
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038920762286860584,
      "loss": 1.3622,
      "theoretical_loss": 3.4547317958603436,
      "tokens_seen": 2029125632
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003891875626880642,
      "loss": 1.333,
      "theoretical_loss": 3.4547225956395797,
      "tokens_seen": 2029191168
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038916750250752257,
      "loss": 1.3986,
      "theoretical_loss": 3.454713395799142,
      "tokens_seen": 2029256704
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.000389147442326981,
      "loss": 1.3664,
      "theoretical_loss": 3.454704196339002,
      "tokens_seen": 2029322240
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00038912738214643935,
      "loss": 1.4102,
      "theoretical_loss": 3.454694997259132,
      "tokens_seen": 2029387776
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0003891073219658977,
      "loss": 1.332,
      "theoretical_loss": 3.4546857985595048,
      "tokens_seen": 2029453312
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038908726178535607,
      "loss": 1.5119,
      "theoretical_loss": 3.4546766002400915,
      "tokens_seen": 2029518848
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038906720160481443,
      "loss": 1.3899,
      "theoretical_loss": 3.4546674023008643,
      "tokens_seen": 2029584384
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038904714142427285,
      "loss": 1.4211,
      "theoretical_loss": 3.4546582047417953,
      "tokens_seen": 2029649920
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003890270812437312,
      "loss": 1.3434,
      "theoretical_loss": 3.454649007562857,
      "tokens_seen": 2029715456
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003890070210631896,
      "loss": 1.4359,
      "theoretical_loss": 3.4546398107640197,
      "tokens_seen": 2029780992
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038898696088264794,
      "loss": 1.4051,
      "theoretical_loss": 3.4546306143452576,
      "tokens_seen": 2029846528
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038896690070210635,
      "loss": 1.4159,
      "theoretical_loss": 3.454621418306542,
      "tokens_seen": 2029912064
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003889468405215647,
      "loss": 1.487,
      "theoretical_loss": 3.4546122226478437,
      "tokens_seen": 2029977600
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003889267803410231,
      "loss": 1.5113,
      "theoretical_loss": 3.4546030273691364,
      "tokens_seen": 2030043136
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038890672016048144,
      "loss": 1.4356,
      "theoretical_loss": 3.4545938324703913,
      "tokens_seen": 2030108672
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003888866599799398,
      "loss": 1.3834,
      "theoretical_loss": 3.4545846379515806,
      "tokens_seen": 2030174208
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003888665997993982,
      "loss": 1.3132,
      "theoretical_loss": 3.4545754438126766,
      "tokens_seen": 2030239744
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003888465396188566,
      "loss": 1.3409,
      "theoretical_loss": 3.4545662500536505,
      "tokens_seen": 2030305280
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038882647943831494,
      "loss": 1.4253,
      "theoretical_loss": 3.454557056674475,
      "tokens_seen": 2030370816
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003888064192577733,
      "loss": 1.4752,
      "theoretical_loss": 3.4545478636751223,
      "tokens_seen": 2030436352
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003887863590772317,
      "loss": 1.5427,
      "theoretical_loss": 3.454538671055564,
      "tokens_seen": 2030501888
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003887662988966901,
      "loss": 1.4424,
      "theoretical_loss": 3.4545294788157728,
      "tokens_seen": 2030567424
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038874623871614845,
      "loss": 1.4211,
      "theoretical_loss": 3.4545202869557197,
      "tokens_seen": 2030632960
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003887261785356068,
      "loss": 1.5005,
      "theoretical_loss": 3.4545110954753775,
      "tokens_seen": 2030698496
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038870611835506517,
      "loss": 1.3371,
      "theoretical_loss": 3.4545019043747183,
      "tokens_seen": 2030764032
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003886860581745236,
      "loss": 1.4411,
      "theoretical_loss": 3.454492713653714,
      "tokens_seen": 2030829568
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038866599799398195,
      "loss": 1.4754,
      "theoretical_loss": 3.4544835233123363,
      "tokens_seen": 2030895104
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003886459378134403,
      "loss": 1.423,
      "theoretical_loss": 3.454474333350558,
      "tokens_seen": 2030960640
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003886258776328987,
      "loss": 1.3863,
      "theoretical_loss": 3.4544651437683505,
      "tokens_seen": 2031026176
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003886058174523571,
      "loss": 1.4623,
      "theoretical_loss": 3.4544559545656863,
      "tokens_seen": 2031091712
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038858575727181545,
      "loss": 1.378,
      "theoretical_loss": 3.4544467657425377,
      "tokens_seen": 2031157248
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003885656970912738,
      "loss": 1.3581,
      "theoretical_loss": 3.454437577298876,
      "tokens_seen": 2031222784
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003885456369107322,
      "loss": 1.3892,
      "theoretical_loss": 3.454428389234674,
      "tokens_seen": 2031288320
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003885255767301906,
      "loss": 1.3629,
      "theoretical_loss": 3.4544192015499027,
      "tokens_seen": 2031353856
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038850551654964896,
      "loss": 1.3068,
      "theoretical_loss": 3.4544100142445355,
      "tokens_seen": 2031419392
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003884854563691073,
      "loss": 1.369,
      "theoretical_loss": 3.4544008273185445,
      "tokens_seen": 2031484928
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003884653961885657,
      "loss": 1.4741,
      "theoretical_loss": 3.4543916407719006,
      "tokens_seen": 2031550464
    },
    {
      "epoch": 0.62,
      "objective/train/advantage_avg": 0.49314558506011963,
      "objective/train/docs_used": 741020,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3117213249206543,
      "objective/train/original_loss": 1.3117215633392334,
      "objective/train/theoretical_loss": 3.454382454604577,
      "objective/train/tokens_used": 2052076000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2441154420375824,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05055570602417,
      "objective/train/weighted_lm_loss": 1.377959966659546,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.454382454604577,
      "tokens_seen": 2031616000
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038844533600802404,
      "loss": 1.3895,
      "theoretical_loss": 3.454382454604577,
      "tokens_seen": 2031616000
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038842527582748246,
      "loss": 1.3591,
      "theoretical_loss": 3.4543732688165454,
      "tokens_seen": 2031681536
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003884052156469408,
      "loss": 1.386,
      "theoretical_loss": 3.4543640834077776,
      "tokens_seen": 2031747072
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003883851554663992,
      "loss": 1.4706,
      "theoretical_loss": 3.454354898378246,
      "tokens_seen": 2031812608
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038836509528585755,
      "loss": 1.3048,
      "theoretical_loss": 3.454345713727923,
      "tokens_seen": 2031878144
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038834503510531596,
      "loss": 1.4027,
      "theoretical_loss": 3.4543365294567803,
      "tokens_seen": 2031943680
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003883249749247743,
      "loss": 1.4545,
      "theoretical_loss": 3.45432734556479,
      "tokens_seen": 2032009216
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003883049147442327,
      "loss": 1.3482,
      "theoretical_loss": 3.454318162051924,
      "tokens_seen": 2032074752
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038828485456369105,
      "loss": 1.4629,
      "theoretical_loss": 3.4543089789181556,
      "tokens_seen": 2032140288
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003882647943831494,
      "loss": 1.4151,
      "theoretical_loss": 3.4542997961634554,
      "tokens_seen": 2032205824
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038824473420260783,
      "loss": 1.2952,
      "theoretical_loss": 3.454290613787797,
      "tokens_seen": 2032271360
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003882246740220662,
      "loss": 1.3451,
      "theoretical_loss": 3.454281431791151,
      "tokens_seen": 2032336896
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038820461384152455,
      "loss": 1.4328,
      "theoretical_loss": 3.4542722501734904,
      "tokens_seen": 2032402432
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003881845536609829,
      "loss": 1.369,
      "theoretical_loss": 3.4542630689347877,
      "tokens_seen": 2032467968
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038816449348044133,
      "loss": 1.4932,
      "theoretical_loss": 3.454253888075014,
      "tokens_seen": 2032533504
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003881444332998997,
      "loss": 1.3803,
      "theoretical_loss": 3.4542447075941425,
      "tokens_seen": 2032599040
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003881243731193581,
      "loss": 1.4204,
      "theoretical_loss": 3.454235527492145,
      "tokens_seen": 2032664576
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038810431293881647,
      "loss": 1.2618,
      "theoretical_loss": 3.454226347768993,
      "tokens_seen": 2032730112
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038808425275827483,
      "loss": 1.318,
      "theoretical_loss": 3.4542171684246594,
      "tokens_seen": 2032795648
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038806419257773325,
      "loss": 1.3819,
      "theoretical_loss": 3.4542079894591162,
      "tokens_seen": 2032861184
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003880441323971916,
      "loss": 1.3811,
      "theoretical_loss": 3.4541988108723354,
      "tokens_seen": 2032926720
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038802407221665,
      "loss": 1.4105,
      "theoretical_loss": 3.4541896326642894,
      "tokens_seen": 2032992256
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038800401203610834,
      "loss": 1.5082,
      "theoretical_loss": 3.45418045483495,
      "tokens_seen": 2033057792
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038798395185556675,
      "loss": 1.4157,
      "theoretical_loss": 3.4541712773842894,
      "tokens_seen": 2033123328
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003879638916750251,
      "loss": 1.456,
      "theoretical_loss": 3.4541621003122804,
      "tokens_seen": 2033188864
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003879438314944835,
      "loss": 1.3554,
      "theoretical_loss": 3.454152923618895,
      "tokens_seen": 2033254400
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038792377131394184,
      "loss": 1.5789,
      "theoretical_loss": 3.454143747304104,
      "tokens_seen": 2033319936
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003879037111334002,
      "loss": 1.5242,
      "theoretical_loss": 3.4541345713678817,
      "tokens_seen": 2033385472
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003878836509528586,
      "loss": 1.4226,
      "theoretical_loss": 3.4541253958101983,
      "tokens_seen": 2033451008
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000387863590772317,
      "loss": 1.3631,
      "theoretical_loss": 3.4541162206310276,
      "tokens_seen": 2033516544
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038784353059177534,
      "loss": 1.4693,
      "theoretical_loss": 3.454107045830341,
      "tokens_seen": 2033582080
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003878234704112337,
      "loss": 1.364,
      "theoretical_loss": 3.454097871408111,
      "tokens_seen": 2033647616
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003878034102306921,
      "loss": 1.4411,
      "theoretical_loss": 3.454088697364309,
      "tokens_seen": 2033713152
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003877833500501505,
      "loss": 1.3548,
      "theoretical_loss": 3.4540795236989084,
      "tokens_seen": 2033778688
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038776328986960885,
      "loss": 1.3357,
      "theoretical_loss": 3.4540703504118806,
      "tokens_seen": 2033844224
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003877432296890672,
      "loss": 1.3334,
      "theoretical_loss": 3.454061177503198,
      "tokens_seen": 2033909760
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038772316950852557,
      "loss": 1.3798,
      "theoretical_loss": 3.454052004972833,
      "tokens_seen": 2033975296
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000387703109327984,
      "loss": 1.3793,
      "theoretical_loss": 3.4540428328207575,
      "tokens_seen": 2034040832
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038768304914744235,
      "loss": 1.4991,
      "theoretical_loss": 3.454033661046944,
      "tokens_seen": 2034106368
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003876629889669007,
      "loss": 1.3612,
      "theoretical_loss": 3.4540244896513643,
      "tokens_seen": 2034171904
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003876429287863591,
      "loss": 1.4133,
      "theoretical_loss": 3.4540153186339912,
      "tokens_seen": 2034237440
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003876228686058175,
      "loss": 1.4453,
      "theoretical_loss": 3.4540061479947965,
      "tokens_seen": 2034302976
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038760280842527585,
      "loss": 1.4764,
      "theoretical_loss": 3.453996977733752,
      "tokens_seen": 2034368512
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003875827482447342,
      "loss": 1.3459,
      "theoretical_loss": 3.453987807850831,
      "tokens_seen": 2034434048
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003875626880641926,
      "loss": 1.3967,
      "theoretical_loss": 3.4539786383460047,
      "tokens_seen": 2034499584
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000387542627883651,
      "loss": 1.363,
      "theoretical_loss": 3.4539694692192464,
      "tokens_seen": 2034565120
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038752256770310936,
      "loss": 1.5082,
      "theoretical_loss": 3.453960300470527,
      "tokens_seen": 2034630656
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003875025075225677,
      "loss": 1.3285,
      "theoretical_loss": 3.45395113209982,
      "tokens_seen": 2034696192
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003874824473420261,
      "loss": 1.3366,
      "theoretical_loss": 3.4539419641070968,
      "tokens_seen": 2034761728
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038746238716148444,
      "loss": 1.3829,
      "theoretical_loss": 3.45393279649233,
      "tokens_seen": 2034827264
    },
    {
      "epoch": 0.62,
      "objective/train/advantage_avg": 0.4909723997116089,
      "objective/train/docs_used": 742098,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4565691947937012,
      "objective/train/original_loss": 1.4565691947937012,
      "objective/train/theoretical_loss": 3.453923629255492,
      "objective/train/tokens_used": 2055352800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24155306816101074,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503250360488892,
      "objective/train/weighted_lm_loss": 1.529902696609497,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.453923629255492,
      "tokens_seen": 2034892800
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038744232698094286,
      "loss": 1.4321,
      "theoretical_loss": 3.453923629255492,
      "tokens_seen": 2034892800
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003874222668004012,
      "loss": 1.4511,
      "theoretical_loss": 3.453914462396555,
      "tokens_seen": 2034958336
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003874022066198596,
      "loss": 1.3463,
      "theoretical_loss": 3.4539052959154906,
      "tokens_seen": 2035023872
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038738214643931795,
      "loss": 1.4616,
      "theoretical_loss": 3.4538961298122715,
      "tokens_seen": 2035089408
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038736208625877636,
      "loss": 1.3264,
      "theoretical_loss": 3.4538869640868706,
      "tokens_seen": 2035154944
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003873420260782347,
      "loss": 1.5311,
      "theoretical_loss": 3.453877798739259,
      "tokens_seen": 2035220480
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003873219658976931,
      "loss": 1.3928,
      "theoretical_loss": 3.4538686337694102,
      "tokens_seen": 2035286016
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038730190571715145,
      "loss": 1.4155,
      "theoretical_loss": 3.4538594691772952,
      "tokens_seen": 2035351552
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003872818455366098,
      "loss": 1.3719,
      "theoretical_loss": 3.453850304962887,
      "tokens_seen": 2035417088
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038726178535606823,
      "loss": 1.4452,
      "theoretical_loss": 3.453841141126158,
      "tokens_seen": 2035482624
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003872417251755266,
      "loss": 1.4728,
      "theoretical_loss": 3.45383197766708,
      "tokens_seen": 2035548160
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038722166499498495,
      "loss": 1.3411,
      "theoretical_loss": 3.453822814585626,
      "tokens_seen": 2035613696
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003872016048144433,
      "loss": 1.3017,
      "theoretical_loss": 3.453813651881767,
      "tokens_seen": 2035679232
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038718154463390173,
      "loss": 1.3111,
      "theoretical_loss": 3.4538044895554765,
      "tokens_seen": 2035744768
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003871614844533601,
      "loss": 1.3944,
      "theoretical_loss": 3.453795327606726,
      "tokens_seen": 2035810304
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038714142427281846,
      "loss": 1.4513,
      "theoretical_loss": 3.4537861660354885,
      "tokens_seen": 2035875840
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003871213640922768,
      "loss": 1.4028,
      "theoretical_loss": 3.453777004841736,
      "tokens_seen": 2035941376
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003871013039117352,
      "loss": 1.3998,
      "theoretical_loss": 3.4537678440254407,
      "tokens_seen": 2036006912
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003870812437311936,
      "loss": 1.4485,
      "theoretical_loss": 3.4537586835865746,
      "tokens_seen": 2036072448
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038706118355065196,
      "loss": 1.3884,
      "theoretical_loss": 3.4537495235251106,
      "tokens_seen": 2036137984
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003870411233701103,
      "loss": 1.4465,
      "theoretical_loss": 3.4537403638410207,
      "tokens_seen": 2036203520
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003870210631895687,
      "loss": 1.4797,
      "theoretical_loss": 3.4537312045342774,
      "tokens_seen": 2036269056
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003870010030090271,
      "loss": 1.4742,
      "theoretical_loss": 3.4537220456048523,
      "tokens_seen": 2036334592
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038698094282848546,
      "loss": 1.4435,
      "theoretical_loss": 3.4537128870527187,
      "tokens_seen": 2036400128
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003869608826479438,
      "loss": 1.4364,
      "theoretical_loss": 3.4537037288778487,
      "tokens_seen": 2036465664
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003869408224674022,
      "loss": 1.4159,
      "theoretical_loss": 3.4536945710802143,
      "tokens_seen": 2036531200
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038692076228686055,
      "loss": 1.3806,
      "theoretical_loss": 3.453685413659788,
      "tokens_seen": 2036596736
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038690070210631896,
      "loss": 1.3204,
      "theoretical_loss": 3.4536762566165415,
      "tokens_seen": 2036662272
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038688064192577733,
      "loss": 1.4151,
      "theoretical_loss": 3.453667099950448,
      "tokens_seen": 2036727808
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003868605817452357,
      "loss": 1.4466,
      "theoretical_loss": 3.4536579436614794,
      "tokens_seen": 2036793344
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038684052156469405,
      "loss": 1.4797,
      "theoretical_loss": 3.4536487877496085,
      "tokens_seen": 2036858880
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038682046138415247,
      "loss": 1.4302,
      "theoretical_loss": 3.453639632214807,
      "tokens_seen": 2036924416
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038680040120361083,
      "loss": 1.394,
      "theoretical_loss": 3.4536304770570476,
      "tokens_seen": 2036989952
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003867803410230692,
      "loss": 1.4648,
      "theoretical_loss": 3.4536213222763026,
      "tokens_seen": 2037055488
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038676028084252755,
      "loss": 1.4622,
      "theoretical_loss": 3.4536121678725444,
      "tokens_seen": 2037121024
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003867402206619859,
      "loss": 1.4003,
      "theoretical_loss": 3.453603013845745,
      "tokens_seen": 2037186560
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038672016048144433,
      "loss": 1.3301,
      "theoretical_loss": 3.4535938601958773,
      "tokens_seen": 2037252096
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003867001003009027,
      "loss": 1.425,
      "theoretical_loss": 3.4535847069229133,
      "tokens_seen": 2037317632
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038668004012036106,
      "loss": 1.4579,
      "theoretical_loss": 3.453575554026825,
      "tokens_seen": 2037383168
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003866599799398194,
      "loss": 1.381,
      "theoretical_loss": 3.4535664015075858,
      "tokens_seen": 2037448704
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038663991975927784,
      "loss": 1.3406,
      "theoretical_loss": 3.4535572493651676,
      "tokens_seen": 2037514240
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038661985957873625,
      "loss": 1.3786,
      "theoretical_loss": 3.4535480975995423,
      "tokens_seen": 2037579776
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003865997993981946,
      "loss": 1.3073,
      "theoretical_loss": 3.4535389462106822,
      "tokens_seen": 2037645312
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000386579739217653,
      "loss": 1.5065,
      "theoretical_loss": 3.453529795198561,
      "tokens_seen": 2037710848
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003865596790371114,
      "loss": 1.4321,
      "theoretical_loss": 3.453520644563149,
      "tokens_seen": 2037776384
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038653961885656976,
      "loss": 1.3953,
      "theoretical_loss": 3.4535114943044203,
      "tokens_seen": 2037841920
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003865195586760281,
      "loss": 1.3495,
      "theoretical_loss": 3.4535023444223465,
      "tokens_seen": 2037907456
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003864994984954865,
      "loss": 1.4472,
      "theoretical_loss": 3.4534931949169003,
      "tokens_seen": 2037972992
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038647943831494484,
      "loss": 1.3473,
      "theoretical_loss": 3.4534840457880542,
      "tokens_seen": 2038038528
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038645937813440326,
      "loss": 1.4314,
      "theoretical_loss": 3.45347489703578,
      "tokens_seen": 2038104064
    },
    {
      "epoch": 0.62,
      "objective/train/advantage_avg": 0.49238401651382446,
      "objective/train/docs_used": 743253,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2639909982681274,
      "objective/train/original_loss": 1.263991117477417,
      "objective/train/theoretical_loss": 3.453465748660051,
      "objective/train/tokens_used": 2058629600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24272356927394867,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050472378730774,
      "objective/train/weighted_lm_loss": 1.3277925252914429,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.453465748660051,
      "tokens_seen": 2038169600
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003864393179538616,
      "loss": 1.3859,
      "theoretical_loss": 3.453465748660051,
      "tokens_seen": 2038169600
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038641925777332,
      "loss": 1.4398,
      "theoretical_loss": 3.4534566006608385,
      "tokens_seen": 2038235136
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038639919759277835,
      "loss": 1.3826,
      "theoretical_loss": 3.453447453038115,
      "tokens_seen": 2038300672
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038637913741223676,
      "loss": 1.3882,
      "theoretical_loss": 3.4534383057918543,
      "tokens_seen": 2038366208
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003863590772316951,
      "loss": 1.3566,
      "theoretical_loss": 3.4534291589220274,
      "tokens_seen": 2038431744
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003863390170511535,
      "loss": 1.3551,
      "theoretical_loss": 3.453420012428607,
      "tokens_seen": 2038497280
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038631895687061185,
      "loss": 1.3672,
      "theoretical_loss": 3.453410866311566,
      "tokens_seen": 2038562816
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003862988966900702,
      "loss": 1.5018,
      "theoretical_loss": 3.4534017205708762,
      "tokens_seen": 2038628352
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038627883650952863,
      "loss": 1.4636,
      "theoretical_loss": 3.4533925752065104,
      "tokens_seen": 2038693888
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000386258776328987,
      "loss": 1.4406,
      "theoretical_loss": 3.453383430218441,
      "tokens_seen": 2038759424
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038623871614844535,
      "loss": 1.4919,
      "theoretical_loss": 3.4533742856066403,
      "tokens_seen": 2038824960
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003862186559679037,
      "loss": 1.2937,
      "theoretical_loss": 3.453365141371081,
      "tokens_seen": 2038890496
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038619859578736213,
      "loss": 1.4432,
      "theoretical_loss": 3.4533559975117347,
      "tokens_seen": 2038956032
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003861785356068205,
      "loss": 1.4811,
      "theoretical_loss": 3.4533468540285748,
      "tokens_seen": 2039021568
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038615847542627886,
      "loss": 1.4225,
      "theoretical_loss": 3.453337710921573,
      "tokens_seen": 2039087104
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003861384152457372,
      "loss": 1.4571,
      "theoretical_loss": 3.4533285681907024,
      "tokens_seen": 2039152640
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003861183550651956,
      "loss": 1.4179,
      "theoretical_loss": 3.453319425835935,
      "tokens_seen": 2039218176
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000386098294884654,
      "loss": 1.379,
      "theoretical_loss": 3.4533102838572436,
      "tokens_seen": 2039283712
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038607823470411236,
      "loss": 1.3441,
      "theoretical_loss": 3.4533011422546,
      "tokens_seen": 2039349248
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003860581745235707,
      "loss": 1.4024,
      "theoretical_loss": 3.453292001027977,
      "tokens_seen": 2039414784
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003860381143430291,
      "loss": 1.3935,
      "theoretical_loss": 3.4532828601773478,
      "tokens_seen": 2039480320
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003860180541624875,
      "loss": 1.3696,
      "theoretical_loss": 3.4532737197026835,
      "tokens_seen": 2039545856
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038599799398194586,
      "loss": 1.3072,
      "theoretical_loss": 3.4532645796039576,
      "tokens_seen": 2039611392
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003859779338014042,
      "loss": 1.3653,
      "theoretical_loss": 3.4532554398811417,
      "tokens_seen": 2039676928
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003859578736208626,
      "loss": 1.4596,
      "theoretical_loss": 3.453246300534209,
      "tokens_seen": 2039742464
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038593781344032095,
      "loss": 1.3965,
      "theoretical_loss": 3.4532371615631323,
      "tokens_seen": 2039808000
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038591775325977937,
      "loss": 1.5146,
      "theoretical_loss": 3.453228022967883,
      "tokens_seen": 2039873536
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038589769307923773,
      "loss": 1.4257,
      "theoretical_loss": 3.4532188847484337,
      "tokens_seen": 2039939072
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003858776328986961,
      "loss": 1.443,
      "theoretical_loss": 3.4532097469047573,
      "tokens_seen": 2040004608
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038585757271815445,
      "loss": 1.3898,
      "theoretical_loss": 3.4532006094368266,
      "tokens_seen": 2040070144
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038583751253761287,
      "loss": 1.5132,
      "theoretical_loss": 3.4531914723446135,
      "tokens_seen": 2040135680
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038581745235707123,
      "loss": 1.399,
      "theoretical_loss": 3.4531823356280906,
      "tokens_seen": 2040201216
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003857973921765296,
      "loss": 1.3763,
      "theoretical_loss": 3.4531731992872303,
      "tokens_seen": 2040266752
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038577733199598796,
      "loss": 1.4848,
      "theoretical_loss": 3.4531640633220055,
      "tokens_seen": 2040332288
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003857572718154463,
      "loss": 1.3176,
      "theoretical_loss": 3.4531549277323883,
      "tokens_seen": 2040397824
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038573721163490473,
      "loss": 1.3506,
      "theoretical_loss": 3.4531457925183515,
      "tokens_seen": 2040463360
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003857171514543631,
      "loss": 1.3689,
      "theoretical_loss": 3.4531366576798668,
      "tokens_seen": 2040528896
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038569709127382146,
      "loss": 1.4737,
      "theoretical_loss": 3.453127523216908,
      "tokens_seen": 2040594432
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003856770310932798,
      "loss": 1.4973,
      "theoretical_loss": 3.4531183891294464,
      "tokens_seen": 2040659968
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038565697091273824,
      "loss": 1.4173,
      "theoretical_loss": 3.4531092554174556,
      "tokens_seen": 2040725504
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003856369107321966,
      "loss": 1.4025,
      "theoretical_loss": 3.4531001220809068,
      "tokens_seen": 2040791040
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038561685055165496,
      "loss": 1.4419,
      "theoretical_loss": 3.4530909891197736,
      "tokens_seen": 2040856576
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003855967903711133,
      "loss": 1.383,
      "theoretical_loss": 3.453081856534028,
      "tokens_seen": 2040922112
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003855767301905717,
      "loss": 1.4427,
      "theoretical_loss": 3.453072724323643,
      "tokens_seen": 2040987648
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003855566700100301,
      "loss": 1.3961,
      "theoretical_loss": 3.453063592488591,
      "tokens_seen": 2041053184
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038553660982948846,
      "loss": 1.3178,
      "theoretical_loss": 3.4530544610288434,
      "tokens_seen": 2041118720
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003855165496489468,
      "loss": 1.3523,
      "theoretical_loss": 3.4530453299443744,
      "tokens_seen": 2041184256
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003854964894684052,
      "loss": 1.4668,
      "theoretical_loss": 3.4530361992351555,
      "tokens_seen": 2041249792
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003854764292878636,
      "loss": 1.366,
      "theoretical_loss": 3.4530270689011595,
      "tokens_seen": 2041315328
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038545636910732197,
      "loss": 1.4262,
      "theoretical_loss": 3.453017938942359,
      "tokens_seen": 2041380864
    },
    {
      "epoch": 0.62,
      "objective/train/advantage_avg": 0.49322783946990967,
      "objective/train/docs_used": 744490,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4896498918533325,
      "objective/train/original_loss": 1.489649772644043,
      "objective/train/theoretical_loss": 3.4530088093587263,
      "objective/train/tokens_used": 2061906400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24370494484901428,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505616664886475,
      "objective/train/weighted_lm_loss": 1.5648424625396729,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4530088093587263,
      "tokens_seen": 2041446400
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038543630892678033,
      "loss": 1.4938,
      "theoretical_loss": 3.4530088093587263,
      "tokens_seen": 2041446400
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003854162487462387,
      "loss": 1.4199,
      "theoretical_loss": 3.4529996801502345,
      "tokens_seen": 2041511936
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038539618856569705,
      "loss": 1.4045,
      "theoretical_loss": 3.4529905513168555,
      "tokens_seen": 2041577472
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038537612838515547,
      "loss": 1.3553,
      "theoretical_loss": 3.452981422858562,
      "tokens_seen": 2041643008
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038535606820461383,
      "loss": 1.4725,
      "theoretical_loss": 3.4529722947753267,
      "tokens_seen": 2041708544
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003853360080240722,
      "loss": 1.4817,
      "theoretical_loss": 3.4529631670671224,
      "tokens_seen": 2041774080
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038531594784353056,
      "loss": 1.3042,
      "theoretical_loss": 3.452954039733921,
      "tokens_seen": 2041839616
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000385295887662989,
      "loss": 1.4833,
      "theoretical_loss": 3.4529449127756955,
      "tokens_seen": 2041905152
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038527582748244734,
      "loss": 1.284,
      "theoretical_loss": 3.452935786192419,
      "tokens_seen": 2041970688
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003852557673019057,
      "loss": 1.3607,
      "theoretical_loss": 3.4529266599840627,
      "tokens_seen": 2042036224
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038523570712136406,
      "loss": 1.2914,
      "theoretical_loss": 3.4529175341505995,
      "tokens_seen": 2042101760
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003852156469408225,
      "loss": 1.3737,
      "theoretical_loss": 3.452908408692003,
      "tokens_seen": 2042167296
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038519558676028084,
      "loss": 1.4433,
      "theoretical_loss": 3.4528992836082453,
      "tokens_seen": 2042232832
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003851755265797392,
      "loss": 1.4704,
      "theoretical_loss": 3.4528901588992986,
      "tokens_seen": 2042298368
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038515546639919756,
      "loss": 1.3209,
      "theoretical_loss": 3.4528810345651357,
      "tokens_seen": 2042363904
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003851354062186559,
      "loss": 1.4271,
      "theoretical_loss": 3.4528719106057295,
      "tokens_seen": 2042429440
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003851153460381144,
      "loss": 1.398,
      "theoretical_loss": 3.4528627870210515,
      "tokens_seen": 2042494976
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038509528585757276,
      "loss": 1.3841,
      "theoretical_loss": 3.4528536638110756,
      "tokens_seen": 2042560512
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003850752256770311,
      "loss": 1.3935,
      "theoretical_loss": 3.4528445409757738,
      "tokens_seen": 2042626048
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003850551654964895,
      "loss": 1.4727,
      "theoretical_loss": 3.452835418515119,
      "tokens_seen": 2042691584
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003850351053159479,
      "loss": 1.3427,
      "theoretical_loss": 3.452826296429083,
      "tokens_seen": 2042757120
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038501504513540626,
      "loss": 1.4449,
      "theoretical_loss": 3.452817174717639,
      "tokens_seen": 2042822656
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003849949849548646,
      "loss": 1.484,
      "theoretical_loss": 3.45280805338076,
      "tokens_seen": 2042888192
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000384974924774323,
      "loss": 1.3558,
      "theoretical_loss": 3.4527989324184176,
      "tokens_seen": 2042953728
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038495486459378135,
      "loss": 1.4029,
      "theoretical_loss": 3.452789811830585,
      "tokens_seen": 2043019264
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038493480441323977,
      "loss": 1.4032,
      "theoretical_loss": 3.452780691617235,
      "tokens_seen": 2043084800
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038491474423269813,
      "loss": 1.2981,
      "theoretical_loss": 3.45277157177834,
      "tokens_seen": 2043150336
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003848946840521565,
      "loss": 1.3239,
      "theoretical_loss": 3.4527624523138725,
      "tokens_seen": 2043215872
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038487462387161485,
      "loss": 1.4238,
      "theoretical_loss": 3.452753333223805,
      "tokens_seen": 2043281408
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038485456369107327,
      "loss": 1.3454,
      "theoretical_loss": 3.452744214508111,
      "tokens_seen": 2043346944
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038483450351053163,
      "loss": 1.413,
      "theoretical_loss": 3.4527350961667613,
      "tokens_seen": 2043412480
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038481444332999,
      "loss": 1.4847,
      "theoretical_loss": 3.45272597819973,
      "tokens_seen": 2043478016
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038479438314944836,
      "loss": 1.3184,
      "theoretical_loss": 3.45271686060699,
      "tokens_seen": 2043543552
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003847743229689067,
      "loss": 1.3839,
      "theoretical_loss": 3.4527077433885127,
      "tokens_seen": 2043609088
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038475426278836513,
      "loss": 1.3636,
      "theoretical_loss": 3.452698626544272,
      "tokens_seen": 2043674624
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003847342026078235,
      "loss": 1.3462,
      "theoretical_loss": 3.4526895100742396,
      "tokens_seen": 2043740160
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038471414242728186,
      "loss": 1.3983,
      "theoretical_loss": 3.4526803939783886,
      "tokens_seen": 2043805696
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003846940822467402,
      "loss": 1.5237,
      "theoretical_loss": 3.452671278256691,
      "tokens_seen": 2043871232
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038467402206619864,
      "loss": 1.3251,
      "theoretical_loss": 3.45266216290912,
      "tokens_seen": 2043936768
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000384653961885657,
      "loss": 1.4966,
      "theoretical_loss": 3.452653047935649,
      "tokens_seen": 2044002304
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038463390170511536,
      "loss": 1.4379,
      "theoretical_loss": 3.4526439333362493,
      "tokens_seen": 2044067840
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003846138415245737,
      "loss": 1.345,
      "theoretical_loss": 3.4526348191108935,
      "tokens_seen": 2044133376
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003845937813440321,
      "loss": 1.3885,
      "theoretical_loss": 3.452625705259556,
      "tokens_seen": 2044198912
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003845737211634905,
      "loss": 1.3813,
      "theoretical_loss": 3.452616591782207,
      "tokens_seen": 2044264448
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038455366098294886,
      "loss": 1.4537,
      "theoretical_loss": 3.4526074786788215,
      "tokens_seen": 2044329984
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038453360080240723,
      "loss": 1.388,
      "theoretical_loss": 3.4525983659493713,
      "tokens_seen": 2044395520
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003845135406218656,
      "loss": 1.4691,
      "theoretical_loss": 3.4525892535938283,
      "tokens_seen": 2044461056
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000384493480441324,
      "loss": 1.4093,
      "theoretical_loss": 3.4525801416121658,
      "tokens_seen": 2044526592
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038447342026078237,
      "loss": 1.4317,
      "theoretical_loss": 3.452571030004357,
      "tokens_seen": 2044592128
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038445336008024073,
      "loss": 1.4787,
      "theoretical_loss": 3.4525619187703738,
      "tokens_seen": 2044657664
    },
    {
      "epoch": 0.62,
      "objective/train/advantage_avg": 0.49355021119117737,
      "objective/train/docs_used": 745676,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.118083119392395,
      "objective/train/original_loss": 1.118083119392395,
      "objective/train/theoretical_loss": 3.4525528079101884,
      "objective/train/tokens_used": 2065183200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24401915073394775,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505955219268799,
      "objective/train/weighted_lm_loss": 1.174604058265686,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4525528079101884,
      "tokens_seen": 2044723200
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003844332998996991,
      "loss": 1.3341,
      "theoretical_loss": 3.4525528079101884,
      "tokens_seen": 2044723200
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038441323971915745,
      "loss": 1.4051,
      "theoretical_loss": 3.452543697423775,
      "tokens_seen": 2044788736
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038439317953861587,
      "loss": 1.3825,
      "theoretical_loss": 3.4525345873111055,
      "tokens_seen": 2044854272
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038437311935807423,
      "loss": 1.4596,
      "theoretical_loss": 3.452525477572152,
      "tokens_seen": 2044919808
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003843530591775326,
      "loss": 1.4012,
      "theoretical_loss": 3.4525163682068882,
      "tokens_seen": 2044985344
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038433299899699096,
      "loss": 1.5012,
      "theoretical_loss": 3.452507259215287,
      "tokens_seen": 2045050880
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003843129388164494,
      "loss": 1.4944,
      "theoretical_loss": 3.45249815059732,
      "tokens_seen": 2045116416
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038429287863590774,
      "loss": 1.4093,
      "theoretical_loss": 3.4524890423529597,
      "tokens_seen": 2045181952
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003842728184553661,
      "loss": 1.4416,
      "theoretical_loss": 3.45247993448218,
      "tokens_seen": 2045247488
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038425275827482446,
      "loss": 1.3506,
      "theoretical_loss": 3.4524708269849533,
      "tokens_seen": 2045313024
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003842326980942829,
      "loss": 1.3836,
      "theoretical_loss": 3.452461719861252,
      "tokens_seen": 2045378560
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038421263791374124,
      "loss": 1.4993,
      "theoretical_loss": 3.4524526131110487,
      "tokens_seen": 2045444096
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003841925777331996,
      "loss": 1.3117,
      "theoretical_loss": 3.4524435067343164,
      "tokens_seen": 2045509632
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038417251755265796,
      "loss": 1.3615,
      "theoretical_loss": 3.452434400731028,
      "tokens_seen": 2045575168
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003841524573721163,
      "loss": 1.3199,
      "theoretical_loss": 3.452425295101156,
      "tokens_seen": 2045640704
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038413239719157474,
      "loss": 1.3473,
      "theoretical_loss": 3.452416189844673,
      "tokens_seen": 2045706240
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003841123370110331,
      "loss": 1.2864,
      "theoretical_loss": 3.452407084961551,
      "tokens_seen": 2045771776
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038409227683049147,
      "loss": 1.3105,
      "theoretical_loss": 3.4523979804517646,
      "tokens_seen": 2045837312
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038407221664994983,
      "loss": 1.3339,
      "theoretical_loss": 3.4523888763152852,
      "tokens_seen": 2045902848
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038405215646940825,
      "loss": 1.3375,
      "theoretical_loss": 3.4523797725520855,
      "tokens_seen": 2045968384
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003840320962888666,
      "loss": 1.3422,
      "theoretical_loss": 3.452370669162139,
      "tokens_seen": 2046033920
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038401203610832497,
      "loss": 1.4033,
      "theoretical_loss": 3.452361566145418,
      "tokens_seen": 2046099456
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038399197592778333,
      "loss": 1.3984,
      "theoretical_loss": 3.452352463501895,
      "tokens_seen": 2046164992
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003839719157472417,
      "loss": 1.3483,
      "theoretical_loss": 3.452343361231543,
      "tokens_seen": 2046230528
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003839518555667001,
      "loss": 1.4514,
      "theoretical_loss": 3.452334259334335,
      "tokens_seen": 2046296064
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003839317953861585,
      "loss": 1.3934,
      "theoretical_loss": 3.452325157810243,
      "tokens_seen": 2046361600
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038391173520561684,
      "loss": 1.3741,
      "theoretical_loss": 3.4523160566592406,
      "tokens_seen": 2046427136
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003838916750250752,
      "loss": 1.4304,
      "theoretical_loss": 3.4523069558813004,
      "tokens_seen": 2046492672
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003838716148445336,
      "loss": 1.4099,
      "theoretical_loss": 3.452297855476395,
      "tokens_seen": 2046558208
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000383851554663992,
      "loss": 1.4467,
      "theoretical_loss": 3.4522887554444965,
      "tokens_seen": 2046623744
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038383149448345034,
      "loss": 1.3483,
      "theoretical_loss": 3.4522796557855786,
      "tokens_seen": 2046689280
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003838114343029087,
      "loss": 1.4041,
      "theoretical_loss": 3.4522705564996143,
      "tokens_seen": 2046754816
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038379137412236706,
      "loss": 1.3461,
      "theoretical_loss": 3.4522614575865753,
      "tokens_seen": 2046820352
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003837713139418255,
      "loss": 1.3563,
      "theoretical_loss": 3.4522523590464353,
      "tokens_seen": 2046885888
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038375125376128384,
      "loss": 1.4171,
      "theoretical_loss": 3.4522432608791664,
      "tokens_seen": 2046951424
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003837311935807422,
      "loss": 1.3473,
      "theoretical_loss": 3.452234163084742,
      "tokens_seen": 2047016960
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038371113340020057,
      "loss": 1.2592,
      "theoretical_loss": 3.452225065663134,
      "tokens_seen": 2047082496
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000383691073219659,
      "loss": 1.4584,
      "theoretical_loss": 3.4522159686143166,
      "tokens_seen": 2047148032
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038367101303911735,
      "loss": 1.3764,
      "theoretical_loss": 3.452206871938261,
      "tokens_seen": 2047213568
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003836509528585757,
      "loss": 1.3257,
      "theoretical_loss": 3.452197775634941,
      "tokens_seen": 2047279104
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003836308926780341,
      "loss": 1.5475,
      "theoretical_loss": 3.4521886797043293,
      "tokens_seen": 2047344640
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003836108324974925,
      "loss": 1.4057,
      "theoretical_loss": 3.4521795841463985,
      "tokens_seen": 2047410176
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003835907723169509,
      "loss": 1.4119,
      "theoretical_loss": 3.4521704889611216,
      "tokens_seen": 2047475712
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038357071213640927,
      "loss": 1.3665,
      "theoretical_loss": 3.4521613941484706,
      "tokens_seen": 2047541248
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038355065195586763,
      "loss": 1.3101,
      "theoretical_loss": 3.4521522997084197,
      "tokens_seen": 2047606784
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000383530591775326,
      "loss": 1.4115,
      "theoretical_loss": 3.4521432056409402,
      "tokens_seen": 2047672320
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003835105315947844,
      "loss": 1.4034,
      "theoretical_loss": 3.4521341119460067,
      "tokens_seen": 2047737856
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038349047141424277,
      "loss": 1.435,
      "theoretical_loss": 3.4521250186235903,
      "tokens_seen": 2047803392
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038347041123370113,
      "loss": 1.4285,
      "theoretical_loss": 3.452115925673665,
      "tokens_seen": 2047868928
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003834503510531595,
      "loss": 1.3007,
      "theoretical_loss": 3.4521068330962024,
      "tokens_seen": 2047934464
    },
    {
      "epoch": 0.62,
      "objective/train/advantage_avg": 0.4959915578365326,
      "objective/train/docs_used": 746888,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2392317056655884,
      "objective/train/original_loss": 1.2392315864562988,
      "objective/train/theoretical_loss": 3.4520977408911766,
      "objective/train/tokens_used": 2068460000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2461855113506317,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508508682250977,
      "objective/train/weighted_lm_loss": 1.3023267984390259,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4520977408911766,
      "tokens_seen": 2048000000
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038343029087261786,
      "loss": 1.2578,
      "theoretical_loss": 3.4520977408911766,
      "tokens_seen": 2048000000
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038341023069207627,
      "loss": 1.3851,
      "theoretical_loss": 3.4520886490585596,
      "tokens_seen": 2048065536
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038339017051153463,
      "loss": 1.2954,
      "theoretical_loss": 3.4520795575983247,
      "tokens_seen": 2048131072
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000383370110330993,
      "loss": 1.5045,
      "theoretical_loss": 3.4520704665104445,
      "tokens_seen": 2048196608
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038335005015045136,
      "loss": 1.4297,
      "theoretical_loss": 3.4520613757948917,
      "tokens_seen": 2048262144
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003833299899699098,
      "loss": 1.5144,
      "theoretical_loss": 3.4520522854516402,
      "tokens_seen": 2048327680
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038330992978936814,
      "loss": 1.4297,
      "theoretical_loss": 3.4520431954806607,
      "tokens_seen": 2048393216
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003832898696088265,
      "loss": 1.3725,
      "theoretical_loss": 3.4520341058819284,
      "tokens_seen": 2048458752
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038326980942828486,
      "loss": 1.3607,
      "theoretical_loss": 3.4520250166554147,
      "tokens_seen": 2048524288
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003832497492477433,
      "loss": 1.509,
      "theoretical_loss": 3.4520159278010922,
      "tokens_seen": 2048589824
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038322968906720164,
      "loss": 1.3956,
      "theoretical_loss": 3.452006839318935,
      "tokens_seen": 2048655360
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038320962888666,
      "loss": 1.3935,
      "theoretical_loss": 3.451997751208915,
      "tokens_seen": 2048720896
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038318956870611836,
      "loss": 1.4034,
      "theoretical_loss": 3.451988663471006,
      "tokens_seen": 2048786432
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003831695085255767,
      "loss": 1.3556,
      "theoretical_loss": 3.4519795761051797,
      "tokens_seen": 2048851968
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038314944834503514,
      "loss": 1.5407,
      "theoretical_loss": 3.45197048911141,
      "tokens_seen": 2048917504
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003831293881644935,
      "loss": 1.4158,
      "theoretical_loss": 3.4519614024896685,
      "tokens_seen": 2048983040
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038310932798395187,
      "loss": 1.5395,
      "theoretical_loss": 3.4519523162399297,
      "tokens_seen": 2049048576
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038308926780341023,
      "loss": 1.3852,
      "theoretical_loss": 3.451943230362165,
      "tokens_seen": 2049114112
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038306920762286865,
      "loss": 1.3998,
      "theoretical_loss": 3.451934144856348,
      "tokens_seen": 2049179648
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000383049147442327,
      "loss": 1.3835,
      "theoretical_loss": 3.4519250597224516,
      "tokens_seen": 2049245184
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038302908726178537,
      "loss": 1.3012,
      "theoretical_loss": 3.4519159749604484,
      "tokens_seen": 2049310720
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038300902708124373,
      "loss": 1.4977,
      "theoretical_loss": 3.451906890570312,
      "tokens_seen": 2049376256
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003829889669007021,
      "loss": 1.4534,
      "theoretical_loss": 3.451897806552014,
      "tokens_seen": 2049441792
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003829689067201605,
      "loss": 1.4067,
      "theoretical_loss": 3.451888722905528,
      "tokens_seen": 2049507328
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003829488465396189,
      "loss": 1.4439,
      "theoretical_loss": 3.4518796396308273,
      "tokens_seen": 2049572864
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038292878635907724,
      "loss": 1.4957,
      "theoretical_loss": 3.4518705567278847,
      "tokens_seen": 2049638400
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003829087261785356,
      "loss": 1.3906,
      "theoretical_loss": 3.451861474196672,
      "tokens_seen": 2049703936
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000382888665997994,
      "loss": 1.4226,
      "theoretical_loss": 3.4518523920371633,
      "tokens_seen": 2049769472
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003828686058174524,
      "loss": 1.4161,
      "theoretical_loss": 3.4518433102493313,
      "tokens_seen": 2049835008
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038284854563691074,
      "loss": 1.4372,
      "theoretical_loss": 3.4518342288331487,
      "tokens_seen": 2049900544
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003828284854563691,
      "loss": 1.4194,
      "theoretical_loss": 3.4518251477885884,
      "tokens_seen": 2049966080
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038280842527582746,
      "loss": 1.4945,
      "theoretical_loss": 3.4518160671156233,
      "tokens_seen": 2050031616
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003827883650952859,
      "loss": 1.4736,
      "theoretical_loss": 3.4518069868142263,
      "tokens_seen": 2050097152
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038276830491474424,
      "loss": 1.4178,
      "theoretical_loss": 3.4517979068843703,
      "tokens_seen": 2050162688
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003827482447342026,
      "loss": 1.538,
      "theoretical_loss": 3.4517888273260287,
      "tokens_seen": 2050228224
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038272818455366097,
      "loss": 1.429,
      "theoretical_loss": 3.4517797481391734,
      "tokens_seen": 2050293760
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003827081243731194,
      "loss": 1.4442,
      "theoretical_loss": 3.4517706693237784,
      "tokens_seen": 2050359296
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038268806419257775,
      "loss": 1.4243,
      "theoretical_loss": 3.4517615908798156,
      "tokens_seen": 2050424832
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003826680040120361,
      "loss": 1.3219,
      "theoretical_loss": 3.4517525128072593,
      "tokens_seen": 2050490368
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038264794383149447,
      "loss": 1.4304,
      "theoretical_loss": 3.4517434351060814,
      "tokens_seen": 2050555904
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038262788365095283,
      "loss": 1.3684,
      "theoretical_loss": 3.451734357776255,
      "tokens_seen": 2050621440
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038260782347041125,
      "loss": 1.413,
      "theoretical_loss": 3.4517252808177528,
      "tokens_seen": 2050686976
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003825877632898696,
      "loss": 1.3415,
      "theoretical_loss": 3.4517162042305483,
      "tokens_seen": 2050752512
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000382567703109328,
      "loss": 1.4091,
      "theoretical_loss": 3.4517071280146148,
      "tokens_seen": 2050818048
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038254764292878634,
      "loss": 1.4164,
      "theoretical_loss": 3.451698052169924,
      "tokens_seen": 2050883584
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038252758274824475,
      "loss": 1.3998,
      "theoretical_loss": 3.45168897669645,
      "tokens_seen": 2050949120
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003825075225677031,
      "loss": 1.3394,
      "theoretical_loss": 3.4516799015941646,
      "tokens_seen": 2051014656
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003824874623871615,
      "loss": 1.4556,
      "theoretical_loss": 3.451670826863042,
      "tokens_seen": 2051080192
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038246740220661984,
      "loss": 1.3847,
      "theoretical_loss": 3.451661752503054,
      "tokens_seen": 2051145728
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003824473420260782,
      "loss": 1.3862,
      "theoretical_loss": 3.451652678514175,
      "tokens_seen": 2051211264
    },
    {
      "epoch": 0.62,
      "objective/train/advantage_avg": 0.49104222655296326,
      "objective/train/docs_used": 748156,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.254499077796936,
      "objective/train/original_loss": 1.2544989585876465,
      "objective/train/theoretical_loss": 3.451643604896377,
      "objective/train/tokens_used": 2071736800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24173857271671295,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050333023071289,
      "objective/train/weighted_lm_loss": 1.3175462484359741,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.451643604896377,
      "tokens_seen": 2051276800
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003824272818455366,
      "loss": 1.4318,
      "theoretical_loss": 3.451643604896377,
      "tokens_seen": 2051276800
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000382407221664995,
      "loss": 1.3698,
      "theoretical_loss": 3.451634531649632,
      "tokens_seen": 2051342336
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038238716148445334,
      "loss": 1.5407,
      "theoretical_loss": 3.4516254587739157,
      "tokens_seen": 2051407872
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003823671013039117,
      "loss": 1.4165,
      "theoretical_loss": 3.4516163862691984,
      "tokens_seen": 2051473408
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003823470411233701,
      "loss": 1.4635,
      "theoretical_loss": 3.4516073141354546,
      "tokens_seen": 2051538944
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003823269809428285,
      "loss": 1.5317,
      "theoretical_loss": 3.451598242372656,
      "tokens_seen": 2051604480
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038230692076228685,
      "loss": 1.4826,
      "theoretical_loss": 3.4515891709807773,
      "tokens_seen": 2051670016
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003822868605817452,
      "loss": 1.4203,
      "theoretical_loss": 3.45158009995979,
      "tokens_seen": 2051735552
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038226680040120357,
      "loss": 1.2949,
      "theoretical_loss": 3.451571029309668,
      "tokens_seen": 2051801088
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000382246740220662,
      "loss": 1.3674,
      "theoretical_loss": 3.451561959030384,
      "tokens_seen": 2051866624
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038222668004012035,
      "loss": 1.3077,
      "theoretical_loss": 3.4515528891219107,
      "tokens_seen": 2051932160
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003822066198595787,
      "loss": 1.3228,
      "theoretical_loss": 3.4515438195842214,
      "tokens_seen": 2051997696
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003821865596790371,
      "loss": 1.5465,
      "theoretical_loss": 3.4515347504172893,
      "tokens_seen": 2052063232
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003821664994984955,
      "loss": 1.4754,
      "theoretical_loss": 3.451525681621087,
      "tokens_seen": 2052128768
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038214643931795385,
      "loss": 1.3928,
      "theoretical_loss": 3.4515166131955874,
      "tokens_seen": 2052194304
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038212637913741227,
      "loss": 1.3722,
      "theoretical_loss": 3.451507545140764,
      "tokens_seen": 2052259840
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038210631895687063,
      "loss": 1.4582,
      "theoretical_loss": 3.4514984774565898,
      "tokens_seen": 2052325376
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038208625877632905,
      "loss": 1.3703,
      "theoretical_loss": 3.4514894101430373,
      "tokens_seen": 2052390912
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003820661985957874,
      "loss": 1.397,
      "theoretical_loss": 3.45148034320008,
      "tokens_seen": 2052456448
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038204613841524577,
      "loss": 1.3476,
      "theoretical_loss": 3.4514712766276907,
      "tokens_seen": 2052521984
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038202607823470413,
      "loss": 1.4827,
      "theoretical_loss": 3.4514622104258423,
      "tokens_seen": 2052587520
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003820060180541625,
      "loss": 1.3694,
      "theoretical_loss": 3.4514531445945082,
      "tokens_seen": 2052653056
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003819859578736209,
      "loss": 1.3841,
      "theoretical_loss": 3.451444079133661,
      "tokens_seen": 2052718592
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003819658976930793,
      "loss": 1.3019,
      "theoretical_loss": 3.451435014043274,
      "tokens_seen": 2052784128
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038194583751253764,
      "loss": 1.3799,
      "theoretical_loss": 3.451425949323321,
      "tokens_seen": 2052849664
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000381925777331996,
      "loss": 1.3182,
      "theoretical_loss": 3.451416884973773,
      "tokens_seen": 2052915200
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003819057171514544,
      "loss": 1.42,
      "theoretical_loss": 3.4514078209946044,
      "tokens_seen": 2052980736
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003818856569709128,
      "loss": 1.4336,
      "theoretical_loss": 3.451398757385789,
      "tokens_seen": 2053046272
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038186559679037114,
      "loss": 1.357,
      "theoretical_loss": 3.451389694147298,
      "tokens_seen": 2053111808
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003818455366098295,
      "loss": 1.3531,
      "theoretical_loss": 3.4513806312791058,
      "tokens_seen": 2053177344
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038182547642928786,
      "loss": 1.4086,
      "theoretical_loss": 3.4513715687811852,
      "tokens_seen": 2053242880
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003818054162487463,
      "loss": 1.4196,
      "theoretical_loss": 3.451362506653509,
      "tokens_seen": 2053308416
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038178535606820464,
      "loss": 1.4384,
      "theoretical_loss": 3.45135344489605,
      "tokens_seen": 2053373952
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000381765295887663,
      "loss": 1.3775,
      "theoretical_loss": 3.4513443835087823,
      "tokens_seen": 2053439488
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038174523570712137,
      "loss": 1.4339,
      "theoretical_loss": 3.451335322491678,
      "tokens_seen": 2053505024
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003817251755265798,
      "loss": 1.4261,
      "theoretical_loss": 3.4513262618447103,
      "tokens_seen": 2053570560
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038170511534603815,
      "loss": 1.4943,
      "theoretical_loss": 3.4513172015678526,
      "tokens_seen": 2053636096
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003816850551654965,
      "loss": 1.3643,
      "theoretical_loss": 3.4513081416610776,
      "tokens_seen": 2053701632
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038166499498495487,
      "loss": 1.3289,
      "theoretical_loss": 3.4512990821243585,
      "tokens_seen": 2053767168
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038164493480441323,
      "loss": 1.3323,
      "theoretical_loss": 3.451290022957669,
      "tokens_seen": 2053832704
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038162487462387165,
      "loss": 1.371,
      "theoretical_loss": 3.451280964160981,
      "tokens_seen": 2053898240
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038160481444333,
      "loss": 1.4024,
      "theoretical_loss": 3.4512719057342682,
      "tokens_seen": 2053963776
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003815847542627884,
      "loss": 1.3582,
      "theoretical_loss": 3.451262847677504,
      "tokens_seen": 2054029312
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038156469408224674,
      "loss": 1.4581,
      "theoretical_loss": 3.4512537899906612,
      "tokens_seen": 2054094848
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038154463390170515,
      "loss": 1.397,
      "theoretical_loss": 3.451244732673713,
      "tokens_seen": 2054160384
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003815245737211635,
      "loss": 1.388,
      "theoretical_loss": 3.451235675726632,
      "tokens_seen": 2054225920
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003815045135406219,
      "loss": 1.377,
      "theoretical_loss": 3.451226619149392,
      "tokens_seen": 2054291456
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038148445336008024,
      "loss": 1.3318,
      "theoretical_loss": 3.4512175629419657,
      "tokens_seen": 2054356992
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003814643931795386,
      "loss": 1.3572,
      "theoretical_loss": 3.451208507104326,
      "tokens_seen": 2054422528
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000381444332998997,
      "loss": 1.404,
      "theoretical_loss": 3.4511994516364464,
      "tokens_seen": 2054488064
    },
    {
      "epoch": 0.62,
      "objective/train/advantage_avg": 0.4941757023334503,
      "objective/train/docs_used": 749325,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.298593282699585,
      "objective/train/original_loss": 1.2985930442810059,
      "objective/train/theoretical_loss": 3.4511903965382995,
      "objective/train/tokens_used": 2075013600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24444127082824707,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050660252571106,
      "objective/train/weighted_lm_loss": 1.364437222480774,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4511903965382995,
      "tokens_seen": 2054553600
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003814242728184554,
      "loss": 1.4475,
      "theoretical_loss": 3.4511903965382995,
      "tokens_seen": 2054553600
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038140421263791374,
      "loss": 1.4023,
      "theoretical_loss": 3.4511813418098596,
      "tokens_seen": 2054619136
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003813841524573721,
      "loss": 1.3506,
      "theoretical_loss": 3.451172287451098,
      "tokens_seen": 2054684672
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003813640922768305,
      "loss": 1.5035,
      "theoretical_loss": 3.4511632334619895,
      "tokens_seen": 2054750208
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003813440320962889,
      "loss": 1.4576,
      "theoretical_loss": 3.451154179842507,
      "tokens_seen": 2054815744
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038132397191574725,
      "loss": 1.3513,
      "theoretical_loss": 3.451145126592622,
      "tokens_seen": 2054881280
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003813039117352056,
      "loss": 1.4648,
      "theoretical_loss": 3.45113607371231,
      "tokens_seen": 2054946816
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038128385155466397,
      "loss": 1.2876,
      "theoretical_loss": 3.451127021201542,
      "tokens_seen": 2055012352
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003812637913741224,
      "loss": 1.4516,
      "theoretical_loss": 3.4511179690602924,
      "tokens_seen": 2055077888
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038124373119358075,
      "loss": 1.3936,
      "theoretical_loss": 3.4511089172885336,
      "tokens_seen": 2055143424
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003812236710130391,
      "loss": 1.4382,
      "theoretical_loss": 3.4510998658862397,
      "tokens_seen": 2055208960
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003812036108324975,
      "loss": 1.3092,
      "theoretical_loss": 3.4510908148533828,
      "tokens_seen": 2055274496
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003811835506519559,
      "loss": 1.3463,
      "theoretical_loss": 3.4510817641899365,
      "tokens_seen": 2055340032
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038116349047141425,
      "loss": 1.4761,
      "theoretical_loss": 3.4510727138958748,
      "tokens_seen": 2055405568
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003811434302908726,
      "loss": 1.4344,
      "theoretical_loss": 3.451063663971169,
      "tokens_seen": 2055471104
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000381123370110331,
      "loss": 1.5058,
      "theoretical_loss": 3.4510546144157934,
      "tokens_seen": 2055536640
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038110330992978934,
      "loss": 1.3852,
      "theoretical_loss": 3.4510455652297214,
      "tokens_seen": 2055602176
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038108324974924776,
      "loss": 1.4124,
      "theoretical_loss": 3.451036516412925,
      "tokens_seen": 2055667712
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003810631895687061,
      "loss": 1.3649,
      "theoretical_loss": 3.4510274679653787,
      "tokens_seen": 2055733248
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003810431293881645,
      "loss": 1.4077,
      "theoretical_loss": 3.4510184198870553,
      "tokens_seen": 2055798784
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038102306920762284,
      "loss": 1.4858,
      "theoretical_loss": 3.451009372177927,
      "tokens_seen": 2055864320
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038100300902708126,
      "loss": 1.3222,
      "theoretical_loss": 3.451000324837968,
      "tokens_seen": 2055929856
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003809829488465396,
      "loss": 1.4663,
      "theoretical_loss": 3.4509912778671517,
      "tokens_seen": 2055995392
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000380962888665998,
      "loss": 1.3891,
      "theoretical_loss": 3.45098223126545,
      "tokens_seen": 2056060928
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038094282848545635,
      "loss": 1.4511,
      "theoretical_loss": 3.4509731850328373,
      "tokens_seen": 2056126464
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038092276830491476,
      "loss": 1.4245,
      "theoretical_loss": 3.450964139169286,
      "tokens_seen": 2056192000
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003809027081243731,
      "loss": 1.2419,
      "theoretical_loss": 3.4509550936747697,
      "tokens_seen": 2056257536
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003808826479438315,
      "loss": 1.4159,
      "theoretical_loss": 3.450946048549261,
      "tokens_seen": 2056323072
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038086258776328985,
      "loss": 1.4529,
      "theoretical_loss": 3.4509370037927343,
      "tokens_seen": 2056388608
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003808425275827482,
      "loss": 1.4286,
      "theoretical_loss": 3.4509279594051616,
      "tokens_seen": 2056454144
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003808224674022066,
      "loss": 1.4827,
      "theoretical_loss": 3.4509189153865165,
      "tokens_seen": 2056519680
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000380802407221665,
      "loss": 1.3478,
      "theoretical_loss": 3.4509098717367723,
      "tokens_seen": 2056585216
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038078234704112335,
      "loss": 1.4531,
      "theoretical_loss": 3.450900828455902,
      "tokens_seen": 2056650752
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003807622868605817,
      "loss": 1.4521,
      "theoretical_loss": 3.450891785543879,
      "tokens_seen": 2056716288
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038074222668004013,
      "loss": 1.4072,
      "theoretical_loss": 3.450882743000677,
      "tokens_seen": 2056781824
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003807221664994985,
      "loss": 1.4709,
      "theoretical_loss": 3.4508737008262678,
      "tokens_seen": 2056847360
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038070210631895685,
      "loss": 1.4678,
      "theoretical_loss": 3.450864659020626,
      "tokens_seen": 2056912896
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003806820461384152,
      "loss": 1.5135,
      "theoretical_loss": 3.450855617583724,
      "tokens_seen": 2056978432
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003806619859578736,
      "loss": 1.4199,
      "theoretical_loss": 3.450846576515535,
      "tokens_seen": 2057043968
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000380641925777332,
      "loss": 1.4468,
      "theoretical_loss": 3.4508375358160333,
      "tokens_seen": 2057109504
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003806218655967904,
      "loss": 1.4611,
      "theoretical_loss": 3.45082849548519,
      "tokens_seen": 2057175040
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003806018054162488,
      "loss": 1.3957,
      "theoretical_loss": 3.4508194555229807,
      "tokens_seen": 2057240576
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038058174523570714,
      "loss": 1.3623,
      "theoretical_loss": 3.4508104159293773,
      "tokens_seen": 2057306112
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038056168505516555,
      "loss": 1.4855,
      "theoretical_loss": 3.4508013767043533,
      "tokens_seen": 2057371648
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003805416248746239,
      "loss": 1.3962,
      "theoretical_loss": 3.450792337847882,
      "tokens_seen": 2057437184
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003805215646940823,
      "loss": 1.486,
      "theoretical_loss": 3.4507832993599363,
      "tokens_seen": 2057502720
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038050150451354064,
      "loss": 1.3877,
      "theoretical_loss": 3.4507742612404897,
      "tokens_seen": 2057568256
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000380481444332999,
      "loss": 1.3592,
      "theoretical_loss": 3.4507652234895154,
      "tokens_seen": 2057633792
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003804613841524574,
      "loss": 1.4227,
      "theoretical_loss": 3.450756186106987,
      "tokens_seen": 2057699328
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003804413239719158,
      "loss": 1.4288,
      "theoretical_loss": 3.450747149092877,
      "tokens_seen": 2057764864
    },
    {
      "epoch": 0.62,
      "objective/train/advantage_avg": 0.49408796429634094,
      "objective/train/docs_used": 750571,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4247468709945679,
      "objective/train/original_loss": 1.4247467517852783,
      "objective/train/theoretical_loss": 3.4507381124471594,
      "objective/train/tokens_used": 2078290400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2443583607673645,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506510734558105,
      "objective/train/weighted_lm_loss": 1.4968202114105225,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.4507381124471594,
      "tokens_seen": 2057830400
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038042126379137414,
      "loss": 1.4923,
      "theoretical_loss": 3.4507381124471594,
      "tokens_seen": 2057830400
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003804012036108325,
      "loss": 1.4171,
      "theoretical_loss": 3.4507290761698064,
      "tokens_seen": 2057895936
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003803811434302909,
      "loss": 1.421,
      "theoretical_loss": 3.4507200402607925,
      "tokens_seen": 2057961472
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003803610832497493,
      "loss": 1.2949,
      "theoretical_loss": 3.450711004720091,
      "tokens_seen": 2058027008
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038034102306920765,
      "loss": 1.4696,
      "theoretical_loss": 3.4507019695476737,
      "tokens_seen": 2058092544
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000380320962888666,
      "loss": 1.4443,
      "theoretical_loss": 3.450692934743515,
      "tokens_seen": 2058158080
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038030090270812437,
      "loss": 1.3748,
      "theoretical_loss": 3.450683900307588,
      "tokens_seen": 2058223616
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003802808425275828,
      "loss": 1.4306,
      "theoretical_loss": 3.450674866239866,
      "tokens_seen": 2058289152
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038026078234704115,
      "loss": 1.4389,
      "theoretical_loss": 3.450665832540322,
      "tokens_seen": 2058354688
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003802407221664995,
      "loss": 1.3926,
      "theoretical_loss": 3.450656799208929,
      "tokens_seen": 2058420224
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003802206619859579,
      "loss": 1.3735,
      "theoretical_loss": 3.4506477662456616,
      "tokens_seen": 2058485760
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003802006018054163,
      "loss": 1.4458,
      "theoretical_loss": 3.4506387336504916,
      "tokens_seen": 2058551296
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038018054162487465,
      "loss": 1.3502,
      "theoretical_loss": 3.450629701423393,
      "tokens_seen": 2058616832
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000380160481444333,
      "loss": 1.485,
      "theoretical_loss": 3.450620669564339,
      "tokens_seen": 2058682368
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003801404212637914,
      "loss": 1.4775,
      "theoretical_loss": 3.4506116380733025,
      "tokens_seen": 2058747904
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038012036108324974,
      "loss": 1.4565,
      "theoretical_loss": 3.450602606950257,
      "tokens_seen": 2058813440
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038010030090270816,
      "loss": 1.4323,
      "theoretical_loss": 3.4505935761951765,
      "tokens_seen": 2058878976
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003800802407221665,
      "loss": 1.4456,
      "theoretical_loss": 3.4505845458080335,
      "tokens_seen": 2058944512
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003800601805416249,
      "loss": 1.442,
      "theoretical_loss": 3.4505755157888016,
      "tokens_seen": 2059010048
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038004012036108324,
      "loss": 1.3406,
      "theoretical_loss": 3.4505664861374536,
      "tokens_seen": 2059075584
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038002006018054166,
      "loss": 1.4157,
      "theoretical_loss": 3.4505574568539634,
      "tokens_seen": 2059141120
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00038,
      "loss": 1.3959,
      "theoretical_loss": 3.4505484279383043,
      "tokens_seen": 2059206656
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003799799398194584,
      "loss": 1.3304,
      "theoretical_loss": 3.450539399390449,
      "tokens_seen": 2059272192
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037995987963891675,
      "loss": 1.3304,
      "theoretical_loss": 3.450530371210372,
      "tokens_seen": 2059337728
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037993981945837516,
      "loss": 1.4441,
      "theoretical_loss": 3.4505213433980453,
      "tokens_seen": 2059403264
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003799197592778335,
      "loss": 1.4655,
      "theoretical_loss": 3.4505123159534428,
      "tokens_seen": 2059468800
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003798996990972919,
      "loss": 1.4227,
      "theoretical_loss": 3.450503288876538,
      "tokens_seen": 2059534336
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037987963891675025,
      "loss": 1.3985,
      "theoretical_loss": 3.450494262167304,
      "tokens_seen": 2059599872
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003798595787362086,
      "loss": 1.5051,
      "theoretical_loss": 3.450485235825714,
      "tokens_seen": 2059665408
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037983951855566703,
      "loss": 1.2677,
      "theoretical_loss": 3.4504762098517414,
      "tokens_seen": 2059730944
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003798194583751254,
      "loss": 1.4736,
      "theoretical_loss": 3.4504671842453596,
      "tokens_seen": 2059796480
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037979939819458375,
      "loss": 1.4295,
      "theoretical_loss": 3.4504581590065424,
      "tokens_seen": 2059862016
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003797793380140421,
      "loss": 1.3979,
      "theoretical_loss": 3.450449134135262,
      "tokens_seen": 2059927552
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037975927783350053,
      "loss": 1.4086,
      "theoretical_loss": 3.4504401096314927,
      "tokens_seen": 2059993088
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003797392176529589,
      "loss": 1.3926,
      "theoretical_loss": 3.4504310854952074,
      "tokens_seen": 2060058624
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037971915747241725,
      "loss": 1.3732,
      "theoretical_loss": 3.45042206172638,
      "tokens_seen": 2060124160
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003796990972918756,
      "loss": 1.409,
      "theoretical_loss": 3.4504130383249834,
      "tokens_seen": 2060189696
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000379679037111334,
      "loss": 1.457,
      "theoretical_loss": 3.450404015290991,
      "tokens_seen": 2060255232
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003796589769307924,
      "loss": 1.5147,
      "theoretical_loss": 3.4503949926243758,
      "tokens_seen": 2060320768
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037963891675025076,
      "loss": 1.2765,
      "theoretical_loss": 3.4503859703251116,
      "tokens_seen": 2060386304
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003796188565697091,
      "loss": 1.3591,
      "theoretical_loss": 3.4503769483931723,
      "tokens_seen": 2060451840
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003795987963891675,
      "loss": 1.4302,
      "theoretical_loss": 3.45036792682853,
      "tokens_seen": 2060517376
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003795787362086259,
      "loss": 1.3153,
      "theoretical_loss": 3.450358905631159,
      "tokens_seen": 2060582912
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037955867602808426,
      "loss": 1.426,
      "theoretical_loss": 3.450349884801032,
      "tokens_seen": 2060648448
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003795386158475426,
      "loss": 1.3563,
      "theoretical_loss": 3.4503408643381235,
      "tokens_seen": 2060713984
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000379518555667001,
      "loss": 1.287,
      "theoretical_loss": 3.450331844242405,
      "tokens_seen": 2060779520
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037949849548645935,
      "loss": 1.4088,
      "theoretical_loss": 3.4503228245138517,
      "tokens_seen": 2060845056
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037947843530591776,
      "loss": 1.3073,
      "theoretical_loss": 3.4503138051524367,
      "tokens_seen": 2060910592
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003794583751253761,
      "loss": 1.4485,
      "theoretical_loss": 3.4503047861581324,
      "tokens_seen": 2060976128
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003794383149448345,
      "loss": 1.4404,
      "theoretical_loss": 3.4502957675309123,
      "tokens_seen": 2061041664
    },
    {
      "epoch": 0.62,
      "objective/train/advantage_avg": 0.4902001917362213,
      "objective/train/docs_used": 751685,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.387594223022461,
      "objective/train/original_loss": 1.387594223022461,
      "objective/train/theoretical_loss": 3.450286749270751,
      "objective/train/tokens_used": 2081567200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24150176346302032,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502477884292603,
      "objective/train/weighted_lm_loss": 1.457322359085083,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.450286749270751,
      "tokens_seen": 2061107200
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037941825476429285,
      "loss": 1.427,
      "theoretical_loss": 3.450286749270751,
      "tokens_seen": 2061107200
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037939819458375127,
      "loss": 1.483,
      "theoretical_loss": 3.4502777313776205,
      "tokens_seen": 2061172736
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037937813440320963,
      "loss": 1.3381,
      "theoretical_loss": 3.4502687138514956,
      "tokens_seen": 2061238272
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000379358074222668,
      "loss": 1.4643,
      "theoretical_loss": 3.450259696692348,
      "tokens_seen": 2061303808
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037933801404212635,
      "loss": 1.3688,
      "theoretical_loss": 3.4502506799001527,
      "tokens_seen": 2061369344
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003793179538615847,
      "loss": 1.4446,
      "theoretical_loss": 3.4502416634748823,
      "tokens_seen": 2061434880
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037929789368104313,
      "loss": 1.4143,
      "theoretical_loss": 3.4502326474165104,
      "tokens_seen": 2061500416
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003792778335005015,
      "loss": 1.3409,
      "theoretical_loss": 3.4502236317250095,
      "tokens_seen": 2061565952
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037925777331995986,
      "loss": 1.5678,
      "theoretical_loss": 3.4502146164003547,
      "tokens_seen": 2061631488
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003792377131394182,
      "loss": 1.4303,
      "theoretical_loss": 3.4502056014425184,
      "tokens_seen": 2061697024
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037921765295887664,
      "loss": 1.4168,
      "theoretical_loss": 3.450196586851474,
      "tokens_seen": 2061762560
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.000379197592778335,
      "loss": 1.4439,
      "theoretical_loss": 3.450187572627195,
      "tokens_seen": 2061828096
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037917753259779336,
      "loss": 1.3811,
      "theoretical_loss": 3.4501785587696547,
      "tokens_seen": 2061893632
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003791574724172517,
      "loss": 1.4831,
      "theoretical_loss": 3.4501695452788272,
      "tokens_seen": 2061959168
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003791374122367101,
      "loss": 1.4153,
      "theoretical_loss": 3.450160532154685,
      "tokens_seen": 2062024704
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037911735205616856,
      "loss": 1.4624,
      "theoretical_loss": 3.4501515193972025,
      "tokens_seen": 2062090240
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003790972918756269,
      "loss": 1.4146,
      "theoretical_loss": 3.4501425070063525,
      "tokens_seen": 2062155776
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003790772316950853,
      "loss": 1.3346,
      "theoretical_loss": 3.4501334949821083,
      "tokens_seen": 2062221312
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037905717151454364,
      "loss": 1.3805,
      "theoretical_loss": 3.4501244833244438,
      "tokens_seen": 2062286848
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00037903711133400206,
      "loss": 1.3838,
      "theoretical_loss": 3.450115472033332,
      "tokens_seen": 2062352384
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003790170511534604,
      "loss": 1.4181,
      "theoretical_loss": 3.4501064611087466,
      "tokens_seen": 2062417920
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0003789969909729188,
      "loss": 1.3604,
      "theoretical_loss": 3.450097450550661,
      "tokens_seen": 2062483456
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037897693079237715,
      "loss": 1.5362,
      "theoretical_loss": 3.450088440359049,
      "tokens_seen": 2062548992
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037895687061183556,
      "loss": 1.4103,
      "theoretical_loss": 3.450079430533883,
      "tokens_seen": 2062614528
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003789368104312939,
      "loss": 1.4126,
      "theoretical_loss": 3.4500704210751376,
      "tokens_seen": 2062680064
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003789167502507523,
      "loss": 1.362,
      "theoretical_loss": 3.4500614119827855,
      "tokens_seen": 2062745600
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037889669007021065,
      "loss": 1.3546,
      "theoretical_loss": 3.450052403256801,
      "tokens_seen": 2062811136
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000378876629889669,
      "loss": 1.2977,
      "theoretical_loss": 3.4500433948971567,
      "tokens_seen": 2062876672
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037885656970912743,
      "loss": 1.3972,
      "theoretical_loss": 3.450034386903827,
      "tokens_seen": 2062942208
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003788365095285858,
      "loss": 1.3867,
      "theoretical_loss": 3.450025379276784,
      "tokens_seen": 2063007744
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037881644934804415,
      "loss": 1.4908,
      "theoretical_loss": 3.450016372016002,
      "tokens_seen": 2063073280
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003787963891675025,
      "loss": 1.3833,
      "theoretical_loss": 3.4500073651214547,
      "tokens_seen": 2063138816
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037877632898696093,
      "loss": 1.5642,
      "theoretical_loss": 3.449998358593115,
      "tokens_seen": 2063204352
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003787562688064193,
      "loss": 1.4246,
      "theoretical_loss": 3.449989352430957,
      "tokens_seen": 2063269888
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037873620862587766,
      "loss": 1.3846,
      "theoretical_loss": 3.4499803466349537,
      "tokens_seen": 2063335424
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000378716148445336,
      "loss": 1.3306,
      "theoretical_loss": 3.449971341205079,
      "tokens_seen": 2063400960
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003786960882647944,
      "loss": 1.4941,
      "theoretical_loss": 3.449962336141306,
      "tokens_seen": 2063466496
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003786760280842528,
      "loss": 1.4992,
      "theoretical_loss": 3.4499533314436084,
      "tokens_seen": 2063532032
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037865596790371116,
      "loss": 1.439,
      "theoretical_loss": 3.449944327111959,
      "tokens_seen": 2063597568
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003786359077231695,
      "loss": 1.324,
      "theoretical_loss": 3.4499353231463328,
      "tokens_seen": 2063663104
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003786158475426279,
      "loss": 1.4692,
      "theoretical_loss": 3.4499263195467016,
      "tokens_seen": 2063728640
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003785957873620863,
      "loss": 1.4235,
      "theoretical_loss": 3.44991731631304,
      "tokens_seen": 2063794176
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037857572718154466,
      "loss": 1.398,
      "theoretical_loss": 3.449908313445321,
      "tokens_seen": 2063859712
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000378555667001003,
      "loss": 1.3105,
      "theoretical_loss": 3.4498993109435188,
      "tokens_seen": 2063925248
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003785356068204614,
      "loss": 1.3941,
      "theoretical_loss": 3.449890308807606,
      "tokens_seen": 2063990784
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037851554663991975,
      "loss": 1.3908,
      "theoretical_loss": 3.4498813070375567,
      "tokens_seen": 2064056320
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037849548645937816,
      "loss": 1.5329,
      "theoretical_loss": 3.4498723056333445,
      "tokens_seen": 2064121856
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003784754262788365,
      "loss": 1.4756,
      "theoretical_loss": 3.4498633045949423,
      "tokens_seen": 2064187392
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003784553660982949,
      "loss": 1.3959,
      "theoretical_loss": 3.449854303922324,
      "tokens_seen": 2064252928
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037843530591775325,
      "loss": 1.5827,
      "theoretical_loss": 3.4498453036154633,
      "tokens_seen": 2064318464
    },
    {
      "epoch": 0.63,
      "objective/train/advantage_avg": 0.49434155225753784,
      "objective/train/docs_used": 752866,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3168742656707764,
      "objective/train/original_loss": 1.316874384880066,
      "objective/train/theoretical_loss": 3.4498363036743336,
      "objective/train/tokens_used": 2084844000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24464908242225647,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506778955459595,
      "objective/train/weighted_lm_loss": 1.383488655090332,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4498363036743336,
      "tokens_seen": 2064384000
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037841524573721167,
      "loss": 1.4906,
      "theoretical_loss": 3.4498363036743336,
      "tokens_seen": 2064384000
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037839518555667003,
      "loss": 1.382,
      "theoretical_loss": 3.449827304098908,
      "tokens_seen": 2064449536
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003783751253761284,
      "loss": 1.3722,
      "theoretical_loss": 3.4498183048891606,
      "tokens_seen": 2064515072
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037835506519558675,
      "loss": 1.3411,
      "theoretical_loss": 3.449809306045065,
      "tokens_seen": 2064580608
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003783350050150451,
      "loss": 1.4379,
      "theoretical_loss": 3.449800307566594,
      "tokens_seen": 2064646144
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037831494483450353,
      "loss": 1.497,
      "theoretical_loss": 3.4497913094537216,
      "tokens_seen": 2064711680
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003782948846539619,
      "loss": 1.4069,
      "theoretical_loss": 3.4497823117064215,
      "tokens_seen": 2064777216
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037827482447342026,
      "loss": 1.3584,
      "theoretical_loss": 3.449773314324667,
      "tokens_seen": 2064842752
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003782547642928786,
      "loss": 1.3859,
      "theoretical_loss": 3.449764317308432,
      "tokens_seen": 2064908288
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037823470411233704,
      "loss": 1.4183,
      "theoretical_loss": 3.44975532065769,
      "tokens_seen": 2064973824
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003782146439317954,
      "loss": 1.5008,
      "theoretical_loss": 3.4497463243724136,
      "tokens_seen": 2065039360
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037819458375125376,
      "loss": 1.4669,
      "theoretical_loss": 3.449737328452578,
      "tokens_seen": 2065104896
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003781745235707121,
      "loss": 1.4343,
      "theoretical_loss": 3.449728332898155,
      "tokens_seen": 2065170432
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003781544633901705,
      "loss": 1.4759,
      "theoretical_loss": 3.4497193377091193,
      "tokens_seen": 2065235968
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003781344032096289,
      "loss": 1.3303,
      "theoretical_loss": 3.4497103428854445,
      "tokens_seen": 2065301504
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037811434302908726,
      "loss": 1.3308,
      "theoretical_loss": 3.4497013484271037,
      "tokens_seen": 2065367040
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003780942828485456,
      "loss": 1.3465,
      "theoretical_loss": 3.4496923543340703,
      "tokens_seen": 2065432576
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000378074222668004,
      "loss": 1.3119,
      "theoretical_loss": 3.449683360606319,
      "tokens_seen": 2065498112
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003780541624874624,
      "loss": 1.5327,
      "theoretical_loss": 3.449674367243822,
      "tokens_seen": 2065563648
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037803410230692077,
      "loss": 1.5092,
      "theoretical_loss": 3.4496653742465533,
      "tokens_seen": 2065629184
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037801404212637913,
      "loss": 1.3769,
      "theoretical_loss": 3.4496563816144867,
      "tokens_seen": 2065694720
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003779939819458375,
      "loss": 1.4405,
      "theoretical_loss": 3.4496473893475956,
      "tokens_seen": 2065760256
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037797392176529585,
      "loss": 1.3559,
      "theoretical_loss": 3.449638397445854,
      "tokens_seen": 2065825792
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037795386158475427,
      "loss": 1.4405,
      "theoretical_loss": 3.449629405909235,
      "tokens_seen": 2065891328
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037793380140421263,
      "loss": 1.3321,
      "theoretical_loss": 3.449620414737713,
      "tokens_seen": 2065956864
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000377913741223671,
      "loss": 1.3152,
      "theoretical_loss": 3.4496114239312603,
      "tokens_seen": 2066022400
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037789368104312936,
      "loss": 1.3927,
      "theoretical_loss": 3.449602433489851,
      "tokens_seen": 2066087936
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003778736208625878,
      "loss": 1.4543,
      "theoretical_loss": 3.4495934434134594,
      "tokens_seen": 2066153472
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037785356068204614,
      "loss": 1.3282,
      "theoretical_loss": 3.449584453702059,
      "tokens_seen": 2066219008
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003778335005015045,
      "loss": 1.4651,
      "theoretical_loss": 3.449575464355622,
      "tokens_seen": 2066284544
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037781344032096286,
      "loss": 1.4529,
      "theoretical_loss": 3.449566475374123,
      "tokens_seen": 2066350080
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003777933801404213,
      "loss": 1.3317,
      "theoretical_loss": 3.449557486757536,
      "tokens_seen": 2066415616
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037777331995987964,
      "loss": 1.3494,
      "theoretical_loss": 3.449548498505834,
      "tokens_seen": 2066481152
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000377753259779338,
      "loss": 1.4517,
      "theoretical_loss": 3.4495395106189912,
      "tokens_seen": 2066546688
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037773319959879636,
      "loss": 1.4923,
      "theoretical_loss": 3.449530523096981,
      "tokens_seen": 2066612224
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003777131394182547,
      "loss": 1.4086,
      "theoretical_loss": 3.449521535939776,
      "tokens_seen": 2066677760
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037769307923771314,
      "loss": 1.5581,
      "theoretical_loss": 3.4495125491473515,
      "tokens_seen": 2066743296
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003776730190571715,
      "loss": 1.3991,
      "theoretical_loss": 3.44950356271968,
      "tokens_seen": 2066808832
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037765295887662987,
      "loss": 1.4291,
      "theoretical_loss": 3.4494945766567358,
      "tokens_seen": 2066874368
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037763289869608823,
      "loss": 1.3896,
      "theoretical_loss": 3.449485590958492,
      "tokens_seen": 2066939904
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003776128385155467,
      "loss": 1.3554,
      "theoretical_loss": 3.449476605624922,
      "tokens_seen": 2067005440
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037759277833500506,
      "loss": 1.4293,
      "theoretical_loss": 3.449467620656,
      "tokens_seen": 2067070976
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003775727181544634,
      "loss": 1.4214,
      "theoretical_loss": 3.4494586360517,
      "tokens_seen": 2067136512
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003775526579739218,
      "loss": 1.3207,
      "theoretical_loss": 3.449449651811995,
      "tokens_seen": 2067202048
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037753259779338015,
      "loss": 1.4405,
      "theoretical_loss": 3.4494406679368583,
      "tokens_seen": 2067267584
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037751253761283857,
      "loss": 1.3759,
      "theoretical_loss": 3.4494316844262647,
      "tokens_seen": 2067333120
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037749247743229693,
      "loss": 1.4938,
      "theoretical_loss": 3.4494227012801866,
      "tokens_seen": 2067398656
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003774724172517553,
      "loss": 1.4165,
      "theoretical_loss": 3.4494137184985982,
      "tokens_seen": 2067464192
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037745235707121365,
      "loss": 1.4663,
      "theoretical_loss": 3.449404736081474,
      "tokens_seen": 2067529728
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037743229689067207,
      "loss": 1.5001,
      "theoretical_loss": 3.449395754028786,
      "tokens_seen": 2067595264
    },
    {
      "epoch": 0.63,
      "objective/train/advantage_avg": 0.48923370242118835,
      "objective/train/docs_used": 753921,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4936641454696655,
      "objective/train/original_loss": 1.4936641454696655,
      "objective/train/theoretical_loss": 3.4493867723405094,
      "objective/train/tokens_used": 2088120800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24002563953399658,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501434803009033,
      "objective/train/weighted_lm_loss": 1.5685746669769287,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.4493867723405094,
      "tokens_seen": 2067660800
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037741223671013043,
      "loss": 1.4731,
      "theoretical_loss": 3.4493867723405094,
      "tokens_seen": 2067660800
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003773921765295888,
      "loss": 1.4584,
      "theoretical_loss": 3.449377791016617,
      "tokens_seen": 2067726336
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037737211634904715,
      "loss": 1.4808,
      "theoretical_loss": 3.4493688100570825,
      "tokens_seen": 2067791872
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003773520561685055,
      "loss": 1.5449,
      "theoretical_loss": 3.44935982946188,
      "tokens_seen": 2067857408
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037733199598796393,
      "loss": 1.4269,
      "theoretical_loss": 3.4493508492309823,
      "tokens_seen": 2067922944
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003773119358074223,
      "loss": 1.4941,
      "theoretical_loss": 3.4493418693643645,
      "tokens_seen": 2067988480
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037729187562688066,
      "loss": 1.347,
      "theoretical_loss": 3.4493328898619993,
      "tokens_seen": 2068054016
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000377271815446339,
      "loss": 1.4122,
      "theoretical_loss": 3.4493239107238605,
      "tokens_seen": 2068119552
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037725175526579744,
      "loss": 1.4179,
      "theoretical_loss": 3.449314931949921,
      "tokens_seen": 2068185088
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003772316950852558,
      "loss": 1.4351,
      "theoretical_loss": 3.4493059535401565,
      "tokens_seen": 2068250624
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037721163490471416,
      "loss": 1.3531,
      "theoretical_loss": 3.449296975494539,
      "tokens_seen": 2068316160
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003771915747241725,
      "loss": 1.4372,
      "theoretical_loss": 3.449287997813043,
      "tokens_seen": 2068381696
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003771715145436309,
      "loss": 1.2813,
      "theoretical_loss": 3.4492790204956414,
      "tokens_seen": 2068447232
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003771514543630893,
      "loss": 1.4034,
      "theoretical_loss": 3.4492700435423087,
      "tokens_seen": 2068512768
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037713139418254766,
      "loss": 1.4115,
      "theoretical_loss": 3.449261066953018,
      "tokens_seen": 2068578304
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000377111334002006,
      "loss": 1.4994,
      "theoretical_loss": 3.449252090727744,
      "tokens_seen": 2068643840
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003770912738214644,
      "loss": 1.3522,
      "theoretical_loss": 3.4492431148664586,
      "tokens_seen": 2068709376
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003770712136409228,
      "loss": 1.4285,
      "theoretical_loss": 3.4492341393691377,
      "tokens_seen": 2068774912
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037705115346038117,
      "loss": 1.4142,
      "theoretical_loss": 3.4492251642357536,
      "tokens_seen": 2068840448
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037703109327983953,
      "loss": 1.4179,
      "theoretical_loss": 3.4492161894662803,
      "tokens_seen": 2068905984
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003770110330992979,
      "loss": 1.4408,
      "theoretical_loss": 3.4492072150606914,
      "tokens_seen": 2068971520
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037699097291875625,
      "loss": 1.4567,
      "theoretical_loss": 3.449198241018961,
      "tokens_seen": 2069037056
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037697091273821467,
      "loss": 1.4418,
      "theoretical_loss": 3.4491892673410627,
      "tokens_seen": 2069102592
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037695085255767303,
      "loss": 1.2966,
      "theoretical_loss": 3.44918029402697,
      "tokens_seen": 2069168128
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003769307923771314,
      "loss": 1.4362,
      "theoretical_loss": 3.4491713210766566,
      "tokens_seen": 2069233664
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037691073219658976,
      "loss": 1.5154,
      "theoretical_loss": 3.4491623484900966,
      "tokens_seen": 2069299200
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003768906720160482,
      "loss": 1.4696,
      "theoretical_loss": 3.449153376267264,
      "tokens_seen": 2069364736
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037687061183550654,
      "loss": 1.4152,
      "theoretical_loss": 3.4491444044081314,
      "tokens_seen": 2069430272
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003768505516549649,
      "loss": 1.3828,
      "theoretical_loss": 3.4491354329126738,
      "tokens_seen": 2069495808
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037683049147442326,
      "loss": 1.4048,
      "theoretical_loss": 3.449126461780864,
      "tokens_seen": 2069561344
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003768104312938817,
      "loss": 1.4357,
      "theoretical_loss": 3.449117491012676,
      "tokens_seen": 2069626880
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037679037111334004,
      "loss": 1.4627,
      "theoretical_loss": 3.4491085206080836,
      "tokens_seen": 2069692416
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003767703109327984,
      "loss": 1.4062,
      "theoretical_loss": 3.4490995505670607,
      "tokens_seen": 2069757952
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037675025075225676,
      "loss": 1.4443,
      "theoretical_loss": 3.449090580889581,
      "tokens_seen": 2069823488
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003767301905717151,
      "loss": 1.2787,
      "theoretical_loss": 3.449081611575618,
      "tokens_seen": 2069889024
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037671013039117354,
      "loss": 1.4244,
      "theoretical_loss": 3.449072642625146,
      "tokens_seen": 2069954560
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003766900702106319,
      "loss": 1.4248,
      "theoretical_loss": 3.449063674038138,
      "tokens_seen": 2070020096
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037667001003009027,
      "loss": 1.3041,
      "theoretical_loss": 3.449054705814569,
      "tokens_seen": 2070085632
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037664994984954863,
      "loss": 1.4376,
      "theoretical_loss": 3.4490457379544113,
      "tokens_seen": 2070151168
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037662988966900705,
      "loss": 1.4643,
      "theoretical_loss": 3.4490367704576395,
      "tokens_seen": 2070216704
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003766098294884654,
      "loss": 1.4449,
      "theoretical_loss": 3.4490278033242268,
      "tokens_seen": 2070282240
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037658976930792377,
      "loss": 1.4826,
      "theoretical_loss": 3.449018836554148,
      "tokens_seen": 2070347776
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037656970912738213,
      "loss": 1.4565,
      "theoretical_loss": 3.4490098701473757,
      "tokens_seen": 2070413312
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003765496489468405,
      "loss": 1.3454,
      "theoretical_loss": 3.449000904103885,
      "tokens_seen": 2070478848
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003765295887662989,
      "loss": 1.3231,
      "theoretical_loss": 3.4489919384236485,
      "tokens_seen": 2070544384
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003765095285857573,
      "loss": 1.5157,
      "theoretical_loss": 3.44898297310664,
      "tokens_seen": 2070609920
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037648946840521564,
      "loss": 1.3938,
      "theoretical_loss": 3.448974008152834,
      "tokens_seen": 2070675456
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000376469408224674,
      "loss": 1.4244,
      "theoretical_loss": 3.448965043562204,
      "tokens_seen": 2070740992
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003764493480441324,
      "loss": 1.3985,
      "theoretical_loss": 3.448956079334724,
      "tokens_seen": 2070806528
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003764292878635908,
      "loss": 1.465,
      "theoretical_loss": 3.448947115470367,
      "tokens_seen": 2070872064
    },
    {
      "epoch": 0.63,
      "objective/train/advantage_avg": 0.4929508566856384,
      "objective/train/docs_used": 755047,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.283677577972412,
      "objective/train/original_loss": 1.283677577972412,
      "objective/train/theoretical_loss": 3.4489381519691085,
      "objective/train/tokens_used": 2091397600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24346697330474854,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505328178405762,
      "objective/train/weighted_lm_loss": 1.3484480381011963,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4489381519691085,
      "tokens_seen": 2070937600
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037640922768304914,
      "loss": 1.3756,
      "theoretical_loss": 3.4489381519691085,
      "tokens_seen": 2070937600
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003763891675025075,
      "loss": 1.4583,
      "theoretical_loss": 3.4489291888309204,
      "tokens_seen": 2071003136
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037636910732196586,
      "loss": 1.4451,
      "theoretical_loss": 3.448920226055777,
      "tokens_seen": 2071068672
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003763490471414243,
      "loss": 1.5204,
      "theoretical_loss": 3.4489112636436525,
      "tokens_seen": 2071134208
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037632898696088264,
      "loss": 1.4154,
      "theoretical_loss": 3.4489023015945213,
      "tokens_seen": 2071199744
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000376308926780341,
      "loss": 1.4512,
      "theoretical_loss": 3.4488933399083557,
      "tokens_seen": 2071265280
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037628886659979937,
      "loss": 1.4041,
      "theoretical_loss": 3.448884378585131,
      "tokens_seen": 2071330816
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003762688064192578,
      "loss": 1.4213,
      "theoretical_loss": 3.44887541762482,
      "tokens_seen": 2071396352
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037624874623871615,
      "loss": 1.3582,
      "theoretical_loss": 3.448866457027397,
      "tokens_seen": 2071461888
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003762286860581745,
      "loss": 1.3782,
      "theoretical_loss": 3.4488574967928356,
      "tokens_seen": 2071527424
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037620862587763287,
      "loss": 1.4162,
      "theoretical_loss": 3.4488485369211093,
      "tokens_seen": 2071592960
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037618856569709123,
      "loss": 1.3594,
      "theoretical_loss": 3.448839577412193,
      "tokens_seen": 2071658496
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037616850551654965,
      "loss": 1.5201,
      "theoretical_loss": 3.4488306182660597,
      "tokens_seen": 2071724032
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000376148445336008,
      "loss": 1.4625,
      "theoretical_loss": 3.4488216594826833,
      "tokens_seen": 2071789568
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037612838515546637,
      "loss": 1.4399,
      "theoretical_loss": 3.448812701062038,
      "tokens_seen": 2071855104
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003761083249749248,
      "loss": 1.492,
      "theoretical_loss": 3.448803743004097,
      "tokens_seen": 2071920640
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003760882647943832,
      "loss": 1.5002,
      "theoretical_loss": 3.448794785308835,
      "tokens_seen": 2071986176
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037606820461384157,
      "loss": 1.3599,
      "theoretical_loss": 3.4487858279762253,
      "tokens_seen": 2072051712
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037604814443329993,
      "loss": 1.5087,
      "theoretical_loss": 3.4487768710062414,
      "tokens_seen": 2072117248
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003760280842527583,
      "loss": 1.4687,
      "theoretical_loss": 3.448767914398858,
      "tokens_seen": 2072182784
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037600802407221665,
      "loss": 1.3427,
      "theoretical_loss": 3.4487589581540483,
      "tokens_seen": 2072248320
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037598796389167507,
      "loss": 1.3409,
      "theoretical_loss": 3.4487500022717867,
      "tokens_seen": 2072313856
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037596790371113343,
      "loss": 1.41,
      "theoretical_loss": 3.4487410467520463,
      "tokens_seen": 2072379392
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003759478435305918,
      "loss": 1.457,
      "theoretical_loss": 3.4487320915948017,
      "tokens_seen": 2072444928
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037592778335005016,
      "loss": 1.3768,
      "theoretical_loss": 3.4487231368000266,
      "tokens_seen": 2072510464
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003759077231695086,
      "loss": 1.4564,
      "theoretical_loss": 3.4487141823676946,
      "tokens_seen": 2072576000
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037588766298896694,
      "loss": 1.2965,
      "theoretical_loss": 3.4487052282977797,
      "tokens_seen": 2072641536
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003758676028084253,
      "loss": 1.3755,
      "theoretical_loss": 3.448696274590256,
      "tokens_seen": 2072707072
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037584754262788366,
      "loss": 1.4592,
      "theoretical_loss": 3.448687321245097,
      "tokens_seen": 2072772608
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003758274824473421,
      "loss": 1.3482,
      "theoretical_loss": 3.448678368262277,
      "tokens_seen": 2072838144
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037580742226680044,
      "loss": 1.3623,
      "theoretical_loss": 3.4486694156417688,
      "tokens_seen": 2072903680
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003757873620862588,
      "loss": 1.3661,
      "theoretical_loss": 3.448660463383548,
      "tokens_seen": 2072969216
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037576730190571716,
      "loss": 1.3932,
      "theoretical_loss": 3.4486515114875873,
      "tokens_seen": 2073034752
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003757472417251755,
      "loss": 1.5689,
      "theoretical_loss": 3.4486425599538606,
      "tokens_seen": 2073100288
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037572718154463394,
      "loss": 1.3971,
      "theoretical_loss": 3.4486336087823424,
      "tokens_seen": 2073165824
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003757071213640923,
      "loss": 1.4041,
      "theoretical_loss": 3.4486246579730064,
      "tokens_seen": 2073231360
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037568706118355067,
      "loss": 1.4833,
      "theoretical_loss": 3.448615707525826,
      "tokens_seen": 2073296896
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037566700100300903,
      "loss": 1.4172,
      "theoretical_loss": 3.4486067574407757,
      "tokens_seen": 2073362432
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037564694082246745,
      "loss": 1.4247,
      "theoretical_loss": 3.4485978077178294,
      "tokens_seen": 2073427968
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003756268806419258,
      "loss": 1.4025,
      "theoretical_loss": 3.4485888583569606,
      "tokens_seen": 2073493504
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037560682046138417,
      "loss": 1.4577,
      "theoretical_loss": 3.448579909358143,
      "tokens_seen": 2073559040
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037558676028084253,
      "loss": 1.609,
      "theoretical_loss": 3.4485709607213515,
      "tokens_seen": 2073624576
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003755667001003009,
      "loss": 1.3604,
      "theoretical_loss": 3.4485620124465592,
      "tokens_seen": 2073690112
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003755466399197593,
      "loss": 1.3721,
      "theoretical_loss": 3.4485530645337397,
      "tokens_seen": 2073755648
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003755265797392177,
      "loss": 1.3591,
      "theoretical_loss": 3.448544116982868,
      "tokens_seen": 2073821184
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037550651955867604,
      "loss": 1.5264,
      "theoretical_loss": 3.4485351697939173,
      "tokens_seen": 2073886720
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003754864593781344,
      "loss": 1.4394,
      "theoretical_loss": 3.448526222966862,
      "tokens_seen": 2073952256
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003754663991975928,
      "loss": 1.3154,
      "theoretical_loss": 3.4485172765016756,
      "tokens_seen": 2074017792
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003754463390170512,
      "loss": 1.4882,
      "theoretical_loss": 3.448508330398332,
      "tokens_seen": 2074083328
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037542627883650954,
      "loss": 1.3998,
      "theoretical_loss": 3.448499384656805,
      "tokens_seen": 2074148864
    },
    {
      "epoch": 0.63,
      "objective/train/advantage_avg": 0.49311262369155884,
      "objective/train/docs_used": 756110,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.306643009185791,
      "objective/train/original_loss": 1.3066428899765015,
      "objective/train/theoretical_loss": 3.4484904392770694,
      "objective/train/tokens_used": 2094674400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24367785453796387,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505499839782715,
      "objective/train/weighted_lm_loss": 1.3725608587265015,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4484904392770694,
      "tokens_seen": 2074214400
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003754062186559679,
      "loss": 1.4504,
      "theoretical_loss": 3.4484904392770694,
      "tokens_seen": 2074214400
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037538615847542626,
      "loss": 1.3159,
      "theoretical_loss": 3.4484814942590987,
      "tokens_seen": 2074279936
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003753660982948847,
      "loss": 1.3875,
      "theoretical_loss": 3.448472549602866,
      "tokens_seen": 2074345472
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037534603811434304,
      "loss": 1.3948,
      "theoretical_loss": 3.448463605308346,
      "tokens_seen": 2074411008
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003753259779338014,
      "loss": 1.324,
      "theoretical_loss": 3.4484546613755134,
      "tokens_seen": 2074476544
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037530591775325977,
      "loss": 1.3629,
      "theoretical_loss": 3.4484457178043404,
      "tokens_seen": 2074542080
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003752858575727182,
      "loss": 1.3171,
      "theoretical_loss": 3.4484367745948026,
      "tokens_seen": 2074607616
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037526579739217655,
      "loss": 1.3868,
      "theoretical_loss": 3.4484278317468724,
      "tokens_seen": 2074673152
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003752457372116349,
      "loss": 1.3395,
      "theoretical_loss": 3.448418889260525,
      "tokens_seen": 2074738688
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037522567703109327,
      "loss": 1.4938,
      "theoretical_loss": 3.448409947135734,
      "tokens_seen": 2074804224
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037520561685055163,
      "loss": 1.5707,
      "theoretical_loss": 3.4484010053724736,
      "tokens_seen": 2074869760
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037518555667001005,
      "loss": 1.4666,
      "theoretical_loss": 3.4483920639707173,
      "tokens_seen": 2074935296
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003751654964894684,
      "loss": 1.3833,
      "theoretical_loss": 3.448383122930439,
      "tokens_seen": 2075000832
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003751454363089268,
      "loss": 1.4305,
      "theoretical_loss": 3.448374182251613,
      "tokens_seen": 2075066368
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037512537612838514,
      "loss": 1.4097,
      "theoretical_loss": 3.448365241934214,
      "tokens_seen": 2075131904
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037510531594784355,
      "loss": 1.3584,
      "theoretical_loss": 3.448356301978214,
      "tokens_seen": 2075197440
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003750852557673019,
      "loss": 1.4278,
      "theoretical_loss": 3.448347362383589,
      "tokens_seen": 2075262976
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003750651955867603,
      "loss": 1.3809,
      "theoretical_loss": 3.4483384231503114,
      "tokens_seen": 2075328512
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037504513540621864,
      "loss": 1.4069,
      "theoretical_loss": 3.4483294842783563,
      "tokens_seen": 2075394048
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000375025075225677,
      "loss": 1.4368,
      "theoretical_loss": 3.4483205457676975,
      "tokens_seen": 2075459584
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003750050150451354,
      "loss": 1.4601,
      "theoretical_loss": 3.4483116076183085,
      "tokens_seen": 2075525120
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003749849548645938,
      "loss": 1.4768,
      "theoretical_loss": 3.4483026698301638,
      "tokens_seen": 2075590656
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037496489468405214,
      "loss": 1.3748,
      "theoretical_loss": 3.4482937324032368,
      "tokens_seen": 2075656192
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003749448345035105,
      "loss": 1.5112,
      "theoretical_loss": 3.4482847953375018,
      "tokens_seen": 2075721728
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003749247743229689,
      "loss": 1.4466,
      "theoretical_loss": 3.4482758586329334,
      "tokens_seen": 2075787264
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003749047141424273,
      "loss": 1.3553,
      "theoretical_loss": 3.4482669222895046,
      "tokens_seen": 2075852800
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037488465396188564,
      "loss": 1.5769,
      "theoretical_loss": 3.4482579863071905,
      "tokens_seen": 2075918336
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000374864593781344,
      "loss": 1.3973,
      "theoretical_loss": 3.4482490506859635,
      "tokens_seen": 2075983872
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037484453360080237,
      "loss": 1.4127,
      "theoretical_loss": 3.4482401154258,
      "tokens_seen": 2076049408
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003748244734202608,
      "loss": 1.4189,
      "theoretical_loss": 3.4482311805266717,
      "tokens_seen": 2076114944
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037480441323971915,
      "loss": 1.4227,
      "theoretical_loss": 3.4482222459885534,
      "tokens_seen": 2076180480
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003747843530591775,
      "loss": 1.3665,
      "theoretical_loss": 3.4482133118114193,
      "tokens_seen": 2076246016
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037476429287863587,
      "loss": 1.4795,
      "theoretical_loss": 3.4482043779952436,
      "tokens_seen": 2076311552
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003747442326980943,
      "loss": 1.4142,
      "theoretical_loss": 3.44819544454,
      "tokens_seen": 2076377088
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037472417251755265,
      "loss": 1.3627,
      "theoretical_loss": 3.4481865114456625,
      "tokens_seen": 2076442624
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000374704112337011,
      "loss": 1.3189,
      "theoretical_loss": 3.4481775787122055,
      "tokens_seen": 2076508160
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003746840521564694,
      "loss": 1.4894,
      "theoretical_loss": 3.448168646339603,
      "tokens_seen": 2076573696
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037466399197592774,
      "loss": 1.3743,
      "theoretical_loss": 3.4481597143278284,
      "tokens_seen": 2076639232
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037464393179538615,
      "loss": 1.3681,
      "theoretical_loss": 3.448150782676856,
      "tokens_seen": 2076704768
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037462387161484457,
      "loss": 1.317,
      "theoretical_loss": 3.4481418513866604,
      "tokens_seen": 2076770304
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037460381143430293,
      "loss": 1.2735,
      "theoretical_loss": 3.448132920457215,
      "tokens_seen": 2076835840
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003745837512537613,
      "loss": 1.4524,
      "theoretical_loss": 3.4481239898884946,
      "tokens_seen": 2076901376
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003745636910732197,
      "loss": 1.2725,
      "theoretical_loss": 3.4481150596804717,
      "tokens_seen": 2076966912
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003745436308926781,
      "loss": 1.3673,
      "theoretical_loss": 3.448106129833122,
      "tokens_seen": 2077032448
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037452357071213644,
      "loss": 1.3939,
      "theoretical_loss": 3.448097200346419,
      "tokens_seen": 2077097984
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003745035105315948,
      "loss": 1.393,
      "theoretical_loss": 3.4480882712203367,
      "tokens_seen": 2077163520
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003744834503510532,
      "loss": 1.3939,
      "theoretical_loss": 3.4480793424548493,
      "tokens_seen": 2077229056
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003744633901705116,
      "loss": 1.4709,
      "theoretical_loss": 3.4480704140499303,
      "tokens_seen": 2077294592
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037444332998996994,
      "loss": 1.2293,
      "theoretical_loss": 3.448061486005554,
      "tokens_seen": 2077360128
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003744232698094283,
      "loss": 1.4024,
      "theoretical_loss": 3.448052558321695,
      "tokens_seen": 2077425664
    },
    {
      "epoch": 0.63,
      "objective/train/advantage_avg": 0.4945654571056366,
      "objective/train/docs_used": 757304,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.416394591331482,
      "objective/train/original_loss": 1.4163947105407715,
      "objective/train/theoretical_loss": 3.4480436309983276,
      "objective/train/tokens_used": 2097951200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24491672217845917,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507017374038696,
      "objective/train/weighted_lm_loss": 1.4880176782608032,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4480436309983276,
      "tokens_seen": 2077491200
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037440320962888666,
      "loss": 1.4144,
      "theoretical_loss": 3.4480436309983276,
      "tokens_seen": 2077491200
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003743831494483451,
      "loss": 1.4001,
      "theoretical_loss": 3.4480347040354244,
      "tokens_seen": 2077556736
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037436308926780344,
      "loss": 1.3442,
      "theoretical_loss": 3.448025777432961,
      "tokens_seen": 2077622272
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003743430290872618,
      "loss": 1.4543,
      "theoretical_loss": 3.44801685119091,
      "tokens_seen": 2077687808
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037432296890672017,
      "loss": 1.5288,
      "theoretical_loss": 3.448007925309247,
      "tokens_seen": 2077753344
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003743029087261786,
      "loss": 1.4363,
      "theoretical_loss": 3.447998999787945,
      "tokens_seen": 2077818880
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037428284854563695,
      "loss": 1.4155,
      "theoretical_loss": 3.4479900746269787,
      "tokens_seen": 2077884416
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003742627883650953,
      "loss": 1.4625,
      "theoretical_loss": 3.447981149826322,
      "tokens_seen": 2077949952
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037424272818455367,
      "loss": 1.4425,
      "theoretical_loss": 3.447972225385949,
      "tokens_seen": 2078015488
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037422266800401203,
      "loss": 1.3585,
      "theoretical_loss": 3.4479633013058337,
      "tokens_seen": 2078081024
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037420260782347045,
      "loss": 1.3657,
      "theoretical_loss": 3.4479543775859502,
      "tokens_seen": 2078146560
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003741825476429288,
      "loss": 1.3266,
      "theoretical_loss": 3.447945454226273,
      "tokens_seen": 2078212096
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003741624874623872,
      "loss": 1.343,
      "theoretical_loss": 3.447936531226776,
      "tokens_seen": 2078277632
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037414242728184554,
      "loss": 1.312,
      "theoretical_loss": 3.447927608587432,
      "tokens_seen": 2078343168
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037412236710130395,
      "loss": 1.3967,
      "theoretical_loss": 3.447918686308217,
      "tokens_seen": 2078408704
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003741023069207623,
      "loss": 1.3527,
      "theoretical_loss": 3.4479097643891046,
      "tokens_seen": 2078474240
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003740822467402207,
      "loss": 1.3549,
      "theoretical_loss": 3.4479008428300686,
      "tokens_seen": 2078539776
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037406218655967904,
      "loss": 1.3879,
      "theoretical_loss": 3.447891921631083,
      "tokens_seen": 2078605312
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003740421263791374,
      "loss": 1.4141,
      "theoretical_loss": 3.447883000792122,
      "tokens_seen": 2078670848
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003740220661985958,
      "loss": 1.4621,
      "theoretical_loss": 3.4478740803131602,
      "tokens_seen": 2078736384
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003740020060180542,
      "loss": 1.3532,
      "theoretical_loss": 3.447865160194171,
      "tokens_seen": 2078801920
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037398194583751254,
      "loss": 1.3567,
      "theoretical_loss": 3.4478562404351294,
      "tokens_seen": 2078867456
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003739618856569709,
      "loss": 1.4213,
      "theoretical_loss": 3.4478473210360088,
      "tokens_seen": 2078932992
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003739418254764293,
      "loss": 1.4771,
      "theoretical_loss": 3.4478384019967834,
      "tokens_seen": 2078998528
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003739217652958877,
      "loss": 1.4426,
      "theoretical_loss": 3.447829483317428,
      "tokens_seen": 2079064064
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037390170511534605,
      "loss": 1.2995,
      "theoretical_loss": 3.4478205649979157,
      "tokens_seen": 2079129600
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003738816449348044,
      "loss": 1.3226,
      "theoretical_loss": 3.447811647038221,
      "tokens_seen": 2079195136
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037386158475426277,
      "loss": 1.3831,
      "theoretical_loss": 3.447802729438319,
      "tokens_seen": 2079260672
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003738415245737212,
      "loss": 1.4435,
      "theoretical_loss": 3.4477938121981824,
      "tokens_seen": 2079326208
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037382146439317955,
      "loss": 1.3102,
      "theoretical_loss": 3.447784895317786,
      "tokens_seen": 2079391744
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003738014042126379,
      "loss": 1.3207,
      "theoretical_loss": 3.4477759787971043,
      "tokens_seen": 2079457280
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037378134403209627,
      "loss": 1.3249,
      "theoretical_loss": 3.447767062636111,
      "tokens_seen": 2079522816
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003737612838515547,
      "loss": 1.3838,
      "theoretical_loss": 3.44775814683478,
      "tokens_seen": 2079588352
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037374122367101305,
      "loss": 1.3526,
      "theoretical_loss": 3.4477492313930864,
      "tokens_seen": 2079653888
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003737211634904714,
      "loss": 1.3819,
      "theoretical_loss": 3.4477403163110036,
      "tokens_seen": 2079719424
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003737011033099298,
      "loss": 1.335,
      "theoretical_loss": 3.4477314015885057,
      "tokens_seen": 2079784960
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037368104312938814,
      "loss": 1.4648,
      "theoretical_loss": 3.447722487225567,
      "tokens_seen": 2079850496
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037366098294884655,
      "loss": 1.3887,
      "theoretical_loss": 3.4477135732221624,
      "tokens_seen": 2079916032
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003736409227683049,
      "loss": 1.3236,
      "theoretical_loss": 3.4477046595782648,
      "tokens_seen": 2079981568
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003736208625877633,
      "loss": 1.2725,
      "theoretical_loss": 3.4476957462938493,
      "tokens_seen": 2080047104
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037360080240722164,
      "loss": 1.3043,
      "theoretical_loss": 3.4476868333688904,
      "tokens_seen": 2080112640
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037358074222668006,
      "loss": 1.3818,
      "theoretical_loss": 3.4476779208033608,
      "tokens_seen": 2080178176
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003735606820461384,
      "loss": 1.3823,
      "theoretical_loss": 3.4476690085972357,
      "tokens_seen": 2080243712
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003735406218655968,
      "loss": 1.4226,
      "theoretical_loss": 3.4476600967504893,
      "tokens_seen": 2080309248
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037352056168505514,
      "loss": 1.4855,
      "theoretical_loss": 3.447651185263096,
      "tokens_seen": 2080374784
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037350050150451356,
      "loss": 1.434,
      "theoretical_loss": 3.447642274135029,
      "tokens_seen": 2080440320
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003734804413239719,
      "loss": 1.4612,
      "theoretical_loss": 3.4476333633662635,
      "tokens_seen": 2080505856
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003734603811434303,
      "loss": 1.3715,
      "theoretical_loss": 3.447624452956773,
      "tokens_seen": 2080571392
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037344032096288865,
      "loss": 1.3345,
      "theoretical_loss": 3.447615542906532,
      "tokens_seen": 2080636928
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000373420260782347,
      "loss": 1.5221,
      "theoretical_loss": 3.447606633215515,
      "tokens_seen": 2080702464
    },
    {
      "epoch": 0.63,
      "objective/train/advantage_avg": 0.4952525496482849,
      "objective/train/docs_used": 758504,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2737497091293335,
      "objective/train/original_loss": 1.273749589920044,
      "objective/train/theoretical_loss": 3.447597723883696,
      "objective/train/tokens_used": 2101228000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24547259509563446,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507732629776,
      "objective/train/weighted_lm_loss": 1.3382982015609741,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.447597723883696,
      "tokens_seen": 2080768000
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003734002006018054,
      "loss": 1.403,
      "theoretical_loss": 3.447597723883696,
      "tokens_seen": 2080768000
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003733801404212638,
      "loss": 1.3636,
      "theoretical_loss": 3.447588814911049,
      "tokens_seen": 2080833536
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037336008024072215,
      "loss": 1.4452,
      "theoretical_loss": 3.4475799062975483,
      "tokens_seen": 2080899072
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003733400200601805,
      "loss": 1.3361,
      "theoretical_loss": 3.447570998043168,
      "tokens_seen": 2080964608
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037331995987963893,
      "loss": 1.5567,
      "theoretical_loss": 3.4475620901478825,
      "tokens_seen": 2081030144
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003732998996990973,
      "loss": 1.355,
      "theoretical_loss": 3.4475531826116663,
      "tokens_seen": 2081095680
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037327983951855565,
      "loss": 1.3996,
      "theoretical_loss": 3.4475442754344927,
      "tokens_seen": 2081161216
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000373259779338014,
      "loss": 1.2742,
      "theoretical_loss": 3.447535368616337,
      "tokens_seen": 2081226752
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003732397191574724,
      "loss": 1.3126,
      "theoretical_loss": 3.447526462157173,
      "tokens_seen": 2081292288
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003732196589769308,
      "loss": 1.3331,
      "theoretical_loss": 3.447517556056974,
      "tokens_seen": 2081357824
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037319959879638916,
      "loss": 1.3929,
      "theoretical_loss": 3.447508650315716,
      "tokens_seen": 2081423360
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003731795386158475,
      "loss": 1.3987,
      "theoretical_loss": 3.4474997449333724,
      "tokens_seen": 2081488896
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003731594784353059,
      "loss": 1.2955,
      "theoretical_loss": 3.447490839909917,
      "tokens_seen": 2081554432
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003731394182547643,
      "loss": 1.4051,
      "theoretical_loss": 3.4474819352453245,
      "tokens_seen": 2081619968
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003731193580742227,
      "loss": 1.4043,
      "theoretical_loss": 3.4474730309395687,
      "tokens_seen": 2081685504
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003730992978936811,
      "loss": 1.3849,
      "theoretical_loss": 3.4474641269926245,
      "tokens_seen": 2081751040
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037307923771313944,
      "loss": 1.5836,
      "theoretical_loss": 3.447455223404466,
      "tokens_seen": 2081816576
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003730591775325978,
      "loss": 1.4596,
      "theoretical_loss": 3.447446320175067,
      "tokens_seen": 2081882112
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003730391173520562,
      "loss": 1.4441,
      "theoretical_loss": 3.4474374173044025,
      "tokens_seen": 2081947648
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003730190571715146,
      "loss": 1.4173,
      "theoretical_loss": 3.4474285147924455,
      "tokens_seen": 2082013184
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037299899699097294,
      "loss": 1.4012,
      "theoretical_loss": 3.4474196126391714,
      "tokens_seen": 2082078720
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003729789368104313,
      "loss": 1.4424,
      "theoretical_loss": 3.4474107108445544,
      "tokens_seen": 2082144256
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003729588766298897,
      "loss": 1.4136,
      "theoretical_loss": 3.4474018094085683,
      "tokens_seen": 2082209792
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003729388164493481,
      "loss": 1.2856,
      "theoretical_loss": 3.4473929083311874,
      "tokens_seen": 2082275328
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037291875626880645,
      "loss": 1.4171,
      "theoretical_loss": 3.4473840076123863,
      "tokens_seen": 2082340864
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003728986960882648,
      "loss": 1.5117,
      "theoretical_loss": 3.447375107252139,
      "tokens_seen": 2082406400
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037287863590772317,
      "loss": 1.4687,
      "theoretical_loss": 3.4473662072504196,
      "tokens_seen": 2082471936
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003728585757271816,
      "loss": 1.3832,
      "theoretical_loss": 3.447357307607203,
      "tokens_seen": 2082537472
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037283851554663995,
      "loss": 1.4159,
      "theoretical_loss": 3.447348408322463,
      "tokens_seen": 2082603008
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003728184553660983,
      "loss": 1.2314,
      "theoretical_loss": 3.4473395093961736,
      "tokens_seen": 2082668544
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003727983951855567,
      "loss": 1.5668,
      "theoretical_loss": 3.4473306108283097,
      "tokens_seen": 2082734080
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003727783350050151,
      "loss": 1.5702,
      "theoretical_loss": 3.4473217126188453,
      "tokens_seen": 2082799616
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037275827482447345,
      "loss": 1.2752,
      "theoretical_loss": 3.4473128147677548,
      "tokens_seen": 2082865152
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003727382146439318,
      "loss": 1.4204,
      "theoretical_loss": 3.447303917275012,
      "tokens_seen": 2082930688
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003727181544633902,
      "loss": 1.3551,
      "theoretical_loss": 3.4472950201405923,
      "tokens_seen": 2082996224
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037269809428284854,
      "loss": 1.367,
      "theoretical_loss": 3.447286123364469,
      "tokens_seen": 2083061760
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037267803410230696,
      "loss": 1.2879,
      "theoretical_loss": 3.4472772269466168,
      "tokens_seen": 2083127296
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003726579739217653,
      "loss": 1.363,
      "theoretical_loss": 3.44726833088701,
      "tokens_seen": 2083192832
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003726379137412237,
      "loss": 1.4036,
      "theoretical_loss": 3.4472594351856225,
      "tokens_seen": 2083258368
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037261785356068204,
      "loss": 1.4722,
      "theoretical_loss": 3.4472505398424294,
      "tokens_seen": 2083323904
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037259779338014046,
      "loss": 1.5124,
      "theoretical_loss": 3.447241644857404,
      "tokens_seen": 2083389440
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003725777331995988,
      "loss": 1.3563,
      "theoretical_loss": 3.447232750230522,
      "tokens_seen": 2083454976
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003725576730190572,
      "loss": 1.3436,
      "theoretical_loss": 3.447223855961756,
      "tokens_seen": 2083520512
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037253761283851554,
      "loss": 1.3702,
      "theoretical_loss": 3.4472149620510812,
      "tokens_seen": 2083586048
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037251755265797396,
      "loss": 1.3924,
      "theoretical_loss": 3.447206068498472,
      "tokens_seen": 2083651584
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003724974924774323,
      "loss": 1.3854,
      "theoretical_loss": 3.447197175303903,
      "tokens_seen": 2083717120
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003724774322968907,
      "loss": 1.4486,
      "theoretical_loss": 3.447188282467348,
      "tokens_seen": 2083782656
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037245737211634905,
      "loss": 1.43,
      "theoretical_loss": 3.447179389988781,
      "tokens_seen": 2083848192
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003724373119358074,
      "loss": 1.4354,
      "theoretical_loss": 3.4471704978681776,
      "tokens_seen": 2083913728
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003724172517552658,
      "loss": 1.4743,
      "theoretical_loss": 3.447161606105511,
      "tokens_seen": 2083979264
    },
    {
      "epoch": 0.63,
      "objective/train/advantage_avg": 0.4949369430541992,
      "objective/train/docs_used": 759560,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.354590654373169,
      "objective/train/original_loss": 1.354590654373169,
      "objective/train/theoretical_loss": 3.4471527147007555,
      "objective/train/tokens_used": 2104504800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24518893659114838,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507402420043945,
      "objective/train/weighted_lm_loss": 1.4232895374298096,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.4471527147007555,
      "tokens_seen": 2084044800
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003723971915747242,
      "loss": 1.555,
      "theoretical_loss": 3.4471527147007555,
      "tokens_seen": 2084044800
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037237713139418255,
      "loss": 1.3388,
      "theoretical_loss": 3.4471438236538865,
      "tokens_seen": 2084110336
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003723570712136409,
      "loss": 1.3439,
      "theoretical_loss": 3.447134932964877,
      "tokens_seen": 2084175872
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037233701103309933,
      "loss": 1.3516,
      "theoretical_loss": 3.4471260426337027,
      "tokens_seen": 2084241408
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003723169508525577,
      "loss": 1.5262,
      "theoretical_loss": 3.447117152660337,
      "tokens_seen": 2084306944
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037229689067201605,
      "loss": 1.3168,
      "theoretical_loss": 3.447108263044754,
      "tokens_seen": 2084372480
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003722768304914744,
      "loss": 1.4252,
      "theoretical_loss": 3.447099373786929,
      "tokens_seen": 2084438016
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003722567703109328,
      "loss": 1.3612,
      "theoretical_loss": 3.447090484886836,
      "tokens_seen": 2084503552
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003722367101303912,
      "loss": 1.555,
      "theoretical_loss": 3.447081596344449,
      "tokens_seen": 2084569088
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037221664994984956,
      "loss": 1.3208,
      "theoretical_loss": 3.4470727081597428,
      "tokens_seen": 2084634624
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003721965897693079,
      "loss": 1.5041,
      "theoretical_loss": 3.4470638203326915,
      "tokens_seen": 2084700160
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003721765295887663,
      "loss": 1.3407,
      "theoretical_loss": 3.44705493286327,
      "tokens_seen": 2084765696
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003721564694082247,
      "loss": 1.4814,
      "theoretical_loss": 3.447046045751451,
      "tokens_seen": 2084831232
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037213640922768306,
      "loss": 1.3165,
      "theoretical_loss": 3.447037158997211,
      "tokens_seen": 2084896768
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003721163490471414,
      "loss": 1.3246,
      "theoretical_loss": 3.4470282726005235,
      "tokens_seen": 2084962304
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003720962888665998,
      "loss": 1.3314,
      "theoretical_loss": 3.4470193865613625,
      "tokens_seen": 2085027840
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037207622868605815,
      "loss": 1.4937,
      "theoretical_loss": 3.447010500879703,
      "tokens_seen": 2085093376
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037205616850551656,
      "loss": 1.3151,
      "theoretical_loss": 3.4470016155555188,
      "tokens_seen": 2085158912
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003720361083249749,
      "loss": 1.3516,
      "theoretical_loss": 3.446992730588785,
      "tokens_seen": 2085224448
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003720160481444333,
      "loss": 1.3097,
      "theoretical_loss": 3.4469838459794753,
      "tokens_seen": 2085289984
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037199598796389165,
      "loss": 1.4241,
      "theoretical_loss": 3.4469749617275642,
      "tokens_seen": 2085355520
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037197592778335007,
      "loss": 1.382,
      "theoretical_loss": 3.446966077833027,
      "tokens_seen": 2085421056
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037195586760280843,
      "loss": 1.2593,
      "theoretical_loss": 3.4469571942958366,
      "tokens_seen": 2085486592
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003719358074222668,
      "loss": 1.3632,
      "theoretical_loss": 3.4469483111159684,
      "tokens_seen": 2085552128
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037191574724172515,
      "loss": 1.4789,
      "theoretical_loss": 3.4469394282933967,
      "tokens_seen": 2085617664
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003718956870611835,
      "loss": 1.4886,
      "theoretical_loss": 3.4469305458280957,
      "tokens_seen": 2085683200
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037187562688064193,
      "loss": 1.4277,
      "theoretical_loss": 3.4469216637200395,
      "tokens_seen": 2085748736
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003718555667001003,
      "loss": 1.4266,
      "theoretical_loss": 3.4469127819692034,
      "tokens_seen": 2085814272
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037183550651955866,
      "loss": 1.4726,
      "theoretical_loss": 3.4469039005755606,
      "tokens_seen": 2085879808
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000371815446339017,
      "loss": 1.3055,
      "theoretical_loss": 3.4468950195390864,
      "tokens_seen": 2085945344
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037179538615847544,
      "loss": 1.4373,
      "theoretical_loss": 3.4468861388597554,
      "tokens_seen": 2086010880
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003717753259779338,
      "loss": 1.5118,
      "theoretical_loss": 3.4468772585375413,
      "tokens_seen": 2086076416
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037175526579739216,
      "loss": 1.4551,
      "theoretical_loss": 3.446868378572419,
      "tokens_seen": 2086141952
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003717352056168505,
      "loss": 1.5045,
      "theoretical_loss": 3.4468594989643626,
      "tokens_seen": 2086207488
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003717151454363089,
      "loss": 1.4059,
      "theoretical_loss": 3.446850619713347,
      "tokens_seen": 2086273024
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003716950852557673,
      "loss": 1.4024,
      "theoretical_loss": 3.4468417408193455,
      "tokens_seen": 2086338560
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037167502507522566,
      "loss": 1.421,
      "theoretical_loss": 3.446832862282334,
      "tokens_seen": 2086404096
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000371654964894684,
      "loss": 1.3873,
      "theoretical_loss": 3.446823984102286,
      "tokens_seen": 2086469632
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003716349047141424,
      "loss": 1.3137,
      "theoretical_loss": 3.4468151062791765,
      "tokens_seen": 2086535168
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037161484453360086,
      "loss": 1.3628,
      "theoretical_loss": 3.4468062288129797,
      "tokens_seen": 2086600704
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003715947843530592,
      "loss": 1.3883,
      "theoretical_loss": 3.44679735170367,
      "tokens_seen": 2086666240
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003715747241725176,
      "loss": 1.4915,
      "theoretical_loss": 3.446788474951221,
      "tokens_seen": 2086731776
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037155466399197595,
      "loss": 1.3281,
      "theoretical_loss": 3.446779598555609,
      "tokens_seen": 2086797312
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037153460381143436,
      "loss": 1.3261,
      "theoretical_loss": 3.446770722516807,
      "tokens_seen": 2086862848
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003715145436308927,
      "loss": 1.3088,
      "theoretical_loss": 3.4467618468347903,
      "tokens_seen": 2086928384
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003714944834503511,
      "loss": 1.3783,
      "theoretical_loss": 3.4467529715095324,
      "tokens_seen": 2086993920
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037147442326980945,
      "loss": 1.41,
      "theoretical_loss": 3.4467440965410088,
      "tokens_seen": 2087059456
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003714543630892678,
      "loss": 1.3341,
      "theoretical_loss": 3.446735221929193,
      "tokens_seen": 2087124992
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037143430290872623,
      "loss": 1.4123,
      "theoretical_loss": 3.44672634767406,
      "tokens_seen": 2087190528
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003714142427281846,
      "loss": 1.304,
      "theoretical_loss": 3.4467174737755846,
      "tokens_seen": 2087256064
    },
    {
      "epoch": 0.63,
      "objective/train/advantage_avg": 0.49172383546829224,
      "objective/train/docs_used": 760861,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2745579481124878,
      "objective/train/original_loss": 1.2745580673217773,
      "objective/train/theoretical_loss": 3.446708600233741,
      "objective/train/tokens_used": 2107781600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2425210177898407,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504053831100464,
      "objective/train/weighted_lm_loss": 1.3387188911437988,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.446708600233741,
      "tokens_seen": 2087321600
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037139418254764295,
      "loss": 1.3772,
      "theoretical_loss": 3.446708600233741,
      "tokens_seen": 2087321600
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003713741223671013,
      "loss": 1.3863,
      "theoretical_loss": 3.446699727048503,
      "tokens_seen": 2087387136
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037135406218655973,
      "loss": 1.4888,
      "theoretical_loss": 3.4466908542198453,
      "tokens_seen": 2087452672
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003713340020060181,
      "loss": 1.3554,
      "theoretical_loss": 3.4466819817477434,
      "tokens_seen": 2087518208
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037131394182547645,
      "loss": 1.4664,
      "theoretical_loss": 3.4466731096321706,
      "tokens_seen": 2087583744
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003712938816449348,
      "loss": 1.3595,
      "theoretical_loss": 3.4466642378731023,
      "tokens_seen": 2087649280
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003712738214643932,
      "loss": 1.4592,
      "theoretical_loss": 3.4466553664705124,
      "tokens_seen": 2087714816
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003712537612838516,
      "loss": 1.3674,
      "theoretical_loss": 3.446646495424375,
      "tokens_seen": 2087780352
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037123370110330996,
      "loss": 1.4481,
      "theoretical_loss": 3.4466376247346657,
      "tokens_seen": 2087845888
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003712136409227683,
      "loss": 1.4156,
      "theoretical_loss": 3.4466287544013583,
      "tokens_seen": 2087911424
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003711935807422267,
      "loss": 1.3151,
      "theoretical_loss": 3.446619884424427,
      "tokens_seen": 2087976960
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003711735205616851,
      "loss": 1.3643,
      "theoretical_loss": 3.446611014803847,
      "tokens_seen": 2088042496
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037115346038114346,
      "loss": 1.4431,
      "theoretical_loss": 3.4466021455395923,
      "tokens_seen": 2088108032
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003711334002006018,
      "loss": 1.4115,
      "theoretical_loss": 3.4465932766316376,
      "tokens_seen": 2088173568
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003711133400200602,
      "loss": 1.4059,
      "theoretical_loss": 3.446584408079958,
      "tokens_seen": 2088239104
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037109327983951855,
      "loss": 1.4153,
      "theoretical_loss": 3.4465755398845266,
      "tokens_seen": 2088304640
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037107321965897696,
      "loss": 1.4723,
      "theoretical_loss": 3.446566672045319,
      "tokens_seen": 2088370176
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003710531594784353,
      "loss": 1.4404,
      "theoretical_loss": 3.4465578045623095,
      "tokens_seen": 2088435712
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003710330992978937,
      "loss": 1.3954,
      "theoretical_loss": 3.4465489374354727,
      "tokens_seen": 2088501248
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037101303911735205,
      "loss": 1.4419,
      "theoretical_loss": 3.4465400706647826,
      "tokens_seen": 2088566784
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037099297893681047,
      "loss": 1.5365,
      "theoretical_loss": 3.4465312042502143,
      "tokens_seen": 2088632320
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037097291875626883,
      "loss": 1.3371,
      "theoretical_loss": 3.446522338191742,
      "tokens_seen": 2088697856
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003709528585757272,
      "loss": 1.3704,
      "theoretical_loss": 3.4465134724893405,
      "tokens_seen": 2088763392
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037093279839518555,
      "loss": 1.4351,
      "theoretical_loss": 3.446504607142984,
      "tokens_seen": 2088828928
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003709127382146439,
      "loss": 1.445,
      "theoretical_loss": 3.446495742152647,
      "tokens_seen": 2088894464
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037089267803410233,
      "loss": 1.3807,
      "theoretical_loss": 3.4464868775183044,
      "tokens_seen": 2088960000
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003708726178535607,
      "loss": 1.3773,
      "theoretical_loss": 3.446478013239931,
      "tokens_seen": 2089025536
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037085255767301906,
      "loss": 1.4237,
      "theoretical_loss": 3.4464691493175,
      "tokens_seen": 2089091072
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003708324974924774,
      "loss": 1.4124,
      "theoretical_loss": 3.4464602857509874,
      "tokens_seen": 2089156608
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037081243731193584,
      "loss": 1.4989,
      "theoretical_loss": 3.446451422540367,
      "tokens_seen": 2089222144
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003707923771313942,
      "loss": 1.3968,
      "theoretical_loss": 3.4464425596856136,
      "tokens_seen": 2089287680
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037077231695085256,
      "loss": 1.5361,
      "theoretical_loss": 3.446433697186702,
      "tokens_seen": 2089353216
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003707522567703109,
      "loss": 1.4084,
      "theoretical_loss": 3.446424835043606,
      "tokens_seen": 2089418752
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003707321965897693,
      "loss": 1.4352,
      "theoretical_loss": 3.4464159732563004,
      "tokens_seen": 2089484288
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003707121364092277,
      "loss": 1.3709,
      "theoretical_loss": 3.446407111824761,
      "tokens_seen": 2089549824
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037069207622868606,
      "loss": 1.5008,
      "theoretical_loss": 3.44639825074896,
      "tokens_seen": 2089615360
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003706720160481444,
      "loss": 1.449,
      "theoretical_loss": 3.446389390028874,
      "tokens_seen": 2089680896
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003706519558676028,
      "loss": 1.3432,
      "theoretical_loss": 3.4463805296644767,
      "tokens_seen": 2089746432
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003706318956870612,
      "loss": 1.3621,
      "theoretical_loss": 3.4463716696557425,
      "tokens_seen": 2089811968
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037061183550651957,
      "loss": 1.4132,
      "theoretical_loss": 3.446362810002647,
      "tokens_seen": 2089877504
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037059177532597793,
      "loss": 1.4114,
      "theoretical_loss": 3.4463539507051633,
      "tokens_seen": 2089943040
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003705717151454363,
      "loss": 1.3677,
      "theoretical_loss": 3.446345091763267,
      "tokens_seen": 2090008576
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037055165496489465,
      "loss": 1.3982,
      "theoretical_loss": 3.4463362331769325,
      "tokens_seen": 2090074112
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037053159478435307,
      "loss": 1.3774,
      "theoretical_loss": 3.446327374946134,
      "tokens_seen": 2090139648
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037051153460381143,
      "loss": 1.4309,
      "theoretical_loss": 3.4463185170708464,
      "tokens_seen": 2090205184
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003704914744232698,
      "loss": 1.3907,
      "theoretical_loss": 3.4463096595510443,
      "tokens_seen": 2090270720
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037047141424272816,
      "loss": 1.3563,
      "theoretical_loss": 3.4463008023867028,
      "tokens_seen": 2090336256
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003704513540621866,
      "loss": 1.4541,
      "theoretical_loss": 3.446291945577795,
      "tokens_seen": 2090401792
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037043129388164494,
      "loss": 1.4389,
      "theoretical_loss": 3.446283089124297,
      "tokens_seen": 2090467328
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003704112337011033,
      "loss": 1.3981,
      "theoretical_loss": 3.4462742330261826,
      "tokens_seen": 2090532864
    },
    {
      "epoch": 0.63,
      "objective/train/advantage_avg": 0.4910328984260559,
      "objective/train/docs_used": 762048,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3989617824554443,
      "objective/train/original_loss": 1.3989615440368652,
      "objective/train/theoretical_loss": 3.4462653772834266,
      "objective/train/tokens_used": 2111058400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24189747869968414,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050333023071289,
      "objective/train/weighted_lm_loss": 1.4693373441696167,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4462653772834266,
      "tokens_seen": 2090598400
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037039117352056166,
      "loss": 1.4779,
      "theoretical_loss": 3.4462653772834266,
      "tokens_seen": 2090598400
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003703711133400201,
      "loss": 1.3711,
      "theoretical_loss": 3.4462565218960037,
      "tokens_seen": 2090663936
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037035105315947844,
      "loss": 1.4156,
      "theoretical_loss": 3.4462476668638886,
      "tokens_seen": 2090729472
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003703309929789368,
      "loss": 1.3568,
      "theoretical_loss": 3.4462388121870555,
      "tokens_seen": 2090795008
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037031093279839516,
      "loss": 1.4956,
      "theoretical_loss": 3.446229957865479,
      "tokens_seen": 2090860544
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003702908726178535,
      "loss": 1.4168,
      "theoretical_loss": 3.4462211038991346,
      "tokens_seen": 2090926080
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037027081243731194,
      "loss": 1.286,
      "theoretical_loss": 3.4462122502879957,
      "tokens_seen": 2090991616
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003702507522567703,
      "loss": 1.5964,
      "theoretical_loss": 3.446203397032038,
      "tokens_seen": 2091057152
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037023069207622867,
      "loss": 1.4151,
      "theoretical_loss": 3.4461945441312354,
      "tokens_seen": 2091122688
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037021063189568703,
      "loss": 1.435,
      "theoretical_loss": 3.4461856915855624,
      "tokens_seen": 2091188224
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037019057171514544,
      "loss": 1.4025,
      "theoretical_loss": 3.4461768393949943,
      "tokens_seen": 2091253760
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003701705115346038,
      "loss": 1.367,
      "theoretical_loss": 3.4461679875595053,
      "tokens_seen": 2091319296
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037015045135406217,
      "loss": 1.3968,
      "theoretical_loss": 3.4461591360790704,
      "tokens_seen": 2091384832
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037013039117352053,
      "loss": 1.3626,
      "theoretical_loss": 3.446150284953663,
      "tokens_seen": 2091450368
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037011033099297895,
      "loss": 1.3172,
      "theoretical_loss": 3.44614143418326,
      "tokens_seen": 2091515904
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037009027081243736,
      "loss": 1.3685,
      "theoretical_loss": 3.4461325837678336,
      "tokens_seen": 2091581440
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003700702106318957,
      "loss": 1.3958,
      "theoretical_loss": 3.44612373370736,
      "tokens_seen": 2091646976
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003700501504513541,
      "loss": 1.4868,
      "theoretical_loss": 3.4461148840018136,
      "tokens_seen": 2091712512
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037003009027081245,
      "loss": 1.3827,
      "theoretical_loss": 3.4461060346511685,
      "tokens_seen": 2091778048
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00037001003009027087,
      "loss": 1.3971,
      "theoretical_loss": 3.4460971856554,
      "tokens_seen": 2091843584
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036998996990972923,
      "loss": 1.42,
      "theoretical_loss": 3.446088337014482,
      "tokens_seen": 2091909120
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003699699097291876,
      "loss": 1.3672,
      "theoretical_loss": 3.4460794887283903,
      "tokens_seen": 2091974656
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036994984954864595,
      "loss": 1.4414,
      "theoretical_loss": 3.4460706407970982,
      "tokens_seen": 2092040192
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003699297893681043,
      "loss": 1.4799,
      "theoretical_loss": 3.446061793220581,
      "tokens_seen": 2092105728
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036990972918756273,
      "loss": 1.3623,
      "theoretical_loss": 3.446052945998814,
      "tokens_seen": 2092171264
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003698896690070211,
      "loss": 1.4023,
      "theoretical_loss": 3.446044099131771,
      "tokens_seen": 2092236800
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036986960882647946,
      "loss": 1.4859,
      "theoretical_loss": 3.446035252619427,
      "tokens_seen": 2092302336
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003698495486459378,
      "loss": 1.2846,
      "theoretical_loss": 3.446026406461756,
      "tokens_seen": 2092367872
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036982948846539624,
      "loss": 1.4627,
      "theoretical_loss": 3.446017560658734,
      "tokens_seen": 2092433408
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003698094282848546,
      "loss": 1.4231,
      "theoretical_loss": 3.4460087152103345,
      "tokens_seen": 2092498944
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036978936810431296,
      "loss": 1.4217,
      "theoretical_loss": 3.4459998701165326,
      "tokens_seen": 2092564480
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003697693079237713,
      "loss": 1.3835,
      "theoretical_loss": 3.445991025377303,
      "tokens_seen": 2092630016
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003697492477432297,
      "loss": 1.351,
      "theoretical_loss": 3.4459821809926208,
      "tokens_seen": 2092695552
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003697291875626881,
      "loss": 1.3854,
      "theoretical_loss": 3.4459733369624597,
      "tokens_seen": 2092761088
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036970912738214646,
      "loss": 1.3289,
      "theoretical_loss": 3.4459644932867954,
      "tokens_seen": 2092826624
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003696890672016048,
      "loss": 1.3819,
      "theoretical_loss": 3.445955649965602,
      "tokens_seen": 2092892160
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003696690070210632,
      "loss": 1.4267,
      "theoretical_loss": 3.4459468069988537,
      "tokens_seen": 2092957696
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003696489468405216,
      "loss": 1.327,
      "theoretical_loss": 3.4459379643865264,
      "tokens_seen": 2093023232
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036962888665997997,
      "loss": 1.3882,
      "theoretical_loss": 3.445929122128594,
      "tokens_seen": 2093088768
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036960882647943833,
      "loss": 1.4086,
      "theoretical_loss": 3.4459202802250317,
      "tokens_seen": 2093154304
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003695887662988967,
      "loss": 1.3897,
      "theoretical_loss": 3.445911438675814,
      "tokens_seen": 2093219840
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036956870611835505,
      "loss": 1.3094,
      "theoretical_loss": 3.445902597480915,
      "tokens_seen": 2093285376
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036954864593781347,
      "loss": 1.2518,
      "theoretical_loss": 3.4458937566403103,
      "tokens_seen": 2093350912
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036952858575727183,
      "loss": 1.3819,
      "theoretical_loss": 3.4458849161539744,
      "tokens_seen": 2093416448
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003695085255767302,
      "loss": 1.334,
      "theoretical_loss": 3.445876076021882,
      "tokens_seen": 2093481984
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036948846539618856,
      "loss": 1.3296,
      "theoretical_loss": 3.4458672362440073,
      "tokens_seen": 2093547520
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.000369468405215647,
      "loss": 1.4845,
      "theoretical_loss": 3.4458583968203254,
      "tokens_seen": 2093613056
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036944834503510534,
      "loss": 1.4596,
      "theoretical_loss": 3.445849557750811,
      "tokens_seen": 2093678592
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003694282848545637,
      "loss": 1.398,
      "theoretical_loss": 3.4458407190354388,
      "tokens_seen": 2093744128
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036940822467402206,
      "loss": 1.3544,
      "theoretical_loss": 3.445831880674184,
      "tokens_seen": 2093809664
    },
    {
      "epoch": 0.63,
      "objective/train/advantage_avg": 0.4930795133113861,
      "objective/train/docs_used": 763180,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.343479037284851,
      "objective/train/original_loss": 1.3434789180755615,
      "objective/train/theoretical_loss": 3.4458230426670204,
      "objective/train/tokens_used": 2114335200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24340446293354034,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505452156066895,
      "objective/train/weighted_lm_loss": 1.4113116264343262,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4458230426670204,
      "tokens_seen": 2093875200
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003693881644934805,
      "loss": 1.3813,
      "theoretical_loss": 3.4458230426670204,
      "tokens_seen": 2093875200
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036936810431293884,
      "loss": 1.3507,
      "theoretical_loss": 3.4458142050139235,
      "tokens_seen": 2093940736
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003693480441323972,
      "loss": 1.4185,
      "theoretical_loss": 3.445805367714868,
      "tokens_seen": 2094006272
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036932798395185556,
      "loss": 1.3563,
      "theoretical_loss": 3.4457965307698277,
      "tokens_seen": 2094071808
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003693079237713139,
      "loss": 1.4013,
      "theoretical_loss": 3.4457876941787786,
      "tokens_seen": 2094137344
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036928786359077234,
      "loss": 1.3965,
      "theoretical_loss": 3.445778857941695,
      "tokens_seen": 2094202880
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003692678034102307,
      "loss": 1.3376,
      "theoretical_loss": 3.445770022058551,
      "tokens_seen": 2094268416
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036924774322968907,
      "loss": 1.432,
      "theoretical_loss": 3.4457611865293223,
      "tokens_seen": 2094333952
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036922768304914743,
      "loss": 1.3476,
      "theoretical_loss": 3.445752351353983,
      "tokens_seen": 2094399488
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036920762286860585,
      "loss": 1.4449,
      "theoretical_loss": 3.4457435165325085,
      "tokens_seen": 2094465024
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003691875626880642,
      "loss": 1.3987,
      "theoretical_loss": 3.445734682064873,
      "tokens_seen": 2094530560
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036916750250752257,
      "loss": 1.3989,
      "theoretical_loss": 3.445725847951051,
      "tokens_seen": 2094596096
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036914744232698093,
      "loss": 1.4166,
      "theoretical_loss": 3.445717014191018,
      "tokens_seen": 2094661632
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003691273821464393,
      "loss": 1.3562,
      "theoretical_loss": 3.445708180784748,
      "tokens_seen": 2094727168
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003691073219658977,
      "loss": 1.3156,
      "theoretical_loss": 3.4456993477322166,
      "tokens_seen": 2094792704
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003690872617853561,
      "loss": 1.3756,
      "theoretical_loss": 3.445690515033398,
      "tokens_seen": 2094858240
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036906720160481444,
      "loss": 1.4432,
      "theoretical_loss": 3.4456816826882672,
      "tokens_seen": 2094923776
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003690471414242728,
      "loss": 1.301,
      "theoretical_loss": 3.445672850696799,
      "tokens_seen": 2094989312
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003690270812437312,
      "loss": 1.406,
      "theoretical_loss": 3.4456640190589676,
      "tokens_seen": 2095054848
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003690070210631896,
      "loss": 1.421,
      "theoretical_loss": 3.445655187774749,
      "tokens_seen": 2095120384
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036898696088264794,
      "loss": 1.4157,
      "theoretical_loss": 3.4456463568441165,
      "tokens_seen": 2095185920
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003689669007021063,
      "loss": 1.4496,
      "theoretical_loss": 3.445637526267046,
      "tokens_seen": 2095251456
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036894684052156466,
      "loss": 1.3908,
      "theoretical_loss": 3.445628696043512,
      "tokens_seen": 2095316992
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0003689267803410231,
      "loss": 1.3875,
      "theoretical_loss": 3.445619866173489,
      "tokens_seen": 2095382528
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.00036890672016048144,
      "loss": 1.3773,
      "theoretical_loss": 3.4456110366569517,
      "tokens_seen": 2095448064
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003688866599799398,
      "loss": 1.3866,
      "theoretical_loss": 3.4456022074938755,
      "tokens_seen": 2095513600
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036886659979939817,
      "loss": 1.3422,
      "theoretical_loss": 3.4455933786842348,
      "tokens_seen": 2095579136
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003688465396188566,
      "loss": 1.5024,
      "theoretical_loss": 3.4455845502280047,
      "tokens_seen": 2095644672
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036882647943831494,
      "loss": 1.4711,
      "theoretical_loss": 3.4455757221251595,
      "tokens_seen": 2095710208
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003688064192577733,
      "loss": 1.4112,
      "theoretical_loss": 3.4455668943756748,
      "tokens_seen": 2095775744
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036878635907723167,
      "loss": 1.4177,
      "theoretical_loss": 3.4455580669795243,
      "tokens_seen": 2095841280
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036876629889669003,
      "loss": 1.3699,
      "theoretical_loss": 3.4455492399366836,
      "tokens_seen": 2095906816
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036874623871614845,
      "loss": 1.4221,
      "theoretical_loss": 3.4455404132471275,
      "tokens_seen": 2095972352
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003687261785356068,
      "loss": 1.3636,
      "theoretical_loss": 3.44553158691083,
      "tokens_seen": 2096037888
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036870611835506517,
      "loss": 1.3673,
      "theoretical_loss": 3.4455227609277674,
      "tokens_seen": 2096103424
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036868605817452353,
      "loss": 1.4592,
      "theoretical_loss": 3.445513935297913,
      "tokens_seen": 2096168960
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036866599799398195,
      "loss": 1.4166,
      "theoretical_loss": 3.4455051100212426,
      "tokens_seen": 2096234496
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003686459378134403,
      "loss": 1.4452,
      "theoretical_loss": 3.4454962850977306,
      "tokens_seen": 2096300032
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003686258776328987,
      "loss": 1.5525,
      "theoretical_loss": 3.445487460527352,
      "tokens_seen": 2096365568
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003686058174523571,
      "loss": 1.3982,
      "theoretical_loss": 3.4454786363100816,
      "tokens_seen": 2096431104
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036858575727181545,
      "loss": 1.4376,
      "theoretical_loss": 3.4454698124458942,
      "tokens_seen": 2096496640
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036856569709127387,
      "loss": 1.4986,
      "theoretical_loss": 3.445460988934765,
      "tokens_seen": 2096562176
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036854563691073223,
      "loss": 1.4178,
      "theoretical_loss": 3.4454521657766675,
      "tokens_seen": 2096627712
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003685255767301906,
      "loss": 1.4786,
      "theoretical_loss": 3.4454433429715783,
      "tokens_seen": 2096693248
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036850551654964896,
      "loss": 1.4451,
      "theoretical_loss": 3.445434520519471,
      "tokens_seen": 2096758784
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003684854563691074,
      "loss": 1.399,
      "theoretical_loss": 3.4454256984203213,
      "tokens_seen": 2096824320
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036846539618856574,
      "loss": 1.4593,
      "theoretical_loss": 3.445416876674104,
      "tokens_seen": 2096889856
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003684453360080241,
      "loss": 1.4081,
      "theoretical_loss": 3.445408055280793,
      "tokens_seen": 2096955392
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036842527582748246,
      "loss": 1.3832,
      "theoretical_loss": 3.445399234240363,
      "tokens_seen": 2097020928
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003684052156469409,
      "loss": 1.3486,
      "theoretical_loss": 3.445390413552791,
      "tokens_seen": 2097086464
    },
    {
      "epoch": 0.64,
      "objective/train/advantage_avg": 0.49380001425743103,
      "objective/train/docs_used": 764413,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4660282135009766,
      "objective/train/original_loss": 1.4660282135009766,
      "objective/train/theoretical_loss": 3.44538159321805,
      "objective/train/tokens_used": 2117612000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24406661093235016,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506207942962646,
      "objective/train/weighted_lm_loss": 1.540139079093933,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 3.44538159321805,
      "tokens_seen": 2097152000
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036838515546639924,
      "loss": 1.4676,
      "theoretical_loss": 3.44538159321805,
      "tokens_seen": 2097152000
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003683650952858576,
      "loss": 1.3738,
      "theoretical_loss": 3.445372773236115,
      "tokens_seen": 2097217536
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036834503510531596,
      "loss": 1.3002,
      "theoretical_loss": 3.445363953606962,
      "tokens_seen": 2097283072
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003683249749247743,
      "loss": 1.3971,
      "theoretical_loss": 3.445355134330564,
      "tokens_seen": 2097348608
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036830491474423274,
      "loss": 1.5433,
      "theoretical_loss": 3.4453463154068977,
      "tokens_seen": 2097414144
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003682848545636911,
      "loss": 1.3629,
      "theoretical_loss": 3.445337496835937,
      "tokens_seen": 2097479680
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036826479438314947,
      "loss": 1.4013,
      "theoretical_loss": 3.445328678617657,
      "tokens_seen": 2097545216
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036824473420260783,
      "loss": 1.4572,
      "theoretical_loss": 3.4453198607520323,
      "tokens_seen": 2097610752
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036822467402206625,
      "loss": 1.5118,
      "theoretical_loss": 3.4453110432390384,
      "tokens_seen": 2097676288
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003682046138415246,
      "loss": 1.3894,
      "theoretical_loss": 3.44530222607865,
      "tokens_seen": 2097741824
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036818455366098297,
      "loss": 1.4487,
      "theoretical_loss": 3.445293409270841,
      "tokens_seen": 2097807360
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036816449348044133,
      "loss": 1.4131,
      "theoretical_loss": 3.4452845928155877,
      "tokens_seen": 2097872896
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003681444332998997,
      "loss": 1.4108,
      "theoretical_loss": 3.4452757767128643,
      "tokens_seen": 2097938432
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003681243731193581,
      "loss": 1.3742,
      "theoretical_loss": 3.4452669609626456,
      "tokens_seen": 2098003968
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003681043129388165,
      "loss": 1.3385,
      "theoretical_loss": 3.4452581455649067,
      "tokens_seen": 2098069504
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036808425275827484,
      "loss": 1.2638,
      "theoretical_loss": 3.445249330519623,
      "tokens_seen": 2098135040
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003680641925777332,
      "loss": 1.3597,
      "theoretical_loss": 3.445240515826768,
      "tokens_seen": 2098200576
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003680441323971916,
      "loss": 1.4984,
      "theoretical_loss": 3.4452317014863176,
      "tokens_seen": 2098266112
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036802407221665,
      "loss": 1.4434,
      "theoretical_loss": 3.4452228874982467,
      "tokens_seen": 2098331648
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036800401203610834,
      "loss": 1.4673,
      "theoretical_loss": 3.4452140738625303,
      "tokens_seen": 2098397184
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003679839518555667,
      "loss": 1.3763,
      "theoretical_loss": 3.4452052605791432,
      "tokens_seen": 2098462720
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036796389167502506,
      "loss": 1.4336,
      "theoretical_loss": 3.44519644764806,
      "tokens_seen": 2098528256
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003679438314944835,
      "loss": 1.3186,
      "theoretical_loss": 3.4451876350692556,
      "tokens_seen": 2098593792
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036792377131394184,
      "loss": 1.2914,
      "theoretical_loss": 3.4451788228427054,
      "tokens_seen": 2098659328
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003679037111334002,
      "loss": 1.5384,
      "theoretical_loss": 3.4451700109683836,
      "tokens_seen": 2098724864
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036788365095285857,
      "loss": 1.4501,
      "theoretical_loss": 3.4451611994462663,
      "tokens_seen": 2098790400
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000367863590772317,
      "loss": 1.4089,
      "theoretical_loss": 3.4451523882763273,
      "tokens_seen": 2098855936
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036784353059177535,
      "loss": 1.375,
      "theoretical_loss": 3.445143577458542,
      "tokens_seen": 2098921472
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003678234704112337,
      "loss": 1.431,
      "theoretical_loss": 3.4451347669928856,
      "tokens_seen": 2098987008
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036780341023069207,
      "loss": 1.3768,
      "theoretical_loss": 3.445125956879332,
      "tokens_seen": 2099052544
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036778335005015043,
      "loss": 1.4638,
      "theoretical_loss": 3.445117147117857,
      "tokens_seen": 2099118080
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036776328986960885,
      "loss": 1.5006,
      "theoretical_loss": 3.4451083377084357,
      "tokens_seen": 2099183616
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003677432296890672,
      "loss": 1.3832,
      "theoretical_loss": 3.4450995286510424,
      "tokens_seen": 2099249152
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036772316950852557,
      "loss": 1.447,
      "theoretical_loss": 3.445090719945653,
      "tokens_seen": 2099314688
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036770310932798393,
      "loss": 1.4478,
      "theoretical_loss": 3.445081911592241,
      "tokens_seen": 2099380224
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036768304914744235,
      "loss": 1.3686,
      "theoretical_loss": 3.4450731035907824,
      "tokens_seen": 2099445760
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003676629889669007,
      "loss": 1.4419,
      "theoretical_loss": 3.445064295941252,
      "tokens_seen": 2099511296
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003676429287863591,
      "loss": 1.4203,
      "theoretical_loss": 3.445055488643624,
      "tokens_seen": 2099576832
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036762286860581744,
      "loss": 1.3743,
      "theoretical_loss": 3.445046681697875,
      "tokens_seen": 2099642368
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003676028084252758,
      "loss": 1.3058,
      "theoretical_loss": 3.4450378751039787,
      "tokens_seen": 2099707904
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003675827482447342,
      "loss": 1.3966,
      "theoretical_loss": 3.4450290688619103,
      "tokens_seen": 2099773440
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003675626880641926,
      "loss": 1.3434,
      "theoretical_loss": 3.4450202629716444,
      "tokens_seen": 2099838976
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036754262788365094,
      "loss": 1.2192,
      "theoretical_loss": 3.4450114574331567,
      "tokens_seen": 2099904512
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003675225677031093,
      "loss": 1.3277,
      "theoretical_loss": 3.445002652246422,
      "tokens_seen": 2099970048
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003675025075225677,
      "loss": 1.3931,
      "theoretical_loss": 3.444993847411415,
      "tokens_seen": 2100035584
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003674824473420261,
      "loss": 1.4038,
      "theoretical_loss": 3.4449850429281104,
      "tokens_seen": 2100101120
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036746238716148444,
      "loss": 1.4864,
      "theoretical_loss": 3.444976238796484,
      "tokens_seen": 2100166656
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003674423269809428,
      "loss": 1.4964,
      "theoretical_loss": 3.44496743501651,
      "tokens_seen": 2100232192
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036742226680040117,
      "loss": 1.3265,
      "theoretical_loss": 3.4449586315881637,
      "tokens_seen": 2100297728
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003674022066198596,
      "loss": 1.4235,
      "theoretical_loss": 3.4449498285114206,
      "tokens_seen": 2100363264
    },
    {
      "epoch": 0.64,
      "objective/train/advantage_avg": 0.4942744970321655,
      "objective/train/docs_used": 765555,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.195361614227295,
      "objective/train/original_loss": 1.1953614950180054,
      "objective/train/theoretical_loss": 3.4449410257862545,
      "objective/train/tokens_used": 2120888800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24456700682640076,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050670862197876,
      "objective/train/weighted_lm_loss": 1.2558636665344238,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.4449410257862545,
      "tokens_seen": 2100428800
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036738214643931795,
      "loss": 1.3353,
      "theoretical_loss": 3.4449410257862545,
      "tokens_seen": 2100428800
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003673620862587763,
      "loss": 1.2327,
      "theoretical_loss": 3.4449322234126414,
      "tokens_seen": 2100494336
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036734202607823467,
      "loss": 1.4036,
      "theoretical_loss": 3.4449234213905564,
      "tokens_seen": 2100559872
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003673219658976931,
      "loss": 1.3751,
      "theoretical_loss": 3.4449146197199734,
      "tokens_seen": 2100625408
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036730190571715145,
      "loss": 1.3491,
      "theoretical_loss": 3.4449058184008683,
      "tokens_seen": 2100690944
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003672818455366098,
      "loss": 1.3274,
      "theoretical_loss": 3.444897017433216,
      "tokens_seen": 2100756480
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003672617853560682,
      "loss": 1.3431,
      "theoretical_loss": 3.4448882168169908,
      "tokens_seen": 2100822016
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036724172517552654,
      "loss": 1.3036,
      "theoretical_loss": 3.444879416552169,
      "tokens_seen": 2100887552
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036722166499498495,
      "loss": 1.3368,
      "theoretical_loss": 3.4448706166387244,
      "tokens_seen": 2100953088
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003672016048144433,
      "loss": 1.3919,
      "theoretical_loss": 3.4448618170766325,
      "tokens_seen": 2101018624
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003671815446339017,
      "loss": 1.3794,
      "theoretical_loss": 3.444853017865869,
      "tokens_seen": 2101084160
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036716148445336004,
      "loss": 1.3562,
      "theoretical_loss": 3.4448442190064075,
      "tokens_seen": 2101149696
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036714142427281846,
      "loss": 1.3417,
      "theoretical_loss": 3.4448354204982237,
      "tokens_seen": 2101215232
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003671213640922768,
      "loss": 1.3189,
      "theoretical_loss": 3.4448266223412927,
      "tokens_seen": 2101280768
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036710130391173524,
      "loss": 1.3373,
      "theoretical_loss": 3.4448178245355896,
      "tokens_seen": 2101346304
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003670812437311936,
      "loss": 1.3707,
      "theoretical_loss": 3.4448090270810896,
      "tokens_seen": 2101411840
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000367061183550652,
      "loss": 1.2713,
      "theoretical_loss": 3.444800229977767,
      "tokens_seen": 2101477376
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003670411233701104,
      "loss": 1.4314,
      "theoretical_loss": 3.444791433225597,
      "tokens_seen": 2101542912
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036702106318956874,
      "loss": 1.3787,
      "theoretical_loss": 3.4447826368245558,
      "tokens_seen": 2101608448
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003670010030090271,
      "loss": 1.4371,
      "theoretical_loss": 3.4447738407746167,
      "tokens_seen": 2101673984
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036698094282848546,
      "loss": 1.3629,
      "theoretical_loss": 3.444765045075756,
      "tokens_seen": 2101739520
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003669608826479439,
      "loss": 1.3061,
      "theoretical_loss": 3.4447562497279485,
      "tokens_seen": 2101805056
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036694082246740224,
      "loss": 1.3865,
      "theoretical_loss": 3.4447474547311683,
      "tokens_seen": 2101870592
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003669207622868606,
      "loss": 1.2737,
      "theoretical_loss": 3.4447386600853918,
      "tokens_seen": 2101936128
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036690070210631897,
      "loss": 1.3613,
      "theoretical_loss": 3.444729865790593,
      "tokens_seen": 2102001664
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003668806419257774,
      "loss": 1.525,
      "theoretical_loss": 3.4447210718467476,
      "tokens_seen": 2102067200
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036686058174523575,
      "loss": 1.3916,
      "theoretical_loss": 3.4447122782538306,
      "tokens_seen": 2102132736
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003668405215646941,
      "loss": 1.3876,
      "theoretical_loss": 3.444703485011817,
      "tokens_seen": 2102198272
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036682046138415247,
      "loss": 1.3495,
      "theoretical_loss": 3.4446946921206814,
      "tokens_seen": 2102263808
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036680040120361083,
      "loss": 1.3762,
      "theoretical_loss": 3.4446858995803993,
      "tokens_seen": 2102329344
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036678034102306925,
      "loss": 1.3257,
      "theoretical_loss": 3.444677107390946,
      "tokens_seen": 2102394880
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003667602808425276,
      "loss": 1.3122,
      "theoretical_loss": 3.444668315552296,
      "tokens_seen": 2102460416
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000366740220661986,
      "loss": 1.3985,
      "theoretical_loss": 3.4446595240644244,
      "tokens_seen": 2102525952
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036672016048144434,
      "loss": 1.3374,
      "theoretical_loss": 3.4446507329273066,
      "tokens_seen": 2102591488
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036670010030090275,
      "loss": 1.2887,
      "theoretical_loss": 3.444641942140918,
      "tokens_seen": 2102657024
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003666800401203611,
      "loss": 1.3601,
      "theoretical_loss": 3.444633151705233,
      "tokens_seen": 2102722560
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003666599799398195,
      "loss": 1.3409,
      "theoretical_loss": 3.444624361620227,
      "tokens_seen": 2102788096
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036663991975927784,
      "loss": 1.3857,
      "theoretical_loss": 3.4446155718858744,
      "tokens_seen": 2102853632
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003666198595787362,
      "loss": 1.3984,
      "theoretical_loss": 3.4446067825021514,
      "tokens_seen": 2102919168
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003665997993981946,
      "loss": 1.3578,
      "theoretical_loss": 3.4445979934690327,
      "tokens_seen": 2102984704
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000366579739217653,
      "loss": 1.3742,
      "theoretical_loss": 3.444589204786493,
      "tokens_seen": 2103050240
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036655967903711134,
      "loss": 1.4285,
      "theoretical_loss": 3.444580416454508,
      "tokens_seen": 2103115776
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003665396188565697,
      "loss": 1.3013,
      "theoretical_loss": 3.444571628473052,
      "tokens_seen": 2103181312
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003665195586760281,
      "loss": 1.4651,
      "theoretical_loss": 3.444562840842101,
      "tokens_seen": 2103246848
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003664994984954865,
      "loss": 1.3441,
      "theoretical_loss": 3.4445540535616295,
      "tokens_seen": 2103312384
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036647943831494484,
      "loss": 1.4568,
      "theoretical_loss": 3.4445452666316125,
      "tokens_seen": 2103377920
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003664593781344032,
      "loss": 1.2206,
      "theoretical_loss": 3.4445364800520255,
      "tokens_seen": 2103443456
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036643931795386157,
      "loss": 1.2948,
      "theoretical_loss": 3.4445276938228435,
      "tokens_seen": 2103508992
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036641925777332,
      "loss": 1.4326,
      "theoretical_loss": 3.4445189079440417,
      "tokens_seen": 2103574528
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036639919759277835,
      "loss": 1.5085,
      "theoretical_loss": 3.4445101224155947,
      "tokens_seen": 2103640064
    },
    {
      "epoch": 0.64,
      "objective/train/advantage_avg": 0.4885838031768799,
      "objective/train/docs_used": 766743,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4047070741653442,
      "objective/train/original_loss": 1.4047071933746338,
      "objective/train/theoretical_loss": 3.444501337237478,
      "objective/train/tokens_used": 2124165600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24002955853939056,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500783920288086,
      "objective/train/weighted_lm_loss": 1.474656581878662,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.444501337237478,
      "tokens_seen": 2103705600
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003663791374122367,
      "loss": 1.5118,
      "theoretical_loss": 3.444501337237478,
      "tokens_seen": 2103705600
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036635907723169507,
      "loss": 1.3766,
      "theoretical_loss": 3.4444925524096672,
      "tokens_seen": 2103771136
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003663390170511535,
      "loss": 1.5416,
      "theoretical_loss": 3.4444837679321365,
      "tokens_seen": 2103836672
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036631895687061185,
      "loss": 1.238,
      "theoretical_loss": 3.444474983804862,
      "tokens_seen": 2103902208
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003662988966900702,
      "loss": 1.3859,
      "theoretical_loss": 3.4444662000278177,
      "tokens_seen": 2103967744
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003662788365095286,
      "loss": 1.2482,
      "theoretical_loss": 3.4444574166009794,
      "tokens_seen": 2104033280
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036625877632898694,
      "loss": 1.2865,
      "theoretical_loss": 3.4444486335243223,
      "tokens_seen": 2104098816
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036623871614844535,
      "loss": 1.3584,
      "theoretical_loss": 3.4444398507978216,
      "tokens_seen": 2104164352
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003662186559679037,
      "loss": 1.3092,
      "theoretical_loss": 3.4444310684214514,
      "tokens_seen": 2104229888
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003661985957873621,
      "loss": 1.2821,
      "theoretical_loss": 3.4444222863951888,
      "tokens_seen": 2104295424
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036617853560682044,
      "loss": 1.3887,
      "theoretical_loss": 3.444413504719007,
      "tokens_seen": 2104360960
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036615847542627886,
      "loss": 1.3343,
      "theoretical_loss": 3.444404723392882,
      "tokens_seen": 2104426496
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003661384152457372,
      "loss": 1.3012,
      "theoretical_loss": 3.444395942416789,
      "tokens_seen": 2104492032
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003661183550651956,
      "loss": 1.3681,
      "theoretical_loss": 3.444387161790703,
      "tokens_seen": 2104557568
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036609829488465394,
      "loss": 1.3404,
      "theoretical_loss": 3.444378381514599,
      "tokens_seen": 2104623104
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036607823470411236,
      "loss": 1.3162,
      "theoretical_loss": 3.4443696015884524,
      "tokens_seen": 2104688640
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003660581745235707,
      "loss": 1.3954,
      "theoretical_loss": 3.4443608220122384,
      "tokens_seen": 2104754176
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003660381143430291,
      "loss": 1.3205,
      "theoretical_loss": 3.444352042785932,
      "tokens_seen": 2104819712
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036601805416248745,
      "loss": 1.3849,
      "theoretical_loss": 3.4443432639095084,
      "tokens_seen": 2104885248
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003659979939819458,
      "loss": 1.3451,
      "theoretical_loss": 3.4443344853829427,
      "tokens_seen": 2104950784
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003659779338014042,
      "loss": 1.4948,
      "theoretical_loss": 3.44432570720621,
      "tokens_seen": 2105016320
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003659578736208626,
      "loss": 1.2196,
      "theoretical_loss": 3.444316929379286,
      "tokens_seen": 2105081856
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036593781344032095,
      "loss": 1.3034,
      "theoretical_loss": 3.4443081519021455,
      "tokens_seen": 2105147392
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003659177532597793,
      "loss": 1.3485,
      "theoretical_loss": 3.4442993747747632,
      "tokens_seen": 2105212928
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036589769307923773,
      "loss": 1.296,
      "theoretical_loss": 3.4442905979971146,
      "tokens_seen": 2105278464
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003658776328986961,
      "loss": 1.3839,
      "theoretical_loss": 3.4442818215691755,
      "tokens_seen": 2105344000
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036585757271815445,
      "loss": 1.451,
      "theoretical_loss": 3.4442730454909203,
      "tokens_seen": 2105409536
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003658375125376128,
      "loss": 1.3671,
      "theoretical_loss": 3.444264269762324,
      "tokens_seen": 2105475072
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003658174523570712,
      "loss": 1.3839,
      "theoretical_loss": 3.4442554943833628,
      "tokens_seen": 2105540608
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003657973921765296,
      "loss": 1.4743,
      "theoretical_loss": 3.4442467193540116,
      "tokens_seen": 2105606144
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036577733199598796,
      "loss": 1.442,
      "theoretical_loss": 3.444237944674245,
      "tokens_seen": 2105671680
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003657572718154463,
      "loss": 1.3538,
      "theoretical_loss": 3.4442291703440384,
      "tokens_seen": 2105737216
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003657372116349047,
      "loss": 1.325,
      "theoretical_loss": 3.444220396363367,
      "tokens_seen": 2105802752
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003657171514543631,
      "loss": 1.2311,
      "theoretical_loss": 3.4442116227322064,
      "tokens_seen": 2105868288
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036569709127382146,
      "loss": 1.4253,
      "theoretical_loss": 3.444202849450531,
      "tokens_seen": 2105933824
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003656770310932798,
      "loss": 1.388,
      "theoretical_loss": 3.4441940765183174,
      "tokens_seen": 2105999360
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003656569709127382,
      "loss": 1.2919,
      "theoretical_loss": 3.444185303935539,
      "tokens_seen": 2106064896
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036563691073219655,
      "loss": 1.3132,
      "theoretical_loss": 3.4441765317021726,
      "tokens_seen": 2106130432
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036561685055165496,
      "loss": 1.3725,
      "theoretical_loss": 3.4441677598181926,
      "tokens_seen": 2106195968
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003655967903711134,
      "loss": 1.3704,
      "theoretical_loss": 3.444158988283574,
      "tokens_seen": 2106261504
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036557673019057174,
      "loss": 1.4077,
      "theoretical_loss": 3.4441502170982927,
      "tokens_seen": 2106327040
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003655566700100301,
      "loss": 1.3199,
      "theoretical_loss": 3.444141446262323,
      "tokens_seen": 2106392576
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003655366098294885,
      "loss": 1.5009,
      "theoretical_loss": 3.4441326757756414,
      "tokens_seen": 2106458112
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003655165496489469,
      "loss": 1.4751,
      "theoretical_loss": 3.4441239056382225,
      "tokens_seen": 2106523648
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036549648946840525,
      "loss": 1.4102,
      "theoretical_loss": 3.444115135850041,
      "tokens_seen": 2106589184
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003654764292878636,
      "loss": 1.352,
      "theoretical_loss": 3.4441063664110727,
      "tokens_seen": 2106654720
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036545636910732197,
      "loss": 1.4857,
      "theoretical_loss": 3.444097597321293,
      "tokens_seen": 2106720256
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003654363089267804,
      "loss": 1.4313,
      "theoretical_loss": 3.4440888285806768,
      "tokens_seen": 2106785792
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036541624874623875,
      "loss": 1.4264,
      "theoretical_loss": 3.444080060189199,
      "tokens_seen": 2106851328
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003653961885656971,
      "loss": 1.4632,
      "theoretical_loss": 3.4440712921468357,
      "tokens_seen": 2106916864
    },
    {
      "epoch": 0.64,
      "objective/train/advantage_avg": 0.49382853507995605,
      "objective/train/docs_used": 767939,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3518633842468262,
      "objective/train/original_loss": 1.3518633842468262,
      "objective/train/theoretical_loss": 3.444062524453561,
      "objective/train/tokens_used": 2127442400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24414005875587463,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506240129470825,
      "objective/train/weighted_lm_loss": 1.4201123714447021,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.444062524453561,
      "tokens_seen": 2106982400
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036537612838515547,
      "loss": 1.4252,
      "theoretical_loss": 3.444062524453561,
      "tokens_seen": 2106982400
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003653560682046139,
      "loss": 1.4315,
      "theoretical_loss": 3.4440537571093515,
      "tokens_seen": 2107047936
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036533600802407225,
      "loss": 1.4314,
      "theoretical_loss": 3.444044990114181,
      "tokens_seen": 2107113472
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003653159478435306,
      "loss": 1.3787,
      "theoretical_loss": 3.444036223468026,
      "tokens_seen": 2107179008
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000365295887662989,
      "loss": 1.4224,
      "theoretical_loss": 3.4440274571708613,
      "tokens_seen": 2107244544
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036527582748244734,
      "loss": 1.5077,
      "theoretical_loss": 3.444018691222662,
      "tokens_seen": 2107310080
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036525576730190575,
      "loss": 1.3146,
      "theoretical_loss": 3.4440099256234036,
      "tokens_seen": 2107375616
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003652357071213641,
      "loss": 1.4845,
      "theoretical_loss": 3.444001160373061,
      "tokens_seen": 2107441152
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003652156469408225,
      "loss": 1.3906,
      "theoretical_loss": 3.44399239547161,
      "tokens_seen": 2107506688
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036519558676028084,
      "loss": 1.3496,
      "theoretical_loss": 3.4439836309190253,
      "tokens_seen": 2107572224
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036517552657973926,
      "loss": 1.4973,
      "theoretical_loss": 3.4439748667152825,
      "tokens_seen": 2107637760
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003651554663991976,
      "loss": 1.3342,
      "theoretical_loss": 3.4439661028603568,
      "tokens_seen": 2107703296
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000365135406218656,
      "loss": 1.3814,
      "theoretical_loss": 3.443957339354223,
      "tokens_seen": 2107768832
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036511534603811434,
      "loss": 1.3896,
      "theoretical_loss": 3.4439485761968576,
      "tokens_seen": 2107834368
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036509528585757276,
      "loss": 1.4389,
      "theoretical_loss": 3.443939813388235,
      "tokens_seen": 2107899904
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003650752256770311,
      "loss": 1.4505,
      "theoretical_loss": 3.44393105092833,
      "tokens_seen": 2107965440
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003650551654964895,
      "loss": 1.3762,
      "theoretical_loss": 3.4439222888171193,
      "tokens_seen": 2108030976
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036503510531594785,
      "loss": 1.3127,
      "theoretical_loss": 3.4439135270545767,
      "tokens_seen": 2108096512
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003650150451354062,
      "loss": 1.3377,
      "theoretical_loss": 3.4439047656406783,
      "tokens_seen": 2108162048
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003649949849548646,
      "loss": 1.4207,
      "theoretical_loss": 3.4438960045753992,
      "tokens_seen": 2108227584
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000364974924774323,
      "loss": 1.4451,
      "theoretical_loss": 3.4438872438587147,
      "tokens_seen": 2108293120
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036495486459378135,
      "loss": 1.4909,
      "theoretical_loss": 3.4438784834906,
      "tokens_seen": 2108358656
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003649348044132397,
      "loss": 1.3863,
      "theoretical_loss": 3.443869723471031,
      "tokens_seen": 2108424192
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036491474423269813,
      "loss": 1.3744,
      "theoretical_loss": 3.4438609637999824,
      "tokens_seen": 2108489728
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003648946840521565,
      "loss": 1.4363,
      "theoretical_loss": 3.4438522044774293,
      "tokens_seen": 2108555264
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036487462387161485,
      "loss": 1.3527,
      "theoretical_loss": 3.4438434455033473,
      "tokens_seen": 2108620800
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003648545636910732,
      "loss": 1.3914,
      "theoretical_loss": 3.443834686877712,
      "tokens_seen": 2108686336
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003648345035105316,
      "loss": 1.3998,
      "theoretical_loss": 3.443825928600498,
      "tokens_seen": 2108751872
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036481444332999,
      "loss": 1.3446,
      "theoretical_loss": 3.443817170671682,
      "tokens_seen": 2108817408
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036479438314944836,
      "loss": 1.4438,
      "theoretical_loss": 3.4438084130912374,
      "tokens_seen": 2108882944
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003647743229689067,
      "loss": 1.3937,
      "theoretical_loss": 3.443799655859141,
      "tokens_seen": 2108948480
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003647542627883651,
      "loss": 1.3429,
      "theoretical_loss": 3.4437908989753674,
      "tokens_seen": 2109014016
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003647342026078235,
      "loss": 1.4185,
      "theoretical_loss": 3.4437821424398924,
      "tokens_seen": 2109079552
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036471414242728186,
      "loss": 1.385,
      "theoretical_loss": 3.4437733862526905,
      "tokens_seen": 2109145088
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003646940822467402,
      "loss": 1.4417,
      "theoretical_loss": 3.443764630413738,
      "tokens_seen": 2109210624
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003646740220661986,
      "loss": 1.4675,
      "theoretical_loss": 3.44375587492301,
      "tokens_seen": 2109276160
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036465396188565695,
      "loss": 1.332,
      "theoretical_loss": 3.4437471197804808,
      "tokens_seen": 2109341696
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036463390170511536,
      "loss": 1.3558,
      "theoretical_loss": 3.4437383649861273,
      "tokens_seen": 2109407232
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003646138415245737,
      "loss": 1.2901,
      "theoretical_loss": 3.4437296105399238,
      "tokens_seen": 2109472768
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003645937813440321,
      "loss": 1.4311,
      "theoretical_loss": 3.4437208564418462,
      "tokens_seen": 2109538304
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036457372116349045,
      "loss": 1.609,
      "theoretical_loss": 3.4437121026918693,
      "tokens_seen": 2109603840
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036455366098294887,
      "loss": 1.4129,
      "theoretical_loss": 3.4437033492899687,
      "tokens_seen": 2109669376
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036453360080240723,
      "loss": 1.3032,
      "theoretical_loss": 3.4436945962361203,
      "tokens_seen": 2109734912
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003645135406218656,
      "loss": 1.496,
      "theoretical_loss": 3.4436858435302984,
      "tokens_seen": 2109800448
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036449348044132395,
      "loss": 1.4047,
      "theoretical_loss": 3.4436770911724794,
      "tokens_seen": 2109865984
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003644734202607823,
      "loss": 1.3382,
      "theoretical_loss": 3.4436683391626373,
      "tokens_seen": 2109931520
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036445336008024073,
      "loss": 1.2737,
      "theoretical_loss": 3.4436595875007487,
      "tokens_seen": 2109997056
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003644332998996991,
      "loss": 1.3443,
      "theoretical_loss": 3.443650836186789,
      "tokens_seen": 2110062592
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036441323971915746,
      "loss": 1.4221,
      "theoretical_loss": 3.443642085220733,
      "tokens_seen": 2110128128
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003643931795386158,
      "loss": 1.4316,
      "theoretical_loss": 3.4436333346025556,
      "tokens_seen": 2110193664
    },
    {
      "epoch": 0.64,
      "objective/train/advantage_avg": 0.49311429262161255,
      "objective/train/docs_used": 769058,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1921217441558838,
      "objective/train/original_loss": 1.1921215057373047,
      "objective/train/theoretical_loss": 3.443624584332233,
      "objective/train/tokens_used": 2130719200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24358122050762177,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505497455596924,
      "objective/train/weighted_lm_loss": 1.2524254322052002,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.443624584332233,
      "tokens_seen": 2110259200
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036437311935807424,
      "loss": 1.3882,
      "theoretical_loss": 3.443624584332233,
      "tokens_seen": 2110259200
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003643530591775326,
      "loss": 1.3775,
      "theoretical_loss": 3.4436158344097407,
      "tokens_seen": 2110324736
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036433299899699096,
      "loss": 1.3945,
      "theoretical_loss": 3.4436070848350533,
      "tokens_seen": 2110390272
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003643129388164493,
      "loss": 1.4116,
      "theoretical_loss": 3.4435983356081468,
      "tokens_seen": 2110455808
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003642928786359077,
      "loss": 1.427,
      "theoretical_loss": 3.4435895867289963,
      "tokens_seen": 2110521344
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003642728184553661,
      "loss": 1.3282,
      "theoretical_loss": 3.443580838197577,
      "tokens_seen": 2110586880
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036425275827482446,
      "loss": 1.5601,
      "theoretical_loss": 3.4435720900138644,
      "tokens_seen": 2110652416
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003642326980942828,
      "loss": 1.3505,
      "theoretical_loss": 3.4435633421778347,
      "tokens_seen": 2110717952
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003642126379137412,
      "loss": 1.286,
      "theoretical_loss": 3.443554594689462,
      "tokens_seen": 2110783488
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003641925777331996,
      "loss": 1.25,
      "theoretical_loss": 3.443545847548722,
      "tokens_seen": 2110849024
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036417251755265797,
      "loss": 1.4217,
      "theoretical_loss": 3.443537100755591,
      "tokens_seen": 2110914560
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036415245737211633,
      "loss": 1.3149,
      "theoretical_loss": 3.4435283543100437,
      "tokens_seen": 2110980096
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003641323971915747,
      "loss": 1.2878,
      "theoretical_loss": 3.443519608212055,
      "tokens_seen": 2111045632
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036411233701103316,
      "loss": 1.3467,
      "theoretical_loss": 3.4435108624616015,
      "tokens_seen": 2111111168
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003640922768304915,
      "loss": 1.3426,
      "theoretical_loss": 3.4435021170586575,
      "tokens_seen": 2111176704
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003640722166499499,
      "loss": 1.2953,
      "theoretical_loss": 3.443493372003199,
      "tokens_seen": 2111242240
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036405215646940825,
      "loss": 1.3098,
      "theoretical_loss": 3.4434846272952013,
      "tokens_seen": 2111307776
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003640320962888666,
      "loss": 1.3796,
      "theoretical_loss": 3.44347588293464,
      "tokens_seen": 2111373312
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000364012036108325,
      "loss": 1.4525,
      "theoretical_loss": 3.4434671389214904,
      "tokens_seen": 2111438848
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003639919759277834,
      "loss": 1.3945,
      "theoretical_loss": 3.4434583952557274,
      "tokens_seen": 2111504384
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036397191574724175,
      "loss": 1.332,
      "theoretical_loss": 3.4434496519373265,
      "tokens_seen": 2111569920
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003639518555667001,
      "loss": 1.4072,
      "theoretical_loss": 3.443440908966264,
      "tokens_seen": 2111635456
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036393179538615853,
      "loss": 1.4604,
      "theoretical_loss": 3.4434321663425145,
      "tokens_seen": 2111700992
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003639117352056169,
      "loss": 1.4209,
      "theoretical_loss": 3.4434234240660535,
      "tokens_seen": 2111766528
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036389167502507525,
      "loss": 1.3856,
      "theoretical_loss": 3.443414682136857,
      "tokens_seen": 2111832064
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003638716148445336,
      "loss": 1.3418,
      "theoretical_loss": 3.4434059405549,
      "tokens_seen": 2111897600
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000363851554663992,
      "loss": 1.3438,
      "theoretical_loss": 3.443397199320158,
      "tokens_seen": 2111963136
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003638314944834504,
      "loss": 1.3954,
      "theoretical_loss": 3.443388458432606,
      "tokens_seen": 2112028672
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036381143430290876,
      "loss": 1.4103,
      "theoretical_loss": 3.4433797178922205,
      "tokens_seen": 2112094208
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003637913741223671,
      "loss": 1.3666,
      "theoretical_loss": 3.443370977698976,
      "tokens_seen": 2112159744
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003637713139418255,
      "loss": 1.4149,
      "theoretical_loss": 3.443362237852848,
      "tokens_seen": 2112225280
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003637512537612839,
      "loss": 1.2367,
      "theoretical_loss": 3.4433534983538125,
      "tokens_seen": 2112290816
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036373119358074226,
      "loss": 1.3563,
      "theoretical_loss": 3.4433447592018447,
      "tokens_seen": 2112356352
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003637111334002006,
      "loss": 1.3986,
      "theoretical_loss": 3.44333602039692,
      "tokens_seen": 2112421888
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000363691073219659,
      "loss": 1.4006,
      "theoretical_loss": 3.4433272819390135,
      "tokens_seen": 2112487424
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036367101303911735,
      "loss": 1.3346,
      "theoretical_loss": 3.4433185438281013,
      "tokens_seen": 2112552960
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036365095285857576,
      "loss": 1.4298,
      "theoretical_loss": 3.443309806064158,
      "tokens_seen": 2112618496
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003636308926780341,
      "loss": 1.4912,
      "theoretical_loss": 3.44330106864716,
      "tokens_seen": 2112684032
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003636108324974925,
      "loss": 1.3253,
      "theoretical_loss": 3.4432923315770823,
      "tokens_seen": 2112749568
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036359077231695085,
      "loss": 1.3141,
      "theoretical_loss": 3.4432835948539005,
      "tokens_seen": 2112815104
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036357071213640927,
      "loss": 1.3794,
      "theoretical_loss": 3.44327485847759,
      "tokens_seen": 2112880640
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036355065195586763,
      "loss": 1.4285,
      "theoretical_loss": 3.443266122448126,
      "tokens_seen": 2112946176
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000363530591775326,
      "loss": 1.3424,
      "theoretical_loss": 3.4432573867654845,
      "tokens_seen": 2113011712
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036351053159478435,
      "loss": 1.4083,
      "theoretical_loss": 3.4432486514296405,
      "tokens_seen": 2113077248
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003634904714142427,
      "loss": 1.3614,
      "theoretical_loss": 3.44323991644057,
      "tokens_seen": 2113142784
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036347041123370113,
      "loss": 1.3817,
      "theoretical_loss": 3.443231181798248,
      "tokens_seen": 2113208320
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003634503510531595,
      "loss": 1.1613,
      "theoretical_loss": 3.44322244750265,
      "tokens_seen": 2113273856
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036343029087261786,
      "loss": 1.3675,
      "theoretical_loss": 3.443213713553752,
      "tokens_seen": 2113339392
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003634102306920762,
      "loss": 1.4662,
      "theoretical_loss": 3.443204979951529,
      "tokens_seen": 2113404928
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036339017051153464,
      "loss": 1.301,
      "theoretical_loss": 3.4431962466959565,
      "tokens_seen": 2113470464
    },
    {
      "epoch": 0.64,
      "objective/train/advantage_avg": 0.49388137459754944,
      "objective/train/docs_used": 770311,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2729945182800293,
      "objective/train/original_loss": 1.2729945182800293,
      "objective/train/theoretical_loss": 3.4431875137870103,
      "objective/train/tokens_used": 2133996000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24430789053440094,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506302118301392,
      "objective/train/weighted_lm_loss": 1.3373852968215942,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4431875137870103,
      "tokens_seen": 2113536000
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000363370110330993,
      "loss": 1.2993,
      "theoretical_loss": 3.4431875137870103,
      "tokens_seen": 2113536000
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036335005015045136,
      "loss": 1.3832,
      "theoretical_loss": 3.443178781224665,
      "tokens_seen": 2113601536
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003633299899699097,
      "loss": 1.3401,
      "theoretical_loss": 3.4431700490088977,
      "tokens_seen": 2113667072
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003633099297893681,
      "loss": 1.3004,
      "theoretical_loss": 3.4431613171396824,
      "tokens_seen": 2113732608
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003632898696088265,
      "loss": 1.426,
      "theoretical_loss": 3.4431525856169953,
      "tokens_seen": 2113798144
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036326980942828486,
      "loss": 1.5185,
      "theoretical_loss": 3.4431438544408124,
      "tokens_seen": 2113863680
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003632497492477432,
      "loss": 1.3483,
      "theoretical_loss": 3.443135123611108,
      "tokens_seen": 2113929216
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003632296890672016,
      "loss": 1.3433,
      "theoretical_loss": 3.443126393127858,
      "tokens_seen": 2113994752
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036320962888666,
      "loss": 1.3919,
      "theoretical_loss": 3.443117662991039,
      "tokens_seen": 2114060288
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036318956870611837,
      "loss": 1.4244,
      "theoretical_loss": 3.443108933200625,
      "tokens_seen": 2114125824
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036316950852557673,
      "loss": 1.3756,
      "theoretical_loss": 3.4431002037565923,
      "tokens_seen": 2114191360
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003631494483450351,
      "loss": 1.3621,
      "theoretical_loss": 3.4430914746589165,
      "tokens_seen": 2114256896
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036312938816449345,
      "loss": 1.303,
      "theoretical_loss": 3.4430827459075726,
      "tokens_seen": 2114322432
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036310932798395187,
      "loss": 1.4634,
      "theoretical_loss": 3.4430740175025365,
      "tokens_seen": 2114387968
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036308926780341023,
      "loss": 1.3302,
      "theoretical_loss": 3.4430652894437834,
      "tokens_seen": 2114453504
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003630692076228686,
      "loss": 1.4993,
      "theoretical_loss": 3.4430565617312894,
      "tokens_seen": 2114519040
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036304914744232696,
      "loss": 1.4887,
      "theoretical_loss": 3.44304783436503,
      "tokens_seen": 2114584576
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036302908726178537,
      "loss": 1.4164,
      "theoretical_loss": 3.44303910734498,
      "tokens_seen": 2114650112
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036300902708124374,
      "loss": 1.4024,
      "theoretical_loss": 3.4430303806711158,
      "tokens_seen": 2114715648
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003629889669007021,
      "loss": 1.3829,
      "theoretical_loss": 3.4430216543434122,
      "tokens_seen": 2114781184
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036296890672016046,
      "loss": 1.4231,
      "theoretical_loss": 3.443012928361845,
      "tokens_seen": 2114846720
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003629488465396188,
      "loss": 1.309,
      "theoretical_loss": 3.4430042027263905,
      "tokens_seen": 2114912256
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036292878635907724,
      "loss": 1.4928,
      "theoretical_loss": 3.442995477437023,
      "tokens_seen": 2114977792
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003629087261785356,
      "loss": 1.4214,
      "theoretical_loss": 3.4429867524937183,
      "tokens_seen": 2115043328
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036288866599799396,
      "loss": 1.3544,
      "theoretical_loss": 3.442978027896453,
      "tokens_seen": 2115108864
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003628686058174523,
      "loss": 1.343,
      "theoretical_loss": 3.442969303645201,
      "tokens_seen": 2115174400
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036284854563691074,
      "loss": 1.3431,
      "theoretical_loss": 3.44296057973994,
      "tokens_seen": 2115239936
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003628284854563691,
      "loss": 1.4721,
      "theoretical_loss": 3.4429518561806436,
      "tokens_seen": 2115305472
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036280842527582747,
      "loss": 1.3114,
      "theoretical_loss": 3.442943132967288,
      "tokens_seen": 2115371008
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036278836509528583,
      "loss": 1.3391,
      "theoretical_loss": 3.4429344100998494,
      "tokens_seen": 2115436544
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036276830491474424,
      "loss": 1.458,
      "theoretical_loss": 3.442925687578302,
      "tokens_seen": 2115502080
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003627482447342026,
      "loss": 1.3367,
      "theoretical_loss": 3.442916965402623,
      "tokens_seen": 2115567616
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036272818455366097,
      "loss": 1.4561,
      "theoretical_loss": 3.442908243572787,
      "tokens_seen": 2115633152
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036270812437311933,
      "loss": 1.5004,
      "theoretical_loss": 3.4428995220887693,
      "tokens_seen": 2115698688
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003626880641925777,
      "loss": 1.2814,
      "theoretical_loss": 3.442890800950546,
      "tokens_seen": 2115764224
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003626680040120361,
      "loss": 1.3374,
      "theoretical_loss": 3.4428820801580935,
      "tokens_seen": 2115829760
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036264794383149447,
      "loss": 1.3371,
      "theoretical_loss": 3.4428733597113856,
      "tokens_seen": 2115895296
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036262788365095283,
      "loss": 1.3097,
      "theoretical_loss": 3.4428646396103986,
      "tokens_seen": 2115960832
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036260782347041125,
      "loss": 1.301,
      "theoretical_loss": 3.442855919855109,
      "tokens_seen": 2116026368
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036258776328986967,
      "loss": 1.3566,
      "theoretical_loss": 3.4428472004454913,
      "tokens_seen": 2116091904
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036256770310932803,
      "loss": 1.2906,
      "theoretical_loss": 3.4428384813815214,
      "tokens_seen": 2116157440
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003625476429287864,
      "loss": 1.3685,
      "theoretical_loss": 3.442829762663175,
      "tokens_seen": 2116222976
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036252758274824475,
      "loss": 1.2487,
      "theoretical_loss": 3.4428210442904277,
      "tokens_seen": 2116288512
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003625075225677031,
      "loss": 1.3621,
      "theoretical_loss": 3.442812326263255,
      "tokens_seen": 2116354048
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036248746238716153,
      "loss": 1.2996,
      "theoretical_loss": 3.4428036085816327,
      "tokens_seen": 2116419584
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003624674022066199,
      "loss": 1.3549,
      "theoretical_loss": 3.4427948912455357,
      "tokens_seen": 2116485120
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036244734202607826,
      "loss": 1.422,
      "theoretical_loss": 3.4427861742549406,
      "tokens_seen": 2116550656
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003624272818455366,
      "loss": 1.3121,
      "theoretical_loss": 3.4427774576098225,
      "tokens_seen": 2116616192
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036240722166499504,
      "loss": 1.4234,
      "theoretical_loss": 3.442768741310157,
      "tokens_seen": 2116681728
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003623871614844534,
      "loss": 1.2652,
      "theoretical_loss": 3.44276002535592,
      "tokens_seen": 2116747264
    },
    {
      "epoch": 0.64,
      "objective/train/advantage_avg": 0.49093812704086304,
      "objective/train/docs_used": 771474,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.259619116783142,
      "objective/train/original_loss": 1.259619116783142,
      "objective/train/theoretical_loss": 3.442751309747086,
      "objective/train/tokens_used": 2137272800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24171867966651917,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503225326538086,
      "objective/train/weighted_lm_loss": 1.3230148553848267,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.442751309747086,
      "tokens_seen": 2116812800
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036236710130391176,
      "loss": 1.3597,
      "theoretical_loss": 3.442751309747086,
      "tokens_seen": 2116812800
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003623470411233701,
      "loss": 1.3396,
      "theoretical_loss": 3.4427425944836325,
      "tokens_seen": 2116878336
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003623269809428285,
      "loss": 1.4154,
      "theoretical_loss": 3.442733879565534,
      "tokens_seen": 2116943872
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003623069207622869,
      "loss": 1.3423,
      "theoretical_loss": 3.442725164992766,
      "tokens_seen": 2117009408
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036228686058174526,
      "loss": 1.4044,
      "theoretical_loss": 3.442716450765304,
      "tokens_seen": 2117074944
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003622668004012036,
      "loss": 1.3613,
      "theoretical_loss": 3.4427077368831247,
      "tokens_seen": 2117140480
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000362246740220662,
      "loss": 1.3555,
      "theoretical_loss": 3.442699023346203,
      "tokens_seen": 2117206016
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003622266800401204,
      "loss": 1.3673,
      "theoretical_loss": 3.4426903101545143,
      "tokens_seen": 2117271552
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036220661985957877,
      "loss": 1.4495,
      "theoretical_loss": 3.4426815973080345,
      "tokens_seen": 2117337088
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036218655967903713,
      "loss": 1.3692,
      "theoretical_loss": 3.4426728848067394,
      "tokens_seen": 2117402624
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003621664994984955,
      "loss": 1.4264,
      "theoretical_loss": 3.4426641726506046,
      "tokens_seen": 2117468160
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036214643931795385,
      "loss": 1.4182,
      "theoretical_loss": 3.4426554608396054,
      "tokens_seen": 2117533696
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036212637913741227,
      "loss": 1.4603,
      "theoretical_loss": 3.4426467493737176,
      "tokens_seen": 2117599232
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036210631895687063,
      "loss": 1.3885,
      "theoretical_loss": 3.442638038252917,
      "tokens_seen": 2117664768
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000362086258776329,
      "loss": 1.4201,
      "theoretical_loss": 3.4426293274771798,
      "tokens_seen": 2117730304
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036206619859578736,
      "loss": 1.3997,
      "theoretical_loss": 3.4426206170464804,
      "tokens_seen": 2117795840
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003620461384152458,
      "loss": 1.2993,
      "theoretical_loss": 3.4426119069607948,
      "tokens_seen": 2117861376
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036202607823470414,
      "loss": 1.3973,
      "theoretical_loss": 3.4426031972201,
      "tokens_seen": 2117926912
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003620060180541625,
      "loss": 1.3644,
      "theoretical_loss": 3.44259448782437,
      "tokens_seen": 2117992448
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036198595787362086,
      "loss": 1.4243,
      "theoretical_loss": 3.442585778773581,
      "tokens_seen": 2118057984
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003619658976930792,
      "loss": 1.3864,
      "theoretical_loss": 3.4425770700677085,
      "tokens_seen": 2118123520
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036194583751253764,
      "loss": 1.2983,
      "theoretical_loss": 3.442568361706729,
      "tokens_seen": 2118189056
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000361925777331996,
      "loss": 1.3053,
      "theoretical_loss": 3.4425596536906173,
      "tokens_seen": 2118254592
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036190571715145436,
      "loss": 1.409,
      "theoretical_loss": 3.442550946019349,
      "tokens_seen": 2118320128
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003618856569709127,
      "loss": 1.3234,
      "theoretical_loss": 3.4425422386929005,
      "tokens_seen": 2118385664
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036186559679037114,
      "loss": 1.3512,
      "theoretical_loss": 3.4425335317112475,
      "tokens_seen": 2118451200
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003618455366098295,
      "loss": 1.3973,
      "theoretical_loss": 3.442524825074365,
      "tokens_seen": 2118516736
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036182547642928787,
      "loss": 1.2865,
      "theoretical_loss": 3.442516118782229,
      "tokens_seen": 2118582272
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036180541624874623,
      "loss": 1.4046,
      "theoretical_loss": 3.4425074128348148,
      "tokens_seen": 2118647808
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036178535606820464,
      "loss": 1.286,
      "theoretical_loss": 3.4424987072320983,
      "tokens_seen": 2118713344
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000361765295887663,
      "loss": 1.463,
      "theoretical_loss": 3.442490001974056,
      "tokens_seen": 2118778880
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036174523570712137,
      "loss": 1.418,
      "theoretical_loss": 3.4424812970606626,
      "tokens_seen": 2118844416
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036172517552657973,
      "loss": 1.3586,
      "theoretical_loss": 3.442472592491894,
      "tokens_seen": 2118909952
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003617051153460381,
      "loss": 1.4296,
      "theoretical_loss": 3.4424638882677265,
      "tokens_seen": 2118975488
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003616850551654965,
      "loss": 1.4349,
      "theoretical_loss": 3.442455184388135,
      "tokens_seen": 2119041024
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036166499498495487,
      "loss": 1.3527,
      "theoretical_loss": 3.4424464808530955,
      "tokens_seen": 2119106560
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036164493480441323,
      "loss": 1.3315,
      "theoretical_loss": 3.442437777662584,
      "tokens_seen": 2119172096
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003616248746238716,
      "loss": 1.3274,
      "theoretical_loss": 3.442429074816576,
      "tokens_seen": 2119237632
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036160481444333,
      "loss": 1.4336,
      "theoretical_loss": 3.4424203723150466,
      "tokens_seen": 2119303168
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003615847542627884,
      "loss": 1.435,
      "theoretical_loss": 3.4424116701579726,
      "tokens_seen": 2119368704
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036156469408224674,
      "loss": 1.4124,
      "theoretical_loss": 3.4424029683453288,
      "tokens_seen": 2119434240
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003615446339017051,
      "loss": 1.3575,
      "theoretical_loss": 3.4423942668770913,
      "tokens_seen": 2119499776
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036152457372116346,
      "loss": 1.376,
      "theoretical_loss": 3.442385565753236,
      "tokens_seen": 2119565312
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003615045135406219,
      "loss": 1.4764,
      "theoretical_loss": 3.4423768649737387,
      "tokens_seen": 2119630848
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036148445336008024,
      "loss": 1.449,
      "theoretical_loss": 3.442368164538575,
      "tokens_seen": 2119696384
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003614643931795386,
      "loss": 1.3483,
      "theoretical_loss": 3.44235946444772,
      "tokens_seen": 2119761920
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036144433299899697,
      "loss": 1.3653,
      "theoretical_loss": 3.4423507647011498,
      "tokens_seen": 2119827456
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003614242728184554,
      "loss": 1.2697,
      "theoretical_loss": 3.44234206529884,
      "tokens_seen": 2119892992
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036140421263791374,
      "loss": 1.4313,
      "theoretical_loss": 3.4423333662407676,
      "tokens_seen": 2119958528
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003613841524573721,
      "loss": 1.382,
      "theoretical_loss": 3.442324667526907,
      "tokens_seen": 2120024064
    },
    {
      "epoch": 0.64,
      "objective/train/advantage_avg": 0.49015626311302185,
      "objective/train/docs_used": 772695,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3296560049057007,
      "objective/train/original_loss": 1.3296557664871216,
      "objective/train/theoretical_loss": 3.442315969157234,
      "objective/train/tokens_used": 2140549600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2408878058195114,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502402782440186,
      "objective/train/weighted_lm_loss": 1.3964051008224487,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.442315969157234,
      "tokens_seen": 2120089600
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036136409227683047,
      "loss": 1.4167,
      "theoretical_loss": 3.442315969157234,
      "tokens_seen": 2120089600
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036134403209628883,
      "loss": 1.3894,
      "theoretical_loss": 3.4423072711317246,
      "tokens_seen": 2120155136
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036132397191574725,
      "loss": 1.3936,
      "theoretical_loss": 3.442298573450355,
      "tokens_seen": 2120220672
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003613039117352056,
      "loss": 1.3962,
      "theoretical_loss": 3.4422898761131,
      "tokens_seen": 2120286208
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036128385155466397,
      "loss": 1.34,
      "theoretical_loss": 3.4422811791199357,
      "tokens_seen": 2120351744
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036126379137412233,
      "loss": 1.424,
      "theoretical_loss": 3.4422724824708384,
      "tokens_seen": 2120417280
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036124373119358075,
      "loss": 1.305,
      "theoretical_loss": 3.4422637861657837,
      "tokens_seen": 2120482816
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003612236710130391,
      "loss": 1.412,
      "theoretical_loss": 3.442255090204746,
      "tokens_seen": 2120548352
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003612036108324975,
      "loss": 1.3772,
      "theoretical_loss": 3.4422463945877038,
      "tokens_seen": 2120613888
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036118355065195584,
      "loss": 1.35,
      "theoretical_loss": 3.4422376993146298,
      "tokens_seen": 2120679424
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003611634904714142,
      "loss": 1.3215,
      "theoretical_loss": 3.442229004385502,
      "tokens_seen": 2120744960
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003611434302908726,
      "loss": 1.4095,
      "theoretical_loss": 3.4422203098002955,
      "tokens_seen": 2120810496
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000361123370110331,
      "loss": 1.2749,
      "theoretical_loss": 3.4422116155589855,
      "tokens_seen": 2120876032
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003611033099297894,
      "loss": 1.328,
      "theoretical_loss": 3.4422029216615484,
      "tokens_seen": 2120941568
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036108324974924776,
      "loss": 1.4698,
      "theoretical_loss": 3.44219422810796,
      "tokens_seen": 2121007104
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003610631895687062,
      "loss": 1.3461,
      "theoretical_loss": 3.4421855348981953,
      "tokens_seen": 2121072640
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036104312938816454,
      "loss": 1.382,
      "theoretical_loss": 3.442176842032231,
      "tokens_seen": 2121138176
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003610230692076229,
      "loss": 1.2899,
      "theoretical_loss": 3.4421681495100422,
      "tokens_seen": 2121203712
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036100300902708126,
      "loss": 1.3216,
      "theoretical_loss": 3.4421594573316057,
      "tokens_seen": 2121269248
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003609829488465396,
      "loss": 1.3781,
      "theoretical_loss": 3.442150765496896,
      "tokens_seen": 2121334784
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036096288866599804,
      "loss": 1.494,
      "theoretical_loss": 3.44214207400589,
      "tokens_seen": 2121400320
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003609428284854564,
      "loss": 1.304,
      "theoretical_loss": 3.4421333828585627,
      "tokens_seen": 2121465856
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036092276830491476,
      "loss": 1.3873,
      "theoretical_loss": 3.44212469205489,
      "tokens_seen": 2121531392
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003609027081243731,
      "loss": 1.3396,
      "theoretical_loss": 3.442116001594848,
      "tokens_seen": 2121596928
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036088264794383154,
      "loss": 1.3884,
      "theoretical_loss": 3.4421073114784124,
      "tokens_seen": 2121662464
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003608625877632899,
      "loss": 1.5521,
      "theoretical_loss": 3.4420986217055587,
      "tokens_seen": 2121728000
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036084252758274827,
      "loss": 1.4716,
      "theoretical_loss": 3.4420899322762635,
      "tokens_seen": 2121793536
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036082246740220663,
      "loss": 1.3434,
      "theoretical_loss": 3.442081243190502,
      "tokens_seen": 2121859072
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036080240722166505,
      "loss": 1.4041,
      "theoretical_loss": 3.4420725544482496,
      "tokens_seen": 2121924608
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003607823470411234,
      "loss": 1.539,
      "theoretical_loss": 3.442063866049483,
      "tokens_seen": 2121990144
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036076228686058177,
      "loss": 1.3554,
      "theoretical_loss": 3.4420551779941775,
      "tokens_seen": 2122055680
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036074222668004013,
      "loss": 1.4534,
      "theoretical_loss": 3.4420464902823094,
      "tokens_seen": 2122121216
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003607221664994985,
      "loss": 1.3402,
      "theoretical_loss": 3.4420378029138536,
      "tokens_seen": 2122186752
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003607021063189569,
      "loss": 1.4344,
      "theoretical_loss": 3.4420291158887864,
      "tokens_seen": 2122252288
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036068204613841527,
      "loss": 1.3933,
      "theoretical_loss": 3.4420204292070844,
      "tokens_seen": 2122317824
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036066198595787364,
      "loss": 1.273,
      "theoretical_loss": 3.442011742868722,
      "tokens_seen": 2122383360
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000360641925777332,
      "loss": 1.4582,
      "theoretical_loss": 3.442003056873676,
      "tokens_seen": 2122448896
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003606218655967904,
      "loss": 1.4249,
      "theoretical_loss": 3.441994371221922,
      "tokens_seen": 2122514432
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003606018054162488,
      "loss": 1.311,
      "theoretical_loss": 3.4419856859134357,
      "tokens_seen": 2122579968
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036058174523570714,
      "loss": 1.4074,
      "theoretical_loss": 3.441977000948193,
      "tokens_seen": 2122645504
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003605616850551655,
      "loss": 1.3522,
      "theoretical_loss": 3.44196831632617,
      "tokens_seen": 2122711040
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036054162487462386,
      "loss": 1.4189,
      "theoretical_loss": 3.4419596320473422,
      "tokens_seen": 2122776576
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003605215646940823,
      "loss": 1.3388,
      "theoretical_loss": 3.4419509481116854,
      "tokens_seen": 2122842112
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036050150451354064,
      "loss": 1.4624,
      "theoretical_loss": 3.4419422645191764,
      "tokens_seen": 2122907648
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000360481444332999,
      "loss": 1.3986,
      "theoretical_loss": 3.4419335812697893,
      "tokens_seen": 2122973184
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036046138415245737,
      "loss": 1.4518,
      "theoretical_loss": 3.441924898363501,
      "tokens_seen": 2123038720
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003604413239719158,
      "loss": 1.4172,
      "theoretical_loss": 3.441916215800288,
      "tokens_seen": 2123104256
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036042126379137414,
      "loss": 1.2882,
      "theoretical_loss": 3.441907533580125,
      "tokens_seen": 2123169792
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003604012036108325,
      "loss": 1.3347,
      "theoretical_loss": 3.4418988517029883,
      "tokens_seen": 2123235328
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036038114343029087,
      "loss": 1.3171,
      "theoretical_loss": 3.441890170168854,
      "tokens_seen": 2123300864
    },
    {
      "epoch": 0.64,
      "objective/train/advantage_avg": 0.4869401156902313,
      "objective/train/docs_used": 773937,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2525248527526855,
      "objective/train/original_loss": 1.2525246143341064,
      "objective/train/theoretical_loss": 3.441881488977697,
      "objective/train/tokens_used": 2143826400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2378321886062622,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0499030351638794,
      "objective/train/weighted_lm_loss": 1.3149054050445557,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.441881488977697,
      "tokens_seen": 2123366400
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036036108324974923,
      "loss": 1.3232,
      "theoretical_loss": 3.441881488977697,
      "tokens_seen": 2123366400
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036034102306920765,
      "loss": 1.2601,
      "theoretical_loss": 3.441872808129494,
      "tokens_seen": 2123431936
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000360320962888666,
      "loss": 1.3613,
      "theoretical_loss": 3.4418641276242212,
      "tokens_seen": 2123497472
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036030090270812437,
      "loss": 1.3262,
      "theoretical_loss": 3.4418554474618537,
      "tokens_seen": 2123563008
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036028084252758273,
      "loss": 1.3015,
      "theoretical_loss": 3.441846767642368,
      "tokens_seen": 2123628544
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036026078234704115,
      "loss": 1.3974,
      "theoretical_loss": 3.4418380881657393,
      "tokens_seen": 2123694080
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003602407221664995,
      "loss": 1.3271,
      "theoretical_loss": 3.4418294090319437,
      "tokens_seen": 2123759616
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003602206619859579,
      "loss": 1.3422,
      "theoretical_loss": 3.441820730240958,
      "tokens_seen": 2123825152
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036020060180541624,
      "loss": 1.4354,
      "theoretical_loss": 3.4418120517927564,
      "tokens_seen": 2123890688
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003601805416248746,
      "loss": 1.461,
      "theoretical_loss": 3.441803373687316,
      "tokens_seen": 2123956224
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000360160481444333,
      "loss": 1.435,
      "theoretical_loss": 3.4417946959246124,
      "tokens_seen": 2124021760
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003601404212637914,
      "loss": 1.3415,
      "theoretical_loss": 3.4417860185046214,
      "tokens_seen": 2124087296
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036012036108324974,
      "loss": 1.4413,
      "theoretical_loss": 3.441777341427319,
      "tokens_seen": 2124152832
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003601003009027081,
      "loss": 1.4744,
      "theoretical_loss": 3.441768664692681,
      "tokens_seen": 2124218368
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003600802407221665,
      "loss": 1.4543,
      "theoretical_loss": 3.4417599883006833,
      "tokens_seen": 2124283904
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003600601805416249,
      "loss": 1.5098,
      "theoretical_loss": 3.4417513122513017,
      "tokens_seen": 2124349440
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00036004012036108324,
      "loss": 1.3952,
      "theoretical_loss": 3.4417426365445127,
      "tokens_seen": 2124414976
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003600200601805416,
      "loss": 1.3107,
      "theoretical_loss": 3.441733961180291,
      "tokens_seen": 2124480512
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035999999999999997,
      "loss": 1.2905,
      "theoretical_loss": 3.441725286158614,
      "tokens_seen": 2124546048
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003599799398194584,
      "loss": 1.4105,
      "theoretical_loss": 3.441716611479456,
      "tokens_seen": 2124611584
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035995987963891675,
      "loss": 1.3275,
      "theoretical_loss": 3.4417079371427945,
      "tokens_seen": 2124677120
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003599398194583751,
      "loss": 1.4166,
      "theoretical_loss": 3.4416992631486045,
      "tokens_seen": 2124742656
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035991975927783347,
      "loss": 1.3871,
      "theoretical_loss": 3.4416905894968624,
      "tokens_seen": 2124808192
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003598996990972919,
      "loss": 1.3484,
      "theoretical_loss": 3.441681916187543,
      "tokens_seen": 2124873728
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035987963891675025,
      "loss": 1.3737,
      "theoretical_loss": 3.441673243220624,
      "tokens_seen": 2124939264
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003598595787362086,
      "loss": 1.2476,
      "theoretical_loss": 3.441664570596079,
      "tokens_seen": 2125004800
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000359839518555667,
      "loss": 1.3448,
      "theoretical_loss": 3.4416558983138863,
      "tokens_seen": 2125070336
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035981945837512534,
      "loss": 1.5149,
      "theoretical_loss": 3.4416472263740205,
      "tokens_seen": 2125135872
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035979939819458375,
      "loss": 1.3296,
      "theoretical_loss": 3.4416385547764583,
      "tokens_seen": 2125201408
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003597793380140421,
      "loss": 1.4161,
      "theoretical_loss": 3.441629883521175,
      "tokens_seen": 2125266944
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003597592778335005,
      "loss": 1.4115,
      "theoretical_loss": 3.4416212126081467,
      "tokens_seen": 2125332480
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035973921765295884,
      "loss": 1.4268,
      "theoretical_loss": 3.441612542037349,
      "tokens_seen": 2125398016
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035971915747241726,
      "loss": 1.5237,
      "theoretical_loss": 3.4416038718087583,
      "tokens_seen": 2125463552
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003596990972918756,
      "loss": 1.3727,
      "theoretical_loss": 3.4415952019223504,
      "tokens_seen": 2125529088
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000359679037111334,
      "loss": 1.3816,
      "theoretical_loss": 3.4415865323781016,
      "tokens_seen": 2125594624
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035965897693079234,
      "loss": 1.3943,
      "theoretical_loss": 3.4415778631759872,
      "tokens_seen": 2125660160
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035963891675025076,
      "loss": 1.4254,
      "theoretical_loss": 3.4415691943159836,
      "tokens_seen": 2125725696
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003596188565697091,
      "loss": 1.4198,
      "theoretical_loss": 3.441560525798067,
      "tokens_seen": 2125791232
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035959879638916754,
      "loss": 1.3585,
      "theoretical_loss": 3.4415518576222124,
      "tokens_seen": 2125856768
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003595787362086259,
      "loss": 1.4356,
      "theoretical_loss": 3.4415431897883964,
      "tokens_seen": 2125922304
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035955867602808426,
      "loss": 1.3541,
      "theoretical_loss": 3.441534522296595,
      "tokens_seen": 2125987840
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003595386158475427,
      "loss": 1.3079,
      "theoretical_loss": 3.441525855146784,
      "tokens_seen": 2126053376
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035951855566700104,
      "loss": 1.3863,
      "theoretical_loss": 3.4415171883389397,
      "tokens_seen": 2126118912
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003594984954864594,
      "loss": 1.2344,
      "theoretical_loss": 3.441508521873038,
      "tokens_seen": 2126184448
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035947843530591777,
      "loss": 1.4021,
      "theoretical_loss": 3.441499855749054,
      "tokens_seen": 2126249984
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003594583751253762,
      "loss": 1.3122,
      "theoretical_loss": 3.441491189966965,
      "tokens_seen": 2126315520
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035943831494483454,
      "loss": 1.3301,
      "theoretical_loss": 3.4414825245267453,
      "tokens_seen": 2126381056
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003594182547642929,
      "loss": 1.3495,
      "theoretical_loss": 3.4414738594283727,
      "tokens_seen": 2126446592
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035939819458375127,
      "loss": 1.3968,
      "theoretical_loss": 3.441465194671822,
      "tokens_seen": 2126512128
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035937813440320963,
      "loss": 1.4868,
      "theoretical_loss": 3.44145653025707,
      "tokens_seen": 2126577664
    },
    {
      "epoch": 0.64,
      "objective/train/advantage_avg": 0.49275076389312744,
      "objective/train/docs_used": 774962,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2388988733291626,
      "objective/train/original_loss": 1.2388988733291626,
      "objective/train/theoretical_loss": 3.441447866184092,
      "objective/train/tokens_used": 2147103200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24353919923305511,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505132675170898,
      "objective/train/weighted_lm_loss": 1.3014239072799683,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.441447866184092,
      "tokens_seen": 2126643200
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035935807422266805,
      "loss": 1.3245,
      "theoretical_loss": 3.441447866184092,
      "tokens_seen": 2126643200
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003593380140421264,
      "loss": 1.385,
      "theoretical_loss": 3.441439202452864,
      "tokens_seen": 2126708736
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035931795386158477,
      "loss": 1.3179,
      "theoretical_loss": 3.4414305390633624,
      "tokens_seen": 2126774272
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035929789368104313,
      "loss": 1.3158,
      "theoretical_loss": 3.4414218760155633,
      "tokens_seen": 2126839808
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035927783350050155,
      "loss": 1.3408,
      "theoretical_loss": 3.441413213309442,
      "tokens_seen": 2126905344
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003592577733199599,
      "loss": 1.3307,
      "theoretical_loss": 3.4414045509449753,
      "tokens_seen": 2126970880
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003592377131394183,
      "loss": 1.3941,
      "theoretical_loss": 3.4413958889221385,
      "tokens_seen": 2127036416
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035921765295887664,
      "loss": 1.3882,
      "theoretical_loss": 3.4413872272409076,
      "tokens_seen": 2127101952
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000359197592778335,
      "loss": 1.3153,
      "theoretical_loss": 3.4413785659012595,
      "tokens_seen": 2127167488
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003591775325977934,
      "loss": 1.3658,
      "theoretical_loss": 3.441369904903169,
      "tokens_seen": 2127233024
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003591574724172518,
      "loss": 1.3103,
      "theoretical_loss": 3.4413612442466133,
      "tokens_seen": 2127298560
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035913741223671014,
      "loss": 1.361,
      "theoretical_loss": 3.4413525839315673,
      "tokens_seen": 2127364096
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003591173520561685,
      "loss": 1.4128,
      "theoretical_loss": 3.4413439239580077,
      "tokens_seen": 2127429632
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003590972918756269,
      "loss": 1.4465,
      "theoretical_loss": 3.441335264325911,
      "tokens_seen": 2127495168
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003590772316950853,
      "loss": 1.3108,
      "theoretical_loss": 3.441326605035252,
      "tokens_seen": 2127560704
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035905717151454364,
      "loss": 1.4209,
      "theoretical_loss": 3.441317946086007,
      "tokens_seen": 2127626240
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000359037111334002,
      "loss": 1.3873,
      "theoretical_loss": 3.441309287478153,
      "tokens_seen": 2127691776
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035901705115346037,
      "loss": 1.3467,
      "theoretical_loss": 3.4413006292116646,
      "tokens_seen": 2127757312
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003589969909729188,
      "loss": 1.4038,
      "theoretical_loss": 3.441291971286519,
      "tokens_seen": 2127822848
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035897693079237715,
      "loss": 1.3899,
      "theoretical_loss": 3.4412833137026917,
      "tokens_seen": 2127888384
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003589568706118355,
      "loss": 1.3314,
      "theoretical_loss": 3.4412746564601586,
      "tokens_seen": 2127953920
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035893681043129387,
      "loss": 1.3112,
      "theoretical_loss": 3.4412659995588966,
      "tokens_seen": 2128019456
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003589167502507523,
      "loss": 1.2997,
      "theoretical_loss": 3.44125734299888,
      "tokens_seen": 2128084992
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035889669007021065,
      "loss": 1.4241,
      "theoretical_loss": 3.441248686780087,
      "tokens_seen": 2128150528
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035889669007021065,
      "loss": 1.6145,
      "theoretical_loss": 3.441240030902492,
      "tokens_seen": 2128216064
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.000358876629889669,
      "loss": 1.3455,
      "theoretical_loss": 3.441231375366072,
      "tokens_seen": 2128281600
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0003588565697091274,
      "loss": 1.3711,
      "theoretical_loss": 3.4412227201708028,
      "tokens_seen": 2128347136
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035883650952858574,
      "loss": 1.3482,
      "theoretical_loss": 3.44121406531666,
      "tokens_seen": 2128412672
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.00035881644934804415,
      "loss": 1.4266,
      "theoretical_loss": 3.44120541080362,
      "tokens_seen": 2128478208
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003587963891675025,
      "loss": 1.3627,
      "theoretical_loss": 3.4411967566316592,
      "tokens_seen": 2128543744
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003587763289869609,
      "loss": 1.5103,
      "theoretical_loss": 3.4411881028007527,
      "tokens_seen": 2128609280
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035875626880641924,
      "loss": 1.29,
      "theoretical_loss": 3.4411794493108774,
      "tokens_seen": 2128674816
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035873620862587766,
      "loss": 1.3897,
      "theoretical_loss": 3.4411707961620093,
      "tokens_seen": 2128740352
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000358716148445336,
      "loss": 1.3482,
      "theoretical_loss": 3.441162143354124,
      "tokens_seen": 2128805888
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003586960882647944,
      "loss": 1.3262,
      "theoretical_loss": 3.4411534908871984,
      "tokens_seen": 2128871424
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035867602808425274,
      "loss": 1.3809,
      "theoretical_loss": 3.441144838761207,
      "tokens_seen": 2128936960
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035865596790371116,
      "loss": 1.3193,
      "theoretical_loss": 3.4411361869761277,
      "tokens_seen": 2129002496
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003586359077231695,
      "loss": 1.4404,
      "theoretical_loss": 3.4411275355319355,
      "tokens_seen": 2129068032
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003586158475426279,
      "loss": 1.483,
      "theoretical_loss": 3.4411188844286067,
      "tokens_seen": 2129133568
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035859578736208625,
      "loss": 1.391,
      "theoretical_loss": 3.4411102336661177,
      "tokens_seen": 2129199104
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003585757271815446,
      "loss": 1.388,
      "theoretical_loss": 3.4411015832444436,
      "tokens_seen": 2129264640
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000358555667001003,
      "loss": 1.4922,
      "theoretical_loss": 3.441092933163562,
      "tokens_seen": 2129330176
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003585356068204614,
      "loss": 1.3294,
      "theoretical_loss": 3.441084283423448,
      "tokens_seen": 2129395712
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035851554663991975,
      "loss": 1.3891,
      "theoretical_loss": 3.4410756340240773,
      "tokens_seen": 2129461248
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003584954864593781,
      "loss": 1.5528,
      "theoretical_loss": 3.441066984965427,
      "tokens_seen": 2129526784
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035847542627883653,
      "loss": 1.2994,
      "theoretical_loss": 3.4410583362474725,
      "tokens_seen": 2129592320
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003584553660982949,
      "loss": 1.4904,
      "theoretical_loss": 3.4410496878701906,
      "tokens_seen": 2129657856
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035843530591775325,
      "loss": 1.2872,
      "theoretical_loss": 3.4410410398335562,
      "tokens_seen": 2129723392
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003584152457372116,
      "loss": 1.2792,
      "theoretical_loss": 3.4410323921375467,
      "tokens_seen": 2129788928
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035839518555667,
      "loss": 1.3966,
      "theoretical_loss": 3.4410237447821377,
      "tokens_seen": 2129854464
    },
    {
      "epoch": 0.65,
      "objective/train/advantage_avg": 0.49363067746162415,
      "objective/train/docs_used": 775953,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2842403650283813,
      "objective/train/original_loss": 1.284240484237671,
      "objective/train/theoretical_loss": 3.441015097767305,
      "objective/train/tokens_used": 2150380000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2439931333065033,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506035089492798,
      "objective/train/weighted_lm_loss": 1.349117636680603,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.441015097767305,
      "tokens_seen": 2129920000
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003583751253761284,
      "loss": 1.4189,
      "theoretical_loss": 3.441015097767305,
      "tokens_seen": 2129920000
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035835506519558676,
      "loss": 1.3802,
      "theoretical_loss": 3.441006451093025,
      "tokens_seen": 2129985536
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003583350050150451,
      "loss": 1.425,
      "theoretical_loss": 3.4409978047592737,
      "tokens_seen": 2130051072
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003583149448345035,
      "loss": 1.5363,
      "theoretical_loss": 3.440989158766027,
      "tokens_seen": 2130116608
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003582948846539619,
      "loss": 1.4126,
      "theoretical_loss": 3.440980513113262,
      "tokens_seen": 2130182144
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035827482447342026,
      "loss": 1.436,
      "theoretical_loss": 3.440971867800953,
      "tokens_seen": 2130247680
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003582547642928786,
      "loss": 1.4254,
      "theoretical_loss": 3.4409632228290783,
      "tokens_seen": 2130313216
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000358234704112337,
      "loss": 1.3866,
      "theoretical_loss": 3.440954578197613,
      "tokens_seen": 2130378752
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035821464393179535,
      "loss": 1.4114,
      "theoretical_loss": 3.4409459339065327,
      "tokens_seen": 2130444288
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035819458375125376,
      "loss": 1.4313,
      "theoretical_loss": 3.440937289955814,
      "tokens_seen": 2130509824
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003581745235707121,
      "loss": 1.4386,
      "theoretical_loss": 3.4409286463454327,
      "tokens_seen": 2130575360
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003581544633901705,
      "loss": 1.5087,
      "theoretical_loss": 3.4409200030753655,
      "tokens_seen": 2130640896
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035813440320962885,
      "loss": 1.4827,
      "theoretical_loss": 3.4409113601455887,
      "tokens_seen": 2130706432
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035811434302908727,
      "loss": 1.339,
      "theoretical_loss": 3.4409027175560776,
      "tokens_seen": 2130771968
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003580942828485457,
      "loss": 1.2952,
      "theoretical_loss": 3.440894075306809,
      "tokens_seen": 2130837504
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035807422266800404,
      "loss": 1.4753,
      "theoretical_loss": 3.4408854333977583,
      "tokens_seen": 2130903040
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003580541624874624,
      "loss": 1.4253,
      "theoretical_loss": 3.4408767918289027,
      "tokens_seen": 2130968576
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035803410230692077,
      "loss": 1.4267,
      "theoretical_loss": 3.4408681506002177,
      "tokens_seen": 2131034112
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003580140421263792,
      "loss": 1.4022,
      "theoretical_loss": 3.440859509711679,
      "tokens_seen": 2131099648
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035799398194583755,
      "loss": 1.2869,
      "theoretical_loss": 3.440850869163264,
      "tokens_seen": 2131165184
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003579739217652959,
      "loss": 1.4006,
      "theoretical_loss": 3.4408422289549483,
      "tokens_seen": 2131230720
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035795386158475427,
      "loss": 1.376,
      "theoretical_loss": 3.440833589086707,
      "tokens_seen": 2131296256
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003579338014042127,
      "loss": 1.3962,
      "theoretical_loss": 3.440824949558518,
      "tokens_seen": 2131361792
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035791374122367105,
      "loss": 1.3115,
      "theoretical_loss": 3.4408163103703564,
      "tokens_seen": 2131427328
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003578936810431294,
      "loss": 1.3318,
      "theoretical_loss": 3.4408076715221982,
      "tokens_seen": 2131492864
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003578736208625878,
      "loss": 1.4281,
      "theoretical_loss": 3.4407990330140206,
      "tokens_seen": 2131558400
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035785356068204614,
      "loss": 1.2965,
      "theoretical_loss": 3.4407903948457985,
      "tokens_seen": 2131623936
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035783350050150455,
      "loss": 1.4172,
      "theoretical_loss": 3.4407817570175094,
      "tokens_seen": 2131689472
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003578134403209629,
      "loss": 1.3039,
      "theoretical_loss": 3.440773119529128,
      "tokens_seen": 2131755008
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003577933801404213,
      "loss": 1.3681,
      "theoretical_loss": 3.440764482380632,
      "tokens_seen": 2131820544
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035777331995987964,
      "loss": 1.3994,
      "theoretical_loss": 3.440755845571996,
      "tokens_seen": 2131886080
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035775325977933806,
      "loss": 1.4493,
      "theoretical_loss": 3.440747209103198,
      "tokens_seen": 2131951616
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003577331995987964,
      "loss": 1.3897,
      "theoretical_loss": 3.440738572974212,
      "tokens_seen": 2132017152
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003577131394182548,
      "loss": 1.4871,
      "theoretical_loss": 3.440729937185016,
      "tokens_seen": 2132082688
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035769307923771314,
      "loss": 1.3813,
      "theoretical_loss": 3.4407213017355858,
      "tokens_seen": 2132148224
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035767301905717156,
      "loss": 1.409,
      "theoretical_loss": 3.440712666625897,
      "tokens_seen": 2132213760
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003576529588766299,
      "loss": 1.4309,
      "theoretical_loss": 3.440704031855926,
      "tokens_seen": 2132279296
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003576328986960883,
      "loss": 1.4186,
      "theoretical_loss": 3.4406953974256496,
      "tokens_seen": 2132344832
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035761283851554665,
      "loss": 1.33,
      "theoretical_loss": 3.440686763335043,
      "tokens_seen": 2132410368
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000357592778335005,
      "loss": 1.4721,
      "theoretical_loss": 3.440678129584083,
      "tokens_seen": 2132475904
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003575727181544634,
      "loss": 1.5311,
      "theoretical_loss": 3.440669496172746,
      "tokens_seen": 2132541440
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003575526579739218,
      "loss": 1.3603,
      "theoretical_loss": 3.440660863101008,
      "tokens_seen": 2132606976
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035753259779338015,
      "loss": 1.404,
      "theoretical_loss": 3.440652230368845,
      "tokens_seen": 2132672512
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003575125376128385,
      "loss": 1.4284,
      "theoretical_loss": 3.440643597976233,
      "tokens_seen": 2132738048
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035749247743229693,
      "loss": 1.478,
      "theoretical_loss": 3.4406349659231488,
      "tokens_seen": 2132803584
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003574724172517553,
      "loss": 1.2782,
      "theoretical_loss": 3.440626334209569,
      "tokens_seen": 2132869120
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035745235707121365,
      "loss": 1.3883,
      "theoretical_loss": 3.4406177028354685,
      "tokens_seen": 2132934656
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000357432296890672,
      "loss": 1.295,
      "theoretical_loss": 3.440609071800824,
      "tokens_seen": 2133000192
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003574122367101304,
      "loss": 1.3368,
      "theoretical_loss": 3.4406004411056124,
      "tokens_seen": 2133065728
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003573921765295888,
      "loss": 1.3155,
      "theoretical_loss": 3.440591810749809,
      "tokens_seen": 2133131264
    },
    {
      "epoch": 0.65,
      "objective/train/advantage_avg": 0.4927752614021301,
      "objective/train/docs_used": 777226,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2377055883407593,
      "objective/train/original_loss": 1.237705945968628,
      "objective/train/theoretical_loss": 3.440583180733391,
      "objective/train/tokens_used": 2153656800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2431272715330124,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050513505935669,
      "objective/train/weighted_lm_loss": 1.3000880479812622,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.440583180733391,
      "tokens_seen": 2133196800
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035737211634904716,
      "loss": 1.3221,
      "theoretical_loss": 3.440583180733391,
      "tokens_seen": 2133196800
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003573520561685055,
      "loss": 1.3356,
      "theoretical_loss": 3.4405745510563337,
      "tokens_seen": 2133262336
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003573319959879639,
      "loss": 1.3483,
      "theoretical_loss": 3.4405659217186138,
      "tokens_seen": 2133327872
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003573119358074223,
      "loss": 1.2838,
      "theoretical_loss": 3.440557292720208,
      "tokens_seen": 2133393408
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035729187562688066,
      "loss": 1.3665,
      "theoretical_loss": 3.440548664061091,
      "tokens_seen": 2133458944
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000357271815446339,
      "loss": 1.3529,
      "theoretical_loss": 3.440540035741241,
      "tokens_seen": 2133524480
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003572517552657974,
      "loss": 1.3137,
      "theoretical_loss": 3.440531407760633,
      "tokens_seen": 2133590016
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035723169508525575,
      "loss": 1.3779,
      "theoretical_loss": 3.4405227801192435,
      "tokens_seen": 2133655552
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035721163490471416,
      "loss": 1.4587,
      "theoretical_loss": 3.440514152817048,
      "tokens_seen": 2133721088
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003571915747241725,
      "loss": 1.3322,
      "theoretical_loss": 3.4405055258540242,
      "tokens_seen": 2133786624
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003571715145436309,
      "loss": 1.4007,
      "theoretical_loss": 3.4404968992301477,
      "tokens_seen": 2133852160
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035715145436308925,
      "loss": 1.388,
      "theoretical_loss": 3.440488272945395,
      "tokens_seen": 2133917696
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035713139418254767,
      "loss": 1.3461,
      "theoretical_loss": 3.4404796469997416,
      "tokens_seen": 2133983232
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035711133400200603,
      "loss": 1.2654,
      "theoretical_loss": 3.440471021393164,
      "tokens_seen": 2134048768
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003570912738214644,
      "loss": 1.3163,
      "theoretical_loss": 3.440462396125639,
      "tokens_seen": 2134114304
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035707121364092275,
      "loss": 1.4072,
      "theoretical_loss": 3.440453771197143,
      "tokens_seen": 2134179840
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003570511534603811,
      "loss": 1.3634,
      "theoretical_loss": 3.440445146607651,
      "tokens_seen": 2134245376
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035703109327983953,
      "loss": 1.3928,
      "theoretical_loss": 3.4404365223571407,
      "tokens_seen": 2134310912
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003570110330992979,
      "loss": 1.4619,
      "theoretical_loss": 3.4404278984455874,
      "tokens_seen": 2134376448
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035699097291875626,
      "loss": 1.4507,
      "theoretical_loss": 3.440419274872968,
      "tokens_seen": 2134441984
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003569709127382146,
      "loss": 1.3505,
      "theoretical_loss": 3.440410651639258,
      "tokens_seen": 2134507520
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035695085255767303,
      "loss": 1.5053,
      "theoretical_loss": 3.440402028744434,
      "tokens_seen": 2134573056
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003569307923771314,
      "loss": 1.3856,
      "theoretical_loss": 3.4403934061884733,
      "tokens_seen": 2134638592
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035691073219658976,
      "loss": 1.3763,
      "theoretical_loss": 3.440384783971351,
      "tokens_seen": 2134704128
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003568906720160481,
      "loss": 1.4744,
      "theoretical_loss": 3.4403761620930435,
      "tokens_seen": 2134769664
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003568706118355065,
      "loss": 1.3688,
      "theoretical_loss": 3.4403675405535274,
      "tokens_seen": 2134835200
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003568505516549649,
      "loss": 1.4084,
      "theoretical_loss": 3.4403589193527786,
      "tokens_seen": 2134900736
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035683049147442326,
      "loss": 1.4335,
      "theoretical_loss": 3.440350298490774,
      "tokens_seen": 2134966272
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003568104312938816,
      "loss": 1.2836,
      "theoretical_loss": 3.4403416779674894,
      "tokens_seen": 2135031808
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035679037111334,
      "loss": 1.4307,
      "theoretical_loss": 3.4403330577829014,
      "tokens_seen": 2135097344
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003567703109327984,
      "loss": 1.4431,
      "theoretical_loss": 3.440324437936986,
      "tokens_seen": 2135162880
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035675025075225677,
      "loss": 1.3708,
      "theoretical_loss": 3.4403158184297196,
      "tokens_seen": 2135228416
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035673019057171513,
      "loss": 1.5171,
      "theoretical_loss": 3.4403071992610785,
      "tokens_seen": 2135293952
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003567101303911735,
      "loss": 1.2895,
      "theoretical_loss": 3.4402985804310395,
      "tokens_seen": 2135359488
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035669007021063185,
      "loss": 1.2993,
      "theoretical_loss": 3.4402899619395786,
      "tokens_seen": 2135425024
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035667001003009027,
      "loss": 1.3216,
      "theoretical_loss": 3.4402813437866713,
      "tokens_seen": 2135490560
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035664994984954863,
      "loss": 1.2165,
      "theoretical_loss": 3.440272725972295,
      "tokens_seen": 2135556096
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000356629889669007,
      "loss": 1.3807,
      "theoretical_loss": 3.440264108496425,
      "tokens_seen": 2135621632
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035660982948846536,
      "loss": 1.306,
      "theoretical_loss": 3.440255491359039,
      "tokens_seen": 2135687168
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003565897693079238,
      "loss": 1.4467,
      "theoretical_loss": 3.440246874560112,
      "tokens_seen": 2135752704
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003565697091273822,
      "loss": 1.4243,
      "theoretical_loss": 3.440238258099621,
      "tokens_seen": 2135818240
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035654964894684055,
      "loss": 1.3784,
      "theoretical_loss": 3.4402296419775427,
      "tokens_seen": 2135883776
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003565295887662989,
      "loss": 1.3108,
      "theoretical_loss": 3.440221026193852,
      "tokens_seen": 2135949312
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035650952858575733,
      "loss": 1.3796,
      "theoretical_loss": 3.4402124107485266,
      "tokens_seen": 2136014848
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003564894684052157,
      "loss": 1.2473,
      "theoretical_loss": 3.4402037956415423,
      "tokens_seen": 2136080384
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035646940822467405,
      "loss": 1.3274,
      "theoretical_loss": 3.4401951808728755,
      "tokens_seen": 2136145920
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003564493480441324,
      "loss": 1.3765,
      "theoretical_loss": 3.4401865664425024,
      "tokens_seen": 2136211456
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003564292878635908,
      "loss": 1.3981,
      "theoretical_loss": 3.4401779523503997,
      "tokens_seen": 2136276992
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003564092276830492,
      "loss": 1.2964,
      "theoretical_loss": 3.4401693385965437,
      "tokens_seen": 2136342528
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035638916750250756,
      "loss": 1.3734,
      "theoretical_loss": 3.44016072518091,
      "tokens_seen": 2136408064
    },
    {
      "epoch": 0.65,
      "objective/train/advantage_avg": 0.49171334505081177,
      "objective/train/docs_used": 778288,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.314366102218628,
      "objective/train/original_loss": 1.314366102218628,
      "objective/train/theoretical_loss": 3.440152112103476,
      "objective/train/tokens_used": 2156933600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24221119284629822,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504025220870972,
      "objective/train/weighted_lm_loss": 1.3805028200149536,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.440152112103476,
      "tokens_seen": 2136473600
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003563691073219659,
      "loss": 1.4138,
      "theoretical_loss": 3.440152112103476,
      "tokens_seen": 2136473600
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003563490471414243,
      "loss": 1.4401,
      "theoretical_loss": 3.4401434993642175,
      "tokens_seen": 2136539136
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003563289869608827,
      "loss": 1.2868,
      "theoretical_loss": 3.4401348869631105,
      "tokens_seen": 2136604672
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035630892678034106,
      "loss": 1.5197,
      "theoretical_loss": 3.440126274900132,
      "tokens_seen": 2136670208
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003562888665997994,
      "loss": 1.2965,
      "theoretical_loss": 3.4401176631752577,
      "tokens_seen": 2136735744
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003562688064192578,
      "loss": 1.369,
      "theoretical_loss": 3.4401090517884647,
      "tokens_seen": 2136801280
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035624874623871615,
      "loss": 1.306,
      "theoretical_loss": 3.4401004407397293,
      "tokens_seen": 2136866816
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035622868605817456,
      "loss": 1.4229,
      "theoretical_loss": 3.440091830029027,
      "tokens_seen": 2136932352
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003562086258776329,
      "loss": 1.3396,
      "theoretical_loss": 3.4400832196563353,
      "tokens_seen": 2136997888
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003561885656970913,
      "loss": 1.4423,
      "theoretical_loss": 3.4400746096216297,
      "tokens_seen": 2137063424
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035616850551654965,
      "loss": 1.3758,
      "theoretical_loss": 3.4400659999248866,
      "tokens_seen": 2137128960
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035614844533600807,
      "loss": 1.355,
      "theoretical_loss": 3.440057390566083,
      "tokens_seen": 2137194496
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035612838515546643,
      "loss": 1.3663,
      "theoretical_loss": 3.440048781545195,
      "tokens_seen": 2137260032
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003561083249749248,
      "loss": 1.484,
      "theoretical_loss": 3.440040172862199,
      "tokens_seen": 2137325568
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035608826479438315,
      "loss": 1.4201,
      "theoretical_loss": 3.4400315645170707,
      "tokens_seen": 2137391104
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003560682046138415,
      "loss": 1.4,
      "theoretical_loss": 3.4400229565097877,
      "tokens_seen": 2137456640
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035604814443329993,
      "loss": 1.403,
      "theoretical_loss": 3.440014348840325,
      "tokens_seen": 2137522176
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003560280842527583,
      "loss": 1.3939,
      "theoretical_loss": 3.44000574150866,
      "tokens_seen": 2137587712
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035600802407221666,
      "loss": 1.3275,
      "theoretical_loss": 3.439997134514769,
      "tokens_seen": 2137653248
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000355987963891675,
      "loss": 1.4101,
      "theoretical_loss": 3.4399885278586284,
      "tokens_seen": 2137718784
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035596790371113344,
      "loss": 1.4498,
      "theoretical_loss": 3.439979921540214,
      "tokens_seen": 2137784320
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003559478435305918,
      "loss": 1.4237,
      "theoretical_loss": 3.439971315559503,
      "tokens_seen": 2137849856
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035592778335005016,
      "loss": 1.3508,
      "theoretical_loss": 3.439962709916471,
      "tokens_seen": 2137915392
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003559077231695085,
      "loss": 1.3857,
      "theoretical_loss": 3.439954104611095,
      "tokens_seen": 2137980928
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003558876629889669,
      "loss": 1.3695,
      "theoretical_loss": 3.4399454996433514,
      "tokens_seen": 2138046464
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003558676028084253,
      "loss": 1.3705,
      "theoretical_loss": 3.439936895013216,
      "tokens_seen": 2138112000
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035584754262788366,
      "loss": 1.3428,
      "theoretical_loss": 3.4399282907206654,
      "tokens_seen": 2138177536
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000355827482447342,
      "loss": 1.4779,
      "theoretical_loss": 3.4399196867656765,
      "tokens_seen": 2138243072
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003558074222668004,
      "loss": 1.4965,
      "theoretical_loss": 3.4399110831482256,
      "tokens_seen": 2138308608
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003557873620862588,
      "loss": 1.389,
      "theoretical_loss": 3.4399024798682887,
      "tokens_seen": 2138374144
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035576730190571717,
      "loss": 1.4164,
      "theoretical_loss": 3.4398938769258427,
      "tokens_seen": 2138439680
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035574724172517553,
      "loss": 1.3472,
      "theoretical_loss": 3.439885274320863,
      "tokens_seen": 2138505216
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003557271815446339,
      "loss": 1.4365,
      "theoretical_loss": 3.4398766720533276,
      "tokens_seen": 2138570752
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035570712136409225,
      "loss": 1.5878,
      "theoretical_loss": 3.439868070123212,
      "tokens_seen": 2138636288
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035568706118355067,
      "loss": 1.3169,
      "theoretical_loss": 3.4398594685304924,
      "tokens_seen": 2138701824
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035566700100300903,
      "loss": 1.4279,
      "theoretical_loss": 3.4398508672751458,
      "tokens_seen": 2138767360
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003556469408224674,
      "loss": 1.4078,
      "theoretical_loss": 3.4398422663571484,
      "tokens_seen": 2138832896
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035562688064192576,
      "loss": 1.3543,
      "theoretical_loss": 3.4398336657764768,
      "tokens_seen": 2138898432
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035560682046138417,
      "loss": 1.3825,
      "theoretical_loss": 3.439825065533107,
      "tokens_seen": 2138963968
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035558676028084253,
      "loss": 1.4388,
      "theoretical_loss": 3.4398164656270156,
      "tokens_seen": 2139029504
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003555667001003009,
      "loss": 1.3189,
      "theoretical_loss": 3.439807866058179,
      "tokens_seen": 2139095040
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035554663991975926,
      "loss": 1.3943,
      "theoretical_loss": 3.439799266826574,
      "tokens_seen": 2139160576
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003555265797392176,
      "loss": 1.4585,
      "theoretical_loss": 3.439790667932177,
      "tokens_seen": 2139226112
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035550651955867604,
      "loss": 1.5142,
      "theoretical_loss": 3.439782069374964,
      "tokens_seen": 2139291648
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003554864593781344,
      "loss": 1.4107,
      "theoretical_loss": 3.439773471154912,
      "tokens_seen": 2139357184
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035546639919759276,
      "loss": 1.4619,
      "theoretical_loss": 3.439764873271997,
      "tokens_seen": 2139422720
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003554463390170511,
      "loss": 1.2768,
      "theoretical_loss": 3.4397562757261957,
      "tokens_seen": 2139488256
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035542627883650954,
      "loss": 1.4788,
      "theoretical_loss": 3.4397476785174845,
      "tokens_seen": 2139553792
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003554062186559679,
      "loss": 1.3858,
      "theoretical_loss": 3.4397390816458397,
      "tokens_seen": 2139619328
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035538615847542627,
      "loss": 1.4704,
      "theoretical_loss": 3.4397304851112382,
      "tokens_seen": 2139684864
    },
    {
      "epoch": 0.65,
      "objective/train/advantage_avg": 0.49559757113456726,
      "objective/train/docs_used": 779442,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3051334619522095,
      "objective/train/original_loss": 1.3051334619522095,
      "objective/train/theoretical_loss": 3.439721888913656,
      "objective/train/tokens_used": 2160210400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24597923457622528,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508103370666504,
      "objective/train/weighted_lm_loss": 1.3713757991790771,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.439721888913656,
      "tokens_seen": 2139750400
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035536609829488463,
      "loss": 1.3958,
      "theoretical_loss": 3.439721888913656,
      "tokens_seen": 2139750400
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035534603811434304,
      "loss": 1.4419,
      "theoretical_loss": 3.439713293053069,
      "tokens_seen": 2139815936
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003553259779338014,
      "loss": 1.26,
      "theoretical_loss": 3.4397046975294554,
      "tokens_seen": 2139881472
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035530591775325977,
      "loss": 1.4408,
      "theoretical_loss": 3.4396961023427908,
      "tokens_seen": 2139947008
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035528585757271813,
      "loss": 1.3664,
      "theoretical_loss": 3.439687507493051,
      "tokens_seen": 2140012544
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003552657973921765,
      "loss": 1.5606,
      "theoretical_loss": 3.439678912980213,
      "tokens_seen": 2140078080
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003552457372116349,
      "loss": 1.3533,
      "theoretical_loss": 3.4396703188042537,
      "tokens_seen": 2140143616
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035522567703109327,
      "loss": 1.3709,
      "theoretical_loss": 3.4396617249651484,
      "tokens_seen": 2140209152
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035520561685055163,
      "loss": 1.2702,
      "theoretical_loss": 3.439653131462875,
      "tokens_seen": 2140274688
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035518555667001,
      "loss": 1.3566,
      "theoretical_loss": 3.4396445382974097,
      "tokens_seen": 2140340224
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003551654964894684,
      "loss": 1.4138,
      "theoretical_loss": 3.4396359454687278,
      "tokens_seen": 2140405760
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003551454363089268,
      "loss": 1.3119,
      "theoretical_loss": 3.4396273529768067,
      "tokens_seen": 2140471296
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035512537612838514,
      "loss": 1.2498,
      "theoretical_loss": 3.4396187608216233,
      "tokens_seen": 2140536832
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003551053159478435,
      "loss": 1.4411,
      "theoretical_loss": 3.4396101690031533,
      "tokens_seen": 2140602368
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003550852557673019,
      "loss": 1.3793,
      "theoretical_loss": 3.4396015775213735,
      "tokens_seen": 2140667904
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035506519558676033,
      "loss": 1.3357,
      "theoretical_loss": 3.439592986376261,
      "tokens_seen": 2140733440
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003550451354062187,
      "loss": 1.3406,
      "theoretical_loss": 3.439584395567791,
      "tokens_seen": 2140798976
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035502507522567706,
      "loss": 1.3995,
      "theoretical_loss": 3.4395758050959406,
      "tokens_seen": 2140864512
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003550050150451354,
      "loss": 1.4214,
      "theoretical_loss": 3.439567214960687,
      "tokens_seen": 2140930048
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035498495486459384,
      "loss": 1.3381,
      "theoretical_loss": 3.439558625162006,
      "tokens_seen": 2140995584
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003549648946840522,
      "loss": 1.3181,
      "theoretical_loss": 3.439550035699874,
      "tokens_seen": 2141061120
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035494483450351056,
      "loss": 1.3709,
      "theoretical_loss": 3.439541446574268,
      "tokens_seen": 2141126656
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003549247743229689,
      "loss": 1.3204,
      "theoretical_loss": 3.439532857785164,
      "tokens_seen": 2141192192
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003549047141424273,
      "loss": 1.39,
      "theoretical_loss": 3.439524269332539,
      "tokens_seen": 2141257728
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003548846539618857,
      "loss": 1.2376,
      "theoretical_loss": 3.4395156812163687,
      "tokens_seen": 2141323264
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035486459378134406,
      "loss": 1.3867,
      "theoretical_loss": 3.439507093436631,
      "tokens_seen": 2141388800
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003548445336008024,
      "loss": 1.4297,
      "theoretical_loss": 3.4394985059933014,
      "tokens_seen": 2141454336
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003548244734202608,
      "loss": 1.5036,
      "theoretical_loss": 3.4394899188863564,
      "tokens_seen": 2141519872
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003548044132397192,
      "loss": 1.3738,
      "theoretical_loss": 3.4394813321157733,
      "tokens_seen": 2141585408
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035478435305917757,
      "loss": 1.2786,
      "theoretical_loss": 3.4394727456815275,
      "tokens_seen": 2141650944
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035476429287863593,
      "loss": 1.3684,
      "theoretical_loss": 3.439464159583597,
      "tokens_seen": 2141716480
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003547442326980943,
      "loss": 1.315,
      "theoretical_loss": 3.4394555738219568,
      "tokens_seen": 2141782016
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035472417251755265,
      "loss": 1.4758,
      "theoretical_loss": 3.4394469883965844,
      "tokens_seen": 2141847552
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035470411233701107,
      "loss": 1.2793,
      "theoretical_loss": 3.4394384033074563,
      "tokens_seen": 2141913088
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035468405215646943,
      "loss": 1.3488,
      "theoretical_loss": 3.4394298185545487,
      "tokens_seen": 2141978624
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003546639919759278,
      "loss": 1.4333,
      "theoretical_loss": 3.4394212341378383,
      "tokens_seen": 2142044160
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035464393179538616,
      "loss": 1.489,
      "theoretical_loss": 3.4394126500573012,
      "tokens_seen": 2142109696
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035462387161484457,
      "loss": 1.3356,
      "theoretical_loss": 3.439404066312915,
      "tokens_seen": 2142175232
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035460381143430293,
      "loss": 1.3902,
      "theoretical_loss": 3.439395482904655,
      "tokens_seen": 2142240768
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003545837512537613,
      "loss": 1.4454,
      "theoretical_loss": 3.439386899832499,
      "tokens_seen": 2142306304
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035456369107321966,
      "loss": 1.2945,
      "theoretical_loss": 3.4393783170964225,
      "tokens_seen": 2142371840
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000354543630892678,
      "loss": 1.2309,
      "theoretical_loss": 3.4393697346964025,
      "tokens_seen": 2142437376
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035452357071213644,
      "loss": 1.3867,
      "theoretical_loss": 3.439361152632416,
      "tokens_seen": 2142502912
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003545035105315948,
      "loss": 1.3863,
      "theoretical_loss": 3.439352570904439,
      "tokens_seen": 2142568448
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035448345035105316,
      "loss": 1.4573,
      "theoretical_loss": 3.439343989512448,
      "tokens_seen": 2142633984
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003544633901705115,
      "loss": 1.2251,
      "theoretical_loss": 3.4393354084564196,
      "tokens_seen": 2142699520
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035444332998996994,
      "loss": 1.4112,
      "theoretical_loss": 3.4393268277363305,
      "tokens_seen": 2142765056
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003544232698094283,
      "loss": 1.4275,
      "theoretical_loss": 3.4393182473521575,
      "tokens_seen": 2142830592
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035440320962888667,
      "loss": 1.3684,
      "theoretical_loss": 3.4393096673038777,
      "tokens_seen": 2142896128
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035438314944834503,
      "loss": 1.5215,
      "theoretical_loss": 3.439301087591466,
      "tokens_seen": 2142961664
    },
    {
      "epoch": 0.65,
      "objective/train/advantage_avg": 0.4951334297657013,
      "objective/train/docs_used": 780500,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1640410423278809,
      "objective/train/original_loss": 1.1640410423278809,
      "objective/train/theoretical_loss": 3.4392925082149004,
      "objective/train/tokens_used": 2163487200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2455301433801651,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507615804672241,
      "objective/train/weighted_lm_loss": 1.223102331161499,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4392925082149004,
      "tokens_seen": 2143027200
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035436308926780344,
      "loss": 1.3532,
      "theoretical_loss": 3.4392925082149004,
      "tokens_seen": 2143027200
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003543430290872618,
      "loss": 1.4315,
      "theoretical_loss": 3.439283929174157,
      "tokens_seen": 2143092736
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035432296890672017,
      "loss": 1.4761,
      "theoretical_loss": 3.439275350469212,
      "tokens_seen": 2143158272
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035430290872617853,
      "loss": 1.2633,
      "theoretical_loss": 3.439266772100043,
      "tokens_seen": 2143223808
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003542828485456369,
      "loss": 1.4442,
      "theoretical_loss": 3.4392581940666256,
      "tokens_seen": 2143289344
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003542627883650953,
      "loss": 1.2896,
      "theoretical_loss": 3.439249616368937,
      "tokens_seen": 2143354880
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035424272818455367,
      "loss": 1.4353,
      "theoretical_loss": 3.4392410390069537,
      "tokens_seen": 2143420416
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035422266800401203,
      "loss": 1.2203,
      "theoretical_loss": 3.4392324619806516,
      "tokens_seen": 2143485952
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003542026078234704,
      "loss": 1.4091,
      "theoretical_loss": 3.4392238852900086,
      "tokens_seen": 2143551488
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003541825476429288,
      "loss": 1.3185,
      "theoretical_loss": 3.439215308935,
      "tokens_seen": 2143617024
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003541624874623872,
      "loss": 1.3945,
      "theoretical_loss": 3.4392067329156033,
      "tokens_seen": 2143682560
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035414242728184554,
      "loss": 1.4565,
      "theoretical_loss": 3.439198157231795,
      "tokens_seen": 2143748096
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003541223671013039,
      "loss": 1.3282,
      "theoretical_loss": 3.4391895818835514,
      "tokens_seen": 2143813632
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035410230692076226,
      "loss": 1.2795,
      "theoretical_loss": 3.4391810068708493,
      "tokens_seen": 2143879168
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003540822467402207,
      "loss": 1.3227,
      "theoretical_loss": 3.439172432193665,
      "tokens_seen": 2143944704
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035406218655967904,
      "loss": 1.3925,
      "theoretical_loss": 3.439163857851975,
      "tokens_seen": 2144010240
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003540421263791374,
      "loss": 1.3013,
      "theoretical_loss": 3.439155283845757,
      "tokens_seen": 2144075776
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035402206619859577,
      "loss": 1.402,
      "theoretical_loss": 3.4391467101749864,
      "tokens_seen": 2144141312
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003540020060180542,
      "loss": 1.4672,
      "theoretical_loss": 3.4391381368396408,
      "tokens_seen": 2144206848
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035398194583751254,
      "loss": 1.4071,
      "theoretical_loss": 3.439129563839696,
      "tokens_seen": 2144272384
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003539618856569709,
      "loss": 1.4085,
      "theoretical_loss": 3.4391209911751286,
      "tokens_seen": 2144337920
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035394182547642927,
      "loss": 1.3684,
      "theoretical_loss": 3.4391124188459163,
      "tokens_seen": 2144403456
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035392176529588763,
      "loss": 1.4489,
      "theoretical_loss": 3.439103846852035,
      "tokens_seen": 2144468992
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035390170511534605,
      "loss": 1.3351,
      "theoretical_loss": 3.439095275193461,
      "tokens_seen": 2144534528
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003538816449348044,
      "loss": 1.2619,
      "theoretical_loss": 3.4390867038701716,
      "tokens_seen": 2144600064
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035386158475426277,
      "loss": 1.3464,
      "theoretical_loss": 3.4390781328821425,
      "tokens_seen": 2144665600
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035384152457372113,
      "loss": 1.2511,
      "theoretical_loss": 3.439069562229352,
      "tokens_seen": 2144731136
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035382146439317955,
      "loss": 1.4257,
      "theoretical_loss": 3.4390609919117745,
      "tokens_seen": 2144796672
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003538014042126379,
      "loss": 1.5225,
      "theoretical_loss": 3.439052421929389,
      "tokens_seen": 2144862208
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003537813440320963,
      "loss": 1.371,
      "theoretical_loss": 3.4390438522821705,
      "tokens_seen": 2144927744
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035376128385155464,
      "loss": 1.3008,
      "theoretical_loss": 3.4390352829700963,
      "tokens_seen": 2144993280
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000353741223671013,
      "loss": 1.3367,
      "theoretical_loss": 3.439026713993143,
      "tokens_seen": 2145058816
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003537211634904714,
      "loss": 1.3325,
      "theoretical_loss": 3.439018145351287,
      "tokens_seen": 2145124352
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003537011033099298,
      "loss": 1.4386,
      "theoretical_loss": 3.439009577044505,
      "tokens_seen": 2145189888
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035368104312938814,
      "loss": 1.4101,
      "theoretical_loss": 3.4390010090727743,
      "tokens_seen": 2145255424
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003536609829488465,
      "loss": 1.3979,
      "theoretical_loss": 3.4389924414360706,
      "tokens_seen": 2145320960
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003536409227683049,
      "loss": 1.3974,
      "theoretical_loss": 3.4389838741343715,
      "tokens_seen": 2145386496
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003536208625877633,
      "loss": 1.3918,
      "theoretical_loss": 3.438975307167653,
      "tokens_seen": 2145452032
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003536008024072217,
      "loss": 1.283,
      "theoretical_loss": 3.438966740535892,
      "tokens_seen": 2145517568
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035358074222668006,
      "loss": 1.3702,
      "theoretical_loss": 3.4389581742390654,
      "tokens_seen": 2145583104
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003535606820461384,
      "loss": 1.43,
      "theoretical_loss": 3.438949608277149,
      "tokens_seen": 2145648640
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035354062186559684,
      "loss": 1.3781,
      "theoretical_loss": 3.4389410426501206,
      "tokens_seen": 2145714176
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003535205616850552,
      "loss": 1.3686,
      "theoretical_loss": 3.438932477357956,
      "tokens_seen": 2145779712
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035350050150451356,
      "loss": 1.3617,
      "theoretical_loss": 3.438923912400633,
      "tokens_seen": 2145845248
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003534804413239719,
      "loss": 1.2793,
      "theoretical_loss": 3.438915347778127,
      "tokens_seen": 2145910784
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035346038114343034,
      "loss": 1.4175,
      "theoretical_loss": 3.438906783490415,
      "tokens_seen": 2145976320
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003534403209628887,
      "loss": 1.3709,
      "theoretical_loss": 3.4388982195374744,
      "tokens_seen": 2146041856
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035342026078234707,
      "loss": 1.4501,
      "theoretical_loss": 3.438889655919281,
      "tokens_seen": 2146107392
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035340020060180543,
      "loss": 1.4604,
      "theoretical_loss": 3.4388810926358127,
      "tokens_seen": 2146172928
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035338014042126384,
      "loss": 1.4015,
      "theoretical_loss": 3.438872529687045,
      "tokens_seen": 2146238464
    },
    {
      "epoch": 0.65,
      "objective/train/advantage_avg": 0.49193495512008667,
      "objective/train/docs_used": 781712,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2377395629882812,
      "objective/train/original_loss": 1.2377395629882812,
      "objective/train/theoretical_loss": 3.4388639670729546,
      "objective/train/tokens_used": 2166764000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24258702993392944,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050426721572876,
      "objective/train/weighted_lm_loss": 1.2999812364578247,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4388639670729546,
      "tokens_seen": 2146304000
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003533600802407222,
      "loss": 1.3294,
      "theoretical_loss": 3.4388639670729546,
      "tokens_seen": 2146304000
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035334002006018057,
      "loss": 1.3394,
      "theoretical_loss": 3.438855404793519,
      "tokens_seen": 2146369536
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035331995987963893,
      "loss": 1.3318,
      "theoretical_loss": 3.4388468428487142,
      "tokens_seen": 2146435072
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003532998996990973,
      "loss": 1.2799,
      "theoretical_loss": 3.4388382812385174,
      "tokens_seen": 2146500608
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003532798395185557,
      "loss": 1.3402,
      "theoretical_loss": 3.4388297199629054,
      "tokens_seen": 2146566144
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035325977933801407,
      "loss": 1.4666,
      "theoretical_loss": 3.4388211590218543,
      "tokens_seen": 2146631680
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035323971915747243,
      "loss": 1.3631,
      "theoretical_loss": 3.4388125984153413,
      "tokens_seen": 2146697216
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003532196589769308,
      "loss": 1.4174,
      "theoretical_loss": 3.438804038143343,
      "tokens_seen": 2146762752
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003531995987963892,
      "loss": 1.3527,
      "theoretical_loss": 3.4387954782058356,
      "tokens_seen": 2146828288
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003531795386158476,
      "loss": 1.3273,
      "theoretical_loss": 3.438786918602797,
      "tokens_seen": 2146893824
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035315947843530594,
      "loss": 1.4234,
      "theoretical_loss": 3.4387783593342025,
      "tokens_seen": 2146959360
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003531394182547643,
      "loss": 1.4358,
      "theoretical_loss": 3.4387698004000304,
      "tokens_seen": 2147024896
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035311935807422266,
      "loss": 1.3981,
      "theoretical_loss": 3.4387612418002558,
      "tokens_seen": 2147090432
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003530992978936811,
      "loss": 1.4547,
      "theoretical_loss": 3.4387526835348567,
      "tokens_seen": 2147155968
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035307923771313944,
      "loss": 1.5062,
      "theoretical_loss": 3.438744125603809,
      "tokens_seen": 2147221504
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003530591775325978,
      "loss": 1.3239,
      "theoretical_loss": 3.4387355680070897,
      "tokens_seen": 2147287040
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035303911735205617,
      "loss": 1.4055,
      "theoretical_loss": 3.4387270107446763,
      "tokens_seen": 2147352576
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003530190571715146,
      "loss": 1.3085,
      "theoretical_loss": 3.438718453816544,
      "tokens_seen": 2147418112
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035299899699097294,
      "loss": 1.381,
      "theoretical_loss": 3.4387098972226706,
      "tokens_seen": 2147483648
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003529789368104313,
      "loss": 1.2744,
      "theoretical_loss": 3.438701340963033,
      "tokens_seen": 2147549184
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035295887662988967,
      "loss": 1.2181,
      "theoretical_loss": 3.4386927850376074,
      "tokens_seen": 2147614720
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035293881644934803,
      "loss": 1.2043,
      "theoretical_loss": 3.4386842294463706,
      "tokens_seen": 2147680256
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035291875626880645,
      "loss": 1.4643,
      "theoretical_loss": 3.4386756741892994,
      "tokens_seen": 2147745792
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003528986960882648,
      "loss": 1.3493,
      "theoretical_loss": 3.4386671192663707,
      "tokens_seen": 2147811328
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035287863590772317,
      "loss": 1.5142,
      "theoretical_loss": 3.438658564677561,
      "tokens_seen": 2147876864
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035285857572718153,
      "loss": 1.413,
      "theoretical_loss": 3.4386500104228475,
      "tokens_seen": 2147942400
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035283851554663995,
      "loss": 1.4924,
      "theoretical_loss": 3.4386414565022063,
      "tokens_seen": 2148007936
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003528184553660983,
      "loss": 1.3908,
      "theoretical_loss": 3.4386329029156153,
      "tokens_seen": 2148073472
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003527983951855567,
      "loss": 1.4379,
      "theoretical_loss": 3.4386243496630495,
      "tokens_seen": 2148139008
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035277833500501504,
      "loss": 1.4792,
      "theoretical_loss": 3.4386157967444877,
      "tokens_seen": 2148204544
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003527582748244734,
      "loss": 1.2298,
      "theoretical_loss": 3.438607244159905,
      "tokens_seen": 2148270080
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003527382146439318,
      "loss": 1.2849,
      "theoretical_loss": 3.438598691909279,
      "tokens_seen": 2148335616
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003527181544633902,
      "loss": 1.4146,
      "theoretical_loss": 3.438590139992586,
      "tokens_seen": 2148401152
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035269809428284854,
      "loss": 1.3299,
      "theoretical_loss": 3.4385815884098037,
      "tokens_seen": 2148466688
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003526780341023069,
      "loss": 1.4048,
      "theoretical_loss": 3.4385730371609076,
      "tokens_seen": 2148532224
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003526579739217653,
      "loss": 1.335,
      "theoretical_loss": 3.438564486245875,
      "tokens_seen": 2148597760
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003526379137412237,
      "loss": 1.4475,
      "theoretical_loss": 3.4385559356646835,
      "tokens_seen": 2148663296
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035261785356068204,
      "loss": 1.3569,
      "theoretical_loss": 3.4385473854173085,
      "tokens_seen": 2148728832
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003525977933801404,
      "loss": 1.4423,
      "theoretical_loss": 3.438538835503728,
      "tokens_seen": 2148794368
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035257773319959877,
      "loss": 1.3545,
      "theoretical_loss": 3.4385302859239184,
      "tokens_seen": 2148859904
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003525576730190572,
      "loss": 1.4471,
      "theoretical_loss": 3.4385217366778558,
      "tokens_seen": 2148925440
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035253761283851555,
      "loss": 1.4004,
      "theoretical_loss": 3.438513187765518,
      "tokens_seen": 2148990976
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003525175526579739,
      "loss": 1.4296,
      "theoretical_loss": 3.438504639186881,
      "tokens_seen": 2149056512
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035249749247743227,
      "loss": 1.3484,
      "theoretical_loss": 3.438496090941922,
      "tokens_seen": 2149122048
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003524774322968907,
      "loss": 1.3514,
      "theoretical_loss": 3.438487543030618,
      "tokens_seen": 2149187584
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035245737211634905,
      "loss": 1.4228,
      "theoretical_loss": 3.4384789954529453,
      "tokens_seen": 2149253120
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003524373119358074,
      "loss": 1.3139,
      "theoretical_loss": 3.4384704482088813,
      "tokens_seen": 2149318656
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003524172517552658,
      "loss": 1.2627,
      "theoretical_loss": 3.438461901298402,
      "tokens_seen": 2149384192
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035239719157472414,
      "loss": 1.4471,
      "theoretical_loss": 3.4384533547214846,
      "tokens_seen": 2149449728
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035237713139418255,
      "loss": 1.3362,
      "theoretical_loss": 3.4384448084781063,
      "tokens_seen": 2149515264
    },
    {
      "epoch": 0.65,
      "objective/train/advantage_avg": 0.49539026618003845,
      "objective/train/docs_used": 782928,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1521631479263306,
      "objective/train/original_loss": 1.152163028717041,
      "objective/train/theoretical_loss": 3.4384362625682434,
      "objective/train/tokens_used": 2170040800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24559660255908966,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507876873016357,
      "objective/train/weighted_lm_loss": 1.2107012271881104,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 3.4384362625682434,
      "tokens_seen": 2149580800
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003523570712136409,
      "loss": 1.2322,
      "theoretical_loss": 3.4384362625682434,
      "tokens_seen": 2149580800
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003523370110330993,
      "loss": 1.2861,
      "theoretical_loss": 3.4384277169918733,
      "tokens_seen": 2149646336
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035231695085255764,
      "loss": 1.2989,
      "theoretical_loss": 3.438419171748972,
      "tokens_seen": 2149711872
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035229689067201606,
      "loss": 1.4503,
      "theoretical_loss": 3.4384106268395174,
      "tokens_seen": 2149777408
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003522768304914744,
      "loss": 1.4069,
      "theoretical_loss": 3.438402082263485,
      "tokens_seen": 2149842944
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003522567703109328,
      "loss": 1.387,
      "theoretical_loss": 3.4383935380208523,
      "tokens_seen": 2149908480
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035223671013039114,
      "loss": 1.2897,
      "theoretical_loss": 3.4383849941115967,
      "tokens_seen": 2149974016
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035221664994984956,
      "loss": 1.4335,
      "theoretical_loss": 3.438376450535694,
      "tokens_seen": 2150039552
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003521965897693079,
      "loss": 1.5318,
      "theoretical_loss": 3.4383679072931215,
      "tokens_seen": 2150105088
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003521765295887663,
      "loss": 1.2666,
      "theoretical_loss": 3.438359364383856,
      "tokens_seen": 2150170624
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035215646940822465,
      "loss": 1.2815,
      "theoretical_loss": 3.438350821807875,
      "tokens_seen": 2150236160
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000352136409227683,
      "loss": 1.4109,
      "theoretical_loss": 3.4383422795651537,
      "tokens_seen": 2150301696
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003521163490471414,
      "loss": 1.4741,
      "theoretical_loss": 3.438333737655671,
      "tokens_seen": 2150367232
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035209628886659984,
      "loss": 1.4307,
      "theoretical_loss": 3.4383251960794015,
      "tokens_seen": 2150432768
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003520762286860582,
      "loss": 1.4475,
      "theoretical_loss": 3.438316654836324,
      "tokens_seen": 2150498304
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035205616850551657,
      "loss": 1.3795,
      "theoretical_loss": 3.4383081139264147,
      "tokens_seen": 2150563840
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000352036108324975,
      "loss": 1.4374,
      "theoretical_loss": 3.43829957334965,
      "tokens_seen": 2150629376
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035201604814443334,
      "loss": 1.3697,
      "theoretical_loss": 3.4382910331060073,
      "tokens_seen": 2150694912
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003519959879638917,
      "loss": 1.3439,
      "theoretical_loss": 3.438282493195463,
      "tokens_seen": 2150760448
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035197592778335007,
      "loss": 1.3542,
      "theoretical_loss": 3.438273953617995,
      "tokens_seen": 2150825984
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035195586760280843,
      "loss": 1.3698,
      "theoretical_loss": 3.4382654143735785,
      "tokens_seen": 2150891520
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035193580742226685,
      "loss": 1.3823,
      "theoretical_loss": 3.4382568754621916,
      "tokens_seen": 2150957056
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003519157472417252,
      "loss": 1.4945,
      "theoretical_loss": 3.4382483368838113,
      "tokens_seen": 2151022592
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035189568706118357,
      "loss": 1.3208,
      "theoretical_loss": 3.4382397986384134,
      "tokens_seen": 2151088128
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035187562688064193,
      "loss": 1.433,
      "theoretical_loss": 3.438231260725975,
      "tokens_seen": 2151153664
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035185556670010035,
      "loss": 1.3653,
      "theoretical_loss": 3.438222723146474,
      "tokens_seen": 2151219200
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003518355065195587,
      "loss": 1.2413,
      "theoretical_loss": 3.438214185899886,
      "tokens_seen": 2151284736
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003518154463390171,
      "loss": 1.4358,
      "theoretical_loss": 3.438205648986189,
      "tokens_seen": 2151350272
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035179538615847544,
      "loss": 1.454,
      "theoretical_loss": 3.4381971124053594,
      "tokens_seen": 2151415808
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003517753259779338,
      "loss": 1.4526,
      "theoretical_loss": 3.4381885761573736,
      "tokens_seen": 2151481344
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003517552657973922,
      "loss": 1.368,
      "theoretical_loss": 3.4381800402422087,
      "tokens_seen": 2151546880
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003517352056168506,
      "loss": 1.4061,
      "theoretical_loss": 3.4381715046598424,
      "tokens_seen": 2151612416
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035171514543630894,
      "loss": 1.4223,
      "theoretical_loss": 3.438162969410251,
      "tokens_seen": 2151677952
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003516950852557673,
      "loss": 1.3985,
      "theoretical_loss": 3.438154434493411,
      "tokens_seen": 2151743488
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003516750250752257,
      "loss": 1.4495,
      "theoretical_loss": 3.4381458999093,
      "tokens_seen": 2151809024
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003516549648946841,
      "loss": 1.4204,
      "theoretical_loss": 3.4381373656578944,
      "tokens_seen": 2151874560
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035163490471414244,
      "loss": 1.3342,
      "theoretical_loss": 3.438128831739171,
      "tokens_seen": 2151940096
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003516148445336008,
      "loss": 1.4676,
      "theoretical_loss": 3.4381202981531076,
      "tokens_seen": 2152005632
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035159478435305917,
      "loss": 1.426,
      "theoretical_loss": 3.4381117648996797,
      "tokens_seen": 2152071168
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003515747241725176,
      "loss": 1.4038,
      "theoretical_loss": 3.438103231978866,
      "tokens_seen": 2152136704
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035155466399197595,
      "loss": 1.4178,
      "theoretical_loss": 3.4380946993906414,
      "tokens_seen": 2152202240
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003515346038114343,
      "loss": 1.348,
      "theoretical_loss": 3.4380861671349843,
      "tokens_seen": 2152267776
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035151454363089267,
      "loss": 1.4294,
      "theoretical_loss": 3.438077635211871,
      "tokens_seen": 2152333312
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003514944834503511,
      "loss": 1.3727,
      "theoretical_loss": 3.4380691036212783,
      "tokens_seen": 2152398848
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035147442326980945,
      "loss": 1.3583,
      "theoretical_loss": 3.4380605723631836,
      "tokens_seen": 2152464384
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003514543630892678,
      "loss": 1.3643,
      "theoretical_loss": 3.4380520414375635,
      "tokens_seen": 2152529920
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003514343029087262,
      "loss": 1.3833,
      "theoretical_loss": 3.4380435108443947,
      "tokens_seen": 2152595456
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035141424272818454,
      "loss": 1.4214,
      "theoretical_loss": 3.438034980583655,
      "tokens_seen": 2152660992
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035139418254764295,
      "loss": 1.4442,
      "theoretical_loss": 3.4380264506553204,
      "tokens_seen": 2152726528
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003513741223671013,
      "loss": 1.3747,
      "theoretical_loss": 3.438017921059368,
      "tokens_seen": 2152792064
    },
    {
      "epoch": 0.65,
      "objective/train/advantage_avg": 0.49731355905532837,
      "objective/train/docs_used": 784071,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.337634801864624,
      "objective/train/original_loss": 1.3376349210739136,
      "objective/train/theoretical_loss": 3.4380093917957755,
      "objective/train/tokens_used": 2173317600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2474249303340912,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0509895086288452,
      "objective/train/weighted_lm_loss": 1.405828833580017,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4380093917957755,
      "tokens_seen": 2152857600
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003513540621865597,
      "loss": 1.4576,
      "theoretical_loss": 3.4380093917957755,
      "tokens_seen": 2152857600
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035133400200601804,
      "loss": 1.3451,
      "theoretical_loss": 3.4380008628645182,
      "tokens_seen": 2152923136
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035131394182547646,
      "loss": 1.3673,
      "theoretical_loss": 3.4379923342655747,
      "tokens_seen": 2152988672
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003512938816449348,
      "loss": 1.3651,
      "theoretical_loss": 3.4379838059989214,
      "tokens_seen": 2153054208
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003512738214643932,
      "loss": 1.5107,
      "theoretical_loss": 3.4379752780645347,
      "tokens_seen": 2153119744
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035125376128385154,
      "loss": 1.2908,
      "theoretical_loss": 3.437966750462392,
      "tokens_seen": 2153185280
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035123370110330996,
      "loss": 1.383,
      "theoretical_loss": 3.437958223192471,
      "tokens_seen": 2153250816
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003512136409227683,
      "loss": 1.4529,
      "theoretical_loss": 3.437949696254747,
      "tokens_seen": 2153316352
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003511935807422267,
      "loss": 1.3913,
      "theoretical_loss": 3.437941169649198,
      "tokens_seen": 2153381888
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035117352056168505,
      "loss": 1.4803,
      "theoretical_loss": 3.4379326433758006,
      "tokens_seen": 2153447424
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003511534603811434,
      "loss": 1.4215,
      "theoretical_loss": 3.4379241174345325,
      "tokens_seen": 2153512960
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003511334002006018,
      "loss": 1.4877,
      "theoretical_loss": 3.4379155918253694,
      "tokens_seen": 2153578496
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003511133400200602,
      "loss": 1.243,
      "theoretical_loss": 3.4379070665482896,
      "tokens_seen": 2153644032
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035109327983951855,
      "loss": 1.2561,
      "theoretical_loss": 3.437898541603269,
      "tokens_seen": 2153709568
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003510732196589769,
      "loss": 1.2839,
      "theoretical_loss": 3.437890016990285,
      "tokens_seen": 2153775104
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035105315947843533,
      "loss": 1.3663,
      "theoretical_loss": 3.437881492709314,
      "tokens_seen": 2153840640
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003510330992978937,
      "loss": 1.3698,
      "theoretical_loss": 3.4378729687603347,
      "tokens_seen": 2153906176
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035101303911735205,
      "loss": 1.3933,
      "theoretical_loss": 3.4378644451433216,
      "tokens_seen": 2153971712
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003509929789368104,
      "loss": 1.4191,
      "theoretical_loss": 3.4378559218582536,
      "tokens_seen": 2154037248
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003509729187562688,
      "loss": 1.3339,
      "theoretical_loss": 3.437847398905107,
      "tokens_seen": 2154102784
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003509528585757272,
      "loss": 1.2398,
      "theoretical_loss": 3.4378388762838585,
      "tokens_seen": 2154168320
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035093279839518556,
      "loss": 1.4197,
      "theoretical_loss": 3.437830353994485,
      "tokens_seen": 2154233856
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003509127382146439,
      "loss": 1.3989,
      "theoretical_loss": 3.4378218320369647,
      "tokens_seen": 2154299392
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003508926780341023,
      "loss": 1.274,
      "theoretical_loss": 3.4378133104112734,
      "tokens_seen": 2154364928
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003508726178535607,
      "loss": 1.4301,
      "theoretical_loss": 3.437804789117388,
      "tokens_seen": 2154430464
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035085255767301906,
      "loss": 1.4165,
      "theoretical_loss": 3.4377962681552865,
      "tokens_seen": 2154496000
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003508324974924774,
      "loss": 1.2855,
      "theoretical_loss": 3.437787747524945,
      "tokens_seen": 2154561536
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003508124373119358,
      "loss": 1.358,
      "theoretical_loss": 3.4377792272263408,
      "tokens_seen": 2154627072
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035079237713139415,
      "loss": 1.3249,
      "theoretical_loss": 3.4377707072594506,
      "tokens_seen": 2154692608
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035077231695085256,
      "loss": 1.4315,
      "theoretical_loss": 3.4377621876242523,
      "tokens_seen": 2154758144
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003507522567703109,
      "loss": 1.4211,
      "theoretical_loss": 3.4377536683207217,
      "tokens_seen": 2154823680
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003507321965897693,
      "loss": 1.4052,
      "theoretical_loss": 3.4377451493488365,
      "tokens_seen": 2154889216
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035071213640922765,
      "loss": 1.3444,
      "theoretical_loss": 3.4377366307085735,
      "tokens_seen": 2154954752
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035069207622868607,
      "loss": 1.3683,
      "theoretical_loss": 3.4377281123999097,
      "tokens_seen": 2155020288
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035067201604814443,
      "loss": 1.4602,
      "theoretical_loss": 3.4377195944228225,
      "tokens_seen": 2155085824
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003506519558676028,
      "loss": 1.2594,
      "theoretical_loss": 3.4377110767772887,
      "tokens_seen": 2155151360
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035063189568706115,
      "loss": 1.3502,
      "theoretical_loss": 3.4377025594632844,
      "tokens_seen": 2155216896
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003506118355065195,
      "loss": 1.4024,
      "theoretical_loss": 3.4376940424807882,
      "tokens_seen": 2155282432
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000350591775325978,
      "loss": 1.4199,
      "theoretical_loss": 3.4376855258297763,
      "tokens_seen": 2155347968
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035057171514543635,
      "loss": 1.3662,
      "theoretical_loss": 3.4376770095102254,
      "tokens_seen": 2155413504
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003505516549648947,
      "loss": 1.3309,
      "theoretical_loss": 3.437668493522113,
      "tokens_seen": 2155479040
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035053159478435307,
      "loss": 1.3285,
      "theoretical_loss": 3.437659977865416,
      "tokens_seen": 2155544576
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003505115346038115,
      "loss": 1.4842,
      "theoretical_loss": 3.4376514625401113,
      "tokens_seen": 2155610112
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035049147442326985,
      "loss": 1.4226,
      "theoretical_loss": 3.437642947546176,
      "tokens_seen": 2155675648
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003504714142427282,
      "loss": 1.4635,
      "theoretical_loss": 3.4376344328835877,
      "tokens_seen": 2155741184
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003504513540621866,
      "loss": 1.294,
      "theoretical_loss": 3.437625918552323,
      "tokens_seen": 2155806720
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035043129388164494,
      "loss": 1.3254,
      "theoretical_loss": 3.437617404552358,
      "tokens_seen": 2155872256
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035041123370110335,
      "loss": 1.369,
      "theoretical_loss": 3.4376088908836717,
      "tokens_seen": 2155937792
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003503911735205617,
      "loss": 1.3986,
      "theoretical_loss": 3.4376003775462394,
      "tokens_seen": 2156003328
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003503711133400201,
      "loss": 1.4694,
      "theoretical_loss": 3.4375918645400385,
      "tokens_seen": 2156068864
    },
    {
      "epoch": 0.65,
      "objective/train/advantage_avg": 0.49280810356140137,
      "objective/train/docs_used": 785270,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3187788724899292,
      "objective/train/original_loss": 1.3187787532806396,
      "objective/train/theoretical_loss": 3.4375833518650465,
      "objective/train/tokens_used": 2176594400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24344173073768616,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05051851272583,
      "objective/train/weighted_lm_loss": 1.3854244947433472,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.4375833518650465,
      "tokens_seen": 2156134400
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035035105315947844,
      "loss": 1.4005,
      "theoretical_loss": 3.4375833518650465,
      "tokens_seen": 2156134400
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035033099297893686,
      "loss": 1.3544,
      "theoretical_loss": 3.4375748395212407,
      "tokens_seen": 2156199936
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003503109327983952,
      "loss": 1.3348,
      "theoretical_loss": 3.4375663275085975,
      "tokens_seen": 2156265472
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003502908726178536,
      "loss": 1.4065,
      "theoretical_loss": 3.437557815827094,
      "tokens_seen": 2156331008
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035027081243731194,
      "loss": 1.3959,
      "theoretical_loss": 3.4375493044767076,
      "tokens_seen": 2156396544
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035025075225677036,
      "loss": 1.3647,
      "theoretical_loss": 3.437540793457415,
      "tokens_seen": 2156462080
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003502306920762287,
      "loss": 1.3594,
      "theoretical_loss": 3.437532282769194,
      "tokens_seen": 2156527616
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003502106318956871,
      "loss": 1.3201,
      "theoretical_loss": 3.4375237724120202,
      "tokens_seen": 2156593152
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035019057171514545,
      "loss": 1.335,
      "theoretical_loss": 3.4375152623858725,
      "tokens_seen": 2156658688
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003501705115346038,
      "loss": 1.4904,
      "theoretical_loss": 3.4375067526907266,
      "tokens_seen": 2156724224
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003501504513540622,
      "loss": 1.3128,
      "theoretical_loss": 3.43749824332656,
      "tokens_seen": 2156789760
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003501303911735206,
      "loss": 1.3544,
      "theoretical_loss": 3.43748973429335,
      "tokens_seen": 2156855296
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035011033099297895,
      "loss": 1.3392,
      "theoretical_loss": 3.437481225591073,
      "tokens_seen": 2156920832
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003500902708124373,
      "loss": 1.4238,
      "theoretical_loss": 3.437472717219707,
      "tokens_seen": 2156986368
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035007021063189573,
      "loss": 1.4227,
      "theoretical_loss": 3.437464209179228,
      "tokens_seen": 2157051904
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003500501504513541,
      "loss": 1.3428,
      "theoretical_loss": 3.4374557014696143,
      "tokens_seen": 2157117440
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00035003009027081245,
      "loss": 1.3227,
      "theoretical_loss": 3.4374471940908418,
      "tokens_seen": 2157182976
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003500100300902708,
      "loss": 1.4215,
      "theoretical_loss": 3.4374386870428886,
      "tokens_seen": 2157248512
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003499899699097292,
      "loss": 1.3125,
      "theoretical_loss": 3.437430180325731,
      "tokens_seen": 2157314048
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003499699097291876,
      "loss": 1.2806,
      "theoretical_loss": 3.4374216739393466,
      "tokens_seen": 2157379584
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034994984954864596,
      "loss": 1.3474,
      "theoretical_loss": 3.4374131678837125,
      "tokens_seen": 2157445120
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003499297893681043,
      "loss": 1.4163,
      "theoretical_loss": 3.437404662158806,
      "tokens_seen": 2157510656
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003499097291875627,
      "loss": 1.3444,
      "theoretical_loss": 3.4373961567646028,
      "tokens_seen": 2157576192
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003498896690070211,
      "loss": 1.3992,
      "theoretical_loss": 3.4373876517010813,
      "tokens_seen": 2157641728
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034986960882647946,
      "loss": 1.3194,
      "theoretical_loss": 3.4373791469682184,
      "tokens_seen": 2157707264
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003498495486459378,
      "loss": 1.2803,
      "theoretical_loss": 3.437370642565991,
      "tokens_seen": 2157772800
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003498294884653962,
      "loss": 1.3342,
      "theoretical_loss": 3.4373621384943767,
      "tokens_seen": 2157838336
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034980942828485455,
      "loss": 1.4852,
      "theoretical_loss": 3.437353634753352,
      "tokens_seen": 2157903872
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034978936810431296,
      "loss": 1.3798,
      "theoretical_loss": 3.437345131342894,
      "tokens_seen": 2157969408
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003497693079237713,
      "loss": 1.4192,
      "theoretical_loss": 3.4373366282629805,
      "tokens_seen": 2158034944
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003497492477432297,
      "loss": 1.3256,
      "theoretical_loss": 3.4373281255135875,
      "tokens_seen": 2158100480
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034972918756268805,
      "loss": 1.4359,
      "theoretical_loss": 3.437319623094693,
      "tokens_seen": 2158166016
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034970912738214647,
      "loss": 1.338,
      "theoretical_loss": 3.437311121006274,
      "tokens_seen": 2158231552
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034968906720160483,
      "loss": 1.4141,
      "theoretical_loss": 3.4373026192483076,
      "tokens_seen": 2158297088
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003496690070210632,
      "loss": 1.3858,
      "theoretical_loss": 3.4372941178207705,
      "tokens_seen": 2158362624
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034964894684052155,
      "loss": 1.3529,
      "theoretical_loss": 3.4372856167236403,
      "tokens_seen": 2158428160
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003496288866599799,
      "loss": 1.403,
      "theoretical_loss": 3.4372771159568942,
      "tokens_seen": 2158493696
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034960882647943833,
      "loss": 1.4074,
      "theoretical_loss": 3.4372686155205088,
      "tokens_seen": 2158559232
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003495887662988967,
      "loss": 1.4704,
      "theoretical_loss": 3.4372601154144617,
      "tokens_seen": 2158624768
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034956870611835506,
      "loss": 1.3485,
      "theoretical_loss": 3.43725161563873,
      "tokens_seen": 2158690304
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003495486459378134,
      "loss": 1.4041,
      "theoretical_loss": 3.43724311619329,
      "tokens_seen": 2158755840
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034952858575727183,
      "loss": 1.3516,
      "theoretical_loss": 3.4372346170781203,
      "tokens_seen": 2158821376
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003495085255767302,
      "loss": 1.3774,
      "theoretical_loss": 3.4372261182931974,
      "tokens_seen": 2158886912
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034948846539618856,
      "loss": 1.3556,
      "theoretical_loss": 3.437217619838498,
      "tokens_seen": 2158952448
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003494684052156469,
      "loss": 1.3941,
      "theoretical_loss": 3.437209121713999,
      "tokens_seen": 2159017984
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003494483450351053,
      "loss": 1.3007,
      "theoretical_loss": 3.437200623919679,
      "tokens_seen": 2159083520
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003494282848545637,
      "loss": 1.3882,
      "theoretical_loss": 3.437192126455514,
      "tokens_seen": 2159149056
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034940822467402206,
      "loss": 1.4621,
      "theoretical_loss": 3.4371836293214812,
      "tokens_seen": 2159214592
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003493881644934804,
      "loss": 1.3999,
      "theoretical_loss": 3.4371751325175586,
      "tokens_seen": 2159280128
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003493681043129388,
      "loss": 1.3861,
      "theoretical_loss": 3.437166636043722,
      "tokens_seen": 2159345664
    },
    {
      "epoch": 0.65,
      "objective/train/advantage_avg": 0.48937565088272095,
      "objective/train/docs_used": 786428,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3131507635116577,
      "objective/train/original_loss": 1.3131505250930786,
      "objective/train/theoretical_loss": 3.4371581398999496,
      "objective/train/tokens_used": 2179871200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24035921692848206,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501593351364136,
      "objective/train/weighted_lm_loss": 1.3792510032653809,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4371581398999496,
      "tokens_seen": 2159411200
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003493480441323972,
      "loss": 1.3195,
      "theoretical_loss": 3.4371581398999496,
      "tokens_seen": 2159411200
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034932798395185557,
      "loss": 1.3252,
      "theoretical_loss": 3.4371496440862184,
      "tokens_seen": 2159476736
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034930792377131393,
      "loss": 1.4151,
      "theoretical_loss": 3.437141148602505,
      "tokens_seen": 2159542272
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003492878635907723,
      "loss": 1.361,
      "theoretical_loss": 3.437132653448787,
      "tokens_seen": 2159607808
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034926780341023065,
      "loss": 1.2817,
      "theoretical_loss": 3.437124158625042,
      "tokens_seen": 2159673344
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034924774322968907,
      "loss": 1.3481,
      "theoretical_loss": 3.4371156641312464,
      "tokens_seen": 2159738880
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034922768304914743,
      "loss": 1.343,
      "theoretical_loss": 3.437107169967378,
      "tokens_seen": 2159804416
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003492076228686058,
      "loss": 1.3999,
      "theoretical_loss": 3.4370986761334135,
      "tokens_seen": 2159869952
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034918756268806416,
      "loss": 1.3761,
      "theoretical_loss": 3.4370901826293303,
      "tokens_seen": 2159935488
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034916750250752257,
      "loss": 1.3319,
      "theoretical_loss": 3.4370816894551055,
      "tokens_seen": 2160001024
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034914744232698093,
      "loss": 1.4465,
      "theoretical_loss": 3.437073196610716,
      "tokens_seen": 2160066560
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003491273821464393,
      "loss": 1.2658,
      "theoretical_loss": 3.43706470409614,
      "tokens_seen": 2160132096
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034910732196589766,
      "loss": 1.3405,
      "theoretical_loss": 3.4370562119113535,
      "tokens_seen": 2160197632
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034908726178535613,
      "loss": 1.3714,
      "theoretical_loss": 3.4370477200563343,
      "tokens_seen": 2160263168
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003490672016048145,
      "loss": 1.3522,
      "theoretical_loss": 3.4370392285310594,
      "tokens_seen": 2160328704
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034904714142427285,
      "loss": 1.3306,
      "theoretical_loss": 3.437030737335506,
      "tokens_seen": 2160394240
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003490270812437312,
      "loss": 1.386,
      "theoretical_loss": 3.437022246469651,
      "tokens_seen": 2160459776
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003490070210631896,
      "loss": 1.3281,
      "theoretical_loss": 3.4370137559334726,
      "tokens_seen": 2160525312
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.000348986960882648,
      "loss": 1.2692,
      "theoretical_loss": 3.437005265726947,
      "tokens_seen": 2160590848
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034896690070210636,
      "loss": 1.3505,
      "theoretical_loss": 3.436996775850052,
      "tokens_seen": 2160656384
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003489468405215647,
      "loss": 1.3848,
      "theoretical_loss": 3.436988286302764,
      "tokens_seen": 2160721920
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003489267803410231,
      "loss": 1.3951,
      "theoretical_loss": 3.4369797970850615,
      "tokens_seen": 2160787456
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003489067201604815,
      "loss": 1.2976,
      "theoretical_loss": 3.4369713081969206,
      "tokens_seen": 2160852992
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034888665997993986,
      "loss": 1.4542,
      "theoretical_loss": 3.436962819638319,
      "tokens_seen": 2160918528
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003488665997993982,
      "loss": 1.4392,
      "theoretical_loss": 3.4369543314092335,
      "tokens_seen": 2160984064
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003488465396188566,
      "loss": 1.4273,
      "theoretical_loss": 3.436945843509642,
      "tokens_seen": 2161049600
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034882647943831495,
      "loss": 1.3186,
      "theoretical_loss": 3.436937355939521,
      "tokens_seen": 2161115136
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034880641925777336,
      "loss": 1.4215,
      "theoretical_loss": 3.436928868698848,
      "tokens_seen": 2161180672
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003487863590772317,
      "loss": 1.3422,
      "theoretical_loss": 3.436920381787601,
      "tokens_seen": 2161246208
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0003487662988966901,
      "loss": 1.5415,
      "theoretical_loss": 3.436911895205756,
      "tokens_seen": 2161311744
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034874623871614845,
      "loss": 1.3691,
      "theoretical_loss": 3.4369034089532904,
      "tokens_seen": 2161377280
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00034872617853560687,
      "loss": 1.3501,
      "theoretical_loss": 3.436894923030182,
      "tokens_seen": 2161442816
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034870611835506523,
      "loss": 1.4456,
      "theoretical_loss": 3.436886437436408,
      "tokens_seen": 2161508352
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003486860581745236,
      "loss": 1.2984,
      "theoretical_loss": 3.4368779521719452,
      "tokens_seen": 2161573888
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034866599799398195,
      "loss": 1.3353,
      "theoretical_loss": 3.436869467236771,
      "tokens_seen": 2161639424
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003486459378134403,
      "loss": 1.2882,
      "theoretical_loss": 3.4368609826308627,
      "tokens_seen": 2161704960
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034862587763289873,
      "loss": 1.4183,
      "theoretical_loss": 3.436852498354198,
      "tokens_seen": 2161770496
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003486058174523571,
      "loss": 1.2473,
      "theoretical_loss": 3.436844014406753,
      "tokens_seen": 2161836032
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034858575727181546,
      "loss": 1.3792,
      "theoretical_loss": 3.4368355307885063,
      "tokens_seen": 2161901568
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003485656970912738,
      "loss": 1.333,
      "theoretical_loss": 3.436827047499434,
      "tokens_seen": 2161967104
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034854563691073223,
      "loss": 1.3931,
      "theoretical_loss": 3.436818564539514,
      "tokens_seen": 2162032640
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003485255767301906,
      "loss": 1.4248,
      "theoretical_loss": 3.4368100819087233,
      "tokens_seen": 2162098176
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034850551654964896,
      "loss": 1.3786,
      "theoretical_loss": 3.436801599607039,
      "tokens_seen": 2162163712
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003484854563691073,
      "loss": 1.5044,
      "theoretical_loss": 3.436793117634439,
      "tokens_seen": 2162229248
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003484653961885657,
      "loss": 1.4432,
      "theoretical_loss": 3.4367846359909002,
      "tokens_seen": 2162294784
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003484453360080241,
      "loss": 1.3484,
      "theoretical_loss": 3.4367761546763993,
      "tokens_seen": 2162360320
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034842527582748246,
      "loss": 1.3586,
      "theoretical_loss": 3.4367676736909143,
      "tokens_seen": 2162425856
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003484052156469408,
      "loss": 1.2696,
      "theoretical_loss": 3.4367591930344226,
      "tokens_seen": 2162491392
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003483851554663992,
      "loss": 1.4039,
      "theoretical_loss": 3.4367507127069006,
      "tokens_seen": 2162556928
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003483650952858576,
      "loss": 1.3897,
      "theoretical_loss": 3.436742232708326,
      "tokens_seen": 2162622464
    },
    {
      "epoch": 0.66,
      "objective/train/advantage_avg": 0.49263080954551697,
      "objective/train/docs_used": 787745,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4425926208496094,
      "objective/train/original_loss": 1.442592740058899,
      "objective/train/theoretical_loss": 3.4367337530386766,
      "objective/train/tokens_used": 2183148000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2432829886674881,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504999160766602,
      "objective/train/weighted_lm_loss": 1.5153738260269165,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.4367337530386766,
      "tokens_seen": 2162688000
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034834503510531597,
      "loss": 1.4962,
      "theoretical_loss": 3.4367337530386766,
      "tokens_seen": 2162688000
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034832497492477433,
      "loss": 1.4782,
      "theoretical_loss": 3.4367252736979292,
      "tokens_seen": 2162753536
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003483049147442327,
      "loss": 1.4277,
      "theoretical_loss": 3.4367167946860606,
      "tokens_seen": 2162819072
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034828485456369105,
      "loss": 1.3298,
      "theoretical_loss": 3.436708316003049,
      "tokens_seen": 2162884608
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034826479438314947,
      "loss": 1.4048,
      "theoretical_loss": 3.436699837648871,
      "tokens_seen": 2162950144
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034824473420260783,
      "loss": 1.5738,
      "theoretical_loss": 3.4366913596235045,
      "tokens_seen": 2163015680
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003482246740220662,
      "loss": 1.4428,
      "theoretical_loss": 3.4366828819269264,
      "tokens_seen": 2163081216
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034820461384152456,
      "loss": 1.4256,
      "theoretical_loss": 3.4366744045591138,
      "tokens_seen": 2163146752
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034818455366098297,
      "loss": 1.3076,
      "theoretical_loss": 3.4366659275200444,
      "tokens_seen": 2163212288
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034816449348044133,
      "loss": 1.4507,
      "theoretical_loss": 3.4366574508096948,
      "tokens_seen": 2163277824
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003481444332998997,
      "loss": 1.3419,
      "theoretical_loss": 3.436648974428043,
      "tokens_seen": 2163343360
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034812437311935806,
      "loss": 1.4349,
      "theoretical_loss": 3.4366404983750667,
      "tokens_seen": 2163408896
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003481043129388164,
      "loss": 1.3777,
      "theoretical_loss": 3.436632022650742,
      "tokens_seen": 2163474432
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034808425275827484,
      "loss": 1.4362,
      "theoretical_loss": 3.436623547255047,
      "tokens_seen": 2163539968
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003480641925777332,
      "loss": 1.4872,
      "theoretical_loss": 3.4366150721879585,
      "tokens_seen": 2163605504
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034804413239719156,
      "loss": 1.4013,
      "theoretical_loss": 3.4366065974494546,
      "tokens_seen": 2163671040
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003480240722166499,
      "loss": 1.3404,
      "theoretical_loss": 3.4365981230395115,
      "tokens_seen": 2163736576
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034800401203610834,
      "loss": 1.3399,
      "theoretical_loss": 3.4365896489581074,
      "tokens_seen": 2163802112
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003479839518555667,
      "loss": 1.428,
      "theoretical_loss": 3.4365811752052196,
      "tokens_seen": 2163867648
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034796389167502506,
      "loss": 1.2807,
      "theoretical_loss": 3.436572701780825,
      "tokens_seen": 2163933184
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034794383149448343,
      "loss": 1.3724,
      "theoretical_loss": 3.436564228684901,
      "tokens_seen": 2163998720
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034792377131394184,
      "loss": 1.3608,
      "theoretical_loss": 3.436555755917425,
      "tokens_seen": 2164064256
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003479037111334002,
      "loss": 1.3672,
      "theoretical_loss": 3.436547283478374,
      "tokens_seen": 2164129792
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034788365095285857,
      "loss": 1.3513,
      "theoretical_loss": 3.436538811367726,
      "tokens_seen": 2164195328
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034786359077231693,
      "loss": 1.3901,
      "theoretical_loss": 3.436530339585458,
      "tokens_seen": 2164260864
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003478435305917753,
      "loss": 1.3253,
      "theoretical_loss": 3.436521868131547,
      "tokens_seen": 2164326400
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003478234704112337,
      "loss": 1.3895,
      "theoretical_loss": 3.436513397005971,
      "tokens_seen": 2164391936
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034780341023069207,
      "loss": 1.4201,
      "theoretical_loss": 3.436504926208707,
      "tokens_seen": 2164457472
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034778335005015043,
      "loss": 1.3836,
      "theoretical_loss": 3.4364964557397317,
      "tokens_seen": 2164523008
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003477632898696088,
      "loss": 1.3734,
      "theoretical_loss": 3.4364879855990234,
      "tokens_seen": 2164588544
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003477432296890672,
      "loss": 1.282,
      "theoretical_loss": 3.436479515786559,
      "tokens_seen": 2164654080
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003477231695085256,
      "loss": 1.397,
      "theoretical_loss": 3.436471046302316,
      "tokens_seen": 2164719616
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034770310932798394,
      "loss": 1.4322,
      "theoretical_loss": 3.436462577146272,
      "tokens_seen": 2164785152
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003476830491474423,
      "loss": 1.4108,
      "theoretical_loss": 3.4364541083184035,
      "tokens_seen": 2164850688
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034766298896690066,
      "loss": 1.3928,
      "theoretical_loss": 3.4364456398186882,
      "tokens_seen": 2164916224
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003476429287863591,
      "loss": 1.3545,
      "theoretical_loss": 3.4364371716471043,
      "tokens_seen": 2164981760
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034762286860581744,
      "loss": 1.3152,
      "theoretical_loss": 3.4364287038036276,
      "tokens_seen": 2165047296
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003476028084252758,
      "loss": 1.4657,
      "theoretical_loss": 3.436420236288237,
      "tokens_seen": 2165112832
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003475827482447342,
      "loss": 1.4216,
      "theoretical_loss": 3.436411769100909,
      "tokens_seen": 2165178368
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034756268806419263,
      "loss": 1.3698,
      "theoretical_loss": 3.436403302241621,
      "tokens_seen": 2165243904
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000347542627883651,
      "loss": 1.3041,
      "theoretical_loss": 3.4363948357103506,
      "tokens_seen": 2165309440
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034752256770310936,
      "loss": 1.4301,
      "theoretical_loss": 3.436386369507075,
      "tokens_seen": 2165374976
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003475025075225677,
      "loss": 1.3835,
      "theoretical_loss": 3.4363779036317714,
      "tokens_seen": 2165440512
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003474824473420261,
      "loss": 1.2919,
      "theoretical_loss": 3.4363694380844176,
      "tokens_seen": 2165506048
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003474623871614845,
      "loss": 1.433,
      "theoretical_loss": 3.436360972864991,
      "tokens_seen": 2165571584
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034744232698094286,
      "loss": 1.3967,
      "theoretical_loss": 3.4363525079734685,
      "tokens_seen": 2165637120
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003474222668004012,
      "loss": 1.3693,
      "theoretical_loss": 3.436344043409828,
      "tokens_seen": 2165702656
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003474022066198596,
      "loss": 1.3133,
      "theoretical_loss": 3.4363355791740458,
      "tokens_seen": 2165768192
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000347382146439318,
      "loss": 1.4617,
      "theoretical_loss": 3.4363271152661006,
      "tokens_seen": 2165833728
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034736208625877637,
      "loss": 1.402,
      "theoretical_loss": 3.436318651685969,
      "tokens_seen": 2165899264
    },
    {
      "epoch": 0.66,
      "objective/train/advantage_avg": 0.49267125129699707,
      "objective/train/docs_used": 788845,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3006417751312256,
      "objective/train/original_loss": 1.3006417751312256,
      "objective/train/theoretical_loss": 3.4363101884336293,
      "objective/train/tokens_used": 2186424800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24353384971618652,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505051612854004,
      "objective/train/weighted_lm_loss": 1.366281270980835,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4363101884336293,
      "tokens_seen": 2165964800
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034734202607823473,
      "loss": 1.407,
      "theoretical_loss": 3.4363101884336293,
      "tokens_seen": 2165964800
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003473219658976931,
      "loss": 1.4142,
      "theoretical_loss": 3.4363017255090575,
      "tokens_seen": 2166030336
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034730190571715145,
      "loss": 1.4349,
      "theoretical_loss": 3.4362932629122325,
      "tokens_seen": 2166095872
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034728184553660987,
      "loss": 1.53,
      "theoretical_loss": 3.43628480064313,
      "tokens_seen": 2166161408
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034726178535606823,
      "loss": 1.5723,
      "theoretical_loss": 3.436276338701729,
      "tokens_seen": 2166226944
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003472417251755266,
      "loss": 1.4615,
      "theoretical_loss": 3.436267877088006,
      "tokens_seen": 2166292480
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034722166499498496,
      "loss": 1.4548,
      "theoretical_loss": 3.436259415801939,
      "tokens_seen": 2166358016
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034720160481444337,
      "loss": 1.3716,
      "theoretical_loss": 3.436250954843504,
      "tokens_seen": 2166423552
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034718154463390173,
      "loss": 1.3286,
      "theoretical_loss": 3.43624249421268,
      "tokens_seen": 2166489088
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003471614844533601,
      "loss": 1.3304,
      "theoretical_loss": 3.436234033909444,
      "tokens_seen": 2166554624
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034714142427281846,
      "loss": 1.3951,
      "theoretical_loss": 3.436225573933773,
      "tokens_seen": 2166620160
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003471213640922768,
      "loss": 1.4828,
      "theoretical_loss": 3.4362171142856446,
      "tokens_seen": 2166685696
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034710130391173524,
      "loss": 1.2735,
      "theoretical_loss": 3.4362086549650366,
      "tokens_seen": 2166751232
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003470812437311936,
      "loss": 1.3763,
      "theoretical_loss": 3.4362001959719253,
      "tokens_seen": 2166816768
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034706118355065196,
      "loss": 1.4124,
      "theoretical_loss": 3.43619173730629,
      "tokens_seen": 2166882304
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003470411233701103,
      "loss": 1.2586,
      "theoretical_loss": 3.4361832789681057,
      "tokens_seen": 2166947840
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034702106318956874,
      "loss": 1.4978,
      "theoretical_loss": 3.4361748209573517,
      "tokens_seen": 2167013376
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003470010030090271,
      "loss": 1.47,
      "theoretical_loss": 3.4361663632740047,
      "tokens_seen": 2167078912
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034698094282848547,
      "loss": 1.3889,
      "theoretical_loss": 3.4361579059180425,
      "tokens_seen": 2167144448
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034696088264794383,
      "loss": 1.4488,
      "theoretical_loss": 3.4361494488894424,
      "tokens_seen": 2167209984
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034694082246740224,
      "loss": 1.3278,
      "theoretical_loss": 3.4361409921881814,
      "tokens_seen": 2167275520
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003469207622868606,
      "loss": 1.387,
      "theoretical_loss": 3.436132535814237,
      "tokens_seen": 2167341056
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034690070210631897,
      "loss": 1.4545,
      "theoretical_loss": 3.4361240797675876,
      "tokens_seen": 2167406592
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034688064192577733,
      "loss": 1.4837,
      "theoretical_loss": 3.4361156240482096,
      "tokens_seen": 2167472128
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003468605817452357,
      "loss": 1.3642,
      "theoretical_loss": 3.4361071686560805,
      "tokens_seen": 2167537664
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003468405215646941,
      "loss": 1.3705,
      "theoretical_loss": 3.4360987135911785,
      "tokens_seen": 2167603200
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034682046138415247,
      "loss": 1.511,
      "theoretical_loss": 3.43609025885348,
      "tokens_seen": 2167668736
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034680040120361083,
      "loss": 1.3046,
      "theoretical_loss": 3.436081804442963,
      "tokens_seen": 2167734272
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003467803410230692,
      "loss": 1.3597,
      "theoretical_loss": 3.4360733503596053,
      "tokens_seen": 2167799808
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003467602808425276,
      "loss": 1.4273,
      "theoretical_loss": 3.4360648966033835,
      "tokens_seen": 2167865344
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000346740220661986,
      "loss": 1.4071,
      "theoretical_loss": 3.436056443174276,
      "tokens_seen": 2167930880
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034672016048144434,
      "loss": 1.471,
      "theoretical_loss": 3.4360479900722596,
      "tokens_seen": 2167996416
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003467001003009027,
      "loss": 1.4435,
      "theoretical_loss": 3.4360395372973125,
      "tokens_seen": 2168061952
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034668004012036106,
      "loss": 1.348,
      "theoretical_loss": 3.436031084849411,
      "tokens_seen": 2168127488
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003466599799398195,
      "loss": 1.4109,
      "theoretical_loss": 3.436022632728533,
      "tokens_seen": 2168193024
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034663991975927784,
      "loss": 1.5279,
      "theoretical_loss": 3.4360141809346567,
      "tokens_seen": 2168258560
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003466198595787362,
      "loss": 1.4264,
      "theoretical_loss": 3.4360057294677584,
      "tokens_seen": 2168324096
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034659979939819456,
      "loss": 1.396,
      "theoretical_loss": 3.4359972783278163,
      "tokens_seen": 2168389632
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000346579739217653,
      "loss": 1.4589,
      "theoretical_loss": 3.4359888275148083,
      "tokens_seen": 2168455168
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034655967903711134,
      "loss": 1.316,
      "theoretical_loss": 3.4359803770287107,
      "tokens_seen": 2168520704
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003465396188565697,
      "loss": 1.375,
      "theoretical_loss": 3.435971926869502,
      "tokens_seen": 2168586240
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034651955867602807,
      "loss": 1.3807,
      "theoretical_loss": 3.435963477037159,
      "tokens_seen": 2168651776
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034649949849548643,
      "loss": 1.3417,
      "theoretical_loss": 3.4359550275316595,
      "tokens_seen": 2168717312
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034647943831494485,
      "loss": 1.3106,
      "theoretical_loss": 3.435946578352981,
      "tokens_seen": 2168782848
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003464593781344032,
      "loss": 1.5016,
      "theoretical_loss": 3.4359381295011007,
      "tokens_seen": 2168848384
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034643931795386157,
      "loss": 1.3682,
      "theoretical_loss": 3.4359296809759963,
      "tokens_seen": 2168913920
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034641925777331993,
      "loss": 1.2999,
      "theoretical_loss": 3.4359212327776456,
      "tokens_seen": 2168979456
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034639919759277835,
      "loss": 1.3437,
      "theoretical_loss": 3.4359127849060256,
      "tokens_seen": 2169044992
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003463791374122367,
      "loss": 1.3338,
      "theoretical_loss": 3.435904337361114,
      "tokens_seen": 2169110528
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003463590772316951,
      "loss": 1.416,
      "theoretical_loss": 3.4358958901428878,
      "tokens_seen": 2169176064
    },
    {
      "epoch": 0.66,
      "objective/train/advantage_avg": 0.4887388050556183,
      "objective/train/docs_used": 790125,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.5084922313690186,
      "objective/train/original_loss": 1.508492112159729,
      "objective/train/theoretical_loss": 3.4358874432513256,
      "objective/train/tokens_used": 2189701600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24017302691936493,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500946044921875,
      "objective/train/weighted_lm_loss": 1.58432936668396,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4358874432513256,
      "tokens_seen": 2169241600
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034633901705115344,
      "loss": 1.4582,
      "theoretical_loss": 3.4358874432513256,
      "tokens_seen": 2169241600
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003463189568706118,
      "loss": 1.3305,
      "theoretical_loss": 3.435878996686404,
      "tokens_seen": 2169307136
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003462988966900702,
      "loss": 1.3061,
      "theoretical_loss": 3.435870550448101,
      "tokens_seen": 2169372672
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003462788365095286,
      "loss": 1.3505,
      "theoretical_loss": 3.435862104536393,
      "tokens_seen": 2169438208
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034625877632898694,
      "loss": 1.4669,
      "theoretical_loss": 3.435853658951259,
      "tokens_seen": 2169503744
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003462387161484453,
      "loss": 1.376,
      "theoretical_loss": 3.435845213692676,
      "tokens_seen": 2169569280
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003462186559679037,
      "loss": 1.3264,
      "theoretical_loss": 3.435836768760621,
      "tokens_seen": 2169634816
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003461985957873621,
      "loss": 1.2782,
      "theoretical_loss": 3.4358283241550724,
      "tokens_seen": 2169700352
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034617853560682044,
      "loss": 1.3156,
      "theoretical_loss": 3.4358198798760067,
      "tokens_seen": 2169765888
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003461584754262788,
      "loss": 1.3899,
      "theoretical_loss": 3.435811435923402,
      "tokens_seen": 2169831424
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034613841524573717,
      "loss": 1.339,
      "theoretical_loss": 3.435802992297236,
      "tokens_seen": 2169896960
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003461183550651956,
      "loss": 1.3577,
      "theoretical_loss": 3.4357945489974857,
      "tokens_seen": 2169962496
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034609829488465395,
      "loss": 1.3798,
      "theoretical_loss": 3.4357861060241293,
      "tokens_seen": 2170028032
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034607823470411236,
      "loss": 1.3227,
      "theoretical_loss": 3.4357776633771433,
      "tokens_seen": 2170093568
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003460581745235707,
      "loss": 1.4447,
      "theoretical_loss": 3.435769221056506,
      "tokens_seen": 2170159104
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034603811434302914,
      "loss": 1.3481,
      "theoretical_loss": 3.4357607790621953,
      "tokens_seen": 2170224640
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003460180541624875,
      "loss": 1.4279,
      "theoretical_loss": 3.4357523373941876,
      "tokens_seen": 2170290176
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034599799398194587,
      "loss": 1.3158,
      "theoretical_loss": 3.4357438960524616,
      "tokens_seen": 2170355712
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034597793380140423,
      "loss": 1.4953,
      "theoretical_loss": 3.435735455036994,
      "tokens_seen": 2170421248
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034595787362086264,
      "loss": 1.3783,
      "theoretical_loss": 3.4357270143477625,
      "tokens_seen": 2170486784
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000345937813440321,
      "loss": 1.4115,
      "theoretical_loss": 3.4357185739847447,
      "tokens_seen": 2170552320
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034591775325977937,
      "loss": 1.3514,
      "theoretical_loss": 3.4357101339479184,
      "tokens_seen": 2170617856
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034589769307923773,
      "loss": 1.3361,
      "theoretical_loss": 3.435701694237261,
      "tokens_seen": 2170683392
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003458776328986961,
      "loss": 1.3872,
      "theoretical_loss": 3.4356932548527497,
      "tokens_seen": 2170748928
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003458575727181545,
      "loss": 1.372,
      "theoretical_loss": 3.4356848157943625,
      "tokens_seen": 2170814464
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034583751253761287,
      "loss": 1.3682,
      "theoretical_loss": 3.435676377062077,
      "tokens_seen": 2170880000
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034581745235707123,
      "loss": 1.3942,
      "theoretical_loss": 3.4356679386558704,
      "tokens_seen": 2170945536
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003457973921765296,
      "loss": 1.3674,
      "theoretical_loss": 3.4356595005757202,
      "tokens_seen": 2171011072
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000345777331995988,
      "loss": 1.5194,
      "theoretical_loss": 3.4356510628216044,
      "tokens_seen": 2171076608
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003457572718154464,
      "loss": 1.4763,
      "theoretical_loss": 3.4356426253935,
      "tokens_seen": 2171142144
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034573721163490474,
      "loss": 1.4959,
      "theoretical_loss": 3.4356341882913854,
      "tokens_seen": 2171207680
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003457171514543631,
      "loss": 1.4047,
      "theoretical_loss": 3.4356257515152375,
      "tokens_seen": 2171273216
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034569709127382146,
      "loss": 1.3869,
      "theoretical_loss": 3.435617315065034,
      "tokens_seen": 2171338752
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003456770310932799,
      "loss": 1.4404,
      "theoretical_loss": 3.4356088789407524,
      "tokens_seen": 2171404288
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034565697091273824,
      "loss": 1.4834,
      "theoretical_loss": 3.435600443142371,
      "tokens_seen": 2171469824
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003456369107321966,
      "loss": 1.3505,
      "theoretical_loss": 3.4355920076698654,
      "tokens_seen": 2171535360
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034561685055165496,
      "loss": 1.3478,
      "theoretical_loss": 3.435583572523216,
      "tokens_seen": 2171600896
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003455967903711134,
      "loss": 1.3712,
      "theoretical_loss": 3.435575137702398,
      "tokens_seen": 2171666432
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034557673019057174,
      "loss": 1.3689,
      "theoretical_loss": 3.4355667032073898,
      "tokens_seen": 2171731968
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003455566700100301,
      "loss": 1.2428,
      "theoretical_loss": 3.4355582690381694,
      "tokens_seen": 2171797504
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034553660982948847,
      "loss": 1.5762,
      "theoretical_loss": 3.4355498351947142,
      "tokens_seen": 2171863040
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034551654964894683,
      "loss": 1.4134,
      "theoretical_loss": 3.4355414016770016,
      "tokens_seen": 2171928576
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034549648946840525,
      "loss": 1.346,
      "theoretical_loss": 3.435532968485009,
      "tokens_seen": 2171994112
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003454764292878636,
      "loss": 1.3119,
      "theoretical_loss": 3.4355245356187143,
      "tokens_seen": 2172059648
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034545636910732197,
      "loss": 1.4712,
      "theoretical_loss": 3.435516103078095,
      "tokens_seen": 2172125184
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034543630892678033,
      "loss": 1.3571,
      "theoretical_loss": 3.435507670863129,
      "tokens_seen": 2172190720
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034541624874623875,
      "loss": 1.5141,
      "theoretical_loss": 3.435499238973793,
      "tokens_seen": 2172256256
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003453961885656971,
      "loss": 1.3832,
      "theoretical_loss": 3.4354908074100656,
      "tokens_seen": 2172321792
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003453761283851555,
      "loss": 1.4329,
      "theoretical_loss": 3.435482376171924,
      "tokens_seen": 2172387328
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034535606820461384,
      "loss": 1.4485,
      "theoretical_loss": 3.435473945259346,
      "tokens_seen": 2172452864
    },
    {
      "epoch": 0.66,
      "objective/train/advantage_avg": 0.49299356341362,
      "objective/train/docs_used": 791273,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2916337251663208,
      "objective/train/original_loss": 1.2916336059570312,
      "objective/train/theoretical_loss": 3.4354655146723085,
      "objective/train/tokens_used": 2192978400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2434101402759552,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050536870956421,
      "objective/train/weighted_lm_loss": 1.3567694425582886,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4354655146723085,
      "tokens_seen": 2172518400
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003453360080240722,
      "loss": 1.3275,
      "theoretical_loss": 3.4354655146723085,
      "tokens_seen": 2172518400
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003453159478435306,
      "loss": 1.4351,
      "theoretical_loss": 3.4354570844107903,
      "tokens_seen": 2172583936
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000345295887662989,
      "loss": 1.2148,
      "theoretical_loss": 3.435448654474768,
      "tokens_seen": 2172649472
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034527582748244734,
      "loss": 1.4877,
      "theoretical_loss": 3.4354402248642195,
      "tokens_seen": 2172715008
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003452557673019057,
      "loss": 1.3506,
      "theoretical_loss": 3.435431795579123,
      "tokens_seen": 2172780544
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003452357071213641,
      "loss": 1.3517,
      "theoretical_loss": 3.435423366619455,
      "tokens_seen": 2172846080
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003452156469408225,
      "loss": 1.4048,
      "theoretical_loss": 3.435414937985194,
      "tokens_seen": 2172911616
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034519558676028084,
      "loss": 1.367,
      "theoretical_loss": 3.4354065096763176,
      "tokens_seen": 2172977152
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003451755265797392,
      "loss": 1.4548,
      "theoretical_loss": 3.4353980816928025,
      "tokens_seen": 2173042688
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034515546639919757,
      "loss": 1.4664,
      "theoretical_loss": 3.4353896540346276,
      "tokens_seen": 2173108224
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000345135406218656,
      "loss": 1.4018,
      "theoretical_loss": 3.43538122670177,
      "tokens_seen": 2173173760
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034511534603811435,
      "loss": 1.4872,
      "theoretical_loss": 3.4353727996942065,
      "tokens_seen": 2173239296
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003450952858575727,
      "loss": 1.4272,
      "theoretical_loss": 3.4353643730119163,
      "tokens_seen": 2173304832
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034507522567703107,
      "loss": 1.439,
      "theoretical_loss": 3.4353559466548758,
      "tokens_seen": 2173370368
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003450551654964895,
      "loss": 1.3641,
      "theoretical_loss": 3.435347520623063,
      "tokens_seen": 2173435904
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034503510531594785,
      "loss": 1.4322,
      "theoretical_loss": 3.435339094916456,
      "tokens_seen": 2173501440
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003450150451354062,
      "loss": 1.4239,
      "theoretical_loss": 3.4353306695350323,
      "tokens_seen": 2173566976
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003449949849548646,
      "loss": 1.3791,
      "theoretical_loss": 3.435322244478769,
      "tokens_seen": 2173632512
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034497492477432294,
      "loss": 1.3734,
      "theoretical_loss": 3.435313819747644,
      "tokens_seen": 2173698048
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034495486459378135,
      "loss": 1.3877,
      "theoretical_loss": 3.4353053953416346,
      "tokens_seen": 2173763584
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003449348044132397,
      "loss": 1.3907,
      "theoretical_loss": 3.435296971260719,
      "tokens_seen": 2173829120
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003449147442326981,
      "loss": 1.3898,
      "theoretical_loss": 3.435288547504875,
      "tokens_seen": 2173894656
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034489468405215644,
      "loss": 1.3224,
      "theoretical_loss": 3.4352801240740805,
      "tokens_seen": 2173960192
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034487462387161486,
      "loss": 1.298,
      "theoretical_loss": 3.435271700968312,
      "tokens_seen": 2174025728
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003448545636910732,
      "loss": 1.2435,
      "theoretical_loss": 3.4352632781875476,
      "tokens_seen": 2174091264
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003448345035105316,
      "loss": 1.3891,
      "theoretical_loss": 3.4352548557317655,
      "tokens_seen": 2174156800
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034481444332998994,
      "loss": 1.3876,
      "theoretical_loss": 3.435246433600943,
      "tokens_seen": 2174222336
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034479438314944836,
      "loss": 1.4246,
      "theoretical_loss": 3.435238011795058,
      "tokens_seen": 2174287872
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003447743229689067,
      "loss": 1.2889,
      "theoretical_loss": 3.4352295903140875,
      "tokens_seen": 2174353408
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003447542627883651,
      "loss": 1.351,
      "theoretical_loss": 3.4352211691580097,
      "tokens_seen": 2174418944
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034473420260782345,
      "loss": 1.2878,
      "theoretical_loss": 3.4352127483268022,
      "tokens_seen": 2174484480
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003447141424272818,
      "loss": 1.3567,
      "theoretical_loss": 3.435204327820443,
      "tokens_seen": 2174550016
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003446940822467402,
      "loss": 1.3695,
      "theoretical_loss": 3.435195907638909,
      "tokens_seen": 2174615552
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003446740220661986,
      "loss": 1.3277,
      "theoretical_loss": 3.435187487782179,
      "tokens_seen": 2174681088
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034465396188565695,
      "loss": 1.2804,
      "theoretical_loss": 3.4351790682502297,
      "tokens_seen": 2174746624
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003446339017051153,
      "loss": 1.3927,
      "theoretical_loss": 3.435170649043039,
      "tokens_seen": 2174812160
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034461384152457373,
      "loss": 1.412,
      "theoretical_loss": 3.435162230160585,
      "tokens_seen": 2174877696
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034459378134403214,
      "loss": 1.3803,
      "theoretical_loss": 3.4351538116028446,
      "tokens_seen": 2174943232
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003445737211634905,
      "loss": 1.4105,
      "theoretical_loss": 3.435145393369796,
      "tokens_seen": 2175008768
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034455366098294887,
      "loss": 1.3639,
      "theoretical_loss": 3.4351369754614174,
      "tokens_seen": 2175074304
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034453360080240723,
      "loss": 1.4054,
      "theoretical_loss": 3.4351285578776856,
      "tokens_seen": 2175139840
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034451354062186565,
      "loss": 1.335,
      "theoretical_loss": 3.4351201406185785,
      "tokens_seen": 2175205376
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000344493480441324,
      "loss": 1.4529,
      "theoretical_loss": 3.4351117236840745,
      "tokens_seen": 2175270912
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034447342026078237,
      "loss": 1.4103,
      "theoretical_loss": 3.4351033070741503,
      "tokens_seen": 2175336448
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034445336008024073,
      "loss": 1.3443,
      "theoretical_loss": 3.4350948907887844,
      "tokens_seen": 2175401984
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034443329989969915,
      "loss": 1.2702,
      "theoretical_loss": 3.435086474827954,
      "tokens_seen": 2175467520
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003444132397191575,
      "loss": 1.2824,
      "theoretical_loss": 3.4350780591916372,
      "tokens_seen": 2175533056
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003443931795386159,
      "loss": 1.3383,
      "theoretical_loss": 3.435069643879811,
      "tokens_seen": 2175598592
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034437311935807424,
      "loss": 1.3871,
      "theoretical_loss": 3.435061228892454,
      "tokens_seen": 2175664128
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003443530591775326,
      "loss": 1.286,
      "theoretical_loss": 3.4350528142295436,
      "tokens_seen": 2175729664
    },
    {
      "epoch": 0.66,
      "objective/train/advantage_avg": 0.49128395318984985,
      "objective/train/docs_used": 792571,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.264043927192688,
      "objective/train/original_loss": 1.2640438079833984,
      "objective/train/theoretical_loss": 3.4350443998910576,
      "objective/train/tokens_used": 2196255200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24196232855319977,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050358533859253,
      "objective/train/weighted_lm_loss": 1.327690601348877,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4350443998910576,
      "tokens_seen": 2175795200
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000344332998996991,
      "loss": 1.3625,
      "theoretical_loss": 3.4350443998910576,
      "tokens_seen": 2175795200
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003443129388164494,
      "loss": 1.3964,
      "theoretical_loss": 3.435035985876973,
      "tokens_seen": 2175860736
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034429287863590774,
      "loss": 1.2726,
      "theoretical_loss": 3.4350275721872685,
      "tokens_seen": 2175926272
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003442728184553661,
      "loss": 1.4215,
      "theoretical_loss": 3.4350191588219214,
      "tokens_seen": 2175991808
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003442527582748245,
      "loss": 1.5011,
      "theoretical_loss": 3.435010745780909,
      "tokens_seen": 2176057344
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003442326980942829,
      "loss": 1.4655,
      "theoretical_loss": 3.4350023330642103,
      "tokens_seen": 2176122880
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034421263791374124,
      "loss": 1.3278,
      "theoretical_loss": 3.434993920671802,
      "tokens_seen": 2176188416
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003441925777331996,
      "loss": 1.3277,
      "theoretical_loss": 3.4349855086036616,
      "tokens_seen": 2176253952
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034417251755265797,
      "loss": 1.265,
      "theoretical_loss": 3.4349770968597677,
      "tokens_seen": 2176319488
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003441524573721164,
      "loss": 1.3686,
      "theoretical_loss": 3.4349686854400976,
      "tokens_seen": 2176385024
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034413239719157475,
      "loss": 1.2918,
      "theoretical_loss": 3.434960274344629,
      "tokens_seen": 2176450560
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003441123370110331,
      "loss": 1.3807,
      "theoretical_loss": 3.4349518635733394,
      "tokens_seen": 2176516096
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034409227683049147,
      "loss": 1.351,
      "theoretical_loss": 3.4349434531262073,
      "tokens_seen": 2176581632
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003440722166499499,
      "loss": 1.3562,
      "theoretical_loss": 3.4349350430032097,
      "tokens_seen": 2176647168
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034405215646940825,
      "loss": 1.356,
      "theoretical_loss": 3.4349266332043245,
      "tokens_seen": 2176712704
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003440320962888666,
      "loss": 1.5332,
      "theoretical_loss": 3.4349182237295297,
      "tokens_seen": 2176778240
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000344012036108325,
      "loss": 1.284,
      "theoretical_loss": 3.4349098145788033,
      "tokens_seen": 2176843776
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034399197592778334,
      "loss": 1.3138,
      "theoretical_loss": 3.4349014057521225,
      "tokens_seen": 2176909312
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034397191574724175,
      "loss": 1.3962,
      "theoretical_loss": 3.4348929972494653,
      "tokens_seen": 2176974848
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003439518555667001,
      "loss": 1.4541,
      "theoretical_loss": 3.4348845890708093,
      "tokens_seen": 2177040384
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003439317953861585,
      "loss": 1.4068,
      "theoretical_loss": 3.434876181216133,
      "tokens_seen": 2177105920
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034391173520561684,
      "loss": 1.3954,
      "theoretical_loss": 3.4348677736854123,
      "tokens_seen": 2177171456
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034389167502507526,
      "loss": 1.4233,
      "theoretical_loss": 3.4348593664786273,
      "tokens_seen": 2177236992
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003438716148445336,
      "loss": 1.2872,
      "theoretical_loss": 3.4348509595957544,
      "tokens_seen": 2177302528
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000343851554663992,
      "loss": 1.2682,
      "theoretical_loss": 3.4348425530367717,
      "tokens_seen": 2177368064
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034383149448345034,
      "loss": 1.4161,
      "theoretical_loss": 3.4348341468016566,
      "tokens_seen": 2177433600
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034381143430290876,
      "loss": 1.298,
      "theoretical_loss": 3.4348257408903873,
      "tokens_seen": 2177499136
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003437913741223671,
      "loss": 1.476,
      "theoretical_loss": 3.4348173353029416,
      "tokens_seen": 2177564672
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003437713139418255,
      "loss": 1.3682,
      "theoretical_loss": 3.4348089300392974,
      "tokens_seen": 2177630208
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034375125376128385,
      "loss": 1.4572,
      "theoretical_loss": 3.434800525099432,
      "tokens_seen": 2177695744
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003437311935807422,
      "loss": 1.2893,
      "theoretical_loss": 3.4347921204833236,
      "tokens_seen": 2177761280
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003437111334002006,
      "loss": 1.3185,
      "theoretical_loss": 3.4347837161909496,
      "tokens_seen": 2177826816
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000343691073219659,
      "loss": 1.4023,
      "theoretical_loss": 3.434775312222288,
      "tokens_seen": 2177892352
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034367101303911735,
      "loss": 1.4691,
      "theoretical_loss": 3.4347669085773167,
      "tokens_seen": 2177957888
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003436509528585757,
      "loss": 1.4516,
      "theoretical_loss": 3.4347585052560134,
      "tokens_seen": 2178023424
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034363089267803413,
      "loss": 1.4502,
      "theoretical_loss": 3.4347501022583558,
      "tokens_seen": 2178088960
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003436108324974925,
      "loss": 1.2438,
      "theoretical_loss": 3.434741699584322,
      "tokens_seen": 2178154496
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034359077231695085,
      "loss": 1.322,
      "theoretical_loss": 3.4347332972338895,
      "tokens_seen": 2178220032
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003435707121364092,
      "loss": 1.3971,
      "theoretical_loss": 3.434724895207036,
      "tokens_seen": 2178285568
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003435506519558676,
      "loss": 1.5656,
      "theoretical_loss": 3.4347164935037395,
      "tokens_seen": 2178351104
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000343530591775326,
      "loss": 1.3294,
      "theoretical_loss": 3.434708092123978,
      "tokens_seen": 2178416640
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034351053159478436,
      "loss": 1.3531,
      "theoretical_loss": 3.434699691067729,
      "tokens_seen": 2178482176
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003434904714142427,
      "loss": 1.3796,
      "theoretical_loss": 3.4346912903349702,
      "tokens_seen": 2178547712
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003434704112337011,
      "loss": 1.4325,
      "theoretical_loss": 3.4346828899256803,
      "tokens_seen": 2178613248
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003434503510531595,
      "loss": 1.4086,
      "theoretical_loss": 3.434674489839836,
      "tokens_seen": 2178678784
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034343029087261786,
      "loss": 1.415,
      "theoretical_loss": 3.4346660900774157,
      "tokens_seen": 2178744320
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003434102306920762,
      "loss": 1.2732,
      "theoretical_loss": 3.4346576906383968,
      "tokens_seen": 2178809856
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003433901705115346,
      "loss": 1.3872,
      "theoretical_loss": 3.4346492915227573,
      "tokens_seen": 2178875392
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034337011033099295,
      "loss": 1.2589,
      "theoretical_loss": 3.4346408927304757,
      "tokens_seen": 2178940928
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034335005015045136,
      "loss": 1.4888,
      "theoretical_loss": 3.4346324942615287,
      "tokens_seen": 2179006464
    },
    {
      "epoch": 0.66,
      "objective/train/advantage_avg": 0.48892244696617126,
      "objective/train/docs_used": 793725,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1502673625946045,
      "objective/train/original_loss": 1.1502673625946045,
      "objective/train/theoretical_loss": 3.434624096115895,
      "objective/train/tokens_used": 2199532000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23986107110977173,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501116514205933,
      "objective/train/weighted_lm_loss": 1.2079797983169556,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.434624096115895,
      "tokens_seen": 2179072000
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003433299899699097,
      "loss": 1.2891,
      "theoretical_loss": 3.434624096115895,
      "tokens_seen": 2179072000
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003433099297893681,
      "loss": 1.3472,
      "theoretical_loss": 3.4346156982935523,
      "tokens_seen": 2179137536
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034328986960882645,
      "loss": 1.3865,
      "theoretical_loss": 3.434607300794478,
      "tokens_seen": 2179203072
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034326980942828487,
      "loss": 1.3314,
      "theoretical_loss": 3.43459890361865,
      "tokens_seen": 2179268608
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034324974924774323,
      "loss": 1.3411,
      "theoretical_loss": 3.4345905067660465,
      "tokens_seen": 2179334144
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003432296890672016,
      "loss": 1.3069,
      "theoretical_loss": 3.4345821102366454,
      "tokens_seen": 2179399680
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034320962888665995,
      "loss": 1.3687,
      "theoretical_loss": 3.434573714030424,
      "tokens_seen": 2179465216
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003431895687061183,
      "loss": 1.3229,
      "theoretical_loss": 3.4345653181473605,
      "tokens_seen": 2179530752
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034316950852557673,
      "loss": 1.4414,
      "theoretical_loss": 3.434556922587433,
      "tokens_seen": 2179596288
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003431494483450351,
      "loss": 1.2861,
      "theoretical_loss": 3.434548527350618,
      "tokens_seen": 2179661824
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034312938816449345,
      "loss": 1.3097,
      "theoretical_loss": 3.4345401324368954,
      "tokens_seen": 2179727360
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003431093279839518,
      "loss": 1.457,
      "theoretical_loss": 3.4345317378462417,
      "tokens_seen": 2179792896
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003430892678034103,
      "loss": 1.2995,
      "theoretical_loss": 3.4345233435786353,
      "tokens_seen": 2179858432
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034306920762286865,
      "loss": 1.4084,
      "theoretical_loss": 3.4345149496340532,
      "tokens_seen": 2179923968
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000343049147442327,
      "loss": 1.3809,
      "theoretical_loss": 3.4345065560124746,
      "tokens_seen": 2179989504
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003430290872617854,
      "loss": 1.3474,
      "theoretical_loss": 3.4344981627138766,
      "tokens_seen": 2180055040
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034300902708124374,
      "loss": 1.3801,
      "theoretical_loss": 3.434489769738237,
      "tokens_seen": 2180120576
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034298896690070215,
      "loss": 1.4128,
      "theoretical_loss": 3.4344813770855334,
      "tokens_seen": 2180186112
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003429689067201605,
      "loss": 1.3179,
      "theoretical_loss": 3.434472984755745,
      "tokens_seen": 2180251648
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003429488465396189,
      "loss": 1.3315,
      "theoretical_loss": 3.4344645927488475,
      "tokens_seen": 2180317184
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034292878635907724,
      "loss": 1.4608,
      "theoretical_loss": 3.4344562010648207,
      "tokens_seen": 2180382720
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034290872617853566,
      "loss": 1.4274,
      "theoretical_loss": 3.434447809703642,
      "tokens_seen": 2180448256
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000342888665997994,
      "loss": 1.3797,
      "theoretical_loss": 3.434439418665289,
      "tokens_seen": 2180513792
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003428686058174524,
      "loss": 1.3207,
      "theoretical_loss": 3.4344310279497394,
      "tokens_seen": 2180579328
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034284854563691074,
      "loss": 1.4099,
      "theoretical_loss": 3.4344226375569713,
      "tokens_seen": 2180644864
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003428284854563691,
      "loss": 1.3665,
      "theoretical_loss": 3.4344142474869623,
      "tokens_seen": 2180710400
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003428084252758275,
      "loss": 1.217,
      "theoretical_loss": 3.434405857739691,
      "tokens_seen": 2180775936
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003427883650952859,
      "loss": 1.2799,
      "theoretical_loss": 3.434397468315135,
      "tokens_seen": 2180841472
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034276830491474425,
      "loss": 1.4218,
      "theoretical_loss": 3.434389079213272,
      "tokens_seen": 2180907008
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003427482447342026,
      "loss": 1.4174,
      "theoretical_loss": 3.4343806904340797,
      "tokens_seen": 2180972544
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000342728184553661,
      "loss": 1.4367,
      "theoretical_loss": 3.434372301977536,
      "tokens_seen": 2181038080
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003427081243731194,
      "loss": 1.3452,
      "theoretical_loss": 3.4343639138436197,
      "tokens_seen": 2181103616
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034268806419257775,
      "loss": 1.3031,
      "theoretical_loss": 3.4343555260323075,
      "tokens_seen": 2181169152
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003426680040120361,
      "loss": 1.46,
      "theoretical_loss": 3.434347138543578,
      "tokens_seen": 2181234688
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034264794383149453,
      "loss": 1.4319,
      "theoretical_loss": 3.4343387513774095,
      "tokens_seen": 2181300224
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003426278836509529,
      "loss": 1.4745,
      "theoretical_loss": 3.4343303645337784,
      "tokens_seen": 2181365760
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034260782347041125,
      "loss": 1.3603,
      "theoretical_loss": 3.4343219780126644,
      "tokens_seen": 2181431296
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003425877632898696,
      "loss": 1.5423,
      "theoretical_loss": 3.434313591814044,
      "tokens_seen": 2181496832
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000342567703109328,
      "loss": 1.4017,
      "theoretical_loss": 3.434305205937896,
      "tokens_seen": 2181562368
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003425476429287864,
      "loss": 1.4729,
      "theoretical_loss": 3.4342968203841977,
      "tokens_seen": 2181627904
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034252758274824476,
      "loss": 1.3876,
      "theoretical_loss": 3.4342884351529275,
      "tokens_seen": 2181693440
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003425075225677031,
      "loss": 1.3895,
      "theoretical_loss": 3.434280050244063,
      "tokens_seen": 2181758976
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003424874623871615,
      "loss": 1.3697,
      "theoretical_loss": 3.4342716656575822,
      "tokens_seen": 2181824512
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003424674022066199,
      "loss": 1.4289,
      "theoretical_loss": 3.434263281393463,
      "tokens_seen": 2181890048
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034244734202607826,
      "loss": 1.4053,
      "theoretical_loss": 3.4342548974516838,
      "tokens_seen": 2181955584
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003424272818455366,
      "loss": 1.3707,
      "theoretical_loss": 3.4342465138322216,
      "tokens_seen": 2182021120
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000342407221664995,
      "loss": 1.3478,
      "theoretical_loss": 3.4342381305350553,
      "tokens_seen": 2182086656
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034238716148445335,
      "loss": 1.5741,
      "theoretical_loss": 3.434229747560162,
      "tokens_seen": 2182152192
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034236710130391176,
      "loss": 1.3484,
      "theoretical_loss": 3.43422136490752,
      "tokens_seen": 2182217728
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003423470411233701,
      "loss": 1.3593,
      "theoretical_loss": 3.4342129825771073,
      "tokens_seen": 2182283264
    },
    {
      "epoch": 0.66,
      "objective/train/advantage_avg": 0.494006872177124,
      "objective/train/docs_used": 794520,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4582791328430176,
      "objective/train/original_loss": 1.4582788944244385,
      "objective/train/theoretical_loss": 3.434204600568902,
      "objective/train/tokens_used": 2202808800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24467585980892181,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506445169448853,
      "objective/train/weighted_lm_loss": 1.5321217775344849,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.434204600568902,
      "tokens_seen": 2182348800
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003423269809428285,
      "loss": 1.4886,
      "theoretical_loss": 3.434204600568902,
      "tokens_seen": 2182348800
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034230692076228685,
      "loss": 1.4287,
      "theoretical_loss": 3.4341962188828816,
      "tokens_seen": 2182414336
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034228686058174527,
      "loss": 1.5175,
      "theoretical_loss": 3.4341878375190245,
      "tokens_seen": 2182479872
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034226680040120363,
      "loss": 1.5197,
      "theoretical_loss": 3.434179456477308,
      "tokens_seen": 2182545408
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000342246740220662,
      "loss": 1.3444,
      "theoretical_loss": 3.434171075757711,
      "tokens_seen": 2182610944
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034222668004012035,
      "loss": 1.3387,
      "theoretical_loss": 3.4341626953602105,
      "tokens_seen": 2182676480
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003422066198595787,
      "loss": 1.3799,
      "theoretical_loss": 3.434154315284785,
      "tokens_seen": 2182742016
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034218655967903713,
      "loss": 1.3793,
      "theoretical_loss": 3.434145935531412,
      "tokens_seen": 2182807552
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003421664994984955,
      "loss": 1.329,
      "theoretical_loss": 3.4341375561000698,
      "tokens_seen": 2182873088
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034214643931795386,
      "loss": 1.3676,
      "theoretical_loss": 3.4341291769907367,
      "tokens_seen": 2182938624
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003421263791374122,
      "loss": 1.2969,
      "theoretical_loss": 3.43412079820339,
      "tokens_seen": 2183004160
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034210631895687063,
      "loss": 1.3462,
      "theoretical_loss": 3.4341124197380077,
      "tokens_seen": 2183069696
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000342086258776329,
      "loss": 1.4094,
      "theoretical_loss": 3.4341040415945683,
      "tokens_seen": 2183135232
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034206619859578736,
      "loss": 1.4115,
      "theoretical_loss": 3.4340956637730495,
      "tokens_seen": 2183200768
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003420461384152457,
      "loss": 1.3749,
      "theoretical_loss": 3.4340872862734293,
      "tokens_seen": 2183266304
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003420260782347041,
      "loss": 1.4142,
      "theoretical_loss": 3.4340789090956854,
      "tokens_seen": 2183331840
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003420060180541625,
      "loss": 1.3799,
      "theoretical_loss": 3.434070532239796,
      "tokens_seen": 2183397376
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034198595787362086,
      "loss": 1.4054,
      "theoretical_loss": 3.434062155705739,
      "tokens_seen": 2183462912
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003419658976930792,
      "loss": 1.4721,
      "theoretical_loss": 3.4340537794934924,
      "tokens_seen": 2183528448
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003419458375125376,
      "loss": 1.2992,
      "theoretical_loss": 3.4340454036030343,
      "tokens_seen": 2183593984
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000341925777331996,
      "loss": 1.2994,
      "theoretical_loss": 3.4340370280343424,
      "tokens_seen": 2183659520
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034190571715145436,
      "loss": 1.362,
      "theoretical_loss": 3.434028652787395,
      "tokens_seen": 2183725056
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003418856569709127,
      "loss": 1.3935,
      "theoretical_loss": 3.43402027786217,
      "tokens_seen": 2183790592
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003418655967903711,
      "loss": 1.2029,
      "theoretical_loss": 3.434011903258645,
      "tokens_seen": 2183856128
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034184553660982945,
      "loss": 1.4075,
      "theoretical_loss": 3.434003528976799,
      "tokens_seen": 2183921664
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034182547642928787,
      "loss": 1.3778,
      "theoretical_loss": 3.433995155016609,
      "tokens_seen": 2183987200
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034180541624874623,
      "loss": 1.4157,
      "theoretical_loss": 3.433986781378053,
      "tokens_seen": 2184052736
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003417853560682046,
      "loss": 1.3673,
      "theoretical_loss": 3.433978408061109,
      "tokens_seen": 2184118272
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034176529588766295,
      "loss": 1.3256,
      "theoretical_loss": 3.433970035065756,
      "tokens_seen": 2184183808
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034174523570712137,
      "loss": 1.4876,
      "theoretical_loss": 3.433961662391971,
      "tokens_seen": 2184249344
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034172517552657973,
      "loss": 1.3357,
      "theoretical_loss": 3.4339532900397325,
      "tokens_seen": 2184314880
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003417051153460381,
      "loss": 1.3049,
      "theoretical_loss": 3.433944918009018,
      "tokens_seen": 2184380416
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034168505516549646,
      "loss": 1.4499,
      "theoretical_loss": 3.4339365462998064,
      "tokens_seen": 2184445952
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003416649949849548,
      "loss": 1.3731,
      "theoretical_loss": 3.4339281749120745,
      "tokens_seen": 2184511488
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034164493480441324,
      "loss": 1.328,
      "theoretical_loss": 3.433919803845801,
      "tokens_seen": 2184577024
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003416248746238716,
      "loss": 1.3117,
      "theoretical_loss": 3.433911433100964,
      "tokens_seen": 2184642560
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034160481444332996,
      "loss": 1.3517,
      "theoretical_loss": 3.4339030626775413,
      "tokens_seen": 2184708096
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003415847542627884,
      "loss": 1.4128,
      "theoretical_loss": 3.433894692575511,
      "tokens_seen": 2184773632
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003415646940822468,
      "loss": 1.3439,
      "theoretical_loss": 3.4338863227948515,
      "tokens_seen": 2184839168
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034154463390170516,
      "loss": 1.3379,
      "theoretical_loss": 3.4338779533355397,
      "tokens_seen": 2184904704
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003415245737211635,
      "loss": 1.3026,
      "theoretical_loss": 3.433869584197555,
      "tokens_seen": 2184970240
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003415045135406219,
      "loss": 1.402,
      "theoretical_loss": 3.4338612153808743,
      "tokens_seen": 2185035776
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003414844533600803,
      "loss": 1.3646,
      "theoretical_loss": 3.4338528468854763,
      "tokens_seen": 2185101312
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034146439317953866,
      "loss": 1.3683,
      "theoretical_loss": 3.4338444787113387,
      "tokens_seen": 2185166848
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000341444332998997,
      "loss": 1.3588,
      "theoretical_loss": 3.43383611085844,
      "tokens_seen": 2185232384
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003414242728184554,
      "loss": 1.3668,
      "theoretical_loss": 3.4338277433267574,
      "tokens_seen": 2185297920
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034140421263791375,
      "loss": 1.3261,
      "theoretical_loss": 3.43381937611627,
      "tokens_seen": 2185363456
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034138415245737216,
      "loss": 1.3599,
      "theoretical_loss": 3.4338110092269547,
      "tokens_seen": 2185428992
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003413640922768305,
      "loss": 1.4053,
      "theoretical_loss": 3.43380264265879,
      "tokens_seen": 2185494528
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003413440320962889,
      "loss": 1.3217,
      "theoretical_loss": 3.4337942764117546,
      "tokens_seen": 2185560064
    },
    {
      "epoch": 0.66,
      "objective/train/advantage_avg": 0.49392616748809814,
      "objective/train/docs_used": 795634,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.235222578048706,
      "objective/train/original_loss": 1.235222339630127,
      "objective/train/theoretical_loss": 3.4337859104858257,
      "objective/train/tokens_used": 2206085600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.244147390127182,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506339073181152,
      "objective/train/weighted_lm_loss": 1.2976487874984741,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 3.4337859104858257,
      "tokens_seen": 2185625600
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034132397191574725,
      "loss": 1.3925,
      "theoretical_loss": 3.4337859104858257,
      "tokens_seen": 2185625600
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034130391173520567,
      "loss": 1.3302,
      "theoretical_loss": 3.4337775448809817,
      "tokens_seen": 2185691136
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034128385155466403,
      "loss": 1.3476,
      "theoretical_loss": 3.433769179597201,
      "tokens_seen": 2185756672
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003412637913741224,
      "loss": 1.3577,
      "theoretical_loss": 3.4337608146344607,
      "tokens_seen": 2185822208
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034124373119358075,
      "loss": 1.2704,
      "theoretical_loss": 3.4337524499927397,
      "tokens_seen": 2185887744
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003412236710130391,
      "loss": 1.4466,
      "theoretical_loss": 3.4337440856720156,
      "tokens_seen": 2185953280
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034120361083249753,
      "loss": 1.4452,
      "theoretical_loss": 3.433735721672267,
      "tokens_seen": 2186018816
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003411835506519559,
      "loss": 1.4931,
      "theoretical_loss": 3.433727357993471,
      "tokens_seen": 2186084352
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034116349047141426,
      "loss": 1.2744,
      "theoretical_loss": 3.4337189946356066,
      "tokens_seen": 2186149888
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003411434302908726,
      "loss": 1.3846,
      "theoretical_loss": 3.433710631598651,
      "tokens_seen": 2186215424
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034112337011033103,
      "loss": 1.3445,
      "theoretical_loss": 3.4337022688825836,
      "tokens_seen": 2186280960
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003411033099297894,
      "loss": 1.5229,
      "theoretical_loss": 3.433693906487381,
      "tokens_seen": 2186346496
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034108324974924776,
      "loss": 1.3441,
      "theoretical_loss": 3.4336855444130223,
      "tokens_seen": 2186412032
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003410631895687061,
      "loss": 1.3839,
      "theoretical_loss": 3.433677182659485,
      "tokens_seen": 2186477568
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003410431293881645,
      "loss": 1.4379,
      "theoretical_loss": 3.4336688212267474,
      "tokens_seen": 2186543104
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003410230692076229,
      "loss": 1.4114,
      "theoretical_loss": 3.4336604601147878,
      "tokens_seen": 2186608640
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034100300902708126,
      "loss": 1.4131,
      "theoretical_loss": 3.4336520993235835,
      "tokens_seen": 2186674176
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003409829488465396,
      "loss": 1.4921,
      "theoretical_loss": 3.4336437388531134,
      "tokens_seen": 2186739712
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000340962888665998,
      "loss": 1.4331,
      "theoretical_loss": 3.4336353787033556,
      "tokens_seen": 2186805248
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003409428284854564,
      "loss": 1.3755,
      "theoretical_loss": 3.4336270188742875,
      "tokens_seen": 2186870784
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034092276830491477,
      "loss": 1.361,
      "theoretical_loss": 3.4336186593658873,
      "tokens_seen": 2186936320
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034090270812437313,
      "loss": 1.4022,
      "theoretical_loss": 3.4336103001781337,
      "tokens_seen": 2187001856
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003408826479438315,
      "loss": 1.3205,
      "theoretical_loss": 3.4336019413110046,
      "tokens_seen": 2187067392
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034086258776328985,
      "loss": 1.4178,
      "theoretical_loss": 3.4335935827644777,
      "tokens_seen": 2187132928
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034084252758274827,
      "loss": 1.3369,
      "theoretical_loss": 3.433585224538531,
      "tokens_seen": 2187198464
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034082246740220663,
      "loss": 1.3514,
      "theoretical_loss": 3.4335768666331434,
      "tokens_seen": 2187264000
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000340802407221665,
      "loss": 1.3312,
      "theoretical_loss": 3.4335685090482926,
      "tokens_seen": 2187329536
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034078234704112335,
      "loss": 1.47,
      "theoretical_loss": 3.4335601517839565,
      "tokens_seen": 2187395072
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034076228686058177,
      "loss": 1.396,
      "theoretical_loss": 3.4335517948401133,
      "tokens_seen": 2187460608
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034074222668004013,
      "loss": 1.5568,
      "theoretical_loss": 3.4335434382167414,
      "tokens_seen": 2187526144
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003407221664994985,
      "loss": 1.3759,
      "theoretical_loss": 3.4335350819138184,
      "tokens_seen": 2187591680
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034070210631895686,
      "loss": 1.4028,
      "theoretical_loss": 3.4335267259313227,
      "tokens_seen": 2187657216
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003406820461384152,
      "loss": 1.3971,
      "theoretical_loss": 3.4335183702692325,
      "tokens_seen": 2187722752
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034066198595787364,
      "loss": 1.3806,
      "theoretical_loss": 3.433510014927526,
      "tokens_seen": 2187788288
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000340641925777332,
      "loss": 1.307,
      "theoretical_loss": 3.433501659906181,
      "tokens_seen": 2187853824
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034062186559679036,
      "loss": 1.4135,
      "theoretical_loss": 3.4334933052051753,
      "tokens_seen": 2187919360
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003406018054162487,
      "loss": 1.4191,
      "theoretical_loss": 3.433484950824488,
      "tokens_seen": 2187984896
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034058174523570714,
      "loss": 1.354,
      "theoretical_loss": 3.4334765967640966,
      "tokens_seen": 2188050432
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003405616850551655,
      "loss": 1.2432,
      "theoretical_loss": 3.433468243023979,
      "tokens_seen": 2188115968
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034054162487462386,
      "loss": 1.333,
      "theoretical_loss": 3.4334598896041144,
      "tokens_seen": 2188181504
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003405215646940822,
      "loss": 1.3529,
      "theoretical_loss": 3.4334515365044798,
      "tokens_seen": 2188247040
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034050150451354064,
      "loss": 1.3594,
      "theoretical_loss": 3.4334431837250534,
      "tokens_seen": 2188312576
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000340481444332999,
      "loss": 1.4027,
      "theoretical_loss": 3.433434831265814,
      "tokens_seen": 2188378112
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034046138415245737,
      "loss": 1.2936,
      "theoretical_loss": 3.433426479126739,
      "tokens_seen": 2188443648
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034044132397191573,
      "loss": 1.3318,
      "theoretical_loss": 3.4334181273078075,
      "tokens_seen": 2188509184
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003404212637913741,
      "loss": 1.4328,
      "theoretical_loss": 3.4334097758089968,
      "tokens_seen": 2188574720
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003404012036108325,
      "loss": 1.5127,
      "theoretical_loss": 3.4334014246302855,
      "tokens_seen": 2188640256
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034038114343029087,
      "loss": 1.4908,
      "theoretical_loss": 3.433393073771651,
      "tokens_seen": 2188705792
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034036108324974923,
      "loss": 1.3659,
      "theoretical_loss": 3.4333847232330728,
      "tokens_seen": 2188771328
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003403410230692076,
      "loss": 1.3814,
      "theoretical_loss": 3.4333763730145277,
      "tokens_seen": 2188836864
    },
    {
      "epoch": 0.66,
      "objective/train/advantage_avg": 0.4887140095233917,
      "objective/train/docs_used": 796788,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3064608573913574,
      "objective/train/original_loss": 1.3064606189727783,
      "objective/train/theoretical_loss": 3.4333680231159946,
      "objective/train/tokens_used": 2209362400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.239866703748703,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500905513763428,
      "objective/train/weighted_lm_loss": 1.3719964027404785,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4333680231159946,
      "tokens_seen": 2188902400
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000340320962888666,
      "loss": 1.3009,
      "theoretical_loss": 3.4333680231159946,
      "tokens_seen": 2188902400
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003403009027081244,
      "loss": 1.3628,
      "theoretical_loss": 3.433359673537452,
      "tokens_seen": 2188967936
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034028084252758274,
      "loss": 1.3849,
      "theoretical_loss": 3.4333513242788767,
      "tokens_seen": 2189033472
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003402607823470411,
      "loss": 1.3837,
      "theoretical_loss": 3.4333429753402482,
      "tokens_seen": 2189099008
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034024072216649946,
      "loss": 1.3366,
      "theoretical_loss": 3.433334626721544,
      "tokens_seen": 2189164544
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003402206619859579,
      "loss": 1.3771,
      "theoretical_loss": 3.4333262784227427,
      "tokens_seen": 2189230080
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034020060180541624,
      "loss": 1.4555,
      "theoretical_loss": 3.4333179304438217,
      "tokens_seen": 2189295616
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003401805416248746,
      "loss": 1.2723,
      "theoretical_loss": 3.43330958278476,
      "tokens_seen": 2189361152
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034016048144433296,
      "loss": 1.4273,
      "theoretical_loss": 3.4333012354455352,
      "tokens_seen": 2189426688
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003401404212637914,
      "loss": 1.4309,
      "theoretical_loss": 3.433292888426126,
      "tokens_seen": 2189492224
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034012036108324974,
      "loss": 1.3993,
      "theoretical_loss": 3.43328454172651,
      "tokens_seen": 2189557760
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003401003009027081,
      "loss": 1.3835,
      "theoretical_loss": 3.4332761953466653,
      "tokens_seen": 2189623296
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003400802407221665,
      "loss": 1.3225,
      "theoretical_loss": 3.433267849286571,
      "tokens_seen": 2189688832
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003400601805416249,
      "loss": 1.5467,
      "theoretical_loss": 3.433259503546205,
      "tokens_seen": 2189754368
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003400401203610833,
      "loss": 1.4372,
      "theoretical_loss": 3.4332511581255445,
      "tokens_seen": 2189819904
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034002006018054166,
      "loss": 1.5189,
      "theoretical_loss": 3.4332428130245685,
      "tokens_seen": 2189885440
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00034,
      "loss": 1.4726,
      "theoretical_loss": 3.4332344682432554,
      "tokens_seen": 2189950976
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003399799398194584,
      "loss": 1.468,
      "theoretical_loss": 3.4332261237815827,
      "tokens_seen": 2190016512
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003399598796389168,
      "loss": 1.3091,
      "theoretical_loss": 3.433217779639529,
      "tokens_seen": 2190082048
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033993981945837517,
      "loss": 1.3894,
      "theoretical_loss": 3.433209435817073,
      "tokens_seen": 2190147584
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033991975927783353,
      "loss": 1.4122,
      "theoretical_loss": 3.4332010923141913,
      "tokens_seen": 2190213120
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003398996990972919,
      "loss": 1.4407,
      "theoretical_loss": 3.433192749130864,
      "tokens_seen": 2190278656
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033987963891675025,
      "loss": 1.3441,
      "theoretical_loss": 3.4331844062670682,
      "tokens_seen": 2190344192
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033985957873620867,
      "loss": 1.2612,
      "theoretical_loss": 3.433176063722782,
      "tokens_seen": 2190409728
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033983951855566703,
      "loss": 1.3064,
      "theoretical_loss": 3.4331677214979845,
      "tokens_seen": 2190475264
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003398194583751254,
      "loss": 1.4637,
      "theoretical_loss": 3.433159379592653,
      "tokens_seen": 2190540800
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033979939819458376,
      "loss": 1.4358,
      "theoretical_loss": 3.4331510380067662,
      "tokens_seen": 2190606336
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033977933801404217,
      "loss": 1.3199,
      "theoretical_loss": 3.4331426967403025,
      "tokens_seen": 2190671872
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033975927783350053,
      "loss": 1.498,
      "theoretical_loss": 3.4331343557932392,
      "tokens_seen": 2190737408
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003397392176529589,
      "loss": 1.3213,
      "theoretical_loss": 3.433126015165555,
      "tokens_seen": 2190802944
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033971915747241726,
      "loss": 1.3009,
      "theoretical_loss": 3.433117674857229,
      "tokens_seen": 2190868480
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003396990972918756,
      "loss": 1.4106,
      "theoretical_loss": 3.433109334868238,
      "tokens_seen": 2190934016
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033967903711133404,
      "loss": 1.3918,
      "theoretical_loss": 3.433100995198561,
      "tokens_seen": 2190999552
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003396589769307924,
      "loss": 1.3841,
      "theoretical_loss": 3.433092655848176,
      "tokens_seen": 2191065088
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033963891675025076,
      "loss": 1.3142,
      "theoretical_loss": 3.4330843168170615,
      "tokens_seen": 2191130624
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003396188565697091,
      "loss": 1.3486,
      "theoretical_loss": 3.4330759781051956,
      "tokens_seen": 2191196160
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033959879638916754,
      "loss": 1.5402,
      "theoretical_loss": 3.433067639712556,
      "tokens_seen": 2191261696
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003395787362086259,
      "loss": 1.3097,
      "theoretical_loss": 3.4330593016391218,
      "tokens_seen": 2191327232
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033955867602808426,
      "loss": 1.3806,
      "theoretical_loss": 3.433050963884871,
      "tokens_seen": 2191392768
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003395386158475426,
      "loss": 1.3194,
      "theoretical_loss": 3.4330426264497813,
      "tokens_seen": 2191458304
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033951855566700104,
      "loss": 1.3939,
      "theoretical_loss": 3.4330342893338313,
      "tokens_seen": 2191523840
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003394984954864594,
      "loss": 1.2978,
      "theoretical_loss": 3.4330259525369993,
      "tokens_seen": 2191589376
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033947843530591777,
      "loss": 1.3225,
      "theoretical_loss": 3.4330176160592636,
      "tokens_seen": 2191654912
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033945837512537613,
      "loss": 1.35,
      "theoretical_loss": 3.4330092799006025,
      "tokens_seen": 2191720448
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003394383149448345,
      "loss": 1.3104,
      "theoretical_loss": 3.4330009440609937,
      "tokens_seen": 2191785984
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003394182547642929,
      "loss": 1.4169,
      "theoretical_loss": 3.432992608540416,
      "tokens_seen": 2191851520
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033939819458375127,
      "loss": 1.4631,
      "theoretical_loss": 3.432984273338848,
      "tokens_seen": 2191917056
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033937813440320963,
      "loss": 1.3108,
      "theoretical_loss": 3.4329759384562664,
      "tokens_seen": 2191982592
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000339358074222668,
      "loss": 1.2864,
      "theoretical_loss": 3.432967603892651,
      "tokens_seen": 2192048128
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003393380140421264,
      "loss": 1.4963,
      "theoretical_loss": 3.43295926964798,
      "tokens_seen": 2192113664
    },
    {
      "epoch": 0.66,
      "objective/train/advantage_avg": 0.494274377822876,
      "objective/train/docs_used": 798413,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3517282009124756,
      "objective/train/original_loss": 1.3517282009124756,
      "objective/train/theoretical_loss": 3.4329509357222303,
      "objective/train/tokens_used": 2212639200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24518908560276031,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506739616394043,
      "objective/train/weighted_lm_loss": 1.420232892036438,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4329509357222303,
      "tokens_seen": 2192179200
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003393179538615848,
      "loss": 1.3773,
      "theoretical_loss": 3.4329509357222303,
      "tokens_seen": 2192179200
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033929789368104314,
      "loss": 1.484,
      "theoretical_loss": 3.4329426021153817,
      "tokens_seen": 2192244736
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003392778335005015,
      "loss": 1.403,
      "theoretical_loss": 3.432934268827412,
      "tokens_seen": 2192310272
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033925777331995986,
      "loss": 1.4063,
      "theoretical_loss": 3.432925935858299,
      "tokens_seen": 2192375808
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003392377131394183,
      "loss": 1.3794,
      "theoretical_loss": 3.432917603208021,
      "tokens_seen": 2192441344
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033921765295887664,
      "loss": 1.4436,
      "theoretical_loss": 3.4329092708765567,
      "tokens_seen": 2192506880
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000339197592778335,
      "loss": 1.2694,
      "theoretical_loss": 3.4329009388638845,
      "tokens_seen": 2192572416
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033917753259779336,
      "loss": 1.3802,
      "theoretical_loss": 3.432892607169982,
      "tokens_seen": 2192637952
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003391574724172518,
      "loss": 1.4306,
      "theoretical_loss": 3.4328842757948284,
      "tokens_seen": 2192703488
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033913741223671014,
      "loss": 1.4257,
      "theoretical_loss": 3.432875944738401,
      "tokens_seen": 2192769024
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003391173520561685,
      "loss": 1.4041,
      "theoretical_loss": 3.4328676140006786,
      "tokens_seen": 2192834560
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033909729187562687,
      "loss": 1.5156,
      "theoretical_loss": 3.4328592835816396,
      "tokens_seen": 2192900096
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033907723169508523,
      "loss": 1.2844,
      "theoretical_loss": 3.432850953481262,
      "tokens_seen": 2192965632
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033905717151454365,
      "loss": 1.3284,
      "theoretical_loss": 3.432842623699524,
      "tokens_seen": 2193031168
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000339037111334002,
      "loss": 1.5247,
      "theoretical_loss": 3.4328342942364043,
      "tokens_seen": 2193096704
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033901705115346037,
      "loss": 1.2779,
      "theoretical_loss": 3.4328259650918813,
      "tokens_seen": 2193162240
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033899699097291873,
      "loss": 1.3909,
      "theoretical_loss": 3.4328176362659324,
      "tokens_seen": 2193227776
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033897693079237715,
      "loss": 1.3452,
      "theoretical_loss": 3.4328093077585367,
      "tokens_seen": 2193293312
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003389568706118355,
      "loss": 1.3518,
      "theoretical_loss": 3.432800979569672,
      "tokens_seen": 2193358848
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003389368104312939,
      "loss": 1.2401,
      "theoretical_loss": 3.4327926516993172,
      "tokens_seen": 2193424384
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033891675025075224,
      "loss": 1.4735,
      "theoretical_loss": 3.43278432414745,
      "tokens_seen": 2193489920
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003388966900702106,
      "loss": 1.3825,
      "theoretical_loss": 3.432775996914049,
      "tokens_seen": 2193555456
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.000338876629889669,
      "loss": 1.4209,
      "theoretical_loss": 3.4327676699990928,
      "tokens_seen": 2193620992
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003388565697091274,
      "loss": 1.353,
      "theoretical_loss": 3.432759343402559,
      "tokens_seen": 2193686528
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033883650952858574,
      "loss": 1.4289,
      "theoretical_loss": 3.4327510171244264,
      "tokens_seen": 2193752064
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003388164493480441,
      "loss": 1.4186,
      "theoretical_loss": 3.4327426911646732,
      "tokens_seen": 2193817600
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003387963891675025,
      "loss": 1.3954,
      "theoretical_loss": 3.432734365523278,
      "tokens_seen": 2193883136
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003387763289869609,
      "loss": 1.4337,
      "theoretical_loss": 3.4327260402002184,
      "tokens_seen": 2193948672
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033875626880641924,
      "loss": 1.3431,
      "theoretical_loss": 3.4327177151954738,
      "tokens_seen": 2194014208
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003387362086258776,
      "loss": 1.4634,
      "theoretical_loss": 3.432709390509021,
      "tokens_seen": 2194079744
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033871614844533597,
      "loss": 1.5083,
      "theoretical_loss": 3.4327010661408397,
      "tokens_seen": 2194145280
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003386960882647944,
      "loss": 1.3305,
      "theoretical_loss": 3.4326927420909077,
      "tokens_seen": 2194210816
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033867602808425275,
      "loss": 1.3906,
      "theoretical_loss": 3.4326844183592033,
      "tokens_seen": 2194276352
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003386559679037111,
      "loss": 1.4692,
      "theoretical_loss": 3.432676094945705,
      "tokens_seen": 2194341888
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.00033863590772316947,
      "loss": 1.3703,
      "theoretical_loss": 3.432667771850391,
      "tokens_seen": 2194407424
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0003386158475426279,
      "loss": 1.4774,
      "theoretical_loss": 3.4326594490732396,
      "tokens_seen": 2194472960
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033859578736208625,
      "loss": 1.3333,
      "theoretical_loss": 3.432651126614229,
      "tokens_seen": 2194538496
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033857572718154467,
      "loss": 1.3878,
      "theoretical_loss": 3.432642804473338,
      "tokens_seen": 2194604032
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033855566700100303,
      "loss": 1.3313,
      "theoretical_loss": 3.4326344826505446,
      "tokens_seen": 2194669568
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033853560682046144,
      "loss": 1.2628,
      "theoretical_loss": 3.4326261611458273,
      "tokens_seen": 2194735104
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003385155466399198,
      "loss": 1.4738,
      "theoretical_loss": 3.4326178399591645,
      "tokens_seen": 2194800640
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033849548645937817,
      "loss": 1.4701,
      "theoretical_loss": 3.432609519090534,
      "tokens_seen": 2194866176
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033847542627883653,
      "loss": 1.3182,
      "theoretical_loss": 3.4326011985399147,
      "tokens_seen": 2194931712
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003384553660982949,
      "loss": 1.4706,
      "theoretical_loss": 3.4325928783072848,
      "tokens_seen": 2194997248
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003384353059177533,
      "loss": 1.3892,
      "theoretical_loss": 3.4325845583926227,
      "tokens_seen": 2195062784
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033841524573721167,
      "loss": 1.4543,
      "theoretical_loss": 3.4325762387959067,
      "tokens_seen": 2195128320
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033839518555667003,
      "loss": 1.3458,
      "theoretical_loss": 3.4325679195171155,
      "tokens_seen": 2195193856
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003383751253761284,
      "loss": 1.3011,
      "theoretical_loss": 3.4325596005562264,
      "tokens_seen": 2195259392
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003383550651955868,
      "loss": 1.3554,
      "theoretical_loss": 3.432551281913219,
      "tokens_seen": 2195324928
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003383350050150452,
      "loss": 1.2632,
      "theoretical_loss": 3.4325429635880713,
      "tokens_seen": 2195390464
    },
    {
      "epoch": 0.67,
      "objective/train/advantage_avg": 0.49310001730918884,
      "objective/train/docs_used": 799213,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3754479885101318,
      "objective/train/original_loss": 1.3754479885101318,
      "objective/train/theoretical_loss": 3.4325346455807613,
      "objective/train/tokens_used": 2215916000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24359363317489624,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505484342575073,
      "objective/train/weighted_lm_loss": 1.4450290203094482,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.4325346455807613,
      "tokens_seen": 2195456000
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033831494483450354,
      "loss": 1.4179,
      "theoretical_loss": 3.4325346455807613,
      "tokens_seen": 2195456000
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003382948846539619,
      "loss": 1.3525,
      "theoretical_loss": 3.4325263278912677,
      "tokens_seen": 2195521536
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033827482447342026,
      "loss": 1.4476,
      "theoretical_loss": 3.4325180105195687,
      "tokens_seen": 2195587072
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003382547642928787,
      "loss": 1.4077,
      "theoretical_loss": 3.4325096934656427,
      "tokens_seen": 2195652608
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033823470411233704,
      "loss": 1.3991,
      "theoretical_loss": 3.4325013767294683,
      "tokens_seen": 2195718144
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003382146439317954,
      "loss": 1.4601,
      "theoretical_loss": 3.4324930603110233,
      "tokens_seen": 2195783680
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033819458375125376,
      "loss": 1.3944,
      "theoretical_loss": 3.432484744210287,
      "tokens_seen": 2195849216
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003381745235707122,
      "loss": 1.4113,
      "theoretical_loss": 3.4324764284272367,
      "tokens_seen": 2195914752
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033815446339017054,
      "loss": 1.4247,
      "theoretical_loss": 3.4324681129618515,
      "tokens_seen": 2195980288
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003381344032096289,
      "loss": 1.5022,
      "theoretical_loss": 3.43245979781411,
      "tokens_seen": 2196045824
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033811434302908727,
      "loss": 1.3083,
      "theoretical_loss": 3.43245148298399,
      "tokens_seen": 2196111360
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033809428284854563,
      "loss": 1.2777,
      "theoretical_loss": 3.43244316847147,
      "tokens_seen": 2196176896
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033807422266800405,
      "loss": 1.4518,
      "theoretical_loss": 3.4324348542765284,
      "tokens_seen": 2196242432
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003380541624874624,
      "loss": 1.4282,
      "theoretical_loss": 3.4324265403991436,
      "tokens_seen": 2196307968
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033803410230692077,
      "loss": 1.4058,
      "theoretical_loss": 3.4324182268392947,
      "tokens_seen": 2196373504
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033801404212637913,
      "loss": 1.4948,
      "theoretical_loss": 3.432409913596959,
      "tokens_seen": 2196439040
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033799398194583755,
      "loss": 1.4696,
      "theoretical_loss": 3.432401600672115,
      "tokens_seen": 2196504576
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003379739217652959,
      "loss": 1.4285,
      "theoretical_loss": 3.432393288064742,
      "tokens_seen": 2196570112
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003379538615847543,
      "loss": 1.412,
      "theoretical_loss": 3.432384975774818,
      "tokens_seen": 2196635648
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033793380140421264,
      "loss": 1.3711,
      "theoretical_loss": 3.432376663802321,
      "tokens_seen": 2196701184
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000337913741223671,
      "loss": 1.3485,
      "theoretical_loss": 3.43236835214723,
      "tokens_seen": 2196766720
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003378936810431294,
      "loss": 1.3381,
      "theoretical_loss": 3.432360040809523,
      "tokens_seen": 2196832256
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003378736208625878,
      "loss": 1.4448,
      "theoretical_loss": 3.4323517297891786,
      "tokens_seen": 2196897792
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033785356068204614,
      "loss": 1.3776,
      "theoretical_loss": 3.4323434190861746,
      "tokens_seen": 2196963328
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003378335005015045,
      "loss": 1.3332,
      "theoretical_loss": 3.4323351087004905,
      "tokens_seen": 2197028864
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003378134403209629,
      "loss": 1.3737,
      "theoretical_loss": 3.432326798632104,
      "tokens_seen": 2197094400
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003377933801404213,
      "loss": 1.3337,
      "theoretical_loss": 3.432318488880994,
      "tokens_seen": 2197159936
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033777331995987964,
      "loss": 1.4017,
      "theoretical_loss": 3.4323101794471382,
      "tokens_seen": 2197225472
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000337753259779338,
      "loss": 1.2222,
      "theoretical_loss": 3.4323018703305155,
      "tokens_seen": 2197291008
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033773319959879637,
      "loss": 1.4201,
      "theoretical_loss": 3.4322935615311043,
      "tokens_seen": 2197356544
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003377131394182548,
      "loss": 1.4347,
      "theoretical_loss": 3.432285253048883,
      "tokens_seen": 2197422080
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033769307923771315,
      "loss": 1.3554,
      "theoretical_loss": 3.43227694488383,
      "tokens_seen": 2197487616
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003376730190571715,
      "loss": 1.2354,
      "theoretical_loss": 3.432268637035924,
      "tokens_seen": 2197553152
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033765295887662987,
      "loss": 1.4411,
      "theoretical_loss": 3.432260329505143,
      "tokens_seen": 2197618688
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003376328986960883,
      "loss": 1.5216,
      "theoretical_loss": 3.4322520222914656,
      "tokens_seen": 2197684224
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033761283851554665,
      "loss": 1.287,
      "theoretical_loss": 3.4322437153948706,
      "tokens_seen": 2197749760
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000337592778335005,
      "loss": 1.4402,
      "theoretical_loss": 3.4322354088153357,
      "tokens_seen": 2197815296
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003375727181544634,
      "loss": 1.3433,
      "theoretical_loss": 3.43222710255284,
      "tokens_seen": 2197880832
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033755265797392174,
      "loss": 1.4512,
      "theoretical_loss": 3.4322187966073616,
      "tokens_seen": 2197946368
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033753259779338015,
      "loss": 1.4612,
      "theoretical_loss": 3.4322104909788793,
      "tokens_seen": 2198011904
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003375125376128385,
      "loss": 1.3788,
      "theoretical_loss": 3.432202185667371,
      "tokens_seen": 2198077440
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003374924774322969,
      "loss": 1.433,
      "theoretical_loss": 3.432193880672816,
      "tokens_seen": 2198142976
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033747241725175524,
      "loss": 1.3371,
      "theoretical_loss": 3.4321855759951916,
      "tokens_seen": 2198208512
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033745235707121366,
      "loss": 1.3775,
      "theoretical_loss": 3.432177271634477,
      "tokens_seen": 2198274048
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000337432296890672,
      "loss": 1.3803,
      "theoretical_loss": 3.432168967590651,
      "tokens_seen": 2198339584
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003374122367101304,
      "loss": 1.3451,
      "theoretical_loss": 3.4321606638636912,
      "tokens_seen": 2198405120
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033739217652958874,
      "loss": 1.379,
      "theoretical_loss": 3.4321523604535766,
      "tokens_seen": 2198470656
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003373721163490471,
      "loss": 1.3479,
      "theoretical_loss": 3.4321440573602855,
      "tokens_seen": 2198536192
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003373520561685055,
      "loss": 1.4945,
      "theoretical_loss": 3.432135754583796,
      "tokens_seen": 2198601728
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003373319959879639,
      "loss": 1.3807,
      "theoretical_loss": 3.432127452124088,
      "tokens_seen": 2198667264
    },
    {
      "epoch": 0.67,
      "objective/train/advantage_avg": 0.4903789460659027,
      "objective/train/docs_used": 800708,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.227108359336853,
      "objective/train/original_loss": 1.227108359336853,
      "objective/train/theoretical_loss": 3.432119149981138,
      "objective/train/tokens_used": 2219192800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24103185534477234,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502631664276123,
      "objective/train/weighted_lm_loss": 1.2887283563613892,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.432119149981138,
      "tokens_seen": 2198732800
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033731193580742225,
      "loss": 1.3943,
      "theoretical_loss": 3.432119149981138,
      "tokens_seen": 2198732800
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003372918756268806,
      "loss": 1.3614,
      "theoretical_loss": 3.4321108481549256,
      "tokens_seen": 2198798336
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000337271815446339,
      "loss": 1.428,
      "theoretical_loss": 3.432102546645429,
      "tokens_seen": 2198863872
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003372517552657974,
      "loss": 1.4381,
      "theoretical_loss": 3.4320942454526273,
      "tokens_seen": 2198929408
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033723169508525575,
      "loss": 1.4204,
      "theoretical_loss": 3.4320859445764977,
      "tokens_seen": 2198994944
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003372116349047141,
      "loss": 1.379,
      "theoretical_loss": 3.43207764401702,
      "tokens_seen": 2199060480
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003371915747241725,
      "loss": 1.3924,
      "theoretical_loss": 3.432069343774172,
      "tokens_seen": 2199126016
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003371715145436309,
      "loss": 1.4253,
      "theoretical_loss": 3.4320610438479324,
      "tokens_seen": 2199191552
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033715145436308925,
      "loss": 1.4442,
      "theoretical_loss": 3.4320527442382796,
      "tokens_seen": 2199257088
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003371313941825476,
      "loss": 1.4662,
      "theoretical_loss": 3.4320444449451917,
      "tokens_seen": 2199322624
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000337111334002006,
      "loss": 1.2991,
      "theoretical_loss": 3.432036145968648,
      "tokens_seen": 2199388160
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003370912738214644,
      "loss": 1.4035,
      "theoretical_loss": 3.432027847308626,
      "tokens_seen": 2199453696
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003370712136409228,
      "loss": 1.4707,
      "theoretical_loss": 3.4320195489651053,
      "tokens_seen": 2199519232
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033705115346038117,
      "loss": 1.3604,
      "theoretical_loss": 3.432011250938064,
      "tokens_seen": 2199584768
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033703109327983953,
      "loss": 1.4046,
      "theoretical_loss": 3.43200295322748,
      "tokens_seen": 2199650304
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033701103309929795,
      "loss": 1.3012,
      "theoretical_loss": 3.4319946558333325,
      "tokens_seen": 2199715840
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003369909729187563,
      "loss": 1.3557,
      "theoretical_loss": 3.4319863587555997,
      "tokens_seen": 2199781376
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003369709127382147,
      "loss": 1.3508,
      "theoretical_loss": 3.4319780619942604,
      "tokens_seen": 2199846912
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033695085255767304,
      "loss": 1.3713,
      "theoretical_loss": 3.431969765549293,
      "tokens_seen": 2199912448
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003369307923771314,
      "loss": 1.3901,
      "theoretical_loss": 3.431961469420675,
      "tokens_seen": 2199977984
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003369107321965898,
      "loss": 1.4059,
      "theoretical_loss": 3.431953173608387,
      "tokens_seen": 2200043520
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003368906720160482,
      "loss": 1.2604,
      "theoretical_loss": 3.4319448781124056,
      "tokens_seen": 2200109056
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033687061183550654,
      "loss": 1.447,
      "theoretical_loss": 3.43193658293271,
      "tokens_seen": 2200174592
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003368505516549649,
      "loss": 1.3825,
      "theoretical_loss": 3.431928288069279,
      "tokens_seen": 2200240128
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003368304914744233,
      "loss": 1.4522,
      "theoretical_loss": 3.431919993522091,
      "tokens_seen": 2200305664
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003368104312938817,
      "loss": 1.3806,
      "theoretical_loss": 3.4319116992911245,
      "tokens_seen": 2200371200
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033679037111334004,
      "loss": 1.4097,
      "theoretical_loss": 3.4319034053763575,
      "tokens_seen": 2200436736
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003367703109327984,
      "loss": 1.3671,
      "theoretical_loss": 3.4318951117777696,
      "tokens_seen": 2200502272
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033675025075225677,
      "loss": 1.3608,
      "theoretical_loss": 3.431886818495338,
      "tokens_seen": 2200567808
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003367301905717152,
      "loss": 1.423,
      "theoretical_loss": 3.4318785255290427,
      "tokens_seen": 2200633344
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033671013039117355,
      "loss": 1.289,
      "theoretical_loss": 3.431870232878861,
      "tokens_seen": 2200698880
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003366900702106319,
      "loss": 1.3187,
      "theoretical_loss": 3.431861940544772,
      "tokens_seen": 2200764416
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033667001003009027,
      "loss": 1.4241,
      "theoretical_loss": 3.4318536485267543,
      "tokens_seen": 2200829952
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003366499498495487,
      "loss": 1.3434,
      "theoretical_loss": 3.4318453568247858,
      "tokens_seen": 2200895488
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033662988966900705,
      "loss": 1.4772,
      "theoretical_loss": 3.431837065438846,
      "tokens_seen": 2200961024
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003366098294884654,
      "loss": 1.3383,
      "theoretical_loss": 3.431828774368913,
      "tokens_seen": 2201026560
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003365897693079238,
      "loss": 1.5097,
      "theoretical_loss": 3.431820483614965,
      "tokens_seen": 2201092096
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033656970912738214,
      "loss": 1.4321,
      "theoretical_loss": 3.431812193176981,
      "tokens_seen": 2201157632
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033654964894684055,
      "loss": 1.3685,
      "theoretical_loss": 3.4318039030549397,
      "tokens_seen": 2201223168
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003365295887662989,
      "loss": 1.385,
      "theoretical_loss": 3.431795613248819,
      "tokens_seen": 2201288704
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003365095285857573,
      "loss": 1.3039,
      "theoretical_loss": 3.431787323758598,
      "tokens_seen": 2201354240
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033648946840521564,
      "loss": 1.4178,
      "theoretical_loss": 3.431779034584255,
      "tokens_seen": 2201419776
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033646940822467406,
      "loss": 1.3645,
      "theoretical_loss": 3.431770745725769,
      "tokens_seen": 2201485312
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003364493480441324,
      "loss": 1.3918,
      "theoretical_loss": 3.431762457183118,
      "tokens_seen": 2201550848
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003364292878635908,
      "loss": 1.2482,
      "theoretical_loss": 3.4317541689562807,
      "tokens_seen": 2201616384
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033640922768304914,
      "loss": 1.34,
      "theoretical_loss": 3.431745881045236,
      "tokens_seen": 2201681920
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003363891675025075,
      "loss": 1.344,
      "theoretical_loss": 3.431737593449962,
      "tokens_seen": 2201747456
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003363691073219659,
      "loss": 1.2418,
      "theoretical_loss": 3.4317293061704377,
      "tokens_seen": 2201812992
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003363490471414243,
      "loss": 1.2957,
      "theoretical_loss": 3.431721019206641,
      "tokens_seen": 2201878528
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033632898696088265,
      "loss": 1.4524,
      "theoretical_loss": 3.4317127325585517,
      "tokens_seen": 2201944064
    },
    {
      "epoch": 0.67,
      "objective/train/advantage_avg": 0.49282798171043396,
      "objective/train/docs_used": 801578,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.368849515914917,
      "objective/train/original_loss": 1.368849515914917,
      "objective/train/theoretical_loss": 3.4317044462261466,
      "objective/train/tokens_used": 2222469600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24322474002838135,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050519347190857,
      "objective/train/weighted_lm_loss": 1.4380278587341309,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.4317044462261466,
      "tokens_seen": 2202009600
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000336308926780341,
      "loss": 1.4194,
      "theoretical_loss": 3.4317044462261466,
      "tokens_seen": 2202009600
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003362888665997994,
      "loss": 1.3278,
      "theoretical_loss": 3.4316961602094063,
      "tokens_seen": 2202075136
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003362688064192578,
      "loss": 1.3633,
      "theoretical_loss": 3.4316878745083077,
      "tokens_seen": 2202140672
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033624874623871615,
      "loss": 1.3629,
      "theoretical_loss": 3.4316795891228304,
      "tokens_seen": 2202206208
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003362286860581745,
      "loss": 1.4495,
      "theoretical_loss": 3.4316713040529523,
      "tokens_seen": 2202271744
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033620862587763293,
      "loss": 1.3395,
      "theoretical_loss": 3.431663019298653,
      "tokens_seen": 2202337280
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003361885656970913,
      "loss": 1.447,
      "theoretical_loss": 3.43165473485991,
      "tokens_seen": 2202402816
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033616850551654965,
      "loss": 1.3917,
      "theoretical_loss": 3.431646450736702,
      "tokens_seen": 2202468352
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000336148445336008,
      "loss": 1.3868,
      "theoretical_loss": 3.4316381669290084,
      "tokens_seen": 2202533888
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003361283851554664,
      "loss": 1.4113,
      "theoretical_loss": 3.431629883436807,
      "tokens_seen": 2202599424
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003361083249749248,
      "loss": 1.3342,
      "theoretical_loss": 3.4316216002600766,
      "tokens_seen": 2202664960
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033608826479438316,
      "loss": 1.3283,
      "theoretical_loss": 3.431613317398796,
      "tokens_seen": 2202730496
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003360682046138415,
      "loss": 1.3545,
      "theoretical_loss": 3.431605034852944,
      "tokens_seen": 2202796032
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003360481444332999,
      "loss": 1.4628,
      "theoretical_loss": 3.4315967526224984,
      "tokens_seen": 2202861568
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003360280842527583,
      "loss": 1.486,
      "theoretical_loss": 3.431588470707439,
      "tokens_seen": 2202927104
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033600802407221666,
      "loss": 1.4213,
      "theoretical_loss": 3.431580189107743,
      "tokens_seen": 2202992640
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000335987963891675,
      "loss": 1.447,
      "theoretical_loss": 3.43157190782339,
      "tokens_seen": 2203058176
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003359679037111334,
      "loss": 1.4098,
      "theoretical_loss": 3.431563626854359,
      "tokens_seen": 2203123712
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033594784353059174,
      "loss": 1.3445,
      "theoretical_loss": 3.4315553462006267,
      "tokens_seen": 2203189248
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033592778335005016,
      "loss": 1.2747,
      "theoretical_loss": 3.431547065862174,
      "tokens_seen": 2203254784
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003359077231695085,
      "loss": 1.3483,
      "theoretical_loss": 3.431538785838978,
      "tokens_seen": 2203320320
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003358876629889669,
      "loss": 1.3435,
      "theoretical_loss": 3.431530506131018,
      "tokens_seen": 2203385856
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033586760280842525,
      "loss": 1.4148,
      "theoretical_loss": 3.4315222267382723,
      "tokens_seen": 2203451392
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033584754262788366,
      "loss": 1.4477,
      "theoretical_loss": 3.4315139476607195,
      "tokens_seen": 2203516928
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000335827482447342,
      "loss": 1.3826,
      "theoretical_loss": 3.4315056688983385,
      "tokens_seen": 2203582464
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003358074222668004,
      "loss": 1.4711,
      "theoretical_loss": 3.431497390451108,
      "tokens_seen": 2203648000
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033578736208625875,
      "loss": 1.2677,
      "theoretical_loss": 3.4314891123190066,
      "tokens_seen": 2203713536
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003357673019057171,
      "loss": 1.417,
      "theoretical_loss": 3.4314808345020125,
      "tokens_seen": 2203779072
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033574724172517553,
      "loss": 1.5195,
      "theoretical_loss": 3.4314725570001046,
      "tokens_seen": 2203844608
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003357271815446339,
      "loss": 1.3257,
      "theoretical_loss": 3.4314642798132615,
      "tokens_seen": 2203910144
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033570712136409225,
      "loss": 1.2994,
      "theoretical_loss": 3.431456002941462,
      "tokens_seen": 2203975680
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003356870611835506,
      "loss": 1.4061,
      "theoretical_loss": 3.4314477263846848,
      "tokens_seen": 2204041216
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033566700100300903,
      "loss": 1.4342,
      "theoretical_loss": 3.431439450142908,
      "tokens_seen": 2204106752
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003356469408224674,
      "loss": 1.3896,
      "theoretical_loss": 3.4314311742161108,
      "tokens_seen": 2204172288
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033562688064192576,
      "loss": 1.404,
      "theoretical_loss": 3.4314228986042714,
      "tokens_seen": 2204237824
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003356068204613841,
      "loss": 1.3957,
      "theoretical_loss": 3.4314146233073695,
      "tokens_seen": 2204303360
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003355867602808425,
      "loss": 1.516,
      "theoretical_loss": 3.4314063483253823,
      "tokens_seen": 2204368896
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033556670010030095,
      "loss": 1.3099,
      "theoretical_loss": 3.4313980736582894,
      "tokens_seen": 2204434432
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003355466399197593,
      "loss": 1.4117,
      "theoretical_loss": 3.4313897993060687,
      "tokens_seen": 2204499968
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003355265797392177,
      "loss": 1.3537,
      "theoretical_loss": 3.4313815252687,
      "tokens_seen": 2204565504
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033550651955867604,
      "loss": 1.347,
      "theoretical_loss": 3.431373251546161,
      "tokens_seen": 2204631040
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033548645937813446,
      "loss": 1.5172,
      "theoretical_loss": 3.4313649781384306,
      "tokens_seen": 2204696576
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003354663991975928,
      "loss": 1.3675,
      "theoretical_loss": 3.4313567050454874,
      "tokens_seen": 2204762112
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003354463390170512,
      "loss": 1.4266,
      "theoretical_loss": 3.4313484322673107,
      "tokens_seen": 2204827648
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033542627883650954,
      "loss": 1.3988,
      "theoretical_loss": 3.431340159803878,
      "tokens_seen": 2204893184
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003354062186559679,
      "loss": 1.4337,
      "theoretical_loss": 3.431331887655169,
      "tokens_seen": 2204958720
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003353861584754263,
      "loss": 1.4059,
      "theoretical_loss": 3.431323615821162,
      "tokens_seen": 2205024256
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003353660982948847,
      "loss": 1.302,
      "theoretical_loss": 3.4313153443018356,
      "tokens_seen": 2205089792
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033534603811434305,
      "loss": 1.37,
      "theoretical_loss": 3.4313070730971686,
      "tokens_seen": 2205155328
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003353259779338014,
      "loss": 1.4186,
      "theoretical_loss": 3.43129880220714,
      "tokens_seen": 2205220864
    },
    {
      "epoch": 0.67,
      "objective/train/advantage_avg": 0.494366854429245,
      "objective/train/docs_used": 802757,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3363842964172363,
      "objective/train/original_loss": 1.3363842964172363,
      "objective/train/theoretical_loss": 3.4312905316317277,
      "objective/train/tokens_used": 2225746400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24504469335079193,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506824254989624,
      "objective/train/weighted_lm_loss": 1.4040369987487793,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4312905316317277,
      "tokens_seen": 2205286400
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003353059177532598,
      "loss": 1.4806,
      "theoretical_loss": 3.4312905316317277,
      "tokens_seen": 2205286400
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003352858575727182,
      "loss": 1.428,
      "theoretical_loss": 3.4312822613709106,
      "tokens_seen": 2205351936
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033526579739217655,
      "loss": 1.376,
      "theoretical_loss": 3.431273991424668,
      "tokens_seen": 2205417472
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003352457372116349,
      "loss": 1.412,
      "theoretical_loss": 3.431265721792978,
      "tokens_seen": 2205483008
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033522567703109333,
      "loss": 1.5271,
      "theoretical_loss": 3.4312574524758195,
      "tokens_seen": 2205548544
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003352056168505517,
      "loss": 1.2899,
      "theoretical_loss": 3.431249183473171,
      "tokens_seen": 2205614080
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033518555667001005,
      "loss": 1.3668,
      "theoretical_loss": 3.431240914785012,
      "tokens_seen": 2205679616
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003351654964894684,
      "loss": 1.2838,
      "theoretical_loss": 3.4312326464113196,
      "tokens_seen": 2205745152
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003351454363089268,
      "loss": 1.343,
      "theoretical_loss": 3.431224378352074,
      "tokens_seen": 2205810688
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003351253761283852,
      "loss": 1.417,
      "theoretical_loss": 3.431216110607253,
      "tokens_seen": 2205876224
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033510531594784356,
      "loss": 1.3649,
      "theoretical_loss": 3.431207843176836,
      "tokens_seen": 2205941760
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003350852557673019,
      "loss": 1.3652,
      "theoretical_loss": 3.431199576060801,
      "tokens_seen": 2206007296
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003350651955867603,
      "loss": 1.3431,
      "theoretical_loss": 3.4311913092591277,
      "tokens_seen": 2206072832
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003350451354062187,
      "loss": 1.4695,
      "theoretical_loss": 3.4311830427717935,
      "tokens_seen": 2206138368
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033502507522567706,
      "loss": 1.4216,
      "theoretical_loss": 3.431174776598778,
      "tokens_seen": 2206203904
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003350050150451354,
      "loss": 1.3752,
      "theoretical_loss": 3.4311665107400597,
      "tokens_seen": 2206269440
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003349849548645938,
      "loss": 1.4799,
      "theoretical_loss": 3.431158245195617,
      "tokens_seen": 2206334976
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033496489468405215,
      "loss": 1.3964,
      "theoretical_loss": 3.4311499799654293,
      "tokens_seen": 2206400512
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033494483450351056,
      "loss": 1.3475,
      "theoretical_loss": 3.431141715049475,
      "tokens_seen": 2206466048
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003349247743229689,
      "loss": 1.3079,
      "theoretical_loss": 3.4311334504477324,
      "tokens_seen": 2206531584
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003349047141424273,
      "loss": 1.3815,
      "theoretical_loss": 3.431125186160181,
      "tokens_seen": 2206597120
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033488465396188565,
      "loss": 1.4344,
      "theoretical_loss": 3.4311169221867988,
      "tokens_seen": 2206662656
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033486459378134406,
      "loss": 1.4452,
      "theoretical_loss": 3.4311086585275645,
      "tokens_seen": 2206728192
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003348445336008024,
      "loss": 1.3434,
      "theoretical_loss": 3.4311003951824577,
      "tokens_seen": 2206793728
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003348244734202608,
      "loss": 1.4183,
      "theoretical_loss": 3.4310921321514565,
      "tokens_seen": 2206859264
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033480441323971915,
      "loss": 1.3281,
      "theoretical_loss": 3.4310838694345396,
      "tokens_seen": 2206924800
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003347843530591775,
      "loss": 1.3369,
      "theoretical_loss": 3.4310756070316857,
      "tokens_seen": 2206990336
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033476429287863593,
      "loss": 1.4971,
      "theoretical_loss": 3.431067344942874,
      "tokens_seen": 2207055872
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003347442326980943,
      "loss": 1.3494,
      "theoretical_loss": 3.4310590831680825,
      "tokens_seen": 2207121408
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033472417251755265,
      "loss": 1.3185,
      "theoretical_loss": 3.4310508217072906,
      "tokens_seen": 2207186944
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000334704112337011,
      "loss": 1.4,
      "theoretical_loss": 3.431042560560477,
      "tokens_seen": 2207252480
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033468405215646943,
      "loss": 1.4414,
      "theoretical_loss": 3.43103429972762,
      "tokens_seen": 2207318016
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003346639919759278,
      "loss": 1.4231,
      "theoretical_loss": 3.4310260392086986,
      "tokens_seen": 2207383552
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033464393179538616,
      "loss": 1.4436,
      "theoretical_loss": 3.4310177790036915,
      "tokens_seen": 2207449088
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003346238716148445,
      "loss": 1.3805,
      "theoretical_loss": 3.431009519112578,
      "tokens_seen": 2207514624
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003346038114343029,
      "loss": 1.4004,
      "theoretical_loss": 3.431001259535336,
      "tokens_seen": 2207580160
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003345837512537613,
      "loss": 1.5329,
      "theoretical_loss": 3.4309930002719446,
      "tokens_seen": 2207645696
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033456369107321966,
      "loss": 1.4434,
      "theoretical_loss": 3.4309847413223826,
      "tokens_seen": 2207711232
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000334543630892678,
      "loss": 1.2807,
      "theoretical_loss": 3.4309764826866287,
      "tokens_seen": 2207776768
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003345235707121364,
      "loss": 1.3582,
      "theoretical_loss": 3.430968224364662,
      "tokens_seen": 2207842304
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003345035105315948,
      "loss": 1.399,
      "theoretical_loss": 3.4309599663564607,
      "tokens_seen": 2207907840
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033448345035105316,
      "loss": 1.3805,
      "theoretical_loss": 3.4309517086620036,
      "tokens_seen": 2207973376
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003344633901705115,
      "loss": 1.4337,
      "theoretical_loss": 3.43094345128127,
      "tokens_seen": 2208038912
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003344433299899699,
      "loss": 1.4559,
      "theoretical_loss": 3.4309351942142383,
      "tokens_seen": 2208104448
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033442326980942825,
      "loss": 1.3706,
      "theoretical_loss": 3.4309269374608875,
      "tokens_seen": 2208169984
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033440320962888667,
      "loss": 1.5556,
      "theoretical_loss": 3.430918681021196,
      "tokens_seen": 2208235520
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033438314944834503,
      "loss": 1.3957,
      "theoretical_loss": 3.430910424895143,
      "tokens_seen": 2208301056
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003343630892678034,
      "loss": 1.3859,
      "theoretical_loss": 3.4309021690827066,
      "tokens_seen": 2208366592
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033434302908726175,
      "loss": 1.4142,
      "theoretical_loss": 3.4308939135838665,
      "tokens_seen": 2208432128
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033432296890672017,
      "loss": 1.4464,
      "theoretical_loss": 3.430885658398601,
      "tokens_seen": 2208497664
    },
    {
      "epoch": 0.67,
      "objective/train/advantage_avg": 0.493489533662796,
      "objective/train/docs_used": 804023,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3930256366729736,
      "objective/train/original_loss": 1.3930253982543945,
      "objective/train/theoretical_loss": 3.4308774035268885,
      "objective/train/tokens_used": 2229023200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2437971532344818,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505882501602173,
      "objective/train/weighted_lm_loss": 1.4634891748428345,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4308774035268885,
      "tokens_seen": 2208563200
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033430290872617853,
      "loss": 1.4283,
      "theoretical_loss": 3.4308774035268885,
      "tokens_seen": 2208563200
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003342828485456369,
      "loss": 1.3331,
      "theoretical_loss": 3.430869148968709,
      "tokens_seen": 2208628736
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033426278836509526,
      "loss": 1.4585,
      "theoretical_loss": 3.4308608947240398,
      "tokens_seen": 2208694272
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003342427281845536,
      "loss": 1.3609,
      "theoretical_loss": 3.4308526407928603,
      "tokens_seen": 2208759808
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033422266800401204,
      "loss": 1.441,
      "theoretical_loss": 3.4308443871751497,
      "tokens_seen": 2208825344
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003342026078234704,
      "loss": 1.332,
      "theoretical_loss": 3.430836133870886,
      "tokens_seen": 2208890880
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033418254764292876,
      "loss": 1.3684,
      "theoretical_loss": 3.430827880880049,
      "tokens_seen": 2208956416
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003341624874623871,
      "loss": 1.2619,
      "theoretical_loss": 3.430819628202617,
      "tokens_seen": 2209021952
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033414242728184554,
      "loss": 1.3625,
      "theoretical_loss": 3.4308113758385685,
      "tokens_seen": 2209087488
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003341223671013039,
      "loss": 1.4287,
      "theoretical_loss": 3.4308031237878827,
      "tokens_seen": 2209153024
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033410230692076226,
      "loss": 1.4133,
      "theoretical_loss": 3.430794872050538,
      "tokens_seen": 2209218560
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003340822467402207,
      "loss": 1.4543,
      "theoretical_loss": 3.430786620626514,
      "tokens_seen": 2209284096
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003340621865596791,
      "loss": 1.3638,
      "theoretical_loss": 3.4307783695157887,
      "tokens_seen": 2209349632
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033404212637913746,
      "loss": 1.4532,
      "theoretical_loss": 3.4307701187183413,
      "tokens_seen": 2209415168
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003340220661985958,
      "loss": 1.4402,
      "theoretical_loss": 3.4307618682341503,
      "tokens_seen": 2209480704
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003340020060180542,
      "loss": 1.3247,
      "theoretical_loss": 3.4307536180631946,
      "tokens_seen": 2209546240
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033398194583751255,
      "loss": 1.3641,
      "theoretical_loss": 3.4307453682054536,
      "tokens_seen": 2209611776
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033396188565697096,
      "loss": 1.4719,
      "theoretical_loss": 3.4307371186609057,
      "tokens_seen": 2209677312
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003339418254764293,
      "loss": 1.3512,
      "theoretical_loss": 3.430728869429529,
      "tokens_seen": 2209742848
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003339217652958877,
      "loss": 1.386,
      "theoretical_loss": 3.4307206205113037,
      "tokens_seen": 2209808384
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033390170511534605,
      "loss": 1.3674,
      "theoretical_loss": 3.430712371906208,
      "tokens_seen": 2209873920
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033388164493480447,
      "loss": 1.3146,
      "theoretical_loss": 3.43070412361422,
      "tokens_seen": 2209939456
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033386158475426283,
      "loss": 1.4302,
      "theoretical_loss": 3.4306958756353194,
      "tokens_seen": 2210004992
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003338415245737212,
      "loss": 1.4282,
      "theoretical_loss": 3.430687627969485,
      "tokens_seen": 2210070528
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033382146439317955,
      "loss": 1.3967,
      "theoretical_loss": 3.4306793806166955,
      "tokens_seen": 2210136064
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003338014042126379,
      "loss": 1.3785,
      "theoretical_loss": 3.4306711335769298,
      "tokens_seen": 2210201600
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033378134403209633,
      "loss": 1.4301,
      "theoretical_loss": 3.4306628868501665,
      "tokens_seen": 2210267136
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003337612838515547,
      "loss": 1.311,
      "theoretical_loss": 3.4306546404363845,
      "tokens_seen": 2210332672
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033374122367101306,
      "loss": 1.355,
      "theoretical_loss": 3.4306463943355627,
      "tokens_seen": 2210398208
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003337211634904714,
      "loss": 1.4255,
      "theoretical_loss": 3.43063814854768,
      "tokens_seen": 2210463744
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033370110330992983,
      "loss": 1.3147,
      "theoretical_loss": 3.4306299030727154,
      "tokens_seen": 2210529280
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003336810431293882,
      "loss": 1.4602,
      "theoretical_loss": 3.430621657910647,
      "tokens_seen": 2210594816
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033366098294884656,
      "loss": 1.3455,
      "theoretical_loss": 3.430613413061455,
      "tokens_seen": 2210660352
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003336409227683049,
      "loss": 1.4488,
      "theoretical_loss": 3.4306051685251173,
      "tokens_seen": 2210725888
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003336208625877633,
      "loss": 1.2681,
      "theoretical_loss": 3.4305969243016126,
      "tokens_seen": 2210791424
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003336008024072217,
      "loss": 1.4555,
      "theoretical_loss": 3.4305886803909202,
      "tokens_seen": 2210856960
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033358074222668006,
      "loss": 1.3882,
      "theoretical_loss": 3.4305804367930186,
      "tokens_seen": 2210922496
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003335606820461384,
      "loss": 1.3366,
      "theoretical_loss": 3.4305721935078872,
      "tokens_seen": 2210988032
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003335406218655968,
      "loss": 1.4355,
      "theoretical_loss": 3.430563950535505,
      "tokens_seen": 2211053568
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003335205616850552,
      "loss": 1.2274,
      "theoretical_loss": 3.4305557078758495,
      "tokens_seen": 2211119104
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033350050150451356,
      "loss": 1.2882,
      "theoretical_loss": 3.4305474655289006,
      "tokens_seen": 2211184640
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003334804413239719,
      "loss": 1.4395,
      "theoretical_loss": 3.4305392234946375,
      "tokens_seen": 2211250176
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003334603811434303,
      "loss": 1.4386,
      "theoretical_loss": 3.4305309817730385,
      "tokens_seen": 2211315712
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033344032096288865,
      "loss": 1.3745,
      "theoretical_loss": 3.430522740364083,
      "tokens_seen": 2211381248
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033342026078234707,
      "loss": 1.3473,
      "theoretical_loss": 3.4305144992677485,
      "tokens_seen": 2211446784
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033340020060180543,
      "loss": 1.4159,
      "theoretical_loss": 3.430506258484016,
      "tokens_seen": 2211512320
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003333801404212638,
      "loss": 1.3768,
      "theoretical_loss": 3.4304980180128624,
      "tokens_seen": 2211577856
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033336008024072215,
      "loss": 1.3568,
      "theoretical_loss": 3.430489777854268,
      "tokens_seen": 2211643392
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033334002006018057,
      "loss": 1.2865,
      "theoretical_loss": 3.4304815380082103,
      "tokens_seen": 2211708928
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033331995987963893,
      "loss": 1.2331,
      "theoretical_loss": 3.4304732984746695,
      "tokens_seen": 2211774464
    },
    {
      "epoch": 0.67,
      "objective/train/advantage_avg": 0.4933943748474121,
      "objective/train/docs_used": 805208,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3697065114974976,
      "objective/train/original_loss": 1.369706392288208,
      "objective/train/theoretical_loss": 3.4304650592536237,
      "objective/train/tokens_used": 2232300000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24368855357170105,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050578236579895,
      "objective/train/weighted_lm_loss": 1.438920259475708,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9834713935852051,
      "theoretical_loss": 3.4304650592536237,
      "tokens_seen": 2211840000
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003332998996990973,
      "loss": 1.4469,
      "theoretical_loss": 3.4304650592536237,
      "tokens_seen": 2211840000
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033327983951855566,
      "loss": 1.3937,
      "theoretical_loss": 3.4304568203450523,
      "tokens_seen": 2211905536
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000333259779338014,
      "loss": 1.2853,
      "theoretical_loss": 3.430448581748934,
      "tokens_seen": 2211971072
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033323971915747244,
      "loss": 1.4471,
      "theoretical_loss": 3.4304403434652473,
      "tokens_seen": 2212036608
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003332196589769308,
      "loss": 1.4757,
      "theoretical_loss": 3.430432105493972,
      "tokens_seen": 2212102144
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033319959879638916,
      "loss": 1.4382,
      "theoretical_loss": 3.430423867835086,
      "tokens_seen": 2212167680
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003331795386158475,
      "loss": 1.3546,
      "theoretical_loss": 3.430415630488569,
      "tokens_seen": 2212233216
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033315947843530594,
      "loss": 1.468,
      "theoretical_loss": 3.4304073934543995,
      "tokens_seen": 2212298752
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003331394182547643,
      "loss": 1.3627,
      "theoretical_loss": 3.4303991567325562,
      "tokens_seen": 2212364288
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033311935807422266,
      "loss": 1.4221,
      "theoretical_loss": 3.430390920323018,
      "tokens_seen": 2212429824
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000333099297893681,
      "loss": 1.4383,
      "theoretical_loss": 3.4303826842257648,
      "tokens_seen": 2212495360
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033307923771313944,
      "loss": 1.5557,
      "theoretical_loss": 3.430374448440774,
      "tokens_seen": 2212560896
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003330591775325978,
      "loss": 1.4006,
      "theoretical_loss": 3.4303662129680257,
      "tokens_seen": 2212626432
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033303911735205617,
      "loss": 1.3156,
      "theoretical_loss": 3.4303579778074984,
      "tokens_seen": 2212691968
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033301905717151453,
      "loss": 1.325,
      "theoretical_loss": 3.430349742959171,
      "tokens_seen": 2212757504
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003329989969909729,
      "loss": 1.4021,
      "theoretical_loss": 3.430341508423022,
      "tokens_seen": 2212823040
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003329789368104313,
      "loss": 1.2502,
      "theoretical_loss": 3.4303332741990316,
      "tokens_seen": 2212888576
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033295887662988967,
      "loss": 1.3656,
      "theoretical_loss": 3.4303250402871774,
      "tokens_seen": 2212954112
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033293881644934803,
      "loss": 1.4373,
      "theoretical_loss": 3.4303168066874385,
      "tokens_seen": 2213019648
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003329187562688064,
      "loss": 1.3845,
      "theoretical_loss": 3.4303085733997944,
      "tokens_seen": 2213085184
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003328986960882648,
      "loss": 1.3743,
      "theoretical_loss": 3.430300340424224,
      "tokens_seen": 2213150720
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003328786359077232,
      "loss": 1.3666,
      "theoretical_loss": 3.430292107760706,
      "tokens_seen": 2213216256
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033285857572718154,
      "loss": 1.4187,
      "theoretical_loss": 3.430283875409219,
      "tokens_seen": 2213281792
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003328385155466399,
      "loss": 1.2512,
      "theoretical_loss": 3.4302756433697423,
      "tokens_seen": 2213347328
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033281845536609826,
      "loss": 1.438,
      "theoretical_loss": 3.430267411642255,
      "tokens_seen": 2213412864
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003327983951855567,
      "loss": 1.3575,
      "theoretical_loss": 3.4302591802267357,
      "tokens_seen": 2213478400
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033277833500501504,
      "loss": 1.3247,
      "theoretical_loss": 3.430250949123163,
      "tokens_seen": 2213543936
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003327582748244734,
      "loss": 1.4311,
      "theoretical_loss": 3.430242718331517,
      "tokens_seen": 2213609472
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033273821464393176,
      "loss": 1.4775,
      "theoretical_loss": 3.4302344878517754,
      "tokens_seen": 2213675008
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003327181544633902,
      "loss": 1.4012,
      "theoretical_loss": 3.4302262576839184,
      "tokens_seen": 2213740544
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033269809428284854,
      "loss": 1.2356,
      "theoretical_loss": 3.430218027827924,
      "tokens_seen": 2213806080
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003326780341023069,
      "loss": 1.3701,
      "theoretical_loss": 3.430209798283771,
      "tokens_seen": 2213871616
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033265797392176527,
      "loss": 1.4427,
      "theoretical_loss": 3.430201569051439,
      "tokens_seen": 2213937152
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033263791374122363,
      "loss": 1.3021,
      "theoretical_loss": 3.4301933401309066,
      "tokens_seen": 2214002688
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033261785356068205,
      "loss": 1.5221,
      "theoretical_loss": 3.430185111522153,
      "tokens_seen": 2214068224
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003325977933801404,
      "loss": 1.2398,
      "theoretical_loss": 3.4301768832251573,
      "tokens_seen": 2214133760
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003325777331995988,
      "loss": 1.2934,
      "theoretical_loss": 3.4301686552398976,
      "tokens_seen": 2214199296
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003325576730190572,
      "loss": 1.3709,
      "theoretical_loss": 3.4301604275663538,
      "tokens_seen": 2214264832
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003325376128385156,
      "loss": 1.4085,
      "theoretical_loss": 3.4301522002045046,
      "tokens_seen": 2214330368
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033251755265797396,
      "loss": 1.3659,
      "theoretical_loss": 3.430143973154329,
      "tokens_seen": 2214395904
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033249749247743233,
      "loss": 1.5054,
      "theoretical_loss": 3.4301357464158055,
      "tokens_seen": 2214461440
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003324774322968907,
      "loss": 1.4388,
      "theoretical_loss": 3.4301275199889134,
      "tokens_seen": 2214526976
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033245737211634905,
      "loss": 1.4169,
      "theoretical_loss": 3.430119293873632,
      "tokens_seen": 2214592512
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033243731193580747,
      "loss": 1.4265,
      "theoretical_loss": 3.4301110680699396,
      "tokens_seen": 2214658048
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033241725175526583,
      "loss": 1.4528,
      "theoretical_loss": 3.4301028425778157,
      "tokens_seen": 2214723584
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003323971915747242,
      "loss": 1.3173,
      "theoretical_loss": 3.4300946173972395,
      "tokens_seen": 2214789120
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033237713139418255,
      "loss": 1.3563,
      "theoretical_loss": 3.4300863925281893,
      "tokens_seen": 2214854656
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033235707121364097,
      "loss": 1.4252,
      "theoretical_loss": 3.4300781679706445,
      "tokens_seen": 2214920192
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033233701103309933,
      "loss": 1.412,
      "theoretical_loss": 3.4300699437245834,
      "tokens_seen": 2214985728
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003323169508525577,
      "loss": 1.3004,
      "theoretical_loss": 3.430061719789986,
      "tokens_seen": 2215051264
    },
    {
      "epoch": 0.67,
      "objective/train/advantage_avg": 0.4921002686023712,
      "objective/train/docs_used": 806281,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.228821039199829,
      "objective/train/original_loss": 1.228821039199829,
      "objective/train/theoretical_loss": 3.430053496166831,
      "objective/train/tokens_used": 2235576800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24294528365135193,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504450798034668,
      "objective/train/weighted_lm_loss": 1.2907449007034302,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.430053496166831,
      "tokens_seen": 2215116800
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033229689067201606,
      "loss": 1.3281,
      "theoretical_loss": 3.430053496166831,
      "tokens_seen": 2215116800
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003322768304914744,
      "loss": 1.4073,
      "theoretical_loss": 3.4300452728550974,
      "tokens_seen": 2215182336
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033225677031093284,
      "loss": 1.4359,
      "theoretical_loss": 3.430037049854764,
      "tokens_seen": 2215247872
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003322367101303912,
      "loss": 1.2979,
      "theoretical_loss": 3.4300288271658093,
      "tokens_seen": 2215313408
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033221664994984956,
      "loss": 1.3764,
      "theoretical_loss": 3.4300206047882136,
      "tokens_seen": 2215378944
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003321965897693079,
      "loss": 1.3253,
      "theoretical_loss": 3.4300123827219546,
      "tokens_seen": 2215444480
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033217652958876634,
      "loss": 1.5492,
      "theoretical_loss": 3.430004160967012,
      "tokens_seen": 2215510016
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003321564694082247,
      "loss": 1.3232,
      "theoretical_loss": 3.429995939523365,
      "tokens_seen": 2215575552
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033213640922768306,
      "loss": 1.4093,
      "theoretical_loss": 3.4299877183909917,
      "tokens_seen": 2215641088
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003321163490471414,
      "loss": 1.3403,
      "theoretical_loss": 3.429979497569872,
      "tokens_seen": 2215706624
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033209628886659984,
      "loss": 1.3591,
      "theoretical_loss": 3.429971277059984,
      "tokens_seen": 2215772160
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003320762286860582,
      "loss": 1.4589,
      "theoretical_loss": 3.4299630568613075,
      "tokens_seen": 2215837696
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033205616850551657,
      "loss": 1.355,
      "theoretical_loss": 3.4299548369738218,
      "tokens_seen": 2215903232
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033203610832497493,
      "loss": 1.4495,
      "theoretical_loss": 3.429946617397505,
      "tokens_seen": 2215968768
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003320160481444333,
      "loss": 1.3117,
      "theoretical_loss": 3.429938398132337,
      "tokens_seen": 2216034304
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003319959879638917,
      "loss": 1.3118,
      "theoretical_loss": 3.4299301791782955,
      "tokens_seen": 2216099840
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033197592778335007,
      "loss": 1.2929,
      "theoretical_loss": 3.429921960535361,
      "tokens_seen": 2216165376
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033195586760280843,
      "loss": 1.3619,
      "theoretical_loss": 3.429913742203512,
      "tokens_seen": 2216230912
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003319358074222668,
      "loss": 1.3803,
      "theoretical_loss": 3.4299055241827268,
      "tokens_seen": 2216296448
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003319157472417252,
      "loss": 1.4153,
      "theoretical_loss": 3.4298973064729856,
      "tokens_seen": 2216361984
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003318956870611836,
      "loss": 1.3404,
      "theoretical_loss": 3.429889089074267,
      "tokens_seen": 2216427520
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033187562688064194,
      "loss": 1.343,
      "theoretical_loss": 3.4298808719865495,
      "tokens_seen": 2216493056
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003318555667001003,
      "loss": 1.3005,
      "theoretical_loss": 3.4298726552098127,
      "tokens_seen": 2216558592
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033183550651955866,
      "loss": 1.352,
      "theoretical_loss": 3.429864438744035,
      "tokens_seen": 2216624128
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003318154463390171,
      "loss": 1.287,
      "theoretical_loss": 3.429856222589197,
      "tokens_seen": 2216689664
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033179538615847544,
      "loss": 1.3882,
      "theoretical_loss": 3.429848006745276,
      "tokens_seen": 2216755200
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003317753259779338,
      "loss": 1.513,
      "theoretical_loss": 3.4298397912122516,
      "tokens_seen": 2216820736
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033175526579739216,
      "loss": 1.337,
      "theoretical_loss": 3.4298315759901032,
      "tokens_seen": 2216886272
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003317352056168506,
      "loss": 1.3222,
      "theoretical_loss": 3.4298233610788094,
      "tokens_seen": 2216951808
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033171514543630894,
      "loss": 1.3281,
      "theoretical_loss": 3.42981514647835,
      "tokens_seen": 2217017344
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003316950852557673,
      "loss": 1.4329,
      "theoretical_loss": 3.429806932188703,
      "tokens_seen": 2217082880
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033167502507522567,
      "loss": 1.3545,
      "theoretical_loss": 3.429798718209848,
      "tokens_seen": 2217148416
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033165496489468403,
      "loss": 1.4287,
      "theoretical_loss": 3.429790504541764,
      "tokens_seen": 2217213952
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033163490471414245,
      "loss": 1.3597,
      "theoretical_loss": 3.4297822911844302,
      "tokens_seen": 2217279488
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003316148445336008,
      "loss": 1.3287,
      "theoretical_loss": 3.429774078137825,
      "tokens_seen": 2217345024
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033159478435305917,
      "loss": 1.3517,
      "theoretical_loss": 3.429765865401929,
      "tokens_seen": 2217410560
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033157472417251753,
      "loss": 1.4168,
      "theoretical_loss": 3.4297576529767193,
      "tokens_seen": 2217476096
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033155466399197595,
      "loss": 1.3434,
      "theoretical_loss": 3.4297494408621767,
      "tokens_seen": 2217541632
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003315346038114343,
      "loss": 1.319,
      "theoretical_loss": 3.429741229058279,
      "tokens_seen": 2217607168
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003315145436308927,
      "loss": 1.5045,
      "theoretical_loss": 3.429733017565006,
      "tokens_seen": 2217672704
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033149448345035104,
      "loss": 1.3921,
      "theoretical_loss": 3.4297248063823362,
      "tokens_seen": 2217738240
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003314744232698094,
      "loss": 1.4301,
      "theoretical_loss": 3.429716595510249,
      "tokens_seen": 2217803776
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003314543630892678,
      "loss": 1.31,
      "theoretical_loss": 3.429708384948724,
      "tokens_seen": 2217869312
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003314343029087262,
      "loss": 1.2931,
      "theoretical_loss": 3.4297001746977394,
      "tokens_seen": 2217934848
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033141424272818454,
      "loss": 1.4339,
      "theoretical_loss": 3.4296919647572746,
      "tokens_seen": 2218000384
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003313941825476429,
      "loss": 1.4243,
      "theoretical_loss": 3.4296837551273085,
      "tokens_seen": 2218065920
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003313741223671013,
      "loss": 1.3871,
      "theoretical_loss": 3.4296755458078207,
      "tokens_seen": 2218131456
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003313540621865597,
      "loss": 1.4208,
      "theoretical_loss": 3.42966733679879,
      "tokens_seen": 2218196992
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033133400200601804,
      "loss": 1.3299,
      "theoretical_loss": 3.4296591281001954,
      "tokens_seen": 2218262528
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003313139418254764,
      "loss": 1.3801,
      "theoretical_loss": 3.429650919712016,
      "tokens_seen": 2218328064
    },
    {
      "epoch": 0.67,
      "objective/train/advantage_avg": 0.49027448892593384,
      "objective/train/docs_used": 807316,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2897812128067017,
      "objective/train/original_loss": 1.289781093597412,
      "objective/train/theoretical_loss": 3.429642711634231,
      "objective/train/tokens_used": 2238853600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24102887511253357,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502526760101318,
      "objective/train/weighted_lm_loss": 1.3545671701431274,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.429642711634231,
      "tokens_seen": 2218393600
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033129388164493477,
      "loss": 1.3395,
      "theoretical_loss": 3.429642711634231,
      "tokens_seen": 2218393600
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003312738214643932,
      "loss": 1.3331,
      "theoretical_loss": 3.4296345038668195,
      "tokens_seen": 2218459136
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033125376128385155,
      "loss": 1.2778,
      "theoretical_loss": 3.42962629640976,
      "tokens_seen": 2218524672
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003312337011033099,
      "loss": 1.3465,
      "theoretical_loss": 3.429618089263033,
      "tokens_seen": 2218590208
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033121364092276827,
      "loss": 1.4475,
      "theoretical_loss": 3.4296098824266164,
      "tokens_seen": 2218655744
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003311935807422267,
      "loss": 1.4056,
      "theoretical_loss": 3.4296016759004893,
      "tokens_seen": 2218721280
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033117352056168505,
      "loss": 1.4492,
      "theoretical_loss": 3.4295934696846313,
      "tokens_seen": 2218786816
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003311534603811434,
      "loss": 1.3008,
      "theoretical_loss": 3.4295852637790216,
      "tokens_seen": 2218852352
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033113340020060177,
      "loss": 1.3465,
      "theoretical_loss": 3.4295770581836384,
      "tokens_seen": 2218917888
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033111334002006013,
      "loss": 1.3051,
      "theoretical_loss": 3.4295688528984623,
      "tokens_seen": 2218983424
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033109327983951855,
      "loss": 1.3815,
      "theoretical_loss": 3.4295606479234713,
      "tokens_seen": 2219048960
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033107321965897697,
      "loss": 1.3705,
      "theoretical_loss": 3.4295524432586446,
      "tokens_seen": 2219114496
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033105315947843533,
      "loss": 1.2744,
      "theoretical_loss": 3.4295442389039614,
      "tokens_seen": 2219180032
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003310330992978937,
      "loss": 1.3853,
      "theoretical_loss": 3.4295360348594013,
      "tokens_seen": 2219245568
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003310130391173521,
      "loss": 1.4409,
      "theoretical_loss": 3.429527831124943,
      "tokens_seen": 2219311104
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033099297893681047,
      "loss": 1.4071,
      "theoretical_loss": 3.4295196277005653,
      "tokens_seen": 2219376640
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033097291875626883,
      "loss": 1.4118,
      "theoretical_loss": 3.4295114245862477,
      "tokens_seen": 2219442176
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003309528585757272,
      "loss": 1.4465,
      "theoretical_loss": 3.4295032217819696,
      "tokens_seen": 2219507712
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003309327983951856,
      "loss": 1.3172,
      "theoretical_loss": 3.42949501928771,
      "tokens_seen": 2219573248
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000330912738214644,
      "loss": 1.3567,
      "theoretical_loss": 3.4294868171034474,
      "tokens_seen": 2219638784
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033089267803410234,
      "loss": 1.2783,
      "theoretical_loss": 3.4294786152291614,
      "tokens_seen": 2219704320
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003308726178535607,
      "loss": 1.3172,
      "theoretical_loss": 3.4294704136648315,
      "tokens_seen": 2219769856
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033085255767301906,
      "loss": 1.428,
      "theoretical_loss": 3.4294622124104364,
      "tokens_seen": 2219835392
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003308324974924775,
      "loss": 1.3566,
      "theoretical_loss": 3.429454011465955,
      "tokens_seen": 2219900928
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033081243731193584,
      "loss": 1.2984,
      "theoretical_loss": 3.4294458108313672,
      "tokens_seen": 2219966464
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003307923771313942,
      "loss": 1.2939,
      "theoretical_loss": 3.4294376105066515,
      "tokens_seen": 2220032000
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033077231695085256,
      "loss": 1.3574,
      "theoretical_loss": 3.429429410491787,
      "tokens_seen": 2220097536
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000330752256770311,
      "loss": 1.2727,
      "theoretical_loss": 3.4294212107867534,
      "tokens_seen": 2220163072
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033073219658976934,
      "loss": 1.3467,
      "theoretical_loss": 3.4294130113915293,
      "tokens_seen": 2220228608
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003307121364092277,
      "loss": 1.3986,
      "theoretical_loss": 3.4294048123060943,
      "tokens_seen": 2220294144
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033069207622868607,
      "loss": 1.4439,
      "theoretical_loss": 3.429396613530427,
      "tokens_seen": 2220359680
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033067201604814443,
      "loss": 1.4459,
      "theoretical_loss": 3.4293884150645075,
      "tokens_seen": 2220425216
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033065195586760285,
      "loss": 1.3352,
      "theoretical_loss": 3.4293802169083136,
      "tokens_seen": 2220490752
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003306318956870612,
      "loss": 1.308,
      "theoretical_loss": 3.4293720190618258,
      "tokens_seen": 2220556288
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033061183550651957,
      "loss": 1.4534,
      "theoretical_loss": 3.429363821525022,
      "tokens_seen": 2220621824
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033059177532597793,
      "loss": 1.3327,
      "theoretical_loss": 3.429355624297883,
      "tokens_seen": 2220687360
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033057171514543635,
      "loss": 1.3779,
      "theoretical_loss": 3.429347427380386,
      "tokens_seen": 2220752896
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003305516549648947,
      "loss": 1.4504,
      "theoretical_loss": 3.429339230772512,
      "tokens_seen": 2220818432
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003305315947843531,
      "loss": 1.3647,
      "theoretical_loss": 3.4293310344742385,
      "tokens_seen": 2220883968
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033051153460381144,
      "loss": 1.2958,
      "theoretical_loss": 3.429322838485546,
      "tokens_seen": 2220949504
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003304914744232698,
      "loss": 1.3906,
      "theoretical_loss": 3.429314642806413,
      "tokens_seen": 2221015040
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003304714142427282,
      "loss": 1.3276,
      "theoretical_loss": 3.4293064474368187,
      "tokens_seen": 2221080576
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003304513540621866,
      "loss": 1.3874,
      "theoretical_loss": 3.429298252376743,
      "tokens_seen": 2221146112
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033043129388164494,
      "loss": 1.3111,
      "theoretical_loss": 3.429290057626164,
      "tokens_seen": 2221211648
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003304112337011033,
      "loss": 1.2994,
      "theoretical_loss": 3.4292818631850612,
      "tokens_seen": 2221277184
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003303911735205617,
      "loss": 1.3558,
      "theoretical_loss": 3.4292736690534142,
      "tokens_seen": 2221342720
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003303711133400201,
      "loss": 1.3314,
      "theoretical_loss": 3.429265475231202,
      "tokens_seen": 2221408256
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033035105315947844,
      "loss": 1.3356,
      "theoretical_loss": 3.4292572817184035,
      "tokens_seen": 2221473792
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003303309929789368,
      "loss": 1.3136,
      "theoretical_loss": 3.4292490885149984,
      "tokens_seen": 2221539328
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033031093279839517,
      "loss": 1.4496,
      "theoretical_loss": 3.4292408956209655,
      "tokens_seen": 2221604864
    },
    {
      "epoch": 0.67,
      "objective/train/advantage_avg": 0.4908287525177002,
      "objective/train/docs_used": 808517,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3176387548446655,
      "objective/train/original_loss": 1.317638874053955,
      "objective/train/theoretical_loss": 3.429232703036284,
      "objective/train/tokens_used": 2242130400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24176858365535736,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050311803817749,
      "objective/train/weighted_lm_loss": 1.3839467763900757,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.429232703036284,
      "tokens_seen": 2221670400
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003302908726178536,
      "loss": 1.2868,
      "theoretical_loss": 3.429232703036284,
      "tokens_seen": 2221670400
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033027081243731195,
      "loss": 1.3253,
      "theoretical_loss": 3.429224510760933,
      "tokens_seen": 2221735936
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003302507522567703,
      "loss": 1.4149,
      "theoretical_loss": 3.4292163187948925,
      "tokens_seen": 2221801472
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033023069207622867,
      "loss": 1.3743,
      "theoretical_loss": 3.429208127138141,
      "tokens_seen": 2221867008
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003302106318956871,
      "loss": 1.4287,
      "theoretical_loss": 3.429199935790657,
      "tokens_seen": 2221932544
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033019057171514545,
      "loss": 1.4252,
      "theoretical_loss": 3.429191744752421,
      "tokens_seen": 2221998080
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003301705115346038,
      "loss": 1.4949,
      "theoretical_loss": 3.4291835540234117,
      "tokens_seen": 2222063616
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003301504513540622,
      "loss": 1.3184,
      "theoretical_loss": 3.4291753636036084,
      "tokens_seen": 2222129152
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033013039117352054,
      "loss": 1.4783,
      "theoretical_loss": 3.4291671734929903,
      "tokens_seen": 2222194688
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033011033099297895,
      "loss": 1.4345,
      "theoretical_loss": 3.4291589836915364,
      "tokens_seen": 2222260224
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003300902708124373,
      "loss": 1.4157,
      "theoretical_loss": 3.429150794199226,
      "tokens_seen": 2222325760
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003300702106318957,
      "loss": 1.4141,
      "theoretical_loss": 3.4291426050160383,
      "tokens_seen": 2222391296
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033005015045135404,
      "loss": 1.3906,
      "theoretical_loss": 3.4291344161419524,
      "tokens_seen": 2222456832
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00033003009027081245,
      "loss": 1.2436,
      "theoretical_loss": 3.4291262275769476,
      "tokens_seen": 2222522368
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003300100300902708,
      "loss": 1.533,
      "theoretical_loss": 3.4291180393210037,
      "tokens_seen": 2222587904
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003299899699097292,
      "loss": 1.2876,
      "theoretical_loss": 3.429109851374099,
      "tokens_seen": 2222653440
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032996990972918754,
      "loss": 1.4025,
      "theoretical_loss": 3.4291016637362137,
      "tokens_seen": 2222718976
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003299498495486459,
      "loss": 1.3275,
      "theoretical_loss": 3.429093476407326,
      "tokens_seen": 2222784512
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003299297893681043,
      "loss": 1.3757,
      "theoretical_loss": 3.4290852893874155,
      "tokens_seen": 2222850048
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003299097291875627,
      "loss": 1.3685,
      "theoretical_loss": 3.429077102676462,
      "tokens_seen": 2222915584
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032988966900702104,
      "loss": 1.3819,
      "theoretical_loss": 3.429068916274444,
      "tokens_seen": 2222981120
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003298696088264794,
      "loss": 1.2785,
      "theoretical_loss": 3.429060730181341,
      "tokens_seen": 2223046656
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003298495486459378,
      "loss": 1.3774,
      "theoretical_loss": 3.4290525443971323,
      "tokens_seen": 2223112192
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003298294884653962,
      "loss": 1.4354,
      "theoretical_loss": 3.429044358921797,
      "tokens_seen": 2223177728
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032980942828485455,
      "loss": 1.3304,
      "theoretical_loss": 3.429036173755314,
      "tokens_seen": 2223243264
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003297893681043129,
      "loss": 1.4672,
      "theoretical_loss": 3.429027988897664,
      "tokens_seen": 2223308800
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003297693079237713,
      "loss": 1.4446,
      "theoretical_loss": 3.4290198043488243,
      "tokens_seen": 2223374336
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003297492477432297,
      "loss": 1.3544,
      "theoretical_loss": 3.4290116201087755,
      "tokens_seen": 2223439872
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032972918756268805,
      "loss": 1.4539,
      "theoretical_loss": 3.429003436177496,
      "tokens_seen": 2223505408
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003297091273821464,
      "loss": 1.3404,
      "theoretical_loss": 3.428995252554966,
      "tokens_seen": 2223570944
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003296890672016048,
      "loss": 1.4883,
      "theoretical_loss": 3.4289870692411637,
      "tokens_seen": 2223636480
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003296690070210632,
      "loss": 1.4233,
      "theoretical_loss": 3.4289788862360693,
      "tokens_seen": 2223702016
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032964894684052155,
      "loss": 1.2966,
      "theoretical_loss": 3.4289707035396613,
      "tokens_seen": 2223767552
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003296288866599799,
      "loss": 1.4487,
      "theoretical_loss": 3.4289625211519192,
      "tokens_seen": 2223833088
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003296088264794383,
      "loss": 1.4999,
      "theoretical_loss": 3.4289543390728228,
      "tokens_seen": 2223898624
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003295887662988967,
      "loss": 1.4303,
      "theoretical_loss": 3.4289461573023505,
      "tokens_seen": 2223964160
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003295687061183551,
      "loss": 1.3935,
      "theoretical_loss": 3.428937975840482,
      "tokens_seen": 2224029696
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003295486459378135,
      "loss": 1.3999,
      "theoretical_loss": 3.4289297946871966,
      "tokens_seen": 2224095232
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032952858575727184,
      "loss": 1.393,
      "theoretical_loss": 3.4289216138424736,
      "tokens_seen": 2224160768
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003295085255767302,
      "loss": 1.3974,
      "theoretical_loss": 3.428913433306292,
      "tokens_seen": 2224226304
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003294884653961886,
      "loss": 1.4983,
      "theoretical_loss": 3.4289052530786313,
      "tokens_seen": 2224291840
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000329468405215647,
      "loss": 1.3855,
      "theoretical_loss": 3.428897073159471,
      "tokens_seen": 2224357376
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032944834503510534,
      "loss": 1.4098,
      "theoretical_loss": 3.42888889354879,
      "tokens_seen": 2224422912
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003294282848545637,
      "loss": 1.4032,
      "theoretical_loss": 3.428880714246567,
      "tokens_seen": 2224488448
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003294082246740221,
      "loss": 1.4328,
      "theoretical_loss": 3.4288725352527827,
      "tokens_seen": 2224553984
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003293881644934805,
      "loss": 1.399,
      "theoretical_loss": 3.4288643565674155,
      "tokens_seen": 2224619520
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032936810431293884,
      "loss": 1.3461,
      "theoretical_loss": 3.4288561781904443,
      "tokens_seen": 2224685056
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003293480441323972,
      "loss": 1.4157,
      "theoretical_loss": 3.4288480001218495,
      "tokens_seen": 2224750592
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032932798395185557,
      "loss": 1.3251,
      "theoretical_loss": 3.42883982236161,
      "tokens_seen": 2224816128
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.000329307923771314,
      "loss": 1.3154,
      "theoretical_loss": 3.428831644909704,
      "tokens_seen": 2224881664
    },
    {
      "epoch": 0.67,
      "objective/train/advantage_avg": 0.4949820041656494,
      "objective/train/docs_used": 809850,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4319509267807007,
      "objective/train/original_loss": 1.4319508075714111,
      "objective/train/theoretical_loss": 3.4288234677661125,
      "objective/train/tokens_used": 2245407200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2452239841222763,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507450103759766,
      "objective/train/weighted_lm_loss": 1.5044344663619995,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4288234677661125,
      "tokens_seen": 2224947200
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032928786359077235,
      "loss": 1.4004,
      "theoretical_loss": 3.4288234677661125,
      "tokens_seen": 2224947200
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003292678034102307,
      "loss": 1.3028,
      "theoretical_loss": 3.4288152909308134,
      "tokens_seen": 2225012736
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032924774322968907,
      "loss": 1.3663,
      "theoretical_loss": 3.428807114403787,
      "tokens_seen": 2225078272
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003292276830491475,
      "loss": 1.2894,
      "theoretical_loss": 3.428798938185012,
      "tokens_seen": 2225143808
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032920762286860585,
      "loss": 1.3394,
      "theoretical_loss": 3.4287907622744678,
      "tokens_seen": 2225209344
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003291875626880642,
      "loss": 1.3109,
      "theoretical_loss": 3.428782586672134,
      "tokens_seen": 2225274880
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003291675025075226,
      "loss": 1.2596,
      "theoretical_loss": 3.428774411377989,
      "tokens_seen": 2225340416
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032914744232698094,
      "loss": 1.5001,
      "theoretical_loss": 3.4287662363920135,
      "tokens_seen": 2225405952
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032912738214643935,
      "loss": 1.4048,
      "theoretical_loss": 3.428758061714186,
      "tokens_seen": 2225471488
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003291073219658977,
      "loss": 1.3897,
      "theoretical_loss": 3.4287498873444857,
      "tokens_seen": 2225537024
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003290872617853561,
      "loss": 1.317,
      "theoretical_loss": 3.4287417132828923,
      "tokens_seen": 2225602560
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032906720160481444,
      "loss": 1.379,
      "theoretical_loss": 3.428733539529385,
      "tokens_seen": 2225668096
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032904714142427286,
      "loss": 1.3778,
      "theoretical_loss": 3.4287253660839427,
      "tokens_seen": 2225733632
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003290270812437312,
      "loss": 1.3201,
      "theoretical_loss": 3.4287171929465456,
      "tokens_seen": 2225799168
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003290070210631896,
      "loss": 1.361,
      "theoretical_loss": 3.428709020117172,
      "tokens_seen": 2225864704
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032898696088264794,
      "loss": 1.3052,
      "theoretical_loss": 3.4287008475958016,
      "tokens_seen": 2225930240
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003289669007021063,
      "loss": 1.4051,
      "theoretical_loss": 3.4286926753824143,
      "tokens_seen": 2225995776
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003289468405215647,
      "loss": 1.3065,
      "theoretical_loss": 3.4286845034769886,
      "tokens_seen": 2226061312
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003289267803410231,
      "loss": 1.3358,
      "theoretical_loss": 3.4286763318795046,
      "tokens_seen": 2226126848
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032890672016048145,
      "loss": 1.3333,
      "theoretical_loss": 3.428668160589941,
      "tokens_seen": 2226192384
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003288866599799398,
      "loss": 1.2579,
      "theoretical_loss": 3.4286599896082772,
      "tokens_seen": 2226257920
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003288665997993982,
      "loss": 1.3024,
      "theoretical_loss": 3.428651818934493,
      "tokens_seen": 2226323456
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003288465396188566,
      "loss": 1.4324,
      "theoretical_loss": 3.428643648568567,
      "tokens_seen": 2226388992
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032882647943831495,
      "loss": 1.4286,
      "theoretical_loss": 3.428635478510479,
      "tokens_seen": 2226454528
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003288064192577733,
      "loss": 1.3377,
      "theoretical_loss": 3.428627308760209,
      "tokens_seen": 2226520064
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003287863590772317,
      "loss": 1.4505,
      "theoretical_loss": 3.428619139317735,
      "tokens_seen": 2226585600
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003287662988966901,
      "loss": 1.4543,
      "theoretical_loss": 3.4286109701830374,
      "tokens_seen": 2226651136
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032874623871614845,
      "loss": 1.4095,
      "theoretical_loss": 3.428602801356095,
      "tokens_seen": 2226716672
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003287261785356068,
      "loss": 1.3586,
      "theoretical_loss": 3.4285946328368873,
      "tokens_seen": 2226782208
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003287061183550652,
      "loss": 1.422,
      "theoretical_loss": 3.4285864646253934,
      "tokens_seen": 2226847744
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003286860581745236,
      "loss": 1.425,
      "theoretical_loss": 3.428578296721593,
      "tokens_seen": 2226913280
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032866599799398195,
      "loss": 1.4159,
      "theoretical_loss": 3.428570129125465,
      "tokens_seen": 2226978816
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003286459378134403,
      "loss": 1.3137,
      "theoretical_loss": 3.42856196183699,
      "tokens_seen": 2227044352
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003286258776328987,
      "loss": 1.4572,
      "theoretical_loss": 3.4285537948561458,
      "tokens_seen": 2227109888
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003286058174523571,
      "loss": 1.351,
      "theoretical_loss": 3.4285456281829125,
      "tokens_seen": 2227175424
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032858575727181546,
      "loss": 1.4392,
      "theoretical_loss": 3.4285374618172693,
      "tokens_seen": 2227240960
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003285656970912738,
      "loss": 1.4845,
      "theoretical_loss": 3.428529295759196,
      "tokens_seen": 2227306496
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0003285456369107322,
      "loss": 1.3716,
      "theoretical_loss": 3.4285211300086713,
      "tokens_seen": 2227372032
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00032852557673019054,
      "loss": 1.4467,
      "theoretical_loss": 3.428512964565675,
      "tokens_seen": 2227437568
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032850551654964896,
      "loss": 1.4676,
      "theoretical_loss": 3.4285047994301863,
      "tokens_seen": 2227503104
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003284854563691073,
      "loss": 1.2762,
      "theoretical_loss": 3.428496634602185,
      "tokens_seen": 2227568640
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003284653961885657,
      "loss": 1.3872,
      "theoretical_loss": 3.42848847008165,
      "tokens_seen": 2227634176
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032844533600802405,
      "loss": 1.3587,
      "theoretical_loss": 3.4284803058685602,
      "tokens_seen": 2227699712
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032842527582748246,
      "loss": 1.3855,
      "theoretical_loss": 3.428472141962896,
      "tokens_seen": 2227765248
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003284052156469408,
      "loss": 1.3842,
      "theoretical_loss": 3.4284639783646362,
      "tokens_seen": 2227830784
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003283851554663992,
      "loss": 1.3192,
      "theoretical_loss": 3.4284558150737605,
      "tokens_seen": 2227896320
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032836509528585755,
      "loss": 1.3637,
      "theoretical_loss": 3.428447652090248,
      "tokens_seen": 2227961856
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003283450351053159,
      "loss": 1.3357,
      "theoretical_loss": 3.428439489414078,
      "tokens_seen": 2228027392
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032832497492477433,
      "loss": 1.4997,
      "theoretical_loss": 3.4284313270452307,
      "tokens_seen": 2228092928
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003283049147442327,
      "loss": 1.4892,
      "theoretical_loss": 3.428423164983684,
      "tokens_seen": 2228158464
    },
    {
      "epoch": 0.68,
      "objective/train/advantage_avg": 0.49216094613075256,
      "objective/train/docs_used": 811140,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2088018655776978,
      "objective/train/original_loss": 1.2088017463684082,
      "objective/train/theoretical_loss": 3.4284150032294187,
      "objective/train/tokens_used": 2248684000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24274730682373047,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504502058029175,
      "objective/train/weighted_lm_loss": 1.2697196006774902,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4284150032294187,
      "tokens_seen": 2228224000
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032828485456369105,
      "loss": 1.261,
      "theoretical_loss": 3.4284150032294187,
      "tokens_seen": 2228224000
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003282647943831494,
      "loss": 1.4092,
      "theoretical_loss": 3.428406841782414,
      "tokens_seen": 2228289536
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032824473420260783,
      "loss": 1.416,
      "theoretical_loss": 3.428398680642648,
      "tokens_seen": 2228355072
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003282246740220662,
      "loss": 1.4566,
      "theoretical_loss": 3.428390519810102,
      "tokens_seen": 2228420608
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032820461384152456,
      "loss": 1.2477,
      "theoretical_loss": 3.428382359284754,
      "tokens_seen": 2228486144
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003281845536609829,
      "loss": 1.4735,
      "theoretical_loss": 3.428374199066584,
      "tokens_seen": 2228551680
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003281644934804413,
      "loss": 1.3143,
      "theoretical_loss": 3.428366039155571,
      "tokens_seen": 2228617216
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003281444332998997,
      "loss": 1.3592,
      "theoretical_loss": 3.4283578795516947,
      "tokens_seen": 2228682752
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032812437311935806,
      "loss": 1.4371,
      "theoretical_loss": 3.4283497202549347,
      "tokens_seen": 2228748288
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003281043129388164,
      "loss": 1.478,
      "theoretical_loss": 3.4283415612652703,
      "tokens_seen": 2228813824
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003280842527582748,
      "loss": 1.3594,
      "theoretical_loss": 3.4283334025826804,
      "tokens_seen": 2228879360
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032806419257773326,
      "loss": 1.4783,
      "theoretical_loss": 3.4283252442071452,
      "tokens_seen": 2228944896
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003280441323971916,
      "loss": 1.4103,
      "theoretical_loss": 3.4283170861386436,
      "tokens_seen": 2229010432
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032802407221665,
      "loss": 1.3998,
      "theoretical_loss": 3.428308928377155,
      "tokens_seen": 2229075968
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032800401203610834,
      "loss": 1.3862,
      "theoretical_loss": 3.4283007709226587,
      "tokens_seen": 2229141504
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003279839518555667,
      "loss": 1.4994,
      "theoretical_loss": 3.428292613775135,
      "tokens_seen": 2229207040
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003279638916750251,
      "loss": 1.3866,
      "theoretical_loss": 3.4282844569345623,
      "tokens_seen": 2229272576
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003279438314944835,
      "loss": 1.3118,
      "theoretical_loss": 3.4282763004009205,
      "tokens_seen": 2229338112
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032792377131394185,
      "loss": 1.3633,
      "theoretical_loss": 3.428268144174189,
      "tokens_seen": 2229403648
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003279037111334002,
      "loss": 1.4843,
      "theoretical_loss": 3.4282599882543474,
      "tokens_seen": 2229469184
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003278836509528586,
      "loss": 1.4397,
      "theoretical_loss": 3.428251832641375,
      "tokens_seen": 2229534720
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000327863590772317,
      "loss": 1.4255,
      "theoretical_loss": 3.428243677335251,
      "tokens_seen": 2229600256
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032784353059177535,
      "loss": 1.4144,
      "theoretical_loss": 3.428235522335955,
      "tokens_seen": 2229665792
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003278234704112337,
      "loss": 1.4657,
      "theoretical_loss": 3.4282273676434665,
      "tokens_seen": 2229731328
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032780341023069213,
      "loss": 1.3759,
      "theoretical_loss": 3.4282192132577647,
      "tokens_seen": 2229796864
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003277833500501505,
      "loss": 1.3018,
      "theoretical_loss": 3.428211059178829,
      "tokens_seen": 2229862400
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032776328986960885,
      "loss": 1.4517,
      "theoretical_loss": 3.4282029054066396,
      "tokens_seen": 2229927936
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003277432296890672,
      "loss": 1.4437,
      "theoretical_loss": 3.428194751941175,
      "tokens_seen": 2229993472
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003277231695085256,
      "loss": 1.3162,
      "theoretical_loss": 3.4281865987824154,
      "tokens_seen": 2230059008
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000327703109327984,
      "loss": 1.4346,
      "theoretical_loss": 3.4281784459303397,
      "tokens_seen": 2230124544
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032768304914744235,
      "loss": 1.3348,
      "theoretical_loss": 3.4281702933849276,
      "tokens_seen": 2230190080
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003276629889669007,
      "loss": 1.4205,
      "theoretical_loss": 3.4281621411461587,
      "tokens_seen": 2230255616
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003276429287863591,
      "loss": 1.2955,
      "theoretical_loss": 3.428153989214012,
      "tokens_seen": 2230321152
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003276228686058175,
      "loss": 1.3584,
      "theoretical_loss": 3.4281458375884677,
      "tokens_seen": 2230386688
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032760280842527586,
      "loss": 1.3359,
      "theoretical_loss": 3.4281376862695043,
      "tokens_seen": 2230452224
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003275827482447342,
      "loss": 1.4081,
      "theoretical_loss": 3.428129535257102,
      "tokens_seen": 2230517760
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003275626880641926,
      "loss": 1.3464,
      "theoretical_loss": 3.4281213845512397,
      "tokens_seen": 2230583296
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032754262788365094,
      "loss": 1.2945,
      "theoretical_loss": 3.4281132341518976,
      "tokens_seen": 2230648832
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032752256770310936,
      "loss": 1.2819,
      "theoretical_loss": 3.4281050840590543,
      "tokens_seen": 2230714368
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003275025075225677,
      "loss": 1.3996,
      "theoretical_loss": 3.42809693427269,
      "tokens_seen": 2230779904
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003274824473420261,
      "loss": 1.3522,
      "theoretical_loss": 3.428088784792784,
      "tokens_seen": 2230845440
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032746238716148445,
      "loss": 1.3687,
      "theoretical_loss": 3.4280806356193154,
      "tokens_seen": 2230910976
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032744232698094286,
      "loss": 1.3586,
      "theoretical_loss": 3.428072486752264,
      "tokens_seen": 2230976512
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003274222668004012,
      "loss": 1.5289,
      "theoretical_loss": 3.4280643381916094,
      "tokens_seen": 2231042048
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003274022066198596,
      "loss": 1.5064,
      "theoretical_loss": 3.4280561899373305,
      "tokens_seen": 2231107584
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032738214643931795,
      "loss": 1.315,
      "theoretical_loss": 3.428048041989408,
      "tokens_seen": 2231173120
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003273620862587763,
      "loss": 1.3604,
      "theoretical_loss": 3.42803989434782,
      "tokens_seen": 2231238656
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032734202607823473,
      "loss": 1.4629,
      "theoretical_loss": 3.4280317470125463,
      "tokens_seen": 2231304192
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003273219658976931,
      "loss": 1.4465,
      "theoretical_loss": 3.428023599983567,
      "tokens_seen": 2231369728
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032730190571715145,
      "loss": 1.4021,
      "theoretical_loss": 3.428015453260861,
      "tokens_seen": 2231435264
    },
    {
      "epoch": 0.68,
      "objective/train/advantage_avg": 0.4927145838737488,
      "objective/train/docs_used": 812344,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1642310619354248,
      "objective/train/original_loss": 1.1642310619354248,
      "objective/train/theoretical_loss": 3.4280073068444086,
      "objective/train/tokens_used": 2251960800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24323010444641113,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505080223083496,
      "objective/train/weighted_lm_loss": 1.222900629043579,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4280073068444086,
      "tokens_seen": 2231500800
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003272818455366098,
      "loss": 1.3416,
      "theoretical_loss": 3.4280073068444086,
      "tokens_seen": 2231500800
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032726178535606823,
      "loss": 1.4499,
      "theoretical_loss": 3.427999160734188,
      "tokens_seen": 2231566336
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003272417251755266,
      "loss": 1.4003,
      "theoretical_loss": 3.4279910149301798,
      "tokens_seen": 2231631872
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032722166499498496,
      "loss": 1.4521,
      "theoretical_loss": 3.4279828694323626,
      "tokens_seen": 2231697408
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003272016048144433,
      "loss": 1.4119,
      "theoretical_loss": 3.427974724240717,
      "tokens_seen": 2231762944
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003271815446339017,
      "loss": 1.3261,
      "theoretical_loss": 3.427966579355222,
      "tokens_seen": 2231828480
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003271614844533601,
      "loss": 1.4135,
      "theoretical_loss": 3.4279584347758565,
      "tokens_seen": 2231894016
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032714142427281846,
      "loss": 1.2416,
      "theoretical_loss": 3.4279502905026007,
      "tokens_seen": 2231959552
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003271213640922768,
      "loss": 1.4009,
      "theoretical_loss": 3.4279421465354343,
      "tokens_seen": 2232025088
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003271013039117352,
      "loss": 1.4041,
      "theoretical_loss": 3.427934002874336,
      "tokens_seen": 2232090624
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003270812437311936,
      "loss": 1.3557,
      "theoretical_loss": 3.4279258595192856,
      "tokens_seen": 2232156160
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032706118355065196,
      "loss": 1.3354,
      "theoretical_loss": 3.427917716470263,
      "tokens_seen": 2232221696
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003270411233701103,
      "loss": 1.3996,
      "theoretical_loss": 3.4279095737272476,
      "tokens_seen": 2232287232
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003270210631895687,
      "loss": 1.2912,
      "theoretical_loss": 3.4279014312902185,
      "tokens_seen": 2232352768
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032700100300902705,
      "loss": 1.3568,
      "theoretical_loss": 3.427893289159156,
      "tokens_seen": 2232418304
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032698094282848547,
      "loss": 1.4517,
      "theoretical_loss": 3.4278851473340386,
      "tokens_seen": 2232483840
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032696088264794383,
      "loss": 1.3607,
      "theoretical_loss": 3.4278770058148464,
      "tokens_seen": 2232549376
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003269408224674022,
      "loss": 1.4136,
      "theoretical_loss": 3.4278688646015594,
      "tokens_seen": 2232614912
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032692076228686055,
      "loss": 1.3558,
      "theoretical_loss": 3.427860723694156,
      "tokens_seen": 2232680448
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032690070210631897,
      "loss": 1.4417,
      "theoretical_loss": 3.4278525830926165,
      "tokens_seen": 2232745984
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032688064192577733,
      "loss": 1.422,
      "theoretical_loss": 3.4278444427969204,
      "tokens_seen": 2232811520
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003268605817452357,
      "loss": 1.3595,
      "theoretical_loss": 3.427836302807047,
      "tokens_seen": 2232877056
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032684052156469406,
      "loss": 1.3797,
      "theoretical_loss": 3.427828163122976,
      "tokens_seen": 2232942592
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003268204613841524,
      "loss": 1.424,
      "theoretical_loss": 3.4278200237446868,
      "tokens_seen": 2233008128
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032680040120361084,
      "loss": 1.3084,
      "theoretical_loss": 3.427811884672159,
      "tokens_seen": 2233073664
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003267803410230692,
      "loss": 1.4608,
      "theoretical_loss": 3.4278037459053725,
      "tokens_seen": 2233139200
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032676028084252756,
      "loss": 1.3742,
      "theoretical_loss": 3.427795607444306,
      "tokens_seen": 2233204736
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003267402206619859,
      "loss": 1.4651,
      "theoretical_loss": 3.4277874692889396,
      "tokens_seen": 2233270272
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032672016048144434,
      "loss": 1.3948,
      "theoretical_loss": 3.427779331439253,
      "tokens_seen": 2233335808
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003267001003009027,
      "loss": 1.4918,
      "theoretical_loss": 3.4277711938952256,
      "tokens_seen": 2233401344
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032668004012036106,
      "loss": 1.3751,
      "theoretical_loss": 3.4277630566568367,
      "tokens_seen": 2233466880
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003266599799398194,
      "loss": 1.367,
      "theoretical_loss": 3.427754919724066,
      "tokens_seen": 2233532416
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032663991975927784,
      "loss": 1.405,
      "theoretical_loss": 3.427746783096893,
      "tokens_seen": 2233597952
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003266198595787362,
      "loss": 1.3254,
      "theoretical_loss": 3.427738646775298,
      "tokens_seen": 2233663488
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032659979939819457,
      "loss": 1.3575,
      "theoretical_loss": 3.4277305107592593,
      "tokens_seen": 2233729024
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032657973921765293,
      "loss": 1.3836,
      "theoretical_loss": 3.427722375048757,
      "tokens_seen": 2233794560
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032655967903711135,
      "loss": 1.396,
      "theoretical_loss": 3.427714239643771,
      "tokens_seen": 2233860096
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032653961885656976,
      "loss": 1.3959,
      "theoretical_loss": 3.4277061045442805,
      "tokens_seen": 2233925632
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003265195586760281,
      "loss": 1.468,
      "theoretical_loss": 3.427697969750265,
      "tokens_seen": 2233991168
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003264994984954865,
      "loss": 1.4232,
      "theoretical_loss": 3.4276898352617047,
      "tokens_seen": 2234056704
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032647943831494485,
      "loss": 1.3883,
      "theoretical_loss": 3.4276817010785785,
      "tokens_seen": 2234122240
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032645937813440326,
      "loss": 1.4854,
      "theoretical_loss": 3.4276735672008662,
      "tokens_seen": 2234187776
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003264393179538616,
      "loss": 1.3708,
      "theoretical_loss": 3.427665433628547,
      "tokens_seen": 2234253312
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032641925777332,
      "loss": 1.4143,
      "theoretical_loss": 3.4276573003616013,
      "tokens_seen": 2234318848
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032639919759277835,
      "loss": 1.462,
      "theoretical_loss": 3.427649167400008,
      "tokens_seen": 2234384384
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003263791374122367,
      "loss": 1.4388,
      "theoretical_loss": 3.4276410347437465,
      "tokens_seen": 2234449920
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032635907723169513,
      "loss": 1.3085,
      "theoretical_loss": 3.427632902392797,
      "tokens_seen": 2234515456
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003263390170511535,
      "loss": 1.2958,
      "theoretical_loss": 3.427624770347139,
      "tokens_seen": 2234580992
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032631895687061185,
      "loss": 1.4735,
      "theoretical_loss": 3.427616638606752,
      "tokens_seen": 2234646528
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003262988966900702,
      "loss": 1.4035,
      "theoretical_loss": 3.4276085071716156,
      "tokens_seen": 2234712064
    },
    {
      "epoch": 0.68,
      "objective/train/advantage_avg": 0.4961683452129364,
      "objective/train/docs_used": 813661,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3265491724014282,
      "objective/train/original_loss": 1.3265490531921387,
      "objective/train/theoretical_loss": 3.427600376041709,
      "objective/train/tokens_used": 2255237600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24633291363716125,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508692264556885,
      "objective/train/weighted_lm_loss": 1.3939355611801147,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 3.427600376041709,
      "tokens_seen": 2234777600
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032627883650952863,
      "loss": 1.4037,
      "theoretical_loss": 3.427600376041709,
      "tokens_seen": 2234777600
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000326258776328987,
      "loss": 1.396,
      "theoretical_loss": 3.4275922452170122,
      "tokens_seen": 2234843136
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032623871614844536,
      "loss": 1.3659,
      "theoretical_loss": 3.427584114697505,
      "tokens_seen": 2234908672
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003262186559679037,
      "loss": 1.3528,
      "theoretical_loss": 3.427575984483166,
      "tokens_seen": 2234974208
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003261985957873621,
      "loss": 1.2618,
      "theoretical_loss": 3.4275678545739763,
      "tokens_seen": 2235039744
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003261785356068205,
      "loss": 1.3862,
      "theoretical_loss": 3.427559724969914,
      "tokens_seen": 2235105280
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032615847542627886,
      "loss": 1.2642,
      "theoretical_loss": 3.42755159567096,
      "tokens_seen": 2235170816
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003261384152457372,
      "loss": 1.4374,
      "theoretical_loss": 3.427543466677093,
      "tokens_seen": 2235236352
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003261183550651956,
      "loss": 1.4858,
      "theoretical_loss": 3.4275353379882927,
      "tokens_seen": 2235301888
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000326098294884654,
      "loss": 1.4257,
      "theoretical_loss": 3.427527209604539,
      "tokens_seen": 2235367424
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032607823470411236,
      "loss": 1.3743,
      "theoretical_loss": 3.4275190815258116,
      "tokens_seen": 2235432960
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003260581745235707,
      "loss": 1.3025,
      "theoretical_loss": 3.42751095375209,
      "tokens_seen": 2235498496
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003260381143430291,
      "loss": 1.441,
      "theoretical_loss": 3.427502826283354,
      "tokens_seen": 2235564032
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032601805416248745,
      "loss": 1.3225,
      "theoretical_loss": 3.427494699119582,
      "tokens_seen": 2235629568
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032599799398194587,
      "loss": 1.3657,
      "theoretical_loss": 3.4274865722607553,
      "tokens_seen": 2235695104
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032597793380140423,
      "loss": 1.494,
      "theoretical_loss": 3.4274784457068526,
      "tokens_seen": 2235760640
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003259578736208626,
      "loss": 1.3533,
      "theoretical_loss": 3.427470319457854,
      "tokens_seen": 2235826176
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032593781344032095,
      "loss": 1.4479,
      "theoretical_loss": 3.4274621935137386,
      "tokens_seen": 2235891712
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032591775325977937,
      "loss": 1.4308,
      "theoretical_loss": 3.4274540678744865,
      "tokens_seen": 2235957248
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032589769307923773,
      "loss": 1.3972,
      "theoretical_loss": 3.4274459425400767,
      "tokens_seen": 2236022784
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003258776328986961,
      "loss": 1.4047,
      "theoretical_loss": 3.4274378175104894,
      "tokens_seen": 2236088320
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032585757271815446,
      "loss": 1.3945,
      "theoretical_loss": 3.427429692785704,
      "tokens_seen": 2236153856
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003258375125376128,
      "loss": 1.3952,
      "theoretical_loss": 3.4274215683657006,
      "tokens_seen": 2236219392
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032581745235707124,
      "loss": 1.2934,
      "theoretical_loss": 3.427413444250458,
      "tokens_seen": 2236284928
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003257973921765296,
      "loss": 1.3608,
      "theoretical_loss": 3.427405320439956,
      "tokens_seen": 2236350464
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032577733199598796,
      "loss": 1.3894,
      "theoretical_loss": 3.427397196934175,
      "tokens_seen": 2236416000
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003257572718154463,
      "loss": 1.3,
      "theoretical_loss": 3.427389073733094,
      "tokens_seen": 2236481536
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032573721163490474,
      "loss": 1.3429,
      "theoretical_loss": 3.4273809508366933,
      "tokens_seen": 2236547072
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003257171514543631,
      "loss": 1.3997,
      "theoretical_loss": 3.4273728282449514,
      "tokens_seen": 2236612608
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032569709127382146,
      "loss": 1.316,
      "theoretical_loss": 3.4273647059578485,
      "tokens_seen": 2236678144
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003256770310932798,
      "loss": 1.4367,
      "theoretical_loss": 3.4273565839753646,
      "tokens_seen": 2236743680
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032565697091273824,
      "loss": 1.3896,
      "theoretical_loss": 3.427348462297479,
      "tokens_seen": 2236809216
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003256369107321966,
      "loss": 1.3423,
      "theoretical_loss": 3.4273403409241716,
      "tokens_seen": 2236874752
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032561685055165497,
      "loss": 1.4773,
      "theoretical_loss": 3.4273322198554217,
      "tokens_seen": 2236940288
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032559679037111333,
      "loss": 1.4146,
      "theoretical_loss": 3.427324099091209,
      "tokens_seen": 2237005824
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003255767301905717,
      "loss": 1.351,
      "theoretical_loss": 3.4273159786315137,
      "tokens_seen": 2237071360
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003255566700100301,
      "loss": 1.2807,
      "theoretical_loss": 3.427307858476315,
      "tokens_seen": 2237136896
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032553660982948847,
      "loss": 1.496,
      "theoretical_loss": 3.4272997386255923,
      "tokens_seen": 2237202432
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032551654964894683,
      "loss": 1.3834,
      "theoretical_loss": 3.4272916190793254,
      "tokens_seen": 2237267968
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003254964894684052,
      "loss": 1.399,
      "theoretical_loss": 3.4272834998374946,
      "tokens_seen": 2237333504
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003254764292878636,
      "loss": 1.4365,
      "theoretical_loss": 3.4272753809000793,
      "tokens_seen": 2237399040
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000325456369107322,
      "loss": 1.4801,
      "theoretical_loss": 3.4272672622670584,
      "tokens_seen": 2237464576
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032543630892678034,
      "loss": 1.4147,
      "theoretical_loss": 3.4272591439384126,
      "tokens_seen": 2237530112
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003254162487462387,
      "loss": 1.2674,
      "theoretical_loss": 3.427251025914121,
      "tokens_seen": 2237595648
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032539618856569706,
      "loss": 1.4278,
      "theoretical_loss": 3.4272429081941636,
      "tokens_seen": 2237661184
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003253761283851555,
      "loss": 1.4407,
      "theoretical_loss": 3.4272347907785194,
      "tokens_seen": 2237726720
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032535606820461384,
      "loss": 1.5186,
      "theoretical_loss": 3.4272266736671693,
      "tokens_seen": 2237792256
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003253360080240722,
      "loss": 1.5052,
      "theoretical_loss": 3.4272185568600917,
      "tokens_seen": 2237857792
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032531594784353056,
      "loss": 1.2891,
      "theoretical_loss": 3.4272104403572667,
      "tokens_seen": 2237923328
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000325295887662989,
      "loss": 1.388,
      "theoretical_loss": 3.427202324158675,
      "tokens_seen": 2237988864
    },
    {
      "epoch": 0.68,
      "objective/train/advantage_avg": 0.4909612536430359,
      "objective/train/docs_used": 814894,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.215299129486084,
      "objective/train/original_loss": 1.215299129486084,
      "objective/train/theoretical_loss": 3.4271942082642948,
      "objective/train/tokens_used": 2258514400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.241785928606987,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503251552581787,
      "objective/train/weighted_lm_loss": 1.2764476537704468,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4271942082642948,
      "tokens_seen": 2238054400
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032527582748244734,
      "loss": 1.2472,
      "theoretical_loss": 3.4271942082642948,
      "tokens_seen": 2238054400
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003252557673019057,
      "loss": 1.4085,
      "theoretical_loss": 3.4271860926741065,
      "tokens_seen": 2238119936
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032523570712136407,
      "loss": 1.4656,
      "theoretical_loss": 3.4271779773880895,
      "tokens_seen": 2238185472
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032521564694082243,
      "loss": 1.4553,
      "theoretical_loss": 3.427169862406224,
      "tokens_seen": 2238251008
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032519558676028084,
      "loss": 1.3567,
      "theoretical_loss": 3.427161747728489,
      "tokens_seen": 2238316544
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003251755265797392,
      "loss": 1.2927,
      "theoretical_loss": 3.427153633354865,
      "tokens_seen": 2238382080
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032515546639919757,
      "loss": 1.335,
      "theoretical_loss": 3.427145519285331,
      "tokens_seen": 2238447616
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032513540621865593,
      "loss": 1.372,
      "theoretical_loss": 3.427137405519867,
      "tokens_seen": 2238513152
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032511534603811435,
      "loss": 1.4608,
      "theoretical_loss": 3.427129292058453,
      "tokens_seen": 2238578688
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003250952858575727,
      "loss": 1.504,
      "theoretical_loss": 3.427121178901068,
      "tokens_seen": 2238644224
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032507522567703107,
      "loss": 1.4073,
      "theoretical_loss": 3.427113066047692,
      "tokens_seen": 2238709760
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003250551654964895,
      "loss": 1.3595,
      "theoretical_loss": 3.427104953498305,
      "tokens_seen": 2238775296
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032503510531594785,
      "loss": 1.5033,
      "theoretical_loss": 3.4270968412528866,
      "tokens_seen": 2238840832
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032501504513540627,
      "loss": 1.4944,
      "theoretical_loss": 3.427088729311417,
      "tokens_seen": 2238906368
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032499498495486463,
      "loss": 1.3373,
      "theoretical_loss": 3.4270806176738744,
      "tokens_seen": 2238971904
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000324974924774323,
      "loss": 1.4148,
      "theoretical_loss": 3.42707250634024,
      "tokens_seen": 2239037440
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032495486459378135,
      "loss": 1.3241,
      "theoretical_loss": 3.4270643953104925,
      "tokens_seen": 2239102976
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032493480441323977,
      "loss": 1.4474,
      "theoretical_loss": 3.4270562845846126,
      "tokens_seen": 2239168512
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032491474423269813,
      "loss": 1.3732,
      "theoretical_loss": 3.4270481741625796,
      "tokens_seen": 2239234048
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003248946840521565,
      "loss": 1.4515,
      "theoretical_loss": 3.427040064044373,
      "tokens_seen": 2239299584
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032487462387161486,
      "loss": 1.2913,
      "theoretical_loss": 3.4270319542299723,
      "tokens_seen": 2239365120
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003248545636910732,
      "loss": 1.447,
      "theoretical_loss": 3.427023844719358,
      "tokens_seen": 2239430656
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032483450351053164,
      "loss": 1.4157,
      "theoretical_loss": 3.4270157355125095,
      "tokens_seen": 2239496192
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032481444332999,
      "loss": 1.4406,
      "theoretical_loss": 3.4270076266094063,
      "tokens_seen": 2239561728
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032479438314944836,
      "loss": 1.3631,
      "theoretical_loss": 3.426999518010028,
      "tokens_seen": 2239627264
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003247743229689067,
      "loss": 1.4448,
      "theoretical_loss": 3.4269914097143555,
      "tokens_seen": 2239692800
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032475426278836514,
      "loss": 1.4654,
      "theoretical_loss": 3.426983301722367,
      "tokens_seen": 2239758336
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003247342026078235,
      "loss": 1.355,
      "theoretical_loss": 3.426975194034043,
      "tokens_seen": 2239823872
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032471414242728186,
      "loss": 1.3786,
      "theoretical_loss": 3.4269670866493636,
      "tokens_seen": 2239889408
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003246940822467402,
      "loss": 1.3726,
      "theoretical_loss": 3.4269589795683073,
      "tokens_seen": 2239954944
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032467402206619864,
      "loss": 1.4132,
      "theoretical_loss": 3.4269508727908553,
      "tokens_seen": 2240020480
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000324653961885657,
      "loss": 1.3988,
      "theoretical_loss": 3.4269427663169867,
      "tokens_seen": 2240086016
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032463390170511537,
      "loss": 1.2776,
      "theoretical_loss": 3.426934660146681,
      "tokens_seen": 2240151552
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032461384152457373,
      "loss": 1.4766,
      "theoretical_loss": 3.4269265542799183,
      "tokens_seen": 2240217088
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003245937813440321,
      "loss": 1.3869,
      "theoretical_loss": 3.426918448716678,
      "tokens_seen": 2240282624
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003245737211634905,
      "loss": 1.4296,
      "theoretical_loss": 3.4269103434569406,
      "tokens_seen": 2240348160
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032455366098294887,
      "loss": 1.3357,
      "theoretical_loss": 3.426902238500685,
      "tokens_seen": 2240413696
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032453360080240723,
      "loss": 1.4709,
      "theoretical_loss": 3.4268941338478918,
      "tokens_seen": 2240479232
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003245135406218656,
      "loss": 1.4014,
      "theoretical_loss": 3.42688602949854,
      "tokens_seen": 2240544768
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000324493480441324,
      "loss": 1.264,
      "theoretical_loss": 3.42687792545261,
      "tokens_seen": 2240610304
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003244734202607824,
      "loss": 1.3921,
      "theoretical_loss": 3.4268698217100804,
      "tokens_seen": 2240675840
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032445336008024074,
      "loss": 1.3718,
      "theoretical_loss": 3.4268617182709327,
      "tokens_seen": 2240741376
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003244332998996991,
      "loss": 1.284,
      "theoretical_loss": 3.426853615135145,
      "tokens_seen": 2240806912
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032441323971915746,
      "loss": 1.3181,
      "theoretical_loss": 3.426845512302698,
      "tokens_seen": 2240872448
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003243931795386159,
      "loss": 1.3434,
      "theoretical_loss": 3.426837409773572,
      "tokens_seen": 2240937984
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032437311935807424,
      "loss": 1.4354,
      "theoretical_loss": 3.4268293075477456,
      "tokens_seen": 2241003520
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003243530591775326,
      "loss": 1.3825,
      "theoretical_loss": 3.426821205625199,
      "tokens_seen": 2241069056
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032433299899699096,
      "loss": 1.3982,
      "theoretical_loss": 3.426813104005912,
      "tokens_seen": 2241134592
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003243129388164494,
      "loss": 1.336,
      "theoretical_loss": 3.4268050026898647,
      "tokens_seen": 2241200128
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032429287863590774,
      "loss": 1.3309,
      "theoretical_loss": 3.4267969016770365,
      "tokens_seen": 2241265664
    },
    {
      "epoch": 0.68,
      "objective/train/advantage_avg": 0.49041199684143066,
      "objective/train/docs_used": 816099,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2903863191604614,
      "objective/train/original_loss": 1.2903861999511719,
      "objective/train/theoretical_loss": 3.4267888009674072,
      "objective/train/tokens_used": 2261791200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24131543934345245,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502679347991943,
      "objective/train/weighted_lm_loss": 1.355191946029663,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4267888009674072,
      "tokens_seen": 2241331200
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003242728184553661,
      "loss": 1.3366,
      "theoretical_loss": 3.4267888009674072,
      "tokens_seen": 2241331200
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032425275827482447,
      "loss": 1.3315,
      "theoretical_loss": 3.4267807005609567,
      "tokens_seen": 2241396736
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032423269809428283,
      "loss": 1.4033,
      "theoretical_loss": 3.4267726004576646,
      "tokens_seen": 2241462272
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032421263791374125,
      "loss": 1.4003,
      "theoretical_loss": 3.4267645006575114,
      "tokens_seen": 2241527808
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003241925777331996,
      "loss": 1.3618,
      "theoretical_loss": 3.4267564011604756,
      "tokens_seen": 2241593344
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032417251755265797,
      "loss": 1.3932,
      "theoretical_loss": 3.4267483019665383,
      "tokens_seen": 2241658880
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032415245737211633,
      "loss": 1.3324,
      "theoretical_loss": 3.4267402030756786,
      "tokens_seen": 2241724416
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032413239719157475,
      "loss": 1.3962,
      "theoretical_loss": 3.4267321044878765,
      "tokens_seen": 2241789952
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003241123370110331,
      "loss": 1.2859,
      "theoretical_loss": 3.426724006203112,
      "tokens_seen": 2241855488
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032409227683049147,
      "loss": 1.3141,
      "theoretical_loss": 3.426715908221364,
      "tokens_seen": 2241921024
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032407221664994984,
      "loss": 1.3263,
      "theoretical_loss": 3.426707810542613,
      "tokens_seen": 2241986560
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003240521564694082,
      "loss": 1.4116,
      "theoretical_loss": 3.4266997131668395,
      "tokens_seen": 2242052096
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003240320962888666,
      "loss": 1.3249,
      "theoretical_loss": 3.426691616094022,
      "tokens_seen": 2242117632
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000324012036108325,
      "loss": 1.4532,
      "theoretical_loss": 3.4266835193241407,
      "tokens_seen": 2242183168
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032399197592778334,
      "loss": 1.351,
      "theoretical_loss": 3.4266754228571763,
      "tokens_seen": 2242248704
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003239719157472417,
      "loss": 1.3902,
      "theoretical_loss": 3.4266673266931074,
      "tokens_seen": 2242314240
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003239518555667001,
      "loss": 1.4312,
      "theoretical_loss": 3.4266592308319144,
      "tokens_seen": 2242379776
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003239317953861585,
      "loss": 1.3922,
      "theoretical_loss": 3.426651135273577,
      "tokens_seen": 2242445312
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032391173520561684,
      "loss": 1.4156,
      "theoretical_loss": 3.426643040018075,
      "tokens_seen": 2242510848
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003238916750250752,
      "loss": 1.3777,
      "theoretical_loss": 3.4266349450653886,
      "tokens_seen": 2242576384
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032387161484453357,
      "loss": 1.3723,
      "theoretical_loss": 3.426626850415497,
      "tokens_seen": 2242641920
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000323851554663992,
      "loss": 1.3013,
      "theoretical_loss": 3.4266187560683803,
      "tokens_seen": 2242707456
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032383149448345034,
      "loss": 1.4547,
      "theoretical_loss": 3.4266106620240184,
      "tokens_seen": 2242772992
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003238114343029087,
      "loss": 1.2969,
      "theoretical_loss": 3.4266025682823913,
      "tokens_seen": 2242838528
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032379137412236707,
      "loss": 1.4013,
      "theoretical_loss": 3.4265944748434785,
      "tokens_seen": 2242904064
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003237713139418255,
      "loss": 1.4342,
      "theoretical_loss": 3.42658638170726,
      "tokens_seen": 2242969600
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032375125376128385,
      "loss": 1.3104,
      "theoretical_loss": 3.4265782888737153,
      "tokens_seen": 2243035136
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003237311935807422,
      "loss": 1.3189,
      "theoretical_loss": 3.4265701963428246,
      "tokens_seen": 2243100672
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032371113340020057,
      "loss": 1.4402,
      "theoretical_loss": 3.4265621041145677,
      "tokens_seen": 2243166208
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032369107321965893,
      "loss": 1.3528,
      "theoretical_loss": 3.426554012188925,
      "tokens_seen": 2243231744
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032367101303911735,
      "loss": 1.2645,
      "theoretical_loss": 3.426545920565875,
      "tokens_seen": 2243297280
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003236509528585757,
      "loss": 1.2917,
      "theoretical_loss": 3.4265378292453983,
      "tokens_seen": 2243362816
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003236308926780341,
      "loss": 1.4025,
      "theoretical_loss": 3.426529738227475,
      "tokens_seen": 2243428352
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032361083249749244,
      "loss": 1.3872,
      "theoretical_loss": 3.426521647512085,
      "tokens_seen": 2243493888
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032359077231695085,
      "loss": 1.3581,
      "theoretical_loss": 3.4265135570992076,
      "tokens_seen": 2243559424
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032357071213640927,
      "loss": 1.3047,
      "theoretical_loss": 3.4265054669888224,
      "tokens_seen": 2243624960
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032355065195586763,
      "loss": 1.2726,
      "theoretical_loss": 3.4264973771809104,
      "tokens_seen": 2243690496
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000323530591775326,
      "loss": 1.4361,
      "theoretical_loss": 3.4264892876754507,
      "tokens_seen": 2243756032
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003235105315947844,
      "loss": 1.2651,
      "theoretical_loss": 3.4264811984724233,
      "tokens_seen": 2243821568
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003234904714142428,
      "loss": 1.4202,
      "theoretical_loss": 3.4264731095718077,
      "tokens_seen": 2243887104
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032347041123370114,
      "loss": 1.3876,
      "theoretical_loss": 3.426465020973584,
      "tokens_seen": 2243952640
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003234503510531595,
      "loss": 1.3938,
      "theoretical_loss": 3.4264569326777323,
      "tokens_seen": 2244018176
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032343029087261786,
      "loss": 1.4039,
      "theoretical_loss": 3.4264488446842325,
      "tokens_seen": 2244083712
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003234102306920763,
      "loss": 1.3875,
      "theoretical_loss": 3.426440756993064,
      "tokens_seen": 2244149248
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032339017051153464,
      "loss": 1.3879,
      "theoretical_loss": 3.4264326696042073,
      "tokens_seen": 2244214784
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000323370110330993,
      "loss": 1.4935,
      "theoretical_loss": 3.426424582517642,
      "tokens_seen": 2244280320
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032335005015045136,
      "loss": 1.5353,
      "theoretical_loss": 3.4264164957333474,
      "tokens_seen": 2244345856
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003233299899699098,
      "loss": 1.3239,
      "theoretical_loss": 3.4264084092513043,
      "tokens_seen": 2244411392
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032330992978936814,
      "loss": 1.3752,
      "theoretical_loss": 3.4264003230714923,
      "tokens_seen": 2244476928
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003232898696088265,
      "loss": 1.2757,
      "theoretical_loss": 3.4263922371938906,
      "tokens_seen": 2244542464
    },
    {
      "epoch": 0.68,
      "objective/train/advantage_avg": 0.4909200072288513,
      "objective/train/docs_used": 817136,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.349873661994934,
      "objective/train/original_loss": 1.349873661994934,
      "objective/train/theoretical_loss": 3.42638415161848,
      "objective/train/tokens_used": 2265068000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24167725443840027,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503205060958862,
      "objective/train/weighted_lm_loss": 1.4177639484405518,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.42638415161848,
      "tokens_seen": 2244608000
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032326980942828487,
      "loss": 1.4494,
      "theoretical_loss": 3.42638415161848,
      "tokens_seen": 2244608000
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032324974924774323,
      "loss": 1.3378,
      "theoretical_loss": 3.42637606634524,
      "tokens_seen": 2244673536
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032322968906720165,
      "loss": 1.315,
      "theoretical_loss": 3.4263679813741503,
      "tokens_seen": 2244739072
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032320962888666,
      "loss": 1.3513,
      "theoretical_loss": 3.426359896705191,
      "tokens_seen": 2244804608
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032318956870611837,
      "loss": 1.3978,
      "theoretical_loss": 3.426351812338342,
      "tokens_seen": 2244870144
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032316950852557673,
      "loss": 1.4322,
      "theoretical_loss": 3.426343728273583,
      "tokens_seen": 2244935680
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032314944834503515,
      "loss": 1.4456,
      "theoretical_loss": 3.4263356445108943,
      "tokens_seen": 2245001216
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003231293881644935,
      "loss": 1.2746,
      "theoretical_loss": 3.4263275610502557,
      "tokens_seen": 2245066752
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003231093279839519,
      "loss": 1.4393,
      "theoretical_loss": 3.4263194778916466,
      "tokens_seen": 2245132288
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032308926780341024,
      "loss": 1.3949,
      "theoretical_loss": 3.4263113950350474,
      "tokens_seen": 2245197824
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003230692076228686,
      "loss": 1.2693,
      "theoretical_loss": 3.426303312480438,
      "tokens_seen": 2245263360
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000323049147442327,
      "loss": 1.4167,
      "theoretical_loss": 3.426295230227798,
      "tokens_seen": 2245328896
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003230290872617854,
      "loss": 1.4884,
      "theoretical_loss": 3.4262871482771073,
      "tokens_seen": 2245394432
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032300902708124374,
      "loss": 1.3894,
      "theoretical_loss": 3.4262790666283456,
      "tokens_seen": 2245459968
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003229889669007021,
      "loss": 1.4,
      "theoretical_loss": 3.426270985281494,
      "tokens_seen": 2245525504
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003229689067201605,
      "loss": 1.4279,
      "theoretical_loss": 3.426262904236531,
      "tokens_seen": 2245591040
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003229488465396189,
      "loss": 1.2768,
      "theoretical_loss": 3.426254823493437,
      "tokens_seen": 2245656576
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032292878635907724,
      "loss": 1.4175,
      "theoretical_loss": 3.4262467430521926,
      "tokens_seen": 2245722112
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003229087261785356,
      "loss": 1.3593,
      "theoretical_loss": 3.426238662912777,
      "tokens_seen": 2245787648
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032288866599799397,
      "loss": 1.4035,
      "theoretical_loss": 3.42623058307517,
      "tokens_seen": 2245853184
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003228686058174524,
      "loss": 1.4006,
      "theoretical_loss": 3.4262225035393516,
      "tokens_seen": 2245918720
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032284854563691074,
      "loss": 1.4144,
      "theoretical_loss": 3.4262144243053023,
      "tokens_seen": 2245984256
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003228284854563691,
      "loss": 1.4044,
      "theoretical_loss": 3.4262063453730014,
      "tokens_seen": 2246049792
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032280842527582747,
      "loss": 1.3326,
      "theoretical_loss": 3.426198266742429,
      "tokens_seen": 2246115328
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003227883650952859,
      "loss": 1.4925,
      "theoretical_loss": 3.426190188413565,
      "tokens_seen": 2246180864
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032276830491474425,
      "loss": 1.3859,
      "theoretical_loss": 3.4261821103863896,
      "tokens_seen": 2246246400
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003227482447342026,
      "loss": 1.3871,
      "theoretical_loss": 3.4261740326608825,
      "tokens_seen": 2246311936
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032272818455366097,
      "loss": 1.2842,
      "theoretical_loss": 3.4261659552370234,
      "tokens_seen": 2246377472
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032270812437311933,
      "loss": 1.4309,
      "theoretical_loss": 3.4261578781147923,
      "tokens_seen": 2246443008
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032268806419257775,
      "loss": 1.3092,
      "theoretical_loss": 3.42614980129417,
      "tokens_seen": 2246508544
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003226680040120361,
      "loss": 1.3179,
      "theoretical_loss": 3.4261417247751353,
      "tokens_seen": 2246574080
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003226479438314945,
      "loss": 1.4664,
      "theoretical_loss": 3.426133648557669,
      "tokens_seen": 2246639616
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032262788365095284,
      "loss": 1.3737,
      "theoretical_loss": 3.4261255726417503,
      "tokens_seen": 2246705152
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032260782347041125,
      "loss": 1.2933,
      "theoretical_loss": 3.4261174970273593,
      "tokens_seen": 2246770688
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003225877632898696,
      "loss": 1.3764,
      "theoretical_loss": 3.4261094217144765,
      "tokens_seen": 2246836224
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000322567703109328,
      "loss": 1.3683,
      "theoretical_loss": 3.4261013467030814,
      "tokens_seen": 2246901760
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032254764292878634,
      "loss": 1.3607,
      "theoretical_loss": 3.426093271993154,
      "tokens_seen": 2246967296
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003225275827482447,
      "loss": 1.5027,
      "theoretical_loss": 3.4260851975846744,
      "tokens_seen": 2247032832
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003225075225677031,
      "loss": 1.4266,
      "theoretical_loss": 3.4260771234776226,
      "tokens_seen": 2247098368
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003224874623871615,
      "loss": 1.3988,
      "theoretical_loss": 3.4260690496719786,
      "tokens_seen": 2247163904
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032246740220661984,
      "loss": 1.2528,
      "theoretical_loss": 3.4260609761677214,
      "tokens_seen": 2247229440
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003224473420260782,
      "loss": 1.4367,
      "theoretical_loss": 3.426052902964832,
      "tokens_seen": 2247294976
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003224272818455366,
      "loss": 1.5235,
      "theoretical_loss": 3.4260448300632906,
      "tokens_seen": 2247360512
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000322407221664995,
      "loss": 1.3526,
      "theoretical_loss": 3.426036757463076,
      "tokens_seen": 2247426048
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032238716148445335,
      "loss": 1.4706,
      "theoretical_loss": 3.4260286851641695,
      "tokens_seen": 2247491584
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003223671013039117,
      "loss": 1.4042,
      "theoretical_loss": 3.42602061316655,
      "tokens_seen": 2247557120
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003223470411233701,
      "loss": 1.403,
      "theoretical_loss": 3.4260125414701976,
      "tokens_seen": 2247622656
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003223269809428285,
      "loss": 1.4038,
      "theoretical_loss": 3.426004470075093,
      "tokens_seen": 2247688192
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032230692076228685,
      "loss": 1.316,
      "theoretical_loss": 3.4259963989812157,
      "tokens_seen": 2247753728
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003222868605817452,
      "loss": 1.3568,
      "theoretical_loss": 3.425988328188545,
      "tokens_seen": 2247819264
    },
    {
      "epoch": 0.68,
      "objective/train/advantage_avg": 0.4955300986766815,
      "objective/train/docs_used": 818241,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2094385623931885,
      "objective/train/original_loss": 1.2094385623931885,
      "objective/train/theoretical_loss": 3.4259802576970624,
      "objective/train/tokens_used": 2268344800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24573001265525818,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050802230834961,
      "objective/train/weighted_lm_loss": 1.2709048986434937,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.4259802576970624,
      "tokens_seen": 2247884800
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003222668004012036,
      "loss": 1.2882,
      "theoretical_loss": 3.4259802576970624,
      "tokens_seen": 2247884800
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000322246740220662,
      "loss": 1.3891,
      "theoretical_loss": 3.425972187506747,
      "tokens_seen": 2247950336
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032222668004012035,
      "loss": 1.3929,
      "theoretical_loss": 3.425964117617578,
      "tokens_seen": 2248015872
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003222066198595787,
      "loss": 1.4113,
      "theoretical_loss": 3.425956048029537,
      "tokens_seen": 2248081408
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003221865596790371,
      "loss": 1.3719,
      "theoretical_loss": 3.425947978742603,
      "tokens_seen": 2248146944
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003221664994984955,
      "loss": 1.3961,
      "theoretical_loss": 3.4259399097567558,
      "tokens_seen": 2248212480
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032214643931795386,
      "loss": 1.3135,
      "theoretical_loss": 3.425931841071976,
      "tokens_seen": 2248278016
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003221263791374122,
      "loss": 1.3619,
      "theoretical_loss": 3.425923772688243,
      "tokens_seen": 2248343552
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003221063189568706,
      "loss": 1.4619,
      "theoretical_loss": 3.425915704605538,
      "tokens_seen": 2248409088
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032208625877632894,
      "loss": 1.2873,
      "theoretical_loss": 3.4259076368238395,
      "tokens_seen": 2248474624
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003220661985957874,
      "loss": 1.3517,
      "theoretical_loss": 3.4258995693431284,
      "tokens_seen": 2248540160
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003220461384152458,
      "loss": 1.4341,
      "theoretical_loss": 3.4258915021633847,
      "tokens_seen": 2248605696
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032202607823470414,
      "loss": 1.4184,
      "theoretical_loss": 3.4258834352845877,
      "tokens_seen": 2248671232
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003220060180541625,
      "loss": 1.4054,
      "theoretical_loss": 3.425875368706718,
      "tokens_seen": 2248736768
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003219859578736209,
      "loss": 1.2825,
      "theoretical_loss": 3.4258673024297552,
      "tokens_seen": 2248802304
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003219658976930793,
      "loss": 1.4454,
      "theoretical_loss": 3.42585923645368,
      "tokens_seen": 2248867840
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032194583751253764,
      "loss": 1.3746,
      "theoretical_loss": 3.425851170778472,
      "tokens_seen": 2248933376
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000321925777331996,
      "loss": 1.391,
      "theoretical_loss": 3.4258431054041107,
      "tokens_seen": 2248998912
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032190571715145437,
      "loss": 1.3894,
      "theoretical_loss": 3.425835040330577,
      "tokens_seen": 2249064448
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003218856569709128,
      "loss": 1.3843,
      "theoretical_loss": 3.4258269755578503,
      "tokens_seen": 2249129984
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032186559679037115,
      "loss": 1.3704,
      "theoretical_loss": 3.4258189110859107,
      "tokens_seen": 2249195520
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003218455366098295,
      "loss": 1.4022,
      "theoretical_loss": 3.4258108469147386,
      "tokens_seen": 2249261056
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032182547642928787,
      "loss": 1.3744,
      "theoretical_loss": 3.425802783044314,
      "tokens_seen": 2249326592
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003218054162487463,
      "loss": 1.284,
      "theoretical_loss": 3.425794719474616,
      "tokens_seen": 2249392128
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032178535606820465,
      "loss": 1.3732,
      "theoretical_loss": 3.425786656205626,
      "tokens_seen": 2249457664
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000321765295887663,
      "loss": 1.3423,
      "theoretical_loss": 3.4257785932373226,
      "tokens_seen": 2249523200
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032174523570712137,
      "loss": 1.4528,
      "theoretical_loss": 3.425770530569687,
      "tokens_seen": 2249588736
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032172517552657974,
      "loss": 1.3496,
      "theoretical_loss": 3.4257624682026986,
      "tokens_seen": 2249654272
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032170511534603815,
      "loss": 1.4133,
      "theoretical_loss": 3.425754406136338,
      "tokens_seen": 2249719808
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003216850551654965,
      "loss": 1.3225,
      "theoretical_loss": 3.4257463443705847,
      "tokens_seen": 2249785344
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003216649949849549,
      "loss": 1.2434,
      "theoretical_loss": 3.425738282905418,
      "tokens_seen": 2249850880
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032164493480441324,
      "loss": 1.3673,
      "theoretical_loss": 3.42573022174082,
      "tokens_seen": 2249916416
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032162487462387165,
      "loss": 1.4261,
      "theoretical_loss": 3.425722160876769,
      "tokens_seen": 2249981952
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032160481444333,
      "loss": 1.4323,
      "theoretical_loss": 3.425714100313246,
      "tokens_seen": 2250047488
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003215847542627884,
      "loss": 1.4413,
      "theoretical_loss": 3.42570604005023,
      "tokens_seen": 2250113024
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032156469408224674,
      "loss": 1.4525,
      "theoretical_loss": 3.425697980087702,
      "tokens_seen": 2250178560
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003215446339017051,
      "loss": 1.3974,
      "theoretical_loss": 3.425689920425642,
      "tokens_seen": 2250244096
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003215245737211635,
      "loss": 1.3823,
      "theoretical_loss": 3.42568186106403,
      "tokens_seen": 2250309632
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003215045135406219,
      "loss": 1.4076,
      "theoretical_loss": 3.425673802002845,
      "tokens_seen": 2250375168
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032148445336008024,
      "loss": 1.338,
      "theoretical_loss": 3.4256657432420683,
      "tokens_seen": 2250440704
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003214643931795386,
      "loss": 1.3869,
      "theoretical_loss": 3.42565768478168,
      "tokens_seen": 2250506240
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000321444332998997,
      "loss": 1.4271,
      "theoretical_loss": 3.425649626621659,
      "tokens_seen": 2250571776
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003214242728184554,
      "loss": 1.4056,
      "theoretical_loss": 3.4256415687619866,
      "tokens_seen": 2250637312
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032140421263791375,
      "loss": 1.4266,
      "theoretical_loss": 3.425633511202642,
      "tokens_seen": 2250702848
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003213841524573721,
      "loss": 1.3434,
      "theoretical_loss": 3.4256254539436055,
      "tokens_seen": 2250768384
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003213640922768305,
      "loss": 1.4833,
      "theoretical_loss": 3.4256173969848573,
      "tokens_seen": 2250833920
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003213440320962889,
      "loss": 1.3896,
      "theoretical_loss": 3.4256093403263774,
      "tokens_seen": 2250899456
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032132397191574725,
      "loss": 1.3668,
      "theoretical_loss": 3.4256012839681462,
      "tokens_seen": 2250964992
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003213039117352056,
      "loss": 1.2948,
      "theoretical_loss": 3.4255932279101433,
      "tokens_seen": 2251030528
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000321283851554664,
      "loss": 1.351,
      "theoretical_loss": 3.4255851721523487,
      "tokens_seen": 2251096064
    },
    {
      "epoch": 0.68,
      "objective/train/advantage_avg": 0.49015873670578003,
      "objective/train/docs_used": 819526,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4248615503311157,
      "objective/train/original_loss": 1.4248615503311157,
      "objective/train/theoretical_loss": 3.4255771166947424,
      "objective/train/tokens_used": 2271621600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24091733992099762,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502405166625977,
      "objective/train/weighted_lm_loss": 1.4966938495635986,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4255771166947424,
      "tokens_seen": 2251161600
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003212637913741224,
      "loss": 1.4788,
      "theoretical_loss": 3.4255771166947424,
      "tokens_seen": 2251161600
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032124373119358075,
      "loss": 1.343,
      "theoretical_loss": 3.4255690615373053,
      "tokens_seen": 2251227136
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003212236710130391,
      "loss": 1.4408,
      "theoretical_loss": 3.4255610066800166,
      "tokens_seen": 2251292672
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003212036108324975,
      "loss": 1.4196,
      "theoretical_loss": 3.425552952122857,
      "tokens_seen": 2251358208
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003211835506519559,
      "loss": 1.4008,
      "theoretical_loss": 3.4255448978658065,
      "tokens_seen": 2251423744
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032116349047141426,
      "loss": 1.3082,
      "theoretical_loss": 3.4255368439088447,
      "tokens_seen": 2251489280
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003211434302908726,
      "loss": 1.4401,
      "theoretical_loss": 3.425528790251952,
      "tokens_seen": 2251554816
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000321123370110331,
      "loss": 1.4313,
      "theoretical_loss": 3.425520736895108,
      "tokens_seen": 2251620352
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032110330992978934,
      "loss": 1.3841,
      "theoretical_loss": 3.425512683838294,
      "tokens_seen": 2251685888
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032108324974924776,
      "loss": 1.3941,
      "theoretical_loss": 3.425504631081489,
      "tokens_seen": 2251751424
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003210631895687061,
      "loss": 1.3772,
      "theoretical_loss": 3.4254965786246734,
      "tokens_seen": 2251816960
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003210431293881645,
      "loss": 1.428,
      "theoretical_loss": 3.425488526467827,
      "tokens_seen": 2251882496
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032102306920762285,
      "loss": 1.4109,
      "theoretical_loss": 3.425480474610931,
      "tokens_seen": 2251948032
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032100300902708126,
      "loss": 1.3326,
      "theoretical_loss": 3.4254724230539644,
      "tokens_seen": 2252013568
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003209829488465396,
      "loss": 1.4397,
      "theoretical_loss": 3.4254643717969073,
      "tokens_seen": 2252079104
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000320962888665998,
      "loss": 1.3711,
      "theoretical_loss": 3.4254563208397406,
      "tokens_seen": 2252144640
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032094282848545635,
      "loss": 1.4055,
      "theoretical_loss": 3.425448270182444,
      "tokens_seen": 2252210176
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003209227683049147,
      "loss": 1.3728,
      "theoretical_loss": 3.4254402198249965,
      "tokens_seen": 2252275712
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032090270812437313,
      "loss": 1.3234,
      "theoretical_loss": 3.42543216976738,
      "tokens_seen": 2252341248
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003208826479438315,
      "loss": 1.3597,
      "theoretical_loss": 3.425424120009574,
      "tokens_seen": 2252406784
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032086258776328985,
      "loss": 1.2998,
      "theoretical_loss": 3.4254160705515577,
      "tokens_seen": 2252472320
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003208425275827482,
      "loss": 1.3129,
      "theoretical_loss": 3.425408021393313,
      "tokens_seen": 2252537856
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032082246740220663,
      "loss": 1.3693,
      "theoretical_loss": 3.425399972534818,
      "tokens_seen": 2252603392
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000320802407221665,
      "loss": 1.2993,
      "theoretical_loss": 3.4253919239760546,
      "tokens_seen": 2252668928
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032078234704112336,
      "loss": 1.3174,
      "theoretical_loss": 3.4253838757170016,
      "tokens_seen": 2252734464
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003207622868605817,
      "loss": 1.3851,
      "theoretical_loss": 3.42537582775764,
      "tokens_seen": 2252800000
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003207422266800401,
      "loss": 1.3277,
      "theoretical_loss": 3.4253677800979494,
      "tokens_seen": 2252865536
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003207221664994985,
      "loss": 1.4115,
      "theoretical_loss": 3.4253597327379097,
      "tokens_seen": 2252931072
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032070210631895686,
      "loss": 1.3658,
      "theoretical_loss": 3.425351685677502,
      "tokens_seen": 2252996608
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003206820461384152,
      "loss": 1.4096,
      "theoretical_loss": 3.4253436389167056,
      "tokens_seen": 2253062144
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003206619859578736,
      "loss": 1.4327,
      "theoretical_loss": 3.425335592455501,
      "tokens_seen": 2253127680
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000320641925777332,
      "loss": 1.3469,
      "theoretical_loss": 3.425327546293868,
      "tokens_seen": 2253193216
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032062186559679036,
      "loss": 1.3278,
      "theoretical_loss": 3.4253195004317867,
      "tokens_seen": 2253258752
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003206018054162487,
      "loss": 1.4288,
      "theoretical_loss": 3.425311454869238,
      "tokens_seen": 2253324288
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003205817452357071,
      "loss": 1.4335,
      "theoretical_loss": 3.4253034096062014,
      "tokens_seen": 2253389824
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003205616850551655,
      "loss": 1.3707,
      "theoretical_loss": 3.425295364642657,
      "tokens_seen": 2253455360
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003205416248746239,
      "loss": 1.5923,
      "theoretical_loss": 3.4252873199785854,
      "tokens_seen": 2253520896
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003205215646940823,
      "loss": 1.4562,
      "theoretical_loss": 3.425279275613966,
      "tokens_seen": 2253586432
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032050150451354064,
      "loss": 1.4268,
      "theoretical_loss": 3.4252712315487797,
      "tokens_seen": 2253651968
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000320481444332999,
      "loss": 1.5164,
      "theoretical_loss": 3.425263187783006,
      "tokens_seen": 2253717504
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003204613841524574,
      "loss": 1.3944,
      "theoretical_loss": 3.4252551443166257,
      "tokens_seen": 2253783040
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003204413239719158,
      "loss": 1.4649,
      "theoretical_loss": 3.4252471011496186,
      "tokens_seen": 2253848576
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032042126379137415,
      "loss": 1.3436,
      "theoretical_loss": 3.4252390582819645,
      "tokens_seen": 2253914112
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003204012036108325,
      "loss": 1.377,
      "theoretical_loss": 3.4252310157136443,
      "tokens_seen": 2253979648
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003203811434302909,
      "loss": 1.3848,
      "theoretical_loss": 3.425222973444638,
      "tokens_seen": 2254045184
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003203610832497493,
      "loss": 1.4048,
      "theoretical_loss": 3.425214931474925,
      "tokens_seen": 2254110720
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032034102306920765,
      "loss": 1.3712,
      "theoretical_loss": 3.4252068898044863,
      "tokens_seen": 2254176256
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000320320962888666,
      "loss": 1.4024,
      "theoretical_loss": 3.425198848433302,
      "tokens_seen": 2254241792
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003203009027081244,
      "loss": 1.299,
      "theoretical_loss": 3.4251908073613517,
      "tokens_seen": 2254307328
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003202808425275828,
      "loss": 1.2897,
      "theoretical_loss": 3.425182766588616,
      "tokens_seen": 2254372864
    },
    {
      "epoch": 0.68,
      "objective/train/advantage_avg": 0.49256476759910583,
      "objective/train/docs_used": 820753,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2347317934036255,
      "objective/train/original_loss": 1.2347315549850464,
      "objective/train/theoretical_loss": 3.425174726115075,
      "objective/train/tokens_used": 2274898400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24284179508686066,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504909753799438,
      "objective/train/weighted_lm_loss": 1.2971256971359253,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.425174726115075,
      "tokens_seen": 2254438400
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032026078234704115,
      "loss": 1.3592,
      "theoretical_loss": 3.425174726115075,
      "tokens_seen": 2254438400
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003202407221664995,
      "loss": 1.4096,
      "theoretical_loss": 3.4251666859407086,
      "tokens_seen": 2254503936
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003202206619859579,
      "loss": 1.3541,
      "theoretical_loss": 3.4251586460654977,
      "tokens_seen": 2254569472
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003202006018054163,
      "loss": 1.3399,
      "theoretical_loss": 3.4251506064894217,
      "tokens_seen": 2254635008
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032018054162487466,
      "loss": 1.3789,
      "theoretical_loss": 3.425142567212461,
      "tokens_seen": 2254700544
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000320160481444333,
      "loss": 1.3343,
      "theoretical_loss": 3.425134528234596,
      "tokens_seen": 2254766080
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003201404212637914,
      "loss": 1.4469,
      "theoretical_loss": 3.4251264895558067,
      "tokens_seen": 2254831616
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032012036108324974,
      "loss": 1.344,
      "theoretical_loss": 3.4251184511760733,
      "tokens_seen": 2254897152
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032010030090270816,
      "loss": 1.308,
      "theoretical_loss": 3.4251104130953762,
      "tokens_seen": 2254962688
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003200802407221665,
      "loss": 1.5071,
      "theoretical_loss": 3.425102375313695,
      "tokens_seen": 2255028224
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003200601805416249,
      "loss": 1.4452,
      "theoretical_loss": 3.4250943378310104,
      "tokens_seen": 2255093760
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032004012036108325,
      "loss": 1.4352,
      "theoretical_loss": 3.4250863006473025,
      "tokens_seen": 2255159296
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032002006018054166,
      "loss": 1.367,
      "theoretical_loss": 3.4250782637625514,
      "tokens_seen": 2255224832
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00032,
      "loss": 1.388,
      "theoretical_loss": 3.4250702271767373,
      "tokens_seen": 2255290368
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003199799398194584,
      "loss": 1.4249,
      "theoretical_loss": 3.4250621908898404,
      "tokens_seen": 2255355904
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031995987963891675,
      "loss": 1.4832,
      "theoretical_loss": 3.425054154901841,
      "tokens_seen": 2255421440
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003199398194583751,
      "loss": 1.2844,
      "theoretical_loss": 3.4250461192127193,
      "tokens_seen": 2255486976
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031991975927783353,
      "loss": 1.4191,
      "theoretical_loss": 3.4250380838224554,
      "tokens_seen": 2255552512
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003198996990972919,
      "loss": 1.3645,
      "theoretical_loss": 3.4250300487310295,
      "tokens_seen": 2255618048
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031987963891675025,
      "loss": 1.4312,
      "theoretical_loss": 3.4250220139384213,
      "tokens_seen": 2255683584
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003198595787362086,
      "loss": 1.349,
      "theoretical_loss": 3.4250139794446124,
      "tokens_seen": 2255749120
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031983951855566703,
      "loss": 1.2587,
      "theoretical_loss": 3.4250059452495814,
      "tokens_seen": 2255814656
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003198194583751254,
      "loss": 1.3069,
      "theoretical_loss": 3.4249979113533096,
      "tokens_seen": 2255880192
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031979939819458376,
      "loss": 1.361,
      "theoretical_loss": 3.424989877755777,
      "tokens_seen": 2255945728
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003197793380140421,
      "loss": 1.3342,
      "theoretical_loss": 3.4249818444569637,
      "tokens_seen": 2256011264
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003197592778335005,
      "loss": 1.2926,
      "theoretical_loss": 3.4249738114568498,
      "tokens_seen": 2256076800
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003197392176529589,
      "loss": 1.3811,
      "theoretical_loss": 3.4249657787554155,
      "tokens_seen": 2256142336
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031971915747241726,
      "loss": 1.3394,
      "theoretical_loss": 3.424957746352641,
      "tokens_seen": 2256207872
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003196990972918756,
      "loss": 1.3597,
      "theoretical_loss": 3.424949714248507,
      "tokens_seen": 2256273408
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000319679037111334,
      "loss": 1.3721,
      "theoretical_loss": 3.4249416824429932,
      "tokens_seen": 2256338944
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003196589769307924,
      "loss": 1.4394,
      "theoretical_loss": 3.42493365093608,
      "tokens_seen": 2256404480
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031963891675025076,
      "loss": 1.3557,
      "theoretical_loss": 3.4249256197277473,
      "tokens_seen": 2256470016
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003196188565697091,
      "loss": 1.3401,
      "theoretical_loss": 3.4249175888179764,
      "tokens_seen": 2256535552
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003195987963891675,
      "loss": 1.409,
      "theoretical_loss": 3.424909558206746,
      "tokens_seen": 2256601088
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031957873620862585,
      "loss": 1.3204,
      "theoretical_loss": 3.4249015278940376,
      "tokens_seen": 2256666624
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031955867602808427,
      "loss": 1.2547,
      "theoretical_loss": 3.4248934978798307,
      "tokens_seen": 2256732160
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031953861584754263,
      "loss": 1.3682,
      "theoretical_loss": 3.424885468164106,
      "tokens_seen": 2256797696
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000319518555667001,
      "loss": 1.4367,
      "theoretical_loss": 3.4248774387468433,
      "tokens_seen": 2256863232
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031949849548645935,
      "loss": 1.3194,
      "theoretical_loss": 3.424869409628023,
      "tokens_seen": 2256928768
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031947843530591777,
      "loss": 1.3677,
      "theoretical_loss": 3.424861380807626,
      "tokens_seen": 2256994304
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031945837512537613,
      "loss": 1.4598,
      "theoretical_loss": 3.4248533522856315,
      "tokens_seen": 2257059840
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003194383149448345,
      "loss": 1.4408,
      "theoretical_loss": 3.4248453240620202,
      "tokens_seen": 2257125376
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031941825476429286,
      "loss": 1.5117,
      "theoretical_loss": 3.424837296136772,
      "tokens_seen": 2257190912
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003193981945837512,
      "loss": 1.4628,
      "theoretical_loss": 3.424829268509868,
      "tokens_seen": 2257256448
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031937813440320964,
      "loss": 1.4182,
      "theoretical_loss": 3.424821241181288,
      "tokens_seen": 2257321984
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000319358074222668,
      "loss": 1.4725,
      "theoretical_loss": 3.424813214151012,
      "tokens_seen": 2257387520
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031933801404212636,
      "loss": 1.4181,
      "theoretical_loss": 3.4248051874190204,
      "tokens_seen": 2257453056
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003193179538615847,
      "loss": 1.3802,
      "theoretical_loss": 3.4247971609852934,
      "tokens_seen": 2257518592
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031929789368104314,
      "loss": 1.3288,
      "theoretical_loss": 3.4247891348498114,
      "tokens_seen": 2257584128
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003192778335005015,
      "loss": 1.453,
      "theoretical_loss": 3.424781109012555,
      "tokens_seen": 2257649664
    },
    {
      "epoch": 0.68,
      "objective/train/advantage_avg": 0.4962832033634186,
      "objective/train/docs_used": 822060,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2130647897720337,
      "objective/train/original_loss": 1.2130649089813232,
      "objective/train/theoretical_loss": 3.4247730834735037,
      "objective/train/tokens_used": 2278175200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2464459091424942,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508813858032227,
      "objective/train/weighted_lm_loss": 1.2747585773468018,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 3.4247730834735037,
      "tokens_seen": 2257715200
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031925777331995986,
      "loss": 1.3246,
      "theoretical_loss": 3.4247730834735037,
      "tokens_seen": 2257715200
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003192377131394182,
      "loss": 1.3368,
      "theoretical_loss": 3.424765058232638,
      "tokens_seen": 2257780736
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003192176529588766,
      "loss": 1.2791,
      "theoretical_loss": 3.424757033289939,
      "tokens_seen": 2257846272
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000319197592778335,
      "loss": 1.3931,
      "theoretical_loss": 3.424749008645386,
      "tokens_seen": 2257911808
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031917753259779337,
      "loss": 1.3976,
      "theoretical_loss": 3.424740984298959,
      "tokens_seen": 2257977344
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031915747241725173,
      "loss": 1.3784,
      "theoretical_loss": 3.4247329602506396,
      "tokens_seen": 2258042880
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003191374122367101,
      "loss": 1.4424,
      "theoretical_loss": 3.424724936500407,
      "tokens_seen": 2258108416
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003191173520561685,
      "loss": 1.4316,
      "theoretical_loss": 3.4247169130482416,
      "tokens_seen": 2258173952
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031909729187562687,
      "loss": 1.3341,
      "theoretical_loss": 3.424708889894124,
      "tokens_seen": 2258239488
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031907723169508523,
      "loss": 1.3933,
      "theoretical_loss": 3.4247008670380343,
      "tokens_seen": 2258305024
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031905717151454365,
      "loss": 1.3035,
      "theoretical_loss": 3.424692844479953,
      "tokens_seen": 2258370560
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031903711133400206,
      "loss": 1.4354,
      "theoretical_loss": 3.42468482221986,
      "tokens_seen": 2258436096
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003190170511534604,
      "loss": 1.4271,
      "theoretical_loss": 3.4246768002577355,
      "tokens_seen": 2258501632
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003189969909729188,
      "loss": 1.4402,
      "theoretical_loss": 3.424668778593561,
      "tokens_seen": 2258567168
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031897693079237715,
      "loss": 1.4026,
      "theoretical_loss": 3.424660757227315,
      "tokens_seen": 2258632704
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003189568706118355,
      "loss": 1.4724,
      "theoretical_loss": 3.424652736158979,
      "tokens_seen": 2258698240
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031893681043129393,
      "loss": 1.4334,
      "theoretical_loss": 3.4246447153885327,
      "tokens_seen": 2258763776
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003189167502507523,
      "loss": 1.5001,
      "theoretical_loss": 3.424636694915957,
      "tokens_seen": 2258829312
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031889669007021065,
      "loss": 1.3724,
      "theoretical_loss": 3.4246286747412316,
      "tokens_seen": 2258894848
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000318876629889669,
      "loss": 1.3818,
      "theoretical_loss": 3.424620654864337,
      "tokens_seen": 2258960384
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031885656970912743,
      "loss": 1.4744,
      "theoretical_loss": 3.4246126352852535,
      "tokens_seen": 2259025920
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003188365095285858,
      "loss": 1.3962,
      "theoretical_loss": 3.424604616003961,
      "tokens_seen": 2259091456
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031881644934804416,
      "loss": 1.4733,
      "theoretical_loss": 3.4245965970204413,
      "tokens_seen": 2259156992
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003187963891675025,
      "loss": 1.3414,
      "theoretical_loss": 3.424588578334673,
      "tokens_seen": 2259222528
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003187763289869609,
      "loss": 1.3891,
      "theoretical_loss": 3.4245805599466372,
      "tokens_seen": 2259288064
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003187562688064193,
      "loss": 1.4011,
      "theoretical_loss": 3.424572541856314,
      "tokens_seen": 2259353600
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031873620862587766,
      "loss": 1.3531,
      "theoretical_loss": 3.4245645240636833,
      "tokens_seen": 2259419136
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.000318716148445336,
      "loss": 1.4259,
      "theoretical_loss": 3.4245565065687265,
      "tokens_seen": 2259484672
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003186960882647944,
      "loss": 1.4783,
      "theoretical_loss": 3.424548489371423,
      "tokens_seen": 2259550208
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003186760280842528,
      "loss": 1.4071,
      "theoretical_loss": 3.4245404724717536,
      "tokens_seen": 2259615744
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031865596790371116,
      "loss": 1.4359,
      "theoretical_loss": 3.4245324558696986,
      "tokens_seen": 2259681280
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003186359077231695,
      "loss": 1.3259,
      "theoretical_loss": 3.424524439565238,
      "tokens_seen": 2259746816
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003186158475426279,
      "loss": 1.298,
      "theoretical_loss": 3.424516423558352,
      "tokens_seen": 2259812352
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031859578736208625,
      "loss": 1.4928,
      "theoretical_loss": 3.4245084078490216,
      "tokens_seen": 2259877888
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031857572718154467,
      "loss": 1.4696,
      "theoretical_loss": 3.4245003924372264,
      "tokens_seen": 2259943424
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031855566700100303,
      "loss": 1.2696,
      "theoretical_loss": 3.424492377322947,
      "tokens_seen": 2260008960
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003185356068204614,
      "loss": 1.325,
      "theoretical_loss": 3.4244843625061643,
      "tokens_seen": 2260074496
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031851554663991975,
      "loss": 1.353,
      "theoretical_loss": 3.4244763479868574,
      "tokens_seen": 2260140032
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031849548645937817,
      "loss": 1.3486,
      "theoretical_loss": 3.424468333765008,
      "tokens_seen": 2260205568
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031847542627883653,
      "loss": 1.2247,
      "theoretical_loss": 3.4244603198405956,
      "tokens_seen": 2260271104
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003184553660982949,
      "loss": 1.5567,
      "theoretical_loss": 3.4244523062136003,
      "tokens_seen": 2260336640
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.00031843530591775326,
      "loss": 1.3729,
      "theoretical_loss": 3.4244442928840035,
      "tokens_seen": 2260402176
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0003184152457372116,
      "loss": 1.4078,
      "theoretical_loss": 3.4244362798517844,
      "tokens_seen": 2260467712
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031839518555667004,
      "loss": 1.247,
      "theoretical_loss": 3.424428267116924,
      "tokens_seen": 2260533248
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003183751253761284,
      "loss": 1.3522,
      "theoretical_loss": 3.4244202546794025,
      "tokens_seen": 2260598784
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031835506519558676,
      "loss": 1.3515,
      "theoretical_loss": 3.4244122425392005,
      "tokens_seen": 2260664320
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003183350050150451,
      "loss": 1.446,
      "theoretical_loss": 3.4244042306962976,
      "tokens_seen": 2260729856
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031831494483450354,
      "loss": 1.4688,
      "theoretical_loss": 3.424396219150675,
      "tokens_seen": 2260795392
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003182948846539619,
      "loss": 1.4083,
      "theoretical_loss": 3.4243882079023127,
      "tokens_seen": 2260860928
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031827482447342026,
      "loss": 1.3466,
      "theoretical_loss": 3.4243801969511907,
      "tokens_seen": 2260926464
    },
    {
      "epoch": 0.69,
      "objective/train/advantage_avg": 0.4904368221759796,
      "objective/train/docs_used": 823330,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2928341627120972,
      "objective/train/original_loss": 1.2928340435028076,
      "objective/train/theoretical_loss": 3.42437218629729,
      "objective/train/tokens_used": 2281452000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2413264513015747,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050270438194275,
      "objective/train/weighted_lm_loss": 1.3576531410217285,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.42437218629729,
      "tokens_seen": 2260992000
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003182547642928786,
      "loss": 1.3811,
      "theoretical_loss": 3.42437218629729,
      "tokens_seen": 2260992000
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000318234704112337,
      "loss": 1.3875,
      "theoretical_loss": 3.4243641759405907,
      "tokens_seen": 2261057536
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003182146439317954,
      "loss": 1.4182,
      "theoretical_loss": 3.424356165881073,
      "tokens_seen": 2261123072
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031819458375125377,
      "loss": 1.4158,
      "theoretical_loss": 3.4243481561187172,
      "tokens_seen": 2261188608
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031817452357071213,
      "loss": 1.4548,
      "theoretical_loss": 3.4243401466535044,
      "tokens_seen": 2261254144
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003181544633901705,
      "loss": 1.285,
      "theoretical_loss": 3.424332137485414,
      "tokens_seen": 2261319680
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003181344032096289,
      "loss": 1.5473,
      "theoretical_loss": 3.4243241286144266,
      "tokens_seen": 2261385216
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031811434302908727,
      "loss": 1.449,
      "theoretical_loss": 3.4243161200405225,
      "tokens_seen": 2261450752
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031809428284854563,
      "loss": 1.5071,
      "theoretical_loss": 3.424308111763683,
      "tokens_seen": 2261516288
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000318074222668004,
      "loss": 1.3628,
      "theoretical_loss": 3.424300103783888,
      "tokens_seen": 2261581824
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003180541624874624,
      "loss": 1.405,
      "theoretical_loss": 3.4242920961011167,
      "tokens_seen": 2261647360
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031803410230692077,
      "loss": 1.3637,
      "theoretical_loss": 3.424284088715351,
      "tokens_seen": 2261712896
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031801404212637913,
      "loss": 1.3735,
      "theoretical_loss": 3.4242760816265707,
      "tokens_seen": 2261778432
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003179939819458375,
      "loss": 1.4712,
      "theoretical_loss": 3.4242680748347563,
      "tokens_seen": 2261843968
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031797392176529586,
      "loss": 1.4124,
      "theoretical_loss": 3.424260068339888,
      "tokens_seen": 2261909504
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003179538615847543,
      "loss": 1.4114,
      "theoretical_loss": 3.424252062141946,
      "tokens_seen": 2261975040
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031793380140421264,
      "loss": 1.4912,
      "theoretical_loss": 3.4242440562409113,
      "tokens_seen": 2262040576
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000317913741223671,
      "loss": 1.4182,
      "theoretical_loss": 3.4242360506367637,
      "tokens_seen": 2262106112
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031789368104312936,
      "loss": 1.3798,
      "theoretical_loss": 3.424228045329484,
      "tokens_seen": 2262171648
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003178736208625878,
      "loss": 1.5067,
      "theoretical_loss": 3.424220040319052,
      "tokens_seen": 2262237184
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031785356068204614,
      "loss": 1.3892,
      "theoretical_loss": 3.424212035605449,
      "tokens_seen": 2262302720
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003178335005015045,
      "loss": 1.4809,
      "theoretical_loss": 3.424204031188655,
      "tokens_seen": 2262368256
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031781344032096287,
      "loss": 1.4291,
      "theoretical_loss": 3.42419602706865,
      "tokens_seen": 2262433792
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031779338014042123,
      "loss": 1.3215,
      "theoretical_loss": 3.4241880232454145,
      "tokens_seen": 2262499328
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031777331995987964,
      "loss": 1.5235,
      "theoretical_loss": 3.4241800197189294,
      "tokens_seen": 2262564864
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000317753259779338,
      "loss": 1.3709,
      "theoretical_loss": 3.424172016489175,
      "tokens_seen": 2262630400
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031773319959879637,
      "loss": 1.4839,
      "theoretical_loss": 3.4241640135561306,
      "tokens_seen": 2262695936
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031771313941825473,
      "loss": 1.342,
      "theoretical_loss": 3.4241560109197784,
      "tokens_seen": 2262761472
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031769307923771315,
      "loss": 1.3614,
      "theoretical_loss": 3.4241480085800977,
      "tokens_seen": 2262827008
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003176730190571715,
      "loss": 1.4877,
      "theoretical_loss": 3.424140006537069,
      "tokens_seen": 2262892544
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031765295887662987,
      "loss": 1.3931,
      "theoretical_loss": 3.424132004790673,
      "tokens_seen": 2262958080
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031763289869608823,
      "loss": 1.4933,
      "theoretical_loss": 3.42412400334089,
      "tokens_seen": 2263023616
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003176128385155466,
      "loss": 1.2973,
      "theoretical_loss": 3.4241160021877004,
      "tokens_seen": 2263089152
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000317592778335005,
      "loss": 1.3875,
      "theoretical_loss": 3.4241080013310845,
      "tokens_seen": 2263154688
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003175727181544634,
      "loss": 1.4282,
      "theoretical_loss": 3.424100000771023,
      "tokens_seen": 2263220224
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003175526579739218,
      "loss": 1.4907,
      "theoretical_loss": 3.424092000507496,
      "tokens_seen": 2263285760
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031753259779338015,
      "loss": 1.4239,
      "theoretical_loss": 3.424084000540484,
      "tokens_seen": 2263351296
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031751253761283857,
      "loss": 1.4454,
      "theoretical_loss": 3.4240760008699676,
      "tokens_seen": 2263416832
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031749247743229693,
      "loss": 1.3319,
      "theoretical_loss": 3.424068001495927,
      "tokens_seen": 2263482368
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003174724172517553,
      "loss": 1.3285,
      "theoretical_loss": 3.4240600024183427,
      "tokens_seen": 2263547904
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031745235707121366,
      "loss": 1.3904,
      "theoretical_loss": 3.424052003637195,
      "tokens_seen": 2263613440
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000317432296890672,
      "loss": 1.4333,
      "theoretical_loss": 3.424044005152465,
      "tokens_seen": 2263678976
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031741223671013044,
      "loss": 1.4783,
      "theoretical_loss": 3.4240360069641325,
      "tokens_seen": 2263744512
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003173921765295888,
      "loss": 1.4025,
      "theoretical_loss": 3.4240280090721775,
      "tokens_seen": 2263810048
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031737211634904716,
      "loss": 1.3444,
      "theoretical_loss": 3.4240200114765815,
      "tokens_seen": 2263875584
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003173520561685055,
      "loss": 1.3971,
      "theoretical_loss": 3.4240120141773245,
      "tokens_seen": 2263941120
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031733199598796394,
      "loss": 1.3721,
      "theoretical_loss": 3.424004017174387,
      "tokens_seen": 2264006656
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003173119358074223,
      "loss": 1.2953,
      "theoretical_loss": 3.423996020467749,
      "tokens_seen": 2264072192
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031729187562688066,
      "loss": 1.3566,
      "theoretical_loss": 3.423988024057391,
      "tokens_seen": 2264137728
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000317271815446339,
      "loss": 1.3857,
      "theoretical_loss": 3.4239800279432946,
      "tokens_seen": 2264203264
    },
    {
      "epoch": 0.69,
      "objective/train/advantage_avg": 0.49498140811920166,
      "objective/train/docs_used": 824641,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.452023983001709,
      "objective/train/original_loss": 1.452023983001709,
      "objective/train/theoretical_loss": 3.4239720321254388,
      "objective/train/tokens_used": 2284728800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24527399241924286,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507452487945557,
      "objective/train/weighted_lm_loss": 1.525787115097046,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4239720321254388,
      "tokens_seen": 2264268800
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003172517552657974,
      "loss": 1.4124,
      "theoretical_loss": 3.4239720321254388,
      "tokens_seen": 2264268800
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003172316950852558,
      "loss": 1.3907,
      "theoretical_loss": 3.423964036603805,
      "tokens_seen": 2264334336
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031721163490471417,
      "loss": 1.5571,
      "theoretical_loss": 3.423956041378373,
      "tokens_seen": 2264399872
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031719157472417253,
      "loss": 1.3512,
      "theoretical_loss": 3.4239480464491234,
      "tokens_seen": 2264465408
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003171715145436309,
      "loss": 1.4581,
      "theoretical_loss": 3.423940051816037,
      "tokens_seen": 2264530944
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003171514543630893,
      "loss": 1.3173,
      "theoretical_loss": 3.423932057479094,
      "tokens_seen": 2264596480
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031713139418254767,
      "loss": 1.3388,
      "theoretical_loss": 3.423924063438275,
      "tokens_seen": 2264662016
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031711133400200603,
      "loss": 1.3001,
      "theoretical_loss": 3.4239160696935604,
      "tokens_seen": 2264727552
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003170912738214644,
      "loss": 1.3634,
      "theoretical_loss": 3.4239080762449303,
      "tokens_seen": 2264793088
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003170712136409228,
      "loss": 1.3638,
      "theoretical_loss": 3.4239000830923656,
      "tokens_seen": 2264858624
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003170511534603812,
      "loss": 1.3961,
      "theoretical_loss": 3.4238920902358467,
      "tokens_seen": 2264924160
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031703109327983954,
      "loss": 1.2733,
      "theoretical_loss": 3.423884097675354,
      "tokens_seen": 2264989696
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003170110330992979,
      "loss": 1.3104,
      "theoretical_loss": 3.4238761054108684,
      "tokens_seen": 2265055232
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031699097291875626,
      "loss": 1.3866,
      "theoretical_loss": 3.42386811344237,
      "tokens_seen": 2265120768
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003169709127382147,
      "loss": 1.5327,
      "theoretical_loss": 3.423860121769839,
      "tokens_seen": 2265186304
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031695085255767304,
      "loss": 1.2981,
      "theoretical_loss": 3.4238521303932563,
      "tokens_seen": 2265251840
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003169307923771314,
      "loss": 1.3501,
      "theoretical_loss": 3.423844139312602,
      "tokens_seen": 2265317376
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031691073219658976,
      "loss": 1.4248,
      "theoretical_loss": 3.423836148527857,
      "tokens_seen": 2265382912
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003168906720160482,
      "loss": 1.4141,
      "theoretical_loss": 3.423828158039001,
      "tokens_seen": 2265448448
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031687061183550654,
      "loss": 1.4613,
      "theoretical_loss": 3.423820167846016,
      "tokens_seen": 2265513984
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003168505516549649,
      "loss": 1.3077,
      "theoretical_loss": 3.423812177948881,
      "tokens_seen": 2265579520
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031683049147442327,
      "loss": 1.4324,
      "theoretical_loss": 3.4238041883475767,
      "tokens_seen": 2265645056
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031681043129388163,
      "loss": 1.5513,
      "theoretical_loss": 3.4237961990420844,
      "tokens_seen": 2265710592
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031679037111334004,
      "loss": 1.4393,
      "theoretical_loss": 3.423788210032384,
      "tokens_seen": 2265776128
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003167703109327984,
      "loss": 1.3369,
      "theoretical_loss": 3.4237802213184563,
      "tokens_seen": 2265841664
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031675025075225677,
      "loss": 1.4119,
      "theoretical_loss": 3.4237722329002813,
      "tokens_seen": 2265907200
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031673019057171513,
      "loss": 1.3615,
      "theoretical_loss": 3.42376424477784,
      "tokens_seen": 2265972736
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031671013039117355,
      "loss": 1.374,
      "theoretical_loss": 3.423756256951113,
      "tokens_seen": 2266038272
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003166900702106319,
      "loss": 1.4213,
      "theoretical_loss": 3.42374826942008,
      "tokens_seen": 2266103808
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031667001003009027,
      "loss": 1.3972,
      "theoretical_loss": 3.4237402821847223,
      "tokens_seen": 2266169344
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031664994984954863,
      "loss": 1.4088,
      "theoretical_loss": 3.4237322952450198,
      "tokens_seen": 2266234880
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000316629889669007,
      "loss": 1.3748,
      "theoretical_loss": 3.4237243086009537,
      "tokens_seen": 2266300416
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003166098294884654,
      "loss": 1.4945,
      "theoretical_loss": 3.423716322252504,
      "tokens_seen": 2266365952
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003165897693079238,
      "loss": 1.3403,
      "theoretical_loss": 3.423708336199651,
      "tokens_seen": 2266431488
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031656970912738214,
      "loss": 1.407,
      "theoretical_loss": 3.423700350442376,
      "tokens_seen": 2266497024
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003165496489468405,
      "loss": 1.3787,
      "theoretical_loss": 3.423692364980659,
      "tokens_seen": 2266562560
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003165295887662989,
      "loss": 1.3631,
      "theoretical_loss": 3.4236843798144805,
      "tokens_seen": 2266628096
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003165095285857573,
      "loss": 1.419,
      "theoretical_loss": 3.423676394943821,
      "tokens_seen": 2266693632
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031648946840521564,
      "loss": 1.3202,
      "theoretical_loss": 3.4236684103686614,
      "tokens_seen": 2266759168
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000316469408224674,
      "loss": 1.4784,
      "theoretical_loss": 3.423660426088982,
      "tokens_seen": 2266824704
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031644934804413237,
      "loss": 1.4319,
      "theoretical_loss": 3.4236524421047627,
      "tokens_seen": 2266890240
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003164292878635908,
      "loss": 1.3844,
      "theoretical_loss": 3.4236444584159846,
      "tokens_seen": 2266955776
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031640922768304914,
      "loss": 1.3554,
      "theoretical_loss": 3.423636475022629,
      "tokens_seen": 2267021312
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003163891675025075,
      "loss": 1.4379,
      "theoretical_loss": 3.423628491924675,
      "tokens_seen": 2267086848
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031636910732196587,
      "loss": 1.38,
      "theoretical_loss": 3.423620509122104,
      "tokens_seen": 2267152384
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003163490471414243,
      "loss": 1.3961,
      "theoretical_loss": 3.423612526614896,
      "tokens_seen": 2267217920
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031632898696088265,
      "loss": 1.3388,
      "theoretical_loss": 3.423604544403032,
      "tokens_seen": 2267283456
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000316308926780341,
      "loss": 1.4995,
      "theoretical_loss": 3.423596562486493,
      "tokens_seen": 2267348992
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031628886659979937,
      "loss": 1.347,
      "theoretical_loss": 3.423588580865258,
      "tokens_seen": 2267414528
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031626880641925773,
      "loss": 1.3374,
      "theoretical_loss": 3.423580599539309,
      "tokens_seen": 2267480064
    },
    {
      "epoch": 0.69,
      "objective/train/advantage_avg": 0.49348893761634827,
      "objective/train/docs_used": 825734,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2497258186340332,
      "objective/train/original_loss": 1.2497258186340332,
      "objective/train/theoretical_loss": 3.4235726185086257,
      "objective/train/tokens_used": 2288005600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24378521740436554,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505882501602173,
      "objective/train/weighted_lm_loss": 1.3128567934036255,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.4235726185086257,
      "tokens_seen": 2267545600
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031624874623871615,
      "loss": 1.308,
      "theoretical_loss": 3.4235726185086257,
      "tokens_seen": 2267545600
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003162286860581745,
      "loss": 1.3286,
      "theoretical_loss": 3.423564637773189,
      "tokens_seen": 2267611136
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003162086258776329,
      "loss": 1.4738,
      "theoretical_loss": 3.4235566573329796,
      "tokens_seen": 2267676672
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031618856569709124,
      "loss": 1.305,
      "theoretical_loss": 3.423548677187978,
      "tokens_seen": 2267742208
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031616850551654965,
      "loss": 1.3972,
      "theoretical_loss": 3.423540697338164,
      "tokens_seen": 2267807744
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000316148445336008,
      "loss": 1.3912,
      "theoretical_loss": 3.423532717783519,
      "tokens_seen": 2267873280
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003161283851554664,
      "loss": 1.335,
      "theoretical_loss": 3.4235247385240233,
      "tokens_seen": 2267938816
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031610832497492474,
      "loss": 1.4665,
      "theoretical_loss": 3.4235167595596576,
      "tokens_seen": 2268004352
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003160882647943831,
      "loss": 1.3739,
      "theoretical_loss": 3.423508780890402,
      "tokens_seen": 2268069888
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003160682046138415,
      "loss": 1.4331,
      "theoretical_loss": 3.4235008025162377,
      "tokens_seen": 2268135424
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031604814443329994,
      "loss": 1.4126,
      "theoretical_loss": 3.4234928244371448,
      "tokens_seen": 2268200960
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003160280842527583,
      "loss": 1.3658,
      "theoretical_loss": 3.423484846653104,
      "tokens_seen": 2268266496
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031600802407221666,
      "loss": 1.4069,
      "theoretical_loss": 3.4234768691640953,
      "tokens_seen": 2268332032
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003159879638916751,
      "loss": 1.2954,
      "theoretical_loss": 3.4234688919701,
      "tokens_seen": 2268397568
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031596790371113344,
      "loss": 1.4191,
      "theoretical_loss": 3.4234609150710993,
      "tokens_seen": 2268463104
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003159478435305918,
      "loss": 1.3781,
      "theoretical_loss": 3.423452938467072,
      "tokens_seen": 2268528640
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031592778335005016,
      "loss": 1.39,
      "theoretical_loss": 3.4234449621580003,
      "tokens_seen": 2268594176
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003159077231695086,
      "loss": 1.3844,
      "theoretical_loss": 3.4234369861438636,
      "tokens_seen": 2268659712
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031588766298896694,
      "loss": 1.3963,
      "theoretical_loss": 3.423429010424643,
      "tokens_seen": 2268725248
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003158676028084253,
      "loss": 1.4011,
      "theoretical_loss": 3.4234210350003194,
      "tokens_seen": 2268790784
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031584754262788367,
      "loss": 1.36,
      "theoretical_loss": 3.4234130598708727,
      "tokens_seen": 2268856320
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031582748244734203,
      "loss": 1.4806,
      "theoretical_loss": 3.4234050850362836,
      "tokens_seen": 2268921856
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031580742226680044,
      "loss": 1.3455,
      "theoretical_loss": 3.4233971104965333,
      "tokens_seen": 2268987392
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003157873620862588,
      "loss": 1.3163,
      "theoretical_loss": 3.423389136251602,
      "tokens_seen": 2269052928
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031576730190571717,
      "loss": 1.3961,
      "theoretical_loss": 3.42338116230147,
      "tokens_seen": 2269118464
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031574724172517553,
      "loss": 1.4683,
      "theoretical_loss": 3.423373188646118,
      "tokens_seen": 2269184000
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031572718154463395,
      "loss": 1.2995,
      "theoretical_loss": 3.423365215285527,
      "tokens_seen": 2269249536
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003157071213640923,
      "loss": 1.3875,
      "theoretical_loss": 3.4233572422196774,
      "tokens_seen": 2269315072
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031568706118355067,
      "loss": 1.3287,
      "theoretical_loss": 3.4233492694485497,
      "tokens_seen": 2269380608
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031566700100300903,
      "loss": 1.3559,
      "theoretical_loss": 3.4233412969721244,
      "tokens_seen": 2269446144
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003156469408224674,
      "loss": 1.3367,
      "theoretical_loss": 3.423333324790382,
      "tokens_seen": 2269511680
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003156268806419258,
      "loss": 1.3291,
      "theoretical_loss": 3.4233253529033036,
      "tokens_seen": 2269577216
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003156068204613842,
      "loss": 1.2884,
      "theoretical_loss": 3.4233173813108695,
      "tokens_seen": 2269642752
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031558676028084254,
      "loss": 1.3385,
      "theoretical_loss": 3.42330941001306,
      "tokens_seen": 2269708288
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003155667001003009,
      "loss": 1.3752,
      "theoretical_loss": 3.4233014390098564,
      "tokens_seen": 2269773824
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003155466399197593,
      "loss": 1.4042,
      "theoretical_loss": 3.423293468301239,
      "tokens_seen": 2269839360
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003155265797392177,
      "loss": 1.3727,
      "theoretical_loss": 3.423285497887188,
      "tokens_seen": 2269904896
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031550651955867604,
      "loss": 1.3389,
      "theoretical_loss": 3.423277527767684,
      "tokens_seen": 2269970432
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003154864593781344,
      "loss": 1.455,
      "theoretical_loss": 3.4232695579427084,
      "tokens_seen": 2270035968
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031546639919759277,
      "loss": 1.3855,
      "theoretical_loss": 3.4232615884122413,
      "tokens_seen": 2270101504
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003154463390170512,
      "loss": 1.4054,
      "theoretical_loss": 3.4232536191762635,
      "tokens_seen": 2270167040
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031542627883650954,
      "loss": 1.4497,
      "theoretical_loss": 3.4232456502347555,
      "tokens_seen": 2270232576
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003154062186559679,
      "loss": 1.3388,
      "theoretical_loss": 3.4232376815876977,
      "tokens_seen": 2270298112
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031538615847542627,
      "loss": 1.5028,
      "theoretical_loss": 3.4232297132350706,
      "tokens_seen": 2270363648
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003153660982948847,
      "loss": 1.3744,
      "theoretical_loss": 3.423221745176856,
      "tokens_seen": 2270429184
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031534603811434305,
      "loss": 1.4274,
      "theoretical_loss": 3.423213777413033,
      "tokens_seen": 2270494720
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003153259779338014,
      "loss": 1.2987,
      "theoretical_loss": 3.423205809943583,
      "tokens_seen": 2270560256
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031530591775325977,
      "loss": 1.3028,
      "theoretical_loss": 3.423197842768487,
      "tokens_seen": 2270625792
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031528585757271813,
      "loss": 1.3464,
      "theoretical_loss": 3.423189875887725,
      "tokens_seen": 2270691328
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031526579739217655,
      "loss": 1.5015,
      "theoretical_loss": 3.4231819093012774,
      "tokens_seen": 2270756864
    },
    {
      "epoch": 0.69,
      "objective/train/advantage_avg": 0.4949108362197876,
      "objective/train/docs_used": 826907,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2672680616378784,
      "objective/train/original_loss": 1.2672679424285889,
      "objective/train/theoretical_loss": 3.4231739430091253,
      "objective/train/tokens_used": 2291282400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24516355991363525,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507375001907349,
      "objective/train/weighted_lm_loss": 1.3315229415893555,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0100501775741577,
      "theoretical_loss": 3.4231739430091253,
      "tokens_seen": 2270822400
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003152457372116349,
      "loss": 1.3697,
      "theoretical_loss": 3.4231739430091253,
      "tokens_seen": 2270822400
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003152256770310933,
      "loss": 1.2895,
      "theoretical_loss": 3.4231659770112497,
      "tokens_seen": 2270887936
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031520561685055164,
      "loss": 1.2126,
      "theoretical_loss": 3.4231580113076303,
      "tokens_seen": 2270953472
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031518555667001005,
      "loss": 1.3516,
      "theoretical_loss": 3.4231500458982485,
      "tokens_seen": 2271019008
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003151654964894684,
      "loss": 1.3891,
      "theoretical_loss": 3.4231420807830846,
      "tokens_seen": 2271084544
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003151454363089268,
      "loss": 1.3843,
      "theoretical_loss": 3.423134115962119,
      "tokens_seen": 2271150080
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031512537612838514,
      "loss": 1.4382,
      "theoretical_loss": 3.4231261514353335,
      "tokens_seen": 2271215616
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003151053159478435,
      "loss": 1.3549,
      "theoretical_loss": 3.423118187202707,
      "tokens_seen": 2271281152
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003150852557673019,
      "loss": 1.3434,
      "theoretical_loss": 3.4231102232642217,
      "tokens_seen": 2271346688
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003150651955867603,
      "loss": 1.3999,
      "theoretical_loss": 3.4231022596198577,
      "tokens_seen": 2271412224
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031504513540621864,
      "loss": 1.2832,
      "theoretical_loss": 3.4230942962695954,
      "tokens_seen": 2271477760
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000315025075225677,
      "loss": 1.4428,
      "theoretical_loss": 3.4230863332134156,
      "tokens_seen": 2271543296
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003150050150451354,
      "loss": 1.4323,
      "theoretical_loss": 3.4230783704512984,
      "tokens_seen": 2271608832
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003149849548645938,
      "loss": 1.5197,
      "theoretical_loss": 3.423070407983226,
      "tokens_seen": 2271674368
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031496489468405215,
      "loss": 1.4533,
      "theoretical_loss": 3.4230624458091774,
      "tokens_seen": 2271739904
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003149448345035105,
      "loss": 1.309,
      "theoretical_loss": 3.423054483929134,
      "tokens_seen": 2271805440
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003149247743229689,
      "loss": 1.3918,
      "theoretical_loss": 3.4230465223430766,
      "tokens_seen": 2271870976
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003149047141424273,
      "loss": 1.3735,
      "theoretical_loss": 3.423038561050986,
      "tokens_seen": 2271936512
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031488465396188565,
      "loss": 1.5956,
      "theoretical_loss": 3.423030600052842,
      "tokens_seen": 2272002048
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000314864593781344,
      "loss": 1.3977,
      "theoretical_loss": 3.423022639348626,
      "tokens_seen": 2272067584
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003148445336008024,
      "loss": 1.3041,
      "theoretical_loss": 3.4230146789383187,
      "tokens_seen": 2272133120
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003148244734202608,
      "loss": 1.3568,
      "theoretical_loss": 3.4230067188219,
      "tokens_seen": 2272198656
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031480441323971915,
      "loss": 1.3131,
      "theoretical_loss": 3.422998758999352,
      "tokens_seen": 2272264192
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003147843530591775,
      "loss": 1.3582,
      "theoretical_loss": 3.4229907994706537,
      "tokens_seen": 2272329728
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003147642928786359,
      "loss": 1.2876,
      "theoretical_loss": 3.422982840235787,
      "tokens_seen": 2272395264
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003147442326980943,
      "loss": 1.3587,
      "theoretical_loss": 3.4229748812947323,
      "tokens_seen": 2272460800
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031472417251755266,
      "loss": 1.2214,
      "theoretical_loss": 3.4229669226474697,
      "tokens_seen": 2272526336
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000314704112337011,
      "loss": 1.2909,
      "theoretical_loss": 3.4229589642939806,
      "tokens_seen": 2272591872
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003146840521564694,
      "loss": 1.4359,
      "theoretical_loss": 3.4229510062342454,
      "tokens_seen": 2272657408
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031466399197592774,
      "loss": 1.3159,
      "theoretical_loss": 3.4229430484682446,
      "tokens_seen": 2272722944
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031464393179538616,
      "loss": 1.5039,
      "theoretical_loss": 3.422935090995959,
      "tokens_seen": 2272788480
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003146238716148445,
      "loss": 1.3704,
      "theoretical_loss": 3.42292713381737,
      "tokens_seen": 2272854016
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003146038114343029,
      "loss": 1.4754,
      "theoretical_loss": 3.4229191769324574,
      "tokens_seen": 2272919552
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031458375125376125,
      "loss": 1.3495,
      "theoretical_loss": 3.4229112203412018,
      "tokens_seen": 2272985088
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003145636910732197,
      "loss": 1.3889,
      "theoretical_loss": 3.4229032640435846,
      "tokens_seen": 2273050624
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003145436308926781,
      "loss": 1.3037,
      "theoretical_loss": 3.4228953080395863,
      "tokens_seen": 2273116160
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031452357071213644,
      "loss": 1.3892,
      "theoretical_loss": 3.422887352329187,
      "tokens_seen": 2273181696
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003145035105315948,
      "loss": 1.3361,
      "theoretical_loss": 3.422879396912368,
      "tokens_seen": 2273247232
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031448345035105317,
      "loss": 1.4123,
      "theoretical_loss": 3.42287144178911,
      "tokens_seen": 2273312768
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003144633901705116,
      "loss": 1.3988,
      "theoretical_loss": 3.4228634869593937,
      "tokens_seen": 2273378304
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031444332998996994,
      "loss": 1.6117,
      "theoretical_loss": 3.4228555324231995,
      "tokens_seen": 2273443840
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003144232698094283,
      "loss": 1.4144,
      "theoretical_loss": 3.4228475781805083,
      "tokens_seen": 2273509376
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031440320962888667,
      "loss": 1.5109,
      "theoretical_loss": 3.4228396242313006,
      "tokens_seen": 2273574912
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003143831494483451,
      "loss": 1.2677,
      "theoretical_loss": 3.4228316705755577,
      "tokens_seen": 2273640448
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031436308926780345,
      "loss": 1.3232,
      "theoretical_loss": 3.4228237172132596,
      "tokens_seen": 2273705984
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003143430290872618,
      "loss": 1.3427,
      "theoretical_loss": 3.4228157641443873,
      "tokens_seen": 2273771520
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031432296890672017,
      "loss": 1.412,
      "theoretical_loss": 3.422807811368922,
      "tokens_seen": 2273837056
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031430290872617853,
      "loss": 1.3272,
      "theoretical_loss": 3.4227998588868433,
      "tokens_seen": 2273902592
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031428284854563695,
      "loss": 1.3319,
      "theoretical_loss": 3.422791906698133,
      "tokens_seen": 2273968128
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003142627883650953,
      "loss": 1.4242,
      "theoretical_loss": 3.4227839548027714,
      "tokens_seen": 2274033664
    },
    {
      "epoch": 0.69,
      "objective/train/advantage_avg": 0.49369320273399353,
      "objective/train/docs_used": 828122,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2761720418930054,
      "objective/train/original_loss": 1.276172161102295,
      "objective/train/theoretical_loss": 3.422776003200739,
      "objective/train/tokens_used": 2294559200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24422746896743774,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506110191345215,
      "objective/train/weighted_lm_loss": 1.3406851291656494,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.422776003200739,
      "tokens_seen": 2274099200
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003142427281845537,
      "loss": 1.3783,
      "theoretical_loss": 3.422776003200739,
      "tokens_seen": 2274099200
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031422266800401204,
      "loss": 1.3334,
      "theoretical_loss": 3.4227680518920174,
      "tokens_seen": 2274164736
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031420260782347045,
      "loss": 1.4531,
      "theoretical_loss": 3.422760100876586,
      "tokens_seen": 2274230272
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003141825476429288,
      "loss": 1.5694,
      "theoretical_loss": 3.4227521501544267,
      "tokens_seen": 2274295808
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003141624874623872,
      "loss": 1.5136,
      "theoretical_loss": 3.4227441997255195,
      "tokens_seen": 2274361344
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031414242728184554,
      "loss": 1.4776,
      "theoretical_loss": 3.4227362495898457,
      "tokens_seen": 2274426880
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003141223671013039,
      "loss": 1.4215,
      "theoretical_loss": 3.4227282997473853,
      "tokens_seen": 2274492416
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003141023069207623,
      "loss": 1.4177,
      "theoretical_loss": 3.42272035019812,
      "tokens_seen": 2274557952
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003140822467402207,
      "loss": 1.4472,
      "theoretical_loss": 3.4227124009420296,
      "tokens_seen": 2274623488
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031406218655967904,
      "loss": 1.3816,
      "theoretical_loss": 3.4227044519790955,
      "tokens_seen": 2274689024
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003140421263791374,
      "loss": 1.3921,
      "theoretical_loss": 3.4226965033092975,
      "tokens_seen": 2274754560
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003140220661985958,
      "loss": 1.2764,
      "theoretical_loss": 3.4226885549326176,
      "tokens_seen": 2274820096
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003140020060180542,
      "loss": 1.3889,
      "theoretical_loss": 3.422680606849036,
      "tokens_seen": 2274885632
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031398194583751255,
      "loss": 1.2462,
      "theoretical_loss": 3.4226726590585335,
      "tokens_seen": 2274951168
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003139618856569709,
      "loss": 1.4759,
      "theoretical_loss": 3.4226647115610906,
      "tokens_seen": 2275016704
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003139418254764293,
      "loss": 1.4125,
      "theoretical_loss": 3.4226567643566885,
      "tokens_seen": 2275082240
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003139217652958877,
      "loss": 1.3301,
      "theoretical_loss": 3.4226488174453076,
      "tokens_seen": 2275147776
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031390170511534605,
      "loss": 1.523,
      "theoretical_loss": 3.4226408708269282,
      "tokens_seen": 2275213312
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003138816449348044,
      "loss": 1.3593,
      "theoretical_loss": 3.4226329245015323,
      "tokens_seen": 2275278848
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003138615847542628,
      "loss": 1.3306,
      "theoretical_loss": 3.4226249784691,
      "tokens_seen": 2275344384
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003138415245737212,
      "loss": 1.4247,
      "theoretical_loss": 3.4226170327296113,
      "tokens_seen": 2275409920
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031382146439317955,
      "loss": 1.3453,
      "theoretical_loss": 3.4226090872830484,
      "tokens_seen": 2275475456
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003138014042126379,
      "loss": 1.5756,
      "theoretical_loss": 3.422601142129391,
      "tokens_seen": 2275540992
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003137813440320963,
      "loss": 1.4467,
      "theoretical_loss": 3.42259319726862,
      "tokens_seen": 2275606528
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003137612838515547,
      "loss": 1.4237,
      "theoretical_loss": 3.422585252700717,
      "tokens_seen": 2275672064
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031374122367101306,
      "loss": 1.3646,
      "theoretical_loss": 3.422577308425662,
      "tokens_seen": 2275737600
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003137211634904714,
      "loss": 1.4563,
      "theoretical_loss": 3.4225693644434356,
      "tokens_seen": 2275803136
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003137011033099298,
      "loss": 1.4122,
      "theoretical_loss": 3.422561420754019,
      "tokens_seen": 2275868672
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031368104312938814,
      "loss": 1.3713,
      "theoretical_loss": 3.422553477357393,
      "tokens_seen": 2275934208
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031366098294884656,
      "loss": 1.4678,
      "theoretical_loss": 3.4225455342535382,
      "tokens_seen": 2275999744
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003136409227683049,
      "loss": 1.3359,
      "theoretical_loss": 3.4225375914424356,
      "tokens_seen": 2276065280
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003136208625877633,
      "loss": 1.3175,
      "theoretical_loss": 3.4225296489240655,
      "tokens_seen": 2276130816
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031360080240722165,
      "loss": 1.4217,
      "theoretical_loss": 3.4225217066984093,
      "tokens_seen": 2276196352
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031358074222668006,
      "loss": 1.3874,
      "theoretical_loss": 3.4225137647654473,
      "tokens_seen": 2276261888
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003135606820461384,
      "loss": 1.2344,
      "theoretical_loss": 3.4225058231251606,
      "tokens_seen": 2276327424
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003135406218655968,
      "loss": 1.3991,
      "theoretical_loss": 3.42249788177753,
      "tokens_seen": 2276392960
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031352056168505515,
      "loss": 1.3287,
      "theoretical_loss": 3.422489940722536,
      "tokens_seen": 2276458496
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003135005015045135,
      "loss": 1.2637,
      "theoretical_loss": 3.4224819999601594,
      "tokens_seen": 2276524032
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031348044132397193,
      "loss": 1.3778,
      "theoretical_loss": 3.422474059490381,
      "tokens_seen": 2276589568
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003134603811434303,
      "loss": 1.4807,
      "theoretical_loss": 3.422466119313182,
      "tokens_seen": 2276655104
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031344032096288865,
      "loss": 1.3757,
      "theoretical_loss": 3.422458179428543,
      "tokens_seen": 2276720640
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000313420260782347,
      "loss": 1.3948,
      "theoretical_loss": 3.4224502398364445,
      "tokens_seen": 2276786176
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031340020060180543,
      "loss": 1.3092,
      "theoretical_loss": 3.422442300536868,
      "tokens_seen": 2276851712
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003133801404212638,
      "loss": 1.3842,
      "theoretical_loss": 3.4224343615297936,
      "tokens_seen": 2276917248
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031336008024072216,
      "loss": 1.308,
      "theoretical_loss": 3.422426422815202,
      "tokens_seen": 2276982784
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003133400200601805,
      "loss": 1.3117,
      "theoretical_loss": 3.422418484393075,
      "tokens_seen": 2277048320
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003133199598796389,
      "loss": 1.4498,
      "theoretical_loss": 3.422410546263392,
      "tokens_seen": 2277113856
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003132998996990973,
      "loss": 1.416,
      "theoretical_loss": 3.422402608426135,
      "tokens_seen": 2277179392
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031327983951855566,
      "loss": 1.379,
      "theoretical_loss": 3.4223946708812845,
      "tokens_seen": 2277244928
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000313259779338014,
      "loss": 1.4014,
      "theoretical_loss": 3.422386733628821,
      "tokens_seen": 2277310464
    },
    {
      "epoch": 0.69,
      "objective/train/advantage_avg": 0.49078336358070374,
      "objective/train/docs_used": 829293,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4776960611343384,
      "objective/train/original_loss": 1.477696180343628,
      "objective/train/theoretical_loss": 3.422378796668726,
      "objective/train/tokens_used": 2297836000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24182593822479248,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503075122833252,
      "objective/train/weighted_lm_loss": 1.55207097530365,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.422378796668726,
      "tokens_seen": 2277376000
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003132397191574724,
      "loss": 1.4715,
      "theoretical_loss": 3.422378796668726,
      "tokens_seen": 2277376000
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003132196589769308,
      "loss": 1.3609,
      "theoretical_loss": 3.422370860000979,
      "tokens_seen": 2277441536
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031319959879638916,
      "loss": 1.4044,
      "theoretical_loss": 3.422362923625562,
      "tokens_seen": 2277507072
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003131795386158475,
      "loss": 1.4344,
      "theoretical_loss": 3.4223549875424557,
      "tokens_seen": 2277572608
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003131594784353059,
      "loss": 1.3368,
      "theoretical_loss": 3.4223470517516406,
      "tokens_seen": 2277638144
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031313941825476425,
      "loss": 1.3628,
      "theoretical_loss": 3.422339116253097,
      "tokens_seen": 2277703680
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031311935807422267,
      "loss": 1.4297,
      "theoretical_loss": 3.422331181046807,
      "tokens_seen": 2277769216
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031309929789368103,
      "loss": 1.3142,
      "theoretical_loss": 3.4223232461327506,
      "tokens_seen": 2277834752
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003130792377131394,
      "loss": 1.3295,
      "theoretical_loss": 3.422315311510909,
      "tokens_seen": 2277900288
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003130591775325978,
      "loss": 1.3847,
      "theoretical_loss": 3.4223073771812627,
      "tokens_seen": 2277965824
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003130391173520562,
      "loss": 1.4862,
      "theoretical_loss": 3.4222994431437925,
      "tokens_seen": 2278031360
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003130190571715146,
      "loss": 1.3895,
      "theoretical_loss": 3.4222915093984794,
      "tokens_seen": 2278096896
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031299899699097295,
      "loss": 1.37,
      "theoretical_loss": 3.4222835759453045,
      "tokens_seen": 2278162432
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003129789368104313,
      "loss": 1.3475,
      "theoretical_loss": 3.4222756427842485,
      "tokens_seen": 2278227968
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003129588766298897,
      "loss": 1.3473,
      "theoretical_loss": 3.4222677099152916,
      "tokens_seen": 2278293504
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003129388164493481,
      "loss": 1.3571,
      "theoretical_loss": 3.4222597773384154,
      "tokens_seen": 2278359040
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031291875626880645,
      "loss": 1.3769,
      "theoretical_loss": 3.422251845053601,
      "tokens_seen": 2278424576
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003128986960882648,
      "loss": 1.3368,
      "theoretical_loss": 3.422243913060828,
      "tokens_seen": 2278490112
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003128786359077232,
      "loss": 1.4543,
      "theoretical_loss": 3.4222359813600782,
      "tokens_seen": 2278555648
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003128585757271816,
      "loss": 1.4204,
      "theoretical_loss": 3.422228049951333,
      "tokens_seen": 2278621184
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031283851554663995,
      "loss": 1.3333,
      "theoretical_loss": 3.422220118834572,
      "tokens_seen": 2278686720
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003128184553660983,
      "loss": 1.3256,
      "theoretical_loss": 3.422212188009776,
      "tokens_seen": 2278752256
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003127983951855567,
      "loss": 1.2848,
      "theoretical_loss": 3.422204257476927,
      "tokens_seen": 2278817792
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003127783350050151,
      "loss": 1.2925,
      "theoretical_loss": 3.422196327236005,
      "tokens_seen": 2278883328
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031275827482447346,
      "loss": 1.3873,
      "theoretical_loss": 3.422188397286991,
      "tokens_seen": 2278948864
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003127382146439318,
      "loss": 1.3346,
      "theoretical_loss": 3.422180467629867,
      "tokens_seen": 2279014400
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003127181544633902,
      "loss": 1.339,
      "theoretical_loss": 3.4221725382646118,
      "tokens_seen": 2279079936
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031269809428284854,
      "loss": 1.3282,
      "theoretical_loss": 3.4221646091912077,
      "tokens_seen": 2279145472
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031267803410230696,
      "loss": 1.3857,
      "theoretical_loss": 3.4221566804096355,
      "tokens_seen": 2279211008
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003126579739217653,
      "loss": 1.4479,
      "theoretical_loss": 3.4221487519198748,
      "tokens_seen": 2279276544
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003126379137412237,
      "loss": 1.3426,
      "theoretical_loss": 3.422140823721908,
      "tokens_seen": 2279342080
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031261785356068205,
      "loss": 1.3655,
      "theoretical_loss": 3.422132895815716,
      "tokens_seen": 2279407616
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031259779338014046,
      "loss": 1.4434,
      "theoretical_loss": 3.4221249682012784,
      "tokens_seen": 2279473152
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003125777331995988,
      "loss": 1.4203,
      "theoretical_loss": 3.422117040878577,
      "tokens_seen": 2279538688
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003125576730190572,
      "loss": 1.4167,
      "theoretical_loss": 3.422109113847592,
      "tokens_seen": 2279604224
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031253761283851555,
      "loss": 1.3287,
      "theoretical_loss": 3.422101187108305,
      "tokens_seen": 2279669760
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003125175526579739,
      "loss": 1.241,
      "theoretical_loss": 3.422093260660697,
      "tokens_seen": 2279735296
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031249749247743233,
      "loss": 1.3214,
      "theoretical_loss": 3.4220853345047475,
      "tokens_seen": 2279800832
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003124774322968907,
      "loss": 1.4579,
      "theoretical_loss": 3.422077408640439,
      "tokens_seen": 2279866368
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031245737211634905,
      "loss": 1.426,
      "theoretical_loss": 3.4220694830677516,
      "tokens_seen": 2279931904
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003124373119358074,
      "loss": 1.4155,
      "theoretical_loss": 3.4220615577866664,
      "tokens_seen": 2279997440
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031241725175526583,
      "loss": 1.4484,
      "theoretical_loss": 3.422053632797164,
      "tokens_seen": 2280062976
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003123971915747242,
      "loss": 1.2741,
      "theoretical_loss": 3.422045708099226,
      "tokens_seen": 2280128512
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031237713139418256,
      "loss": 1.5207,
      "theoretical_loss": 3.4220377836928324,
      "tokens_seen": 2280194048
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003123570712136409,
      "loss": 1.3931,
      "theoretical_loss": 3.4220298595779646,
      "tokens_seen": 2280259584
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003123370110330993,
      "loss": 1.4053,
      "theoretical_loss": 3.4220219357546036,
      "tokens_seen": 2280325120
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003123169508525577,
      "loss": 1.3851,
      "theoretical_loss": 3.42201401222273,
      "tokens_seen": 2280390656
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031229689067201606,
      "loss": 1.3544,
      "theoretical_loss": 3.4220060889823243,
      "tokens_seen": 2280456192
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003122768304914744,
      "loss": 1.4619,
      "theoretical_loss": 3.4219981660333687,
      "tokens_seen": 2280521728
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003122567703109328,
      "loss": 1.3967,
      "theoretical_loss": 3.4219902433758427,
      "tokens_seen": 2280587264
    },
    {
      "epoch": 0.69,
      "objective/train/advantage_avg": 0.4929426610469818,
      "objective/train/docs_used": 830439,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.244478702545166,
      "objective/train/original_loss": 1.244478702545166,
      "objective/train/theoretical_loss": 3.421982321009728,
      "objective/train/tokens_used": 2301112800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24374686181545258,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505335330963135,
      "objective/train/weighted_lm_loss": 1.3073673248291016,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.421982321009728,
      "tokens_seen": 2280652800
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003122367101303912,
      "loss": 1.3022,
      "theoretical_loss": 3.421982321009728,
      "tokens_seen": 2280652800
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031221664994984956,
      "loss": 1.3219,
      "theoretical_loss": 3.421974398935005,
      "tokens_seen": 2280718336
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003121965897693079,
      "loss": 1.3828,
      "theoretical_loss": 3.4219664771516554,
      "tokens_seen": 2280783872
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003121765295887663,
      "loss": 1.52,
      "theoretical_loss": 3.4219585556596597,
      "tokens_seen": 2280849408
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031215646940822465,
      "loss": 1.3886,
      "theoretical_loss": 3.4219506344589985,
      "tokens_seen": 2280914944
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031213640922768307,
      "loss": 1.3285,
      "theoretical_loss": 3.4219427135496527,
      "tokens_seen": 2280980480
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031211634904714143,
      "loss": 1.3975,
      "theoretical_loss": 3.4219347929316037,
      "tokens_seen": 2281046016
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003120962888665998,
      "loss": 1.3789,
      "theoretical_loss": 3.4219268726048324,
      "tokens_seen": 2281111552
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031207622868605815,
      "loss": 1.3675,
      "theoretical_loss": 3.421918952569319,
      "tokens_seen": 2281177088
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031205616850551657,
      "loss": 1.1691,
      "theoretical_loss": 3.421911032825046,
      "tokens_seen": 2281242624
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031203610832497493,
      "loss": 1.5312,
      "theoretical_loss": 3.421903113371992,
      "tokens_seen": 2281308160
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003120160481444333,
      "loss": 1.3427,
      "theoretical_loss": 3.42189519421014,
      "tokens_seen": 2281373696
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031199598796389166,
      "loss": 1.296,
      "theoretical_loss": 3.4218872753394702,
      "tokens_seen": 2281439232
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031197592778335,
      "loss": 1.4101,
      "theoretical_loss": 3.421879356759963,
      "tokens_seen": 2281504768
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031195586760280843,
      "loss": 1.3136,
      "theoretical_loss": 3.4218714384716,
      "tokens_seen": 2281570304
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003119358074222668,
      "loss": 1.4504,
      "theoretical_loss": 3.421863520474362,
      "tokens_seen": 2281635840
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031191574724172516,
      "loss": 1.3344,
      "theoretical_loss": 3.4218556027682294,
      "tokens_seen": 2281701376
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003118956870611835,
      "loss": 1.3885,
      "theoretical_loss": 3.421847685353184,
      "tokens_seen": 2281766912
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031187562688064194,
      "loss": 1.3332,
      "theoretical_loss": 3.4218397682292063,
      "tokens_seen": 2281832448
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003118555667001003,
      "loss": 1.3939,
      "theoretical_loss": 3.421831851396277,
      "tokens_seen": 2281897984
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031183550651955866,
      "loss": 1.2264,
      "theoretical_loss": 3.4218239348543777,
      "tokens_seen": 2281963520
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000311815446339017,
      "loss": 1.5172,
      "theoretical_loss": 3.4218160186034887,
      "tokens_seen": 2282029056
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003117953861584754,
      "loss": 1.3477,
      "theoretical_loss": 3.421808102643591,
      "tokens_seen": 2282094592
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003117753259779338,
      "loss": 1.4253,
      "theoretical_loss": 3.421800186974666,
      "tokens_seen": 2282160128
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031175526579739217,
      "loss": 1.3326,
      "theoretical_loss": 3.4217922715966944,
      "tokens_seen": 2282225664
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031173520561685053,
      "loss": 1.4003,
      "theoretical_loss": 3.421784356509657,
      "tokens_seen": 2282291200
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003117151454363089,
      "loss": 1.3737,
      "theoretical_loss": 3.4217764417135355,
      "tokens_seen": 2282356736
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003116950852557673,
      "loss": 1.321,
      "theoretical_loss": 3.4217685272083096,
      "tokens_seen": 2282422272
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031167502507522567,
      "loss": 1.3898,
      "theoretical_loss": 3.421760612993961,
      "tokens_seen": 2282487808
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031165496489468403,
      "loss": 1.4494,
      "theoretical_loss": 3.4217526990704705,
      "tokens_seen": 2282553344
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003116349047141424,
      "loss": 1.4597,
      "theoretical_loss": 3.4217447854378196,
      "tokens_seen": 2282618880
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003116148445336008,
      "loss": 1.2572,
      "theoretical_loss": 3.4217368720959884,
      "tokens_seen": 2282684416
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031159478435305917,
      "loss": 1.3474,
      "theoretical_loss": 3.421728959044958,
      "tokens_seen": 2282749952
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031157472417251753,
      "loss": 1.488,
      "theoretical_loss": 3.42172104628471,
      "tokens_seen": 2282815488
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031155466399197595,
      "loss": 1.2965,
      "theoretical_loss": 3.421713133815225,
      "tokens_seen": 2282881024
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003115346038114343,
      "loss": 1.3817,
      "theoretical_loss": 3.421705221636484,
      "tokens_seen": 2282946560
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031151454363089273,
      "loss": 1.3064,
      "theoretical_loss": 3.4216973097484678,
      "tokens_seen": 2283012096
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003114944834503511,
      "loss": 1.3708,
      "theoretical_loss": 3.421689398151157,
      "tokens_seen": 2283077632
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031147442326980945,
      "loss": 1.3282,
      "theoretical_loss": 3.421681486844534,
      "tokens_seen": 2283143168
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003114543630892678,
      "loss": 1.3591,
      "theoretical_loss": 3.4216735758285783,
      "tokens_seen": 2283208704
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031143430290872623,
      "loss": 1.468,
      "theoretical_loss": 3.4216656651032715,
      "tokens_seen": 2283274240
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003114142427281846,
      "loss": 1.4033,
      "theoretical_loss": 3.4216577546685945,
      "tokens_seen": 2283339776
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031139418254764296,
      "loss": 1.2622,
      "theoretical_loss": 3.4216498445245285,
      "tokens_seen": 2283405312
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003113741223671013,
      "loss": 1.3025,
      "theoretical_loss": 3.421641934671054,
      "tokens_seen": 2283470848
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003113540621865597,
      "loss": 1.4621,
      "theoretical_loss": 3.421634025108152,
      "tokens_seen": 2283536384
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003113340020060181,
      "loss": 1.3079,
      "theoretical_loss": 3.4216261158358043,
      "tokens_seen": 2283601920
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031131394182547646,
      "loss": 1.3864,
      "theoretical_loss": 3.4216182068539913,
      "tokens_seen": 2283667456
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003112938816449348,
      "loss": 1.2977,
      "theoretical_loss": 3.421610298162694,
      "tokens_seen": 2283732992
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003112738214643932,
      "loss": 1.4677,
      "theoretical_loss": 3.421602389761893,
      "tokens_seen": 2283798528
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003112537612838516,
      "loss": 1.4025,
      "theoretical_loss": 3.42159448165157,
      "tokens_seen": 2283864064
    },
    {
      "epoch": 0.69,
      "objective/train/advantage_avg": 0.49503567814826965,
      "objective/train/docs_used": 831582,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3050607442855835,
      "objective/train/original_loss": 1.305060625076294,
      "objective/train/theoretical_loss": 3.4215865738317057,
      "objective/train/tokens_used": 2304389600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24526900053024292,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507506132125854,
      "objective/train/weighted_lm_loss": 1.3712611198425293,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4215865738317057,
      "tokens_seen": 2283929600
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031123370110330996,
      "loss": 1.3342,
      "theoretical_loss": 3.4215865738317057,
      "tokens_seen": 2283929600
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003112136409227683,
      "loss": 1.4263,
      "theoretical_loss": 3.421578666302281,
      "tokens_seen": 2283995136
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003111935807422267,
      "loss": 1.3365,
      "theoretical_loss": 3.421570759063277,
      "tokens_seen": 2284060672
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031117352056168505,
      "loss": 1.47,
      "theoretical_loss": 3.4215628521146746,
      "tokens_seen": 2284126208
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031115346038114347,
      "loss": 1.3738,
      "theoretical_loss": 3.4215549454564553,
      "tokens_seen": 2284191744
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031113340020060183,
      "loss": 1.3537,
      "theoretical_loss": 3.4215470390885994,
      "tokens_seen": 2284257280
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003111133400200602,
      "loss": 1.3286,
      "theoretical_loss": 3.4215391330110885,
      "tokens_seen": 2284322816
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031109327983951855,
      "loss": 1.4532,
      "theoretical_loss": 3.4215312272239027,
      "tokens_seen": 2284388352
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031107321965897697,
      "loss": 1.3058,
      "theoretical_loss": 3.4215233217270242,
      "tokens_seen": 2284453888
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031105315947843533,
      "loss": 1.4543,
      "theoretical_loss": 3.4215154165204336,
      "tokens_seen": 2284519424
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003110330992978937,
      "loss": 1.4009,
      "theoretical_loss": 3.421507511604111,
      "tokens_seen": 2284584960
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031101303911735206,
      "loss": 1.3538,
      "theoretical_loss": 3.421499606978039,
      "tokens_seen": 2284650496
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003109929789368104,
      "loss": 1.3537,
      "theoretical_loss": 3.4214917026421974,
      "tokens_seen": 2284716032
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031097291875626883,
      "loss": 1.3002,
      "theoretical_loss": 3.4214837985965674,
      "tokens_seen": 2284781568
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003109528585757272,
      "loss": 1.3716,
      "theoretical_loss": 3.42147589484113,
      "tokens_seen": 2284847104
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031093279839518556,
      "loss": 1.5796,
      "theoretical_loss": 3.4214679913758674,
      "tokens_seen": 2284912640
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003109127382146439,
      "loss": 1.318,
      "theoretical_loss": 3.4214600882007593,
      "tokens_seen": 2284978176
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031089267803410234,
      "loss": 1.3933,
      "theoretical_loss": 3.4214521853157867,
      "tokens_seen": 2285043712
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003108726178535607,
      "loss": 1.3674,
      "theoretical_loss": 3.4214442827209313,
      "tokens_seen": 2285109248
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031085255767301906,
      "loss": 1.3915,
      "theoretical_loss": 3.421436380416174,
      "tokens_seen": 2285174784
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003108324974924774,
      "loss": 1.288,
      "theoretical_loss": 3.4214284784014954,
      "tokens_seen": 2285240320
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003108124373119358,
      "loss": 1.2827,
      "theoretical_loss": 3.4214205766768773,
      "tokens_seen": 2285305856
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003107923771313942,
      "loss": 1.2857,
      "theoretical_loss": 3.4214126752423,
      "tokens_seen": 2285371392
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031077231695085257,
      "loss": 1.3166,
      "theoretical_loss": 3.4214047740977445,
      "tokens_seen": 2285436928
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031075225677031093,
      "loss": 1.3248,
      "theoretical_loss": 3.4213968732431925,
      "tokens_seen": 2285502464
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003107321965897693,
      "loss": 1.3975,
      "theoretical_loss": 3.4213889726786246,
      "tokens_seen": 2285568000
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003107121364092277,
      "loss": 1.3663,
      "theoretical_loss": 3.4213810724040217,
      "tokens_seen": 2285633536
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031069207622868607,
      "loss": 1.2463,
      "theoretical_loss": 3.4213731724193654,
      "tokens_seen": 2285699072
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031067201604814443,
      "loss": 1.3579,
      "theoretical_loss": 3.4213652727246364,
      "tokens_seen": 2285764608
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003106519558676028,
      "loss": 1.4049,
      "theoretical_loss": 3.4213573733198155,
      "tokens_seen": 2285830144
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003106318956870612,
      "loss": 1.3787,
      "theoretical_loss": 3.421349474204884,
      "tokens_seen": 2285895680
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031061183550651957,
      "loss": 1.446,
      "theoretical_loss": 3.4213415753798233,
      "tokens_seen": 2285961216
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031059177532597793,
      "loss": 1.3821,
      "theoretical_loss": 3.421333676844614,
      "tokens_seen": 2286026752
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003105717151454363,
      "loss": 1.3038,
      "theoretical_loss": 3.421325778599237,
      "tokens_seen": 2286092288
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031055165496489466,
      "loss": 1.3529,
      "theoretical_loss": 3.4213178806436737,
      "tokens_seen": 2286157824
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003105315947843531,
      "loss": 1.3267,
      "theoretical_loss": 3.4213099829779052,
      "tokens_seen": 2286223360
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031051153460381144,
      "loss": 1.3236,
      "theoretical_loss": 3.4213020856019125,
      "tokens_seen": 2286288896
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003104914744232698,
      "loss": 1.3167,
      "theoretical_loss": 3.4212941885156765,
      "tokens_seen": 2286354432
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031047141424272816,
      "loss": 1.3004,
      "theoretical_loss": 3.4212862917191784,
      "tokens_seen": 2286419968
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003104513540621866,
      "loss": 1.3595,
      "theoretical_loss": 3.4212783952123993,
      "tokens_seen": 2286485504
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031043129388164494,
      "loss": 1.4255,
      "theoretical_loss": 3.4212704989953204,
      "tokens_seen": 2286551040
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003104112337011033,
      "loss": 1.3155,
      "theoretical_loss": 3.421262603067922,
      "tokens_seen": 2286616576
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031039117352056167,
      "loss": 1.3587,
      "theoretical_loss": 3.4212547074301862,
      "tokens_seen": 2286682112
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031037111334002003,
      "loss": 1.2312,
      "theoretical_loss": 3.421246812082093,
      "tokens_seen": 2286747648
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031035105315947844,
      "loss": 1.3473,
      "theoretical_loss": 3.421238917023625,
      "tokens_seen": 2286813184
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003103309929789368,
      "loss": 1.3868,
      "theoretical_loss": 3.4212310222547617,
      "tokens_seen": 2286878720
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031031093279839517,
      "loss": 1.3714,
      "theoretical_loss": 3.4212231277754848,
      "tokens_seen": 2286944256
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031029087261785353,
      "loss": 1.3336,
      "theoretical_loss": 3.4212152335857757,
      "tokens_seen": 2287009792
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031027081243731195,
      "loss": 1.4964,
      "theoretical_loss": 3.4212073396856155,
      "tokens_seen": 2287075328
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003102507522567703,
      "loss": 1.2711,
      "theoretical_loss": 3.4211994460749846,
      "tokens_seen": 2287140864
    },
    {
      "epoch": 0.69,
      "objective/train/advantage_avg": 0.4920370578765869,
      "objective/train/docs_used": 832763,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3145228624343872,
      "objective/train/original_loss": 1.3145227432250977,
      "objective/train/theoretical_loss": 3.4211915527538643,
      "objective/train/tokens_used": 2307666400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24250473082065582,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504364967346191,
      "objective/train/weighted_lm_loss": 1.3808619976043701,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.4211915527538643,
      "tokens_seen": 2287206400
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031023069207622867,
      "loss": 1.2985,
      "theoretical_loss": 3.4211915527538643,
      "tokens_seen": 2287206400
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031021063189568703,
      "loss": 1.4687,
      "theoretical_loss": 3.421183659722236,
      "tokens_seen": 2287271936
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003101905717151454,
      "loss": 1.2941,
      "theoretical_loss": 3.421175766980081,
      "tokens_seen": 2287337472
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003101705115346038,
      "loss": 1.2451,
      "theoretical_loss": 3.4211678745273795,
      "tokens_seen": 2287403008
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003101504513540622,
      "loss": 1.3824,
      "theoretical_loss": 3.4211599823641134,
      "tokens_seen": 2287468544
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031013039117352054,
      "loss": 1.3893,
      "theoretical_loss": 3.4211520904902635,
      "tokens_seen": 2287534080
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003101103309929789,
      "loss": 1.3434,
      "theoretical_loss": 3.421144198905811,
      "tokens_seen": 2287599616
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003100902708124373,
      "loss": 1.382,
      "theoretical_loss": 3.4211363076107366,
      "tokens_seen": 2287665152
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003100702106318957,
      "loss": 1.4226,
      "theoretical_loss": 3.421128416605022,
      "tokens_seen": 2287730688
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003100501504513541,
      "loss": 1.3648,
      "theoretical_loss": 3.421120525888648,
      "tokens_seen": 2287796224
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00031003009027081246,
      "loss": 1.4173,
      "theoretical_loss": 3.4211126354615957,
      "tokens_seen": 2287861760
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003100100300902708,
      "loss": 1.3784,
      "theoretical_loss": 3.421104745323846,
      "tokens_seen": 2287927296
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030998996990972924,
      "loss": 1.3232,
      "theoretical_loss": 3.4210968554753807,
      "tokens_seen": 2287992832
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003099699097291876,
      "loss": 1.3509,
      "theoretical_loss": 3.42108896591618,
      "tokens_seen": 2288058368
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030994984954864596,
      "loss": 1.3023,
      "theoretical_loss": 3.4210810766462254,
      "tokens_seen": 2288123904
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003099297893681043,
      "loss": 1.461,
      "theoretical_loss": 3.4210731876654985,
      "tokens_seen": 2288189440
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030990972918756274,
      "loss": 1.2928,
      "theoretical_loss": 3.42106529897398,
      "tokens_seen": 2288254976
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003098896690070211,
      "loss": 1.4474,
      "theoretical_loss": 3.4210574105716502,
      "tokens_seen": 2288320512
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030986960882647946,
      "loss": 1.3511,
      "theoretical_loss": 3.4210495224584916,
      "tokens_seen": 2288386048
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003098495486459378,
      "loss": 1.3927,
      "theoretical_loss": 3.4210416346344847,
      "tokens_seen": 2288451584
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003098294884653962,
      "loss": 1.3465,
      "theoretical_loss": 3.4210337470996106,
      "tokens_seen": 2288517120
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003098094282848546,
      "loss": 1.369,
      "theoretical_loss": 3.4210258598538505,
      "tokens_seen": 2288582656
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030978936810431297,
      "loss": 1.4399,
      "theoretical_loss": 3.421017972897185,
      "tokens_seen": 2288648192
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030976930792377133,
      "loss": 1.3564,
      "theoretical_loss": 3.4210100862295962,
      "tokens_seen": 2288713728
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003097492477432297,
      "loss": 1.3528,
      "theoretical_loss": 3.4210021998510647,
      "tokens_seen": 2288779264
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003097291875626881,
      "loss": 1.4112,
      "theoretical_loss": 3.4209943137615717,
      "tokens_seen": 2288844800
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030970912738214647,
      "loss": 1.45,
      "theoretical_loss": 3.420986427961098,
      "tokens_seen": 2288910336
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030968906720160483,
      "loss": 1.2767,
      "theoretical_loss": 3.4209785424496255,
      "tokens_seen": 2288975872
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003096690070210632,
      "loss": 1.363,
      "theoretical_loss": 3.4209706572271346,
      "tokens_seen": 2289041408
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003096489468405216,
      "loss": 1.3756,
      "theoretical_loss": 3.420962772293607,
      "tokens_seen": 2289106944
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030962888665997997,
      "loss": 1.4083,
      "theoretical_loss": 3.4209548876490232,
      "tokens_seen": 2289172480
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030960882647943833,
      "loss": 1.4067,
      "theoretical_loss": 3.4209470032933647,
      "tokens_seen": 2289238016
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003095887662988967,
      "loss": 1.3547,
      "theoretical_loss": 3.4209391192266128,
      "tokens_seen": 2289303552
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030956870611835506,
      "loss": 1.3741,
      "theoretical_loss": 3.4209312354487484,
      "tokens_seen": 2289369088
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003095486459378135,
      "loss": 1.4218,
      "theoretical_loss": 3.4209233519597526,
      "tokens_seen": 2289434624
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030952858575727184,
      "loss": 1.387,
      "theoretical_loss": 3.4209154687596066,
      "tokens_seen": 2289500160
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003095085255767302,
      "loss": 1.3806,
      "theoretical_loss": 3.420907585848292,
      "tokens_seen": 2289565696
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030948846539618856,
      "loss": 1.4758,
      "theoretical_loss": 3.420899703225789,
      "tokens_seen": 2289631232
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.000309468405215647,
      "loss": 1.3786,
      "theoretical_loss": 3.4208918208920798,
      "tokens_seen": 2289696768
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030944834503510534,
      "loss": 1.3206,
      "theoretical_loss": 3.420883938847145,
      "tokens_seen": 2289762304
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003094282848545637,
      "loss": 1.3761,
      "theoretical_loss": 3.4208760570909655,
      "tokens_seen": 2289827840
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030940822467402207,
      "loss": 1.3553,
      "theoretical_loss": 3.4208681756235233,
      "tokens_seen": 2289893376
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030938816449348043,
      "loss": 1.2362,
      "theoretical_loss": 3.4208602944447986,
      "tokens_seen": 2289958912
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030936810431293884,
      "loss": 1.4189,
      "theoretical_loss": 3.420852413554773,
      "tokens_seen": 2290024448
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003093480441323972,
      "loss": 1.369,
      "theoretical_loss": 3.420844532953428,
      "tokens_seen": 2290089984
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030932798395185557,
      "loss": 1.4543,
      "theoretical_loss": 3.420836652640744,
      "tokens_seen": 2290155520
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030930792377131393,
      "loss": 1.3835,
      "theoretical_loss": 3.420828772616703,
      "tokens_seen": 2290221056
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030928786359077235,
      "loss": 1.3864,
      "theoretical_loss": 3.4208208928812853,
      "tokens_seen": 2290286592
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003092678034102307,
      "loss": 1.3857,
      "theoretical_loss": 3.420813013434473,
      "tokens_seen": 2290352128
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030924774322968907,
      "loss": 1.3097,
      "theoretical_loss": 3.4208051342762467,
      "tokens_seen": 2290417664
    },
    {
      "epoch": 0.69,
      "objective/train/advantage_avg": 0.492944598197937,
      "objective/train/docs_used": 833966,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3662586212158203,
      "objective/train/original_loss": 1.3662586212158203,
      "objective/train/theoretical_loss": 3.4207972554065877,
      "objective/train/tokens_used": 2310943200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24356944859027863,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505326986312866,
      "objective/train/weighted_lm_loss": 1.4353595972061157,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4207972554065877,
      "tokens_seen": 2290483200
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030922768304914743,
      "loss": 1.3591,
      "theoretical_loss": 3.4207972554065877,
      "tokens_seen": 2290483200
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003092076228686058,
      "loss": 1.4669,
      "theoretical_loss": 3.4207893768254767,
      "tokens_seen": 2290548736
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003091875626880642,
      "loss": 1.279,
      "theoretical_loss": 3.4207814985328957,
      "tokens_seen": 2290614272
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003091675025075226,
      "loss": 1.3654,
      "theoretical_loss": 3.4207736205288257,
      "tokens_seen": 2290679808
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030914744232698094,
      "loss": 1.3705,
      "theoretical_loss": 3.4207657428132476,
      "tokens_seen": 2290745344
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003091273821464393,
      "loss": 1.3169,
      "theoretical_loss": 3.4207578653861423,
      "tokens_seen": 2290810880
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003091073219658977,
      "loss": 1.3774,
      "theoretical_loss": 3.4207499882474917,
      "tokens_seen": 2290876416
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003090872617853561,
      "loss": 1.2949,
      "theoretical_loss": 3.420742111397277,
      "tokens_seen": 2290941952
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030906720160481444,
      "loss": 1.3158,
      "theoretical_loss": 3.4207342348354786,
      "tokens_seen": 2291007488
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003090471414242728,
      "loss": 1.3156,
      "theoretical_loss": 3.420726358562078,
      "tokens_seen": 2291073024
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030902708124373116,
      "loss": 1.3331,
      "theoretical_loss": 3.4207184825770565,
      "tokens_seen": 2291138560
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003090070210631896,
      "loss": 1.3236,
      "theoretical_loss": 3.4207106068803954,
      "tokens_seen": 2291204096
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030898696088264794,
      "loss": 1.3608,
      "theoretical_loss": 3.420702731472076,
      "tokens_seen": 2291269632
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003089669007021063,
      "loss": 1.2699,
      "theoretical_loss": 3.420694856352079,
      "tokens_seen": 2291335168
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030894684052156467,
      "loss": 1.3839,
      "theoretical_loss": 3.420686981520386,
      "tokens_seen": 2291400704
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003089267803410231,
      "loss": 1.4065,
      "theoretical_loss": 3.420679106976978,
      "tokens_seen": 2291466240
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030890672016048145,
      "loss": 1.3791,
      "theoretical_loss": 3.4206712327218365,
      "tokens_seen": 2291531776
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003088866599799398,
      "loss": 1.3205,
      "theoretical_loss": 3.4206633587549424,
      "tokens_seen": 2291597312
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030886659979939817,
      "loss": 1.3926,
      "theoretical_loss": 3.420655485076277,
      "tokens_seen": 2291662848
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030884653961885653,
      "loss": 1.3452,
      "theoretical_loss": 3.4206476116858218,
      "tokens_seen": 2291728384
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030882647943831495,
      "loss": 1.2668,
      "theoretical_loss": 3.4206397385835574,
      "tokens_seen": 2291793920
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003088064192577733,
      "loss": 1.3524,
      "theoretical_loss": 3.4206318657694657,
      "tokens_seen": 2291859456
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003087863590772317,
      "loss": 1.4492,
      "theoretical_loss": 3.420623993243527,
      "tokens_seen": 2291924992
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030876629889669004,
      "loss": 1.2779,
      "theoretical_loss": 3.4206161210057235,
      "tokens_seen": 2291990528
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030874623871614845,
      "loss": 1.3587,
      "theoretical_loss": 3.4206082490560354,
      "tokens_seen": 2292056064
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003087261785356068,
      "loss": 1.445,
      "theoretical_loss": 3.420600377394445,
      "tokens_seen": 2292121600
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003087061183550652,
      "loss": 1.3899,
      "theoretical_loss": 3.420592506020933,
      "tokens_seen": 2292187136
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030868605817452354,
      "loss": 1.3859,
      "theoretical_loss": 3.4205846349354805,
      "tokens_seen": 2292252672
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003086659979939819,
      "loss": 1.3383,
      "theoretical_loss": 3.420576764138069,
      "tokens_seen": 2292318208
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003086459378134403,
      "loss": 1.3844,
      "theoretical_loss": 3.4205688936286793,
      "tokens_seen": 2292383744
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003086258776328987,
      "loss": 1.243,
      "theoretical_loss": 3.420561023407293,
      "tokens_seen": 2292449280
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030860581745235704,
      "loss": 1.3938,
      "theoretical_loss": 3.4205531534738913,
      "tokens_seen": 2292514816
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003085857572718154,
      "loss": 1.3929,
      "theoretical_loss": 3.4205452838284556,
      "tokens_seen": 2292580352
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003085656970912738,
      "loss": 1.3688,
      "theoretical_loss": 3.4205374144709664,
      "tokens_seen": 2292645888
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030854563691073224,
      "loss": 1.4293,
      "theoretical_loss": 3.420529545401406,
      "tokens_seen": 2292711424
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003085255767301906,
      "loss": 1.3807,
      "theoretical_loss": 3.420521676619755,
      "tokens_seen": 2292776960
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030850551654964896,
      "loss": 1.2998,
      "theoretical_loss": 3.4205138081259943,
      "tokens_seen": 2292842496
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003084854563691074,
      "loss": 1.4735,
      "theoretical_loss": 3.4205059399201057,
      "tokens_seen": 2292908032
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030846539618856574,
      "loss": 1.4137,
      "theoretical_loss": 3.42049807200207,
      "tokens_seen": 2292973568
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003084453360080241,
      "loss": 1.4925,
      "theoretical_loss": 3.420490204371869,
      "tokens_seen": 2293039104
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030842527582748247,
      "loss": 1.275,
      "theoretical_loss": 3.4204823370294837,
      "tokens_seen": 2293104640
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030840521564694083,
      "loss": 1.23,
      "theoretical_loss": 3.420474469974895,
      "tokens_seen": 2293170176
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030838515546639924,
      "loss": 1.4715,
      "theoretical_loss": 3.420466603208085,
      "tokens_seen": 2293235712
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0003083650952858576,
      "loss": 1.3694,
      "theoretical_loss": 3.420458736729034,
      "tokens_seen": 2293301248
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030834503510531597,
      "loss": 1.2555,
      "theoretical_loss": 3.4204508705377243,
      "tokens_seen": 2293366784
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030832497492477433,
      "loss": 1.3015,
      "theoretical_loss": 3.420443004634136,
      "tokens_seen": 2293432320
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00030830491474423275,
      "loss": 1.3101,
      "theoretical_loss": 3.4204351390182506,
      "tokens_seen": 2293497856
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003082848545636911,
      "loss": 1.4133,
      "theoretical_loss": 3.4204272736900503,
      "tokens_seen": 2293563392
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030826479438314947,
      "loss": 1.3506,
      "theoretical_loss": 3.420419408649515,
      "tokens_seen": 2293628928
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030824473420260783,
      "loss": 1.4653,
      "theoretical_loss": 3.420411543896627,
      "tokens_seen": 2293694464
    },
    {
      "debugging/Compilability": 1.0,
      "debugging/distinct-1-grams": 0.7361949066418735,
      "debugging/entropy-1-grams": 5.412433223050153,
      "debugging/length": 455.6666666666667,
      "debugging/num_segments": 15,
      "debugging/raw_token_scores_avg": 0.004264768213033676,
      "debugging/raw_token_scores_std": 0.017437797039747238,
      "debugging/score": 0.0066926720884253575,
      "debugging/score_std": 0.006256546529562537,
      "epoch": 0.7,
      "objective/train/advantage_avg": 0.49573466181755066,
      "objective/train/docs_used": 835219,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2304052114486694,
      "objective/train/original_loss": 1.2304052114486694,
      "objective/train/theoretical_loss": 3.4204036794313675,
      "objective/train/tokens_used": 2314220000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2460571676492691,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508244037628174,
      "objective/train/weighted_lm_loss": 1.2928634881973267,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.4204036794313675,
      "tokens_seen": 2293760000
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003082246740220662,
      "loss": 1.3434,
      "theoretical_loss": 3.4204036794313675,
      "tokens_seen": 2293760000
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003082046138415246,
      "loss": 1.3543,
      "theoretical_loss": 3.420395815253717,
      "tokens_seen": 2293825536
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000308184553660983,
      "loss": 1.4361,
      "theoretical_loss": 3.4203879513636575,
      "tokens_seen": 2293891072
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030816449348044134,
      "loss": 1.3896,
      "theoretical_loss": 3.4203800877611696,
      "tokens_seen": 2293956608
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003081444332998997,
      "loss": 1.3811,
      "theoretical_loss": 3.4203722244462353,
      "tokens_seen": 2294022144
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003081243731193581,
      "loss": 1.4169,
      "theoretical_loss": 3.4203643614188355,
      "tokens_seen": 2294087680
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003081043129388165,
      "loss": 1.4167,
      "theoretical_loss": 3.420356498678952,
      "tokens_seen": 2294153216
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030808425275827484,
      "loss": 1.3924,
      "theoretical_loss": 3.420348636226565,
      "tokens_seen": 2294218752
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003080641925777332,
      "loss": 1.5187,
      "theoretical_loss": 3.4203407740616565,
      "tokens_seen": 2294284288
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030804413239719157,
      "loss": 1.4593,
      "theoretical_loss": 3.4203329121842074,
      "tokens_seen": 2294349824
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030802407221665,
      "loss": 1.5008,
      "theoretical_loss": 3.4203250505941996,
      "tokens_seen": 2294415360
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030800401203610834,
      "loss": 1.4531,
      "theoretical_loss": 3.4203171892916138,
      "tokens_seen": 2294480896
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003079839518555667,
      "loss": 1.3566,
      "theoretical_loss": 3.4203093282764314,
      "tokens_seen": 2294546432
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030796389167502507,
      "loss": 1.4046,
      "theoretical_loss": 3.420301467548634,
      "tokens_seen": 2294611968
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003079438314944835,
      "loss": 1.3721,
      "theoretical_loss": 3.4202936071082024,
      "tokens_seen": 2294677504
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030792377131394185,
      "loss": 1.342,
      "theoretical_loss": 3.4202857469551184,
      "tokens_seen": 2294743040
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003079037111334002,
      "loss": 1.2842,
      "theoretical_loss": 3.420277887089363,
      "tokens_seen": 2294808576
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030788365095285857,
      "loss": 1.3903,
      "theoretical_loss": 3.4202700275109175,
      "tokens_seen": 2294874112
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030786359077231693,
      "loss": 1.2613,
      "theoretical_loss": 3.420262168219763,
      "tokens_seen": 2294939648
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030784353059177535,
      "loss": 1.2762,
      "theoretical_loss": 3.4202543092158812,
      "tokens_seen": 2295005184
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003078234704112337,
      "loss": 1.3645,
      "theoretical_loss": 3.420246450499253,
      "tokens_seen": 2295070720
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003078034102306921,
      "loss": 1.3321,
      "theoretical_loss": 3.42023859206986,
      "tokens_seen": 2295136256
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030778335005015044,
      "loss": 1.3202,
      "theoretical_loss": 3.4202307339276836,
      "tokens_seen": 2295201792
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030776328986960885,
      "loss": 1.3126,
      "theoretical_loss": 3.420222876072705,
      "tokens_seen": 2295267328
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003077432296890672,
      "loss": 1.3587,
      "theoretical_loss": 3.420215018504905,
      "tokens_seen": 2295332864
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003077231695085256,
      "loss": 1.255,
      "theoretical_loss": 3.420207161224266,
      "tokens_seen": 2295398400
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030770310932798394,
      "loss": 1.3318,
      "theoretical_loss": 3.420199304230768,
      "tokens_seen": 2295463936
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003076830491474423,
      "loss": 1.2513,
      "theoretical_loss": 3.420191447524393,
      "tokens_seen": 2295529472
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003076629889669007,
      "loss": 1.3609,
      "theoretical_loss": 3.4201835911051224,
      "tokens_seen": 2295595008
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003076429287863591,
      "loss": 1.4206,
      "theoretical_loss": 3.420175734972937,
      "tokens_seen": 2295660544
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030762286860581744,
      "loss": 1.4473,
      "theoretical_loss": 3.420167879127819,
      "tokens_seen": 2295726080
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003076028084252758,
      "loss": 1.4294,
      "theoretical_loss": 3.420160023569749,
      "tokens_seen": 2295791616
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003075827482447342,
      "loss": 1.3789,
      "theoretical_loss": 3.4201521682987086,
      "tokens_seen": 2295857152
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003075626880641926,
      "loss": 1.3655,
      "theoretical_loss": 3.4201443133146787,
      "tokens_seen": 2295922688
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030754262788365095,
      "loss": 1.3681,
      "theoretical_loss": 3.420136458617641,
      "tokens_seen": 2295988224
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003075225677031093,
      "loss": 1.4608,
      "theoretical_loss": 3.4201286042075774,
      "tokens_seen": 2296053760
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003075025075225677,
      "loss": 1.456,
      "theoretical_loss": 3.420120750084468,
      "tokens_seen": 2296119296
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003074824473420261,
      "loss": 1.4858,
      "theoretical_loss": 3.4201128962482947,
      "tokens_seen": 2296184832
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030746238716148445,
      "loss": 1.4546,
      "theoretical_loss": 3.420105042699039,
      "tokens_seen": 2296250368
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003074423269809428,
      "loss": 1.2431,
      "theoretical_loss": 3.420097189436682,
      "tokens_seen": 2296315904
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003074222668004012,
      "loss": 1.3747,
      "theoretical_loss": 3.4200893364612055,
      "tokens_seen": 2296381440
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003074022066198596,
      "loss": 1.4054,
      "theoretical_loss": 3.4200814837725897,
      "tokens_seen": 2296446976
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030738214643931795,
      "loss": 1.3788,
      "theoretical_loss": 3.4200736313708173,
      "tokens_seen": 2296512512
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003073620862587763,
      "loss": 1.4571,
      "theoretical_loss": 3.420065779255869,
      "tokens_seen": 2296578048
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003073420260782347,
      "loss": 1.469,
      "theoretical_loss": 3.4200579274277256,
      "tokens_seen": 2296643584
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003073219658976931,
      "loss": 1.3495,
      "theoretical_loss": 3.4200500758863694,
      "tokens_seen": 2296709120
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030730190571715146,
      "loss": 1.3058,
      "theoretical_loss": 3.420042224631781,
      "tokens_seen": 2296774656
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003072818455366098,
      "loss": 1.4324,
      "theoretical_loss": 3.4200343736639427,
      "tokens_seen": 2296840192
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003072617853560682,
      "loss": 1.4521,
      "theoretical_loss": 3.420026522982835,
      "tokens_seen": 2296905728
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030724172517552654,
      "loss": 1.3256,
      "theoretical_loss": 3.4200186725884394,
      "tokens_seen": 2296971264
    },
    {
      "epoch": 0.7,
      "objective/train/advantage_avg": 0.49237775802612305,
      "objective/train/docs_used": 836410,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2474843263626099,
      "objective/train/original_loss": 1.2474844455718994,
      "objective/train/theoretical_loss": 3.4200108224807373,
      "objective/train/tokens_used": 2317496800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24292311072349548,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050472617149353,
      "objective/train/weighted_lm_loss": 1.3104928731918335,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4200108224807373,
      "tokens_seen": 2297036800
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030722166499498496,
      "loss": 1.3437,
      "theoretical_loss": 3.4200108224807373,
      "tokens_seen": 2297036800
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003072016048144433,
      "loss": 1.3806,
      "theoretical_loss": 3.42000297265971,
      "tokens_seen": 2297102336
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003071815446339017,
      "loss": 1.3825,
      "theoretical_loss": 3.4199951231253394,
      "tokens_seen": 2297167872
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030716148445336005,
      "loss": 1.336,
      "theoretical_loss": 3.419987273877606,
      "tokens_seen": 2297233408
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030714142427281846,
      "loss": 1.3821,
      "theoretical_loss": 3.419979424916491,
      "tokens_seen": 2297298944
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003071213640922768,
      "loss": 1.3902,
      "theoretical_loss": 3.419971576241977,
      "tokens_seen": 2297364480
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003071013039117352,
      "loss": 1.4158,
      "theoretical_loss": 3.4199637278540447,
      "tokens_seen": 2297430016
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030708124373119355,
      "loss": 1.3074,
      "theoretical_loss": 3.419955879752675,
      "tokens_seen": 2297495552
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003070611835506519,
      "loss": 1.4089,
      "theoretical_loss": 3.41994803193785,
      "tokens_seen": 2297561088
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003070411233701104,
      "loss": 1.3015,
      "theoretical_loss": 3.4199401844095507,
      "tokens_seen": 2297626624
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030702106318956874,
      "loss": 1.3459,
      "theoretical_loss": 3.4199323371677584,
      "tokens_seen": 2297692160
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003070010030090271,
      "loss": 1.313,
      "theoretical_loss": 3.4199244902124546,
      "tokens_seen": 2297757696
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030698094282848547,
      "loss": 1.3809,
      "theoretical_loss": 3.419916643543621,
      "tokens_seen": 2297823232
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003069608826479439,
      "loss": 1.4001,
      "theoretical_loss": 3.419908797161238,
      "tokens_seen": 2297888768
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030694082246740225,
      "loss": 1.3614,
      "theoretical_loss": 3.419900951065288,
      "tokens_seen": 2297954304
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003069207622868606,
      "loss": 1.3175,
      "theoretical_loss": 3.4198931052557517,
      "tokens_seen": 2298019840
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030690070210631897,
      "loss": 1.4417,
      "theoretical_loss": 3.419885259732611,
      "tokens_seen": 2298085376
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030688064192577733,
      "loss": 1.3926,
      "theoretical_loss": 3.419877414495847,
      "tokens_seen": 2298150912
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030686058174523575,
      "loss": 1.331,
      "theoretical_loss": 3.419869569545441,
      "tokens_seen": 2298216448
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003068405215646941,
      "loss": 1.2478,
      "theoretical_loss": 3.4198617248813745,
      "tokens_seen": 2298281984
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003068204613841525,
      "loss": 1.4573,
      "theoretical_loss": 3.419853880503629,
      "tokens_seen": 2298347520
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030680040120361084,
      "loss": 1.3692,
      "theoretical_loss": 3.4198460364121854,
      "tokens_seen": 2298413056
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030678034102306925,
      "loss": 1.3446,
      "theoretical_loss": 3.419838192607026,
      "tokens_seen": 2298478592
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003067602808425276,
      "loss": 1.3516,
      "theoretical_loss": 3.419830349088131,
      "tokens_seen": 2298544128
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000306740220661986,
      "loss": 1.3658,
      "theoretical_loss": 3.419822505855483,
      "tokens_seen": 2298609664
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030672016048144434,
      "loss": 1.2789,
      "theoretical_loss": 3.4198146629090624,
      "tokens_seen": 2298675200
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003067001003009027,
      "loss": 1.2195,
      "theoretical_loss": 3.4198068202488514,
      "tokens_seen": 2298740736
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003066800401203611,
      "loss": 1.5117,
      "theoretical_loss": 3.419798977874831,
      "tokens_seen": 2298806272
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003066599799398195,
      "loss": 1.3519,
      "theoretical_loss": 3.4197911357869826,
      "tokens_seen": 2298871808
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030663991975927784,
      "loss": 1.2717,
      "theoretical_loss": 3.4197832939852875,
      "tokens_seen": 2298937344
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003066198595787362,
      "loss": 1.4256,
      "theoretical_loss": 3.419775452469727,
      "tokens_seen": 2299002880
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003065997993981946,
      "loss": 1.3395,
      "theoretical_loss": 3.4197676112402826,
      "tokens_seen": 2299068416
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000306579739217653,
      "loss": 1.3471,
      "theoretical_loss": 3.4197597702969365,
      "tokens_seen": 2299133952
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030655967903711135,
      "loss": 1.3882,
      "theoretical_loss": 3.419751929639669,
      "tokens_seen": 2299199488
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003065396188565697,
      "loss": 1.3683,
      "theoretical_loss": 3.4197440892684616,
      "tokens_seen": 2299265024
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003065195586760281,
      "loss": 1.3869,
      "theoretical_loss": 3.4197362491832966,
      "tokens_seen": 2299330560
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003064994984954865,
      "loss": 1.3953,
      "theoretical_loss": 3.4197284093841547,
      "tokens_seen": 2299396096
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030647943831494485,
      "loss": 1.4322,
      "theoretical_loss": 3.419720569871018,
      "tokens_seen": 2299461632
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003064593781344032,
      "loss": 1.2573,
      "theoretical_loss": 3.4197127306438664,
      "tokens_seen": 2299527168
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003064393179538616,
      "loss": 1.402,
      "theoretical_loss": 3.4197048917026827,
      "tokens_seen": 2299592704
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030641925777332,
      "loss": 1.3666,
      "theoretical_loss": 3.419697053047448,
      "tokens_seen": 2299658240
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030639919759277835,
      "loss": 1.4689,
      "theoretical_loss": 3.4196892146781437,
      "tokens_seen": 2299723776
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003063791374122367,
      "loss": 1.4502,
      "theoretical_loss": 3.4196813765947507,
      "tokens_seen": 2299789312
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003063590772316951,
      "loss": 1.3922,
      "theoretical_loss": 3.4196735387972517,
      "tokens_seen": 2299854848
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003063390170511535,
      "loss": 1.3652,
      "theoretical_loss": 3.4196657012856266,
      "tokens_seen": 2299920384
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030631895687061186,
      "loss": 1.4926,
      "theoretical_loss": 3.419657864059858,
      "tokens_seen": 2299985920
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003062988966900702,
      "loss": 1.4016,
      "theoretical_loss": 3.4196500271199266,
      "tokens_seen": 2300051456
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003062788365095286,
      "loss": 1.4524,
      "theoretical_loss": 3.4196421904658143,
      "tokens_seen": 2300116992
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030625877632898694,
      "loss": 1.407,
      "theoretical_loss": 3.419634354097502,
      "tokens_seen": 2300182528
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030623871614844536,
      "loss": 1.3116,
      "theoretical_loss": 3.4196265180149714,
      "tokens_seen": 2300248064
    },
    {
      "epoch": 0.7,
      "objective/train/advantage_avg": 0.4923454821109772,
      "objective/train/docs_used": 837515,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.379377841949463,
      "objective/train/original_loss": 1.3793777227401733,
      "objective/train/theoretical_loss": 3.4196186822182044,
      "objective/train/tokens_used": 2320773600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24345660209655762,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504721403121948,
      "objective/train/weighted_lm_loss": 1.4490232467651367,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4196186822182044,
      "tokens_seen": 2300313600
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003062186559679037,
      "loss": 1.452,
      "theoretical_loss": 3.4196186822182044,
      "tokens_seen": 2300313600
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003061985957873621,
      "loss": 1.3066,
      "theoretical_loss": 3.419610846707182,
      "tokens_seen": 2300379136
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030617853560682045,
      "loss": 1.3079,
      "theoretical_loss": 3.4196030114818856,
      "tokens_seen": 2300444672
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030615847542627886,
      "loss": 1.388,
      "theoretical_loss": 3.419595176542297,
      "tokens_seen": 2300510208
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003061384152457372,
      "loss": 1.3907,
      "theoretical_loss": 3.419587341888397,
      "tokens_seen": 2300575744
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003061183550651956,
      "loss": 1.447,
      "theoretical_loss": 3.4195795075201674,
      "tokens_seen": 2300641280
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030609829488465395,
      "loss": 1.2164,
      "theoretical_loss": 3.4195716734375896,
      "tokens_seen": 2300706816
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003060782347041123,
      "loss": 1.4316,
      "theoretical_loss": 3.4195638396406456,
      "tokens_seen": 2300772352
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030605817452357073,
      "loss": 1.2311,
      "theoretical_loss": 3.4195560061293158,
      "tokens_seen": 2300837888
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003060381143430291,
      "loss": 1.4386,
      "theoretical_loss": 3.419548172903583,
      "tokens_seen": 2300903424
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030601805416248745,
      "loss": 1.3838,
      "theoretical_loss": 3.4195403399634268,
      "tokens_seen": 2300968960
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003059979939819458,
      "loss": 1.4082,
      "theoretical_loss": 3.4195325073088307,
      "tokens_seen": 2301034496
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030597793380140423,
      "loss": 1.3384,
      "theoretical_loss": 3.4195246749397743,
      "tokens_seen": 2301100032
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003059578736208626,
      "loss": 1.353,
      "theoretical_loss": 3.4195168428562406,
      "tokens_seen": 2301165568
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030593781344032096,
      "loss": 1.3503,
      "theoretical_loss": 3.41950901105821,
      "tokens_seen": 2301231104
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003059177532597793,
      "loss": 1.2695,
      "theoretical_loss": 3.4195011795456645,
      "tokens_seen": 2301296640
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003058976930792377,
      "loss": 1.3423,
      "theoretical_loss": 3.4194933483185856,
      "tokens_seen": 2301362176
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003058776328986961,
      "loss": 1.4033,
      "theoretical_loss": 3.419485517376954,
      "tokens_seen": 2301427712
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030585757271815446,
      "loss": 1.3148,
      "theoretical_loss": 3.4194776867207524,
      "tokens_seen": 2301493248
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003058375125376128,
      "loss": 1.3432,
      "theoretical_loss": 3.4194698563499615,
      "tokens_seen": 2301558784
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003058174523570712,
      "loss": 1.3854,
      "theoretical_loss": 3.4194620262645627,
      "tokens_seen": 2301624320
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003057973921765296,
      "loss": 1.4436,
      "theoretical_loss": 3.4194541964645375,
      "tokens_seen": 2301689856
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030577733199598796,
      "loss": 1.4189,
      "theoretical_loss": 3.419446366949868,
      "tokens_seen": 2301755392
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003057572718154463,
      "loss": 1.3607,
      "theoretical_loss": 3.419438537720535,
      "tokens_seen": 2301820928
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003057372116349047,
      "loss": 1.4496,
      "theoretical_loss": 3.4194307087765203,
      "tokens_seen": 2301886464
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030571715145436305,
      "loss": 1.4107,
      "theoretical_loss": 3.4194228801178053,
      "tokens_seen": 2301952000
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030569709127382147,
      "loss": 1.348,
      "theoretical_loss": 3.419415051744371,
      "tokens_seen": 2302017536
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030567703109327983,
      "loss": 1.5813,
      "theoretical_loss": 3.4194072236562,
      "tokens_seen": 2302083072
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003056569709127382,
      "loss": 1.4374,
      "theoretical_loss": 3.4193993958532722,
      "tokens_seen": 2302148608
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030563691073219655,
      "loss": 1.3,
      "theoretical_loss": 3.4193915683355707,
      "tokens_seen": 2302214144
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030561685055165497,
      "loss": 1.4269,
      "theoretical_loss": 3.4193837411030765,
      "tokens_seen": 2302279680
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030559679037111333,
      "loss": 1.3077,
      "theoretical_loss": 3.4193759141557702,
      "tokens_seen": 2302345216
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003055767301905717,
      "loss": 1.3244,
      "theoretical_loss": 3.4193680874936345,
      "tokens_seen": 2302410752
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030555667001003006,
      "loss": 1.3705,
      "theoretical_loss": 3.4193602611166503,
      "tokens_seen": 2302476288
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003055366098294885,
      "loss": 1.3226,
      "theoretical_loss": 3.4193524350247992,
      "tokens_seen": 2302541824
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003055165496489469,
      "loss": 1.4855,
      "theoretical_loss": 3.4193446092180624,
      "tokens_seen": 2302607360
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030549648946840525,
      "loss": 1.3635,
      "theoretical_loss": 3.419336783696422,
      "tokens_seen": 2302672896
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003054764292878636,
      "loss": 1.3851,
      "theoretical_loss": 3.419328958459859,
      "tokens_seen": 2302738432
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000305456369107322,
      "loss": 1.4469,
      "theoretical_loss": 3.4193211335083546,
      "tokens_seen": 2302803968
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003054363089267804,
      "loss": 1.3543,
      "theoretical_loss": 3.4193133088418914,
      "tokens_seen": 2302869504
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030541624874623875,
      "loss": 1.3249,
      "theoretical_loss": 3.41930548446045,
      "tokens_seen": 2302935040
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003053961885656971,
      "loss": 1.3625,
      "theoretical_loss": 3.4192976603640126,
      "tokens_seen": 2303000576
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003053761283851555,
      "loss": 1.4538,
      "theoretical_loss": 3.4192898365525597,
      "tokens_seen": 2303066112
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003053560682046139,
      "loss": 1.4336,
      "theoretical_loss": 3.4192820130260735,
      "tokens_seen": 2303131648
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030533600802407226,
      "loss": 1.4389,
      "theoretical_loss": 3.419274189784536,
      "tokens_seen": 2303197184
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003053159478435306,
      "loss": 1.3371,
      "theoretical_loss": 3.4192663668279275,
      "tokens_seen": 2303262720
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000305295887662989,
      "loss": 1.4551,
      "theoretical_loss": 3.4192585441562304,
      "tokens_seen": 2303328256
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030527582748244734,
      "loss": 1.3857,
      "theoretical_loss": 3.419250721769426,
      "tokens_seen": 2303393792
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030525576730190576,
      "loss": 1.4229,
      "theoretical_loss": 3.4192428996674957,
      "tokens_seen": 2303459328
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003052357071213641,
      "loss": 1.4191,
      "theoretical_loss": 3.419235077850421,
      "tokens_seen": 2303524864
    },
    {
      "epoch": 0.7,
      "objective/train/advantage_avg": 0.4902251660823822,
      "objective/train/docs_used": 838679,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2884494066238403,
      "objective/train/original_loss": 1.2884492874145508,
      "objective/train/theoretical_loss": 3.419227256318184,
      "objective/train/tokens_used": 2324050400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24112531542778015,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502482652664185,
      "objective/train/weighted_lm_loss": 1.3531256914138794,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.419227256318184,
      "tokens_seen": 2303590400
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003052156469408225,
      "loss": 1.3873,
      "theoretical_loss": 3.419227256318184,
      "tokens_seen": 2303590400
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030519558676028085,
      "loss": 1.4293,
      "theoretical_loss": 3.419219435070765,
      "tokens_seen": 2303655936
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030517552657973926,
      "loss": 1.5237,
      "theoretical_loss": 3.4192116141081463,
      "tokens_seen": 2303721472
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003051554663991976,
      "loss": 1.4015,
      "theoretical_loss": 3.4192037934303103,
      "tokens_seen": 2303787008
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000305135406218656,
      "loss": 1.3993,
      "theoretical_loss": 3.4191959730372368,
      "tokens_seen": 2303852544
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030511534603811435,
      "loss": 1.4101,
      "theoretical_loss": 3.4191881529289083,
      "tokens_seen": 2303918080
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003050952858575727,
      "loss": 1.4105,
      "theoretical_loss": 3.4191803331053063,
      "tokens_seen": 2303983616
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030507522567703113,
      "loss": 1.396,
      "theoretical_loss": 3.419172513566412,
      "tokens_seen": 2304049152
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003050551654964895,
      "loss": 1.4005,
      "theoretical_loss": 3.419164694312208,
      "tokens_seen": 2304114688
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030503510531594785,
      "loss": 1.3728,
      "theoretical_loss": 3.4191568753426744,
      "tokens_seen": 2304180224
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003050150451354062,
      "loss": 1.4312,
      "theoretical_loss": 3.419149056657793,
      "tokens_seen": 2304245760
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030499498495486463,
      "loss": 1.3049,
      "theoretical_loss": 3.4191412382575463,
      "tokens_seen": 2304311296
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000304974924774323,
      "loss": 1.3577,
      "theoretical_loss": 3.419133420141915,
      "tokens_seen": 2304376832
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030495486459378136,
      "loss": 1.328,
      "theoretical_loss": 3.419125602310881,
      "tokens_seen": 2304442368
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003049348044132397,
      "loss": 1.3736,
      "theoretical_loss": 3.4191177847644254,
      "tokens_seen": 2304507904
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003049147442326981,
      "loss": 1.4462,
      "theoretical_loss": 3.4191099675025303,
      "tokens_seen": 2304573440
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003048946840521565,
      "loss": 1.378,
      "theoretical_loss": 3.419102150525177,
      "tokens_seen": 2304638976
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030487462387161486,
      "loss": 1.3713,
      "theoretical_loss": 3.419094333832347,
      "tokens_seen": 2304704512
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003048545636910732,
      "loss": 1.4072,
      "theoretical_loss": 3.419086517424022,
      "tokens_seen": 2304770048
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003048345035105316,
      "loss": 1.388,
      "theoretical_loss": 3.4190787013001835,
      "tokens_seen": 2304835584
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030481444332999,
      "loss": 1.3932,
      "theoretical_loss": 3.4190708854608127,
      "tokens_seen": 2304901120
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030479438314944836,
      "loss": 1.3805,
      "theoretical_loss": 3.419063069905892,
      "tokens_seen": 2304966656
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003047743229689067,
      "loss": 1.376,
      "theoretical_loss": 3.4190552546354023,
      "tokens_seen": 2305032192
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003047542627883651,
      "loss": 1.2854,
      "theoretical_loss": 3.4190474396493253,
      "tokens_seen": 2305097728
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030473420260782345,
      "loss": 1.3617,
      "theoretical_loss": 3.4190396249476427,
      "tokens_seen": 2305163264
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030471414242728187,
      "loss": 1.4906,
      "theoretical_loss": 3.4190318105303357,
      "tokens_seen": 2305228800
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030469408224674023,
      "loss": 1.4368,
      "theoretical_loss": 3.419023996397386,
      "tokens_seen": 2305294336
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003046740220661986,
      "loss": 1.2445,
      "theoretical_loss": 3.4190161825487757,
      "tokens_seen": 2305359872
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030465396188565695,
      "loss": 1.4304,
      "theoretical_loss": 3.4190083689844855,
      "tokens_seen": 2305425408
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030463390170511537,
      "loss": 1.3449,
      "theoretical_loss": 3.419000555704498,
      "tokens_seen": 2305490944
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030461384152457373,
      "loss": 1.3734,
      "theoretical_loss": 3.418992742708794,
      "tokens_seen": 2305556480
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003045937813440321,
      "loss": 1.4533,
      "theoretical_loss": 3.4189849299973547,
      "tokens_seen": 2305622016
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030457372116349046,
      "loss": 1.3074,
      "theoretical_loss": 3.4189771175701633,
      "tokens_seen": 2305687552
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003045536609829488,
      "loss": 1.3945,
      "theoretical_loss": 3.4189693054271997,
      "tokens_seen": 2305753088
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030453360080240723,
      "loss": 1.2508,
      "theoretical_loss": 3.418961493568446,
      "tokens_seen": 2305818624
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003045135406218656,
      "loss": 1.3562,
      "theoretical_loss": 3.418953681993884,
      "tokens_seen": 2305884160
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030449348044132396,
      "loss": 1.4379,
      "theoretical_loss": 3.4189458707034954,
      "tokens_seen": 2305949696
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003044734202607823,
      "loss": 1.4055,
      "theoretical_loss": 3.418938059697261,
      "tokens_seen": 2306015232
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030445336008024074,
      "loss": 1.3618,
      "theoretical_loss": 3.4189302489751636,
      "tokens_seen": 2306080768
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003044332998996991,
      "loss": 1.4449,
      "theoretical_loss": 3.418922438537184,
      "tokens_seen": 2306146304
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030441323971915746,
      "loss": 1.3543,
      "theoretical_loss": 3.418914628383304,
      "tokens_seen": 2306211840
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003043931795386158,
      "loss": 1.4587,
      "theoretical_loss": 3.418906818513505,
      "tokens_seen": 2306277376
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003043731193580742,
      "loss": 1.2663,
      "theoretical_loss": 3.418899008927769,
      "tokens_seen": 2306342912
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003043530591775326,
      "loss": 1.3244,
      "theoretical_loss": 3.418891199626077,
      "tokens_seen": 2306408448
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030433299899699097,
      "loss": 1.3915,
      "theoretical_loss": 3.418883390608411,
      "tokens_seen": 2306473984
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030431293881644933,
      "loss": 1.3411,
      "theoretical_loss": 3.4188755818747527,
      "tokens_seen": 2306539520
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003042928786359077,
      "loss": 1.386,
      "theoretical_loss": 3.4188677734250836,
      "tokens_seen": 2306605056
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003042728184553661,
      "loss": 1.4363,
      "theoretical_loss": 3.4188599652593847,
      "tokens_seen": 2306670592
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030425275827482447,
      "loss": 1.2977,
      "theoretical_loss": 3.4188521573776383,
      "tokens_seen": 2306736128
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030423269809428283,
      "loss": 1.3745,
      "theoretical_loss": 3.4188443497798264,
      "tokens_seen": 2306801664
    },
    {
      "epoch": 0.7,
      "objective/train/advantage_avg": 0.4904576241970062,
      "objective/train/docs_used": 839893,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3918921947479248,
      "objective/train/original_loss": 1.3918921947479248,
      "objective/train/theoretical_loss": 3.4188365424659293,
      "objective/train/tokens_used": 2327327200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24111886322498322,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502715110778809,
      "objective/train/weighted_lm_loss": 1.461655855178833,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4188365424659293,
      "tokens_seen": 2306867200
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003042126379137412,
      "loss": 1.446,
      "theoretical_loss": 3.4188365424659293,
      "tokens_seen": 2306867200
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003041925777331996,
      "loss": 1.262,
      "theoretical_loss": 3.41882873543593,
      "tokens_seen": 2306932736
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030417251755265797,
      "loss": 1.4661,
      "theoretical_loss": 3.418820928689809,
      "tokens_seen": 2306998272
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030415245737211633,
      "loss": 1.2961,
      "theoretical_loss": 3.4188131222275486,
      "tokens_seen": 2307063808
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003041323971915747,
      "loss": 1.4536,
      "theoretical_loss": 3.4188053160491303,
      "tokens_seen": 2307129344
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030411233701103306,
      "loss": 1.335,
      "theoretical_loss": 3.418797510154535,
      "tokens_seen": 2307194880
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003040922768304915,
      "loss": 1.3981,
      "theoretical_loss": 3.418789704543746,
      "tokens_seen": 2307260416
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030407221664994984,
      "loss": 1.3765,
      "theoretical_loss": 3.418781899216743,
      "tokens_seen": 2307325952
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030405215646940825,
      "loss": 1.3885,
      "theoretical_loss": 3.418774094173509,
      "tokens_seen": 2307391488
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003040320962888666,
      "loss": 1.3489,
      "theoretical_loss": 3.418766289414025,
      "tokens_seen": 2307457024
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030401203610832503,
      "loss": 1.2577,
      "theoretical_loss": 3.418758484938272,
      "tokens_seen": 2307522560
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003039919759277834,
      "loss": 1.3738,
      "theoretical_loss": 3.4187506807462333,
      "tokens_seen": 2307588096
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030397191574724176,
      "loss": 1.473,
      "theoretical_loss": 3.4187428768378894,
      "tokens_seen": 2307653632
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003039518555667001,
      "loss": 1.3896,
      "theoretical_loss": 3.418735073213222,
      "tokens_seen": 2307719168
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003039317953861585,
      "loss": 1.2849,
      "theoretical_loss": 3.4187272698722126,
      "tokens_seen": 2307784704
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003039117352056169,
      "loss": 1.4246,
      "theoretical_loss": 3.4187194668148435,
      "tokens_seen": 2307850240
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030389167502507526,
      "loss": 1.4696,
      "theoretical_loss": 3.4187116640410955,
      "tokens_seen": 2307915776
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003038716148445336,
      "loss": 1.2533,
      "theoretical_loss": 3.418703861550951,
      "tokens_seen": 2307981312
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000303851554663992,
      "loss": 1.3325,
      "theoretical_loss": 3.4186960593443914,
      "tokens_seen": 2308046848
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003038314944834504,
      "loss": 1.4639,
      "theoretical_loss": 3.418688257421398,
      "tokens_seen": 2308112384
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030381143430290876,
      "loss": 1.4017,
      "theoretical_loss": 3.418680455781953,
      "tokens_seen": 2308177920
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003037913741223671,
      "loss": 1.3229,
      "theoretical_loss": 3.418672654426037,
      "tokens_seen": 2308243456
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003037713139418255,
      "loss": 1.3701,
      "theoretical_loss": 3.418664853353633,
      "tokens_seen": 2308308992
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030375125376128385,
      "loss": 1.3841,
      "theoretical_loss": 3.418657052564722,
      "tokens_seen": 2308374528
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030373119358074227,
      "loss": 1.4385,
      "theoretical_loss": 3.4186492520592853,
      "tokens_seen": 2308440064
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030371113340020063,
      "loss": 1.4372,
      "theoretical_loss": 3.418641451837305,
      "tokens_seen": 2308505600
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000303691073219659,
      "loss": 1.3943,
      "theoretical_loss": 3.418633651898763,
      "tokens_seen": 2308571136
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030367101303911735,
      "loss": 1.5641,
      "theoretical_loss": 3.4186258522436406,
      "tokens_seen": 2308636672
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030365095285857577,
      "loss": 1.48,
      "theoretical_loss": 3.4186180528719188,
      "tokens_seen": 2308702208
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030363089267803413,
      "loss": 1.3414,
      "theoretical_loss": 3.418610253783581,
      "tokens_seen": 2308767744
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003036108324974925,
      "loss": 1.409,
      "theoretical_loss": 3.418602454978607,
      "tokens_seen": 2308833280
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030359077231695086,
      "loss": 1.4416,
      "theoretical_loss": 3.4185946564569796,
      "tokens_seen": 2308898816
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003035707121364092,
      "loss": 1.4761,
      "theoretical_loss": 3.41858685821868,
      "tokens_seen": 2308964352
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030355065195586763,
      "loss": 1.335,
      "theoretical_loss": 3.4185790602636903,
      "tokens_seen": 2309029888
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000303530591775326,
      "loss": 1.3836,
      "theoretical_loss": 3.4185712625919917,
      "tokens_seen": 2309095424
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030351053159478436,
      "loss": 1.4626,
      "theoretical_loss": 3.418563465203566,
      "tokens_seen": 2309160960
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003034904714142427,
      "loss": 1.4621,
      "theoretical_loss": 3.418555668098395,
      "tokens_seen": 2309226496
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030347041123370114,
      "loss": 1.4003,
      "theoretical_loss": 3.41854787127646,
      "tokens_seen": 2309292032
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003034503510531595,
      "loss": 1.4173,
      "theoretical_loss": 3.4185400747377432,
      "tokens_seen": 2309357568
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030343029087261786,
      "loss": 1.4928,
      "theoretical_loss": 3.4185322784822256,
      "tokens_seen": 2309423104
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003034102306920762,
      "loss": 1.363,
      "theoretical_loss": 3.41852448250989,
      "tokens_seen": 2309488640
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003033901705115346,
      "loss": 1.356,
      "theoretical_loss": 3.418516686820717,
      "tokens_seen": 2309554176
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000303370110330993,
      "loss": 1.3793,
      "theoretical_loss": 3.4185088914146884,
      "tokens_seen": 2309619712
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030335005015045137,
      "loss": 1.3563,
      "theoretical_loss": 3.4185010962917866,
      "tokens_seen": 2309685248
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030332998996990973,
      "loss": 1.3141,
      "theoretical_loss": 3.4184933014519925,
      "tokens_seen": 2309750784
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003033099297893681,
      "loss": 1.4037,
      "theoretical_loss": 3.4184855068952884,
      "tokens_seen": 2309816320
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003032898696088265,
      "loss": 1.3442,
      "theoretical_loss": 3.4184777126216552,
      "tokens_seen": 2309881856
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030326980942828487,
      "loss": 1.3768,
      "theoretical_loss": 3.4184699186310756,
      "tokens_seen": 2309947392
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030324974924774323,
      "loss": 1.2578,
      "theoretical_loss": 3.4184621249235305,
      "tokens_seen": 2310012928
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003032296890672016,
      "loss": 1.3712,
      "theoretical_loss": 3.418454331499002,
      "tokens_seen": 2310078464
    },
    {
      "epoch": 0.7,
      "objective/train/advantage_avg": 0.4919644892215729,
      "objective/train/docs_used": 841155,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.208345890045166,
      "objective/train/original_loss": 1.208345890045166,
      "objective/train/theoretical_loss": 3.418446538357472,
      "objective/train/tokens_used": 2330604000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24263885617256165,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504299402236938,
      "objective/train/weighted_lm_loss": 1.2690961360931396,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.418446538357472,
      "tokens_seen": 2310144000
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030320962888666,
      "loss": 1.4004,
      "theoretical_loss": 3.418446538357472,
      "tokens_seen": 2310144000
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030318956870611837,
      "loss": 1.2856,
      "theoretical_loss": 3.4184387454989213,
      "tokens_seen": 2310209536
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030316950852557673,
      "loss": 1.3888,
      "theoretical_loss": 3.418430952923332,
      "tokens_seen": 2310275072
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003031494483450351,
      "loss": 1.4336,
      "theoretical_loss": 3.4184231606306867,
      "tokens_seen": 2310340608
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030312938816449346,
      "loss": 1.3479,
      "theoretical_loss": 3.418415368620966,
      "tokens_seen": 2310406144
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003031093279839519,
      "loss": 1.424,
      "theoretical_loss": 3.4184075768941518,
      "tokens_seen": 2310471680
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030308926780341024,
      "loss": 1.4724,
      "theoretical_loss": 3.418399785450226,
      "tokens_seen": 2310537216
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003030692076228686,
      "loss": 1.4256,
      "theoretical_loss": 3.4183919942891707,
      "tokens_seen": 2310602752
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030304914744232696,
      "loss": 1.2866,
      "theoretical_loss": 3.418384203410967,
      "tokens_seen": 2310668288
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003030290872617854,
      "loss": 1.3105,
      "theoretical_loss": 3.4183764128155962,
      "tokens_seen": 2310733824
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030300902708124374,
      "loss": 1.3713,
      "theoretical_loss": 3.418368622503041,
      "tokens_seen": 2310799360
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003029889669007021,
      "loss": 1.3881,
      "theoretical_loss": 3.418360832473283,
      "tokens_seen": 2310864896
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030296890672016046,
      "loss": 1.2866,
      "theoretical_loss": 3.418353042726303,
      "tokens_seen": 2310930432
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003029488465396188,
      "loss": 1.2643,
      "theoretical_loss": 3.4183452532620837,
      "tokens_seen": 2310995968
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030292878635907724,
      "loss": 1.511,
      "theoretical_loss": 3.4183374640806066,
      "tokens_seen": 2311061504
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003029087261785356,
      "loss": 1.4102,
      "theoretical_loss": 3.418329675181853,
      "tokens_seen": 2311127040
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030288866599799397,
      "loss": 1.5318,
      "theoretical_loss": 3.4183218865658054,
      "tokens_seen": 2311192576
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030286860581745233,
      "loss": 1.216,
      "theoretical_loss": 3.4183140982324445,
      "tokens_seen": 2311258112
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030284854563691075,
      "loss": 1.3955,
      "theoretical_loss": 3.4183063101817526,
      "tokens_seen": 2311323648
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003028284854563691,
      "loss": 1.4759,
      "theoretical_loss": 3.4182985224137115,
      "tokens_seen": 2311389184
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030280842527582747,
      "loss": 1.4091,
      "theoretical_loss": 3.418290734928303,
      "tokens_seen": 2311454720
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030278836509528583,
      "loss": 1.4603,
      "theoretical_loss": 3.4182829477255083,
      "tokens_seen": 2311520256
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003027683049147442,
      "loss": 1.3503,
      "theoretical_loss": 3.41827516080531,
      "tokens_seen": 2311585792
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003027482447342026,
      "loss": 1.4859,
      "theoretical_loss": 3.4182673741676886,
      "tokens_seen": 2311651328
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000302728184553661,
      "loss": 1.4432,
      "theoretical_loss": 3.418259587812627,
      "tokens_seen": 2311716864
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030270812437311934,
      "loss": 1.4626,
      "theoretical_loss": 3.418251801740106,
      "tokens_seen": 2311782400
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003026880641925777,
      "loss": 1.3918,
      "theoretical_loss": 3.418244015950108,
      "tokens_seen": 2311847936
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003026680040120361,
      "loss": 1.3597,
      "theoretical_loss": 3.4182362304426146,
      "tokens_seen": 2311913472
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003026479438314945,
      "loss": 1.3771,
      "theoretical_loss": 3.4182284452176077,
      "tokens_seen": 2311979008
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030262788365095284,
      "loss": 1.3939,
      "theoretical_loss": 3.4182206602750687,
      "tokens_seen": 2312044544
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003026078234704112,
      "loss": 1.3058,
      "theoretical_loss": 3.418212875614979,
      "tokens_seen": 2312110080
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030258776328986956,
      "loss": 1.4653,
      "theoretical_loss": 3.4182050912373216,
      "tokens_seen": 2312175616
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000302567703109328,
      "loss": 1.3446,
      "theoretical_loss": 3.4181973071420773,
      "tokens_seen": 2312241152
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003025476429287864,
      "loss": 1.2939,
      "theoretical_loss": 3.418189523329228,
      "tokens_seen": 2312306688
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030252758274824476,
      "loss": 1.3933,
      "theoretical_loss": 3.4181817397987553,
      "tokens_seen": 2312372224
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003025075225677031,
      "loss": 1.3766,
      "theoretical_loss": 3.418173956550641,
      "tokens_seen": 2312437760
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030248746238716154,
      "loss": 1.4017,
      "theoretical_loss": 3.418166173584867,
      "tokens_seen": 2312503296
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003024674022066199,
      "loss": 1.4378,
      "theoretical_loss": 3.4181583909014153,
      "tokens_seen": 2312568832
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030244734202607826,
      "loss": 1.3323,
      "theoretical_loss": 3.4181506085002673,
      "tokens_seen": 2312634368
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003024272818455366,
      "loss": 1.3612,
      "theoretical_loss": 3.4181428263814047,
      "tokens_seen": 2312699904
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000302407221664995,
      "loss": 1.3921,
      "theoretical_loss": 3.4181350445448095,
      "tokens_seen": 2312765440
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003023871614844534,
      "loss": 1.3503,
      "theoretical_loss": 3.4181272629904638,
      "tokens_seen": 2312830976
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030236710130391177,
      "loss": 1.4336,
      "theoretical_loss": 3.4181194817183487,
      "tokens_seen": 2312896512
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030234704112337013,
      "loss": 1.3249,
      "theoretical_loss": 3.4181117007284456,
      "tokens_seen": 2312962048
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003023269809428285,
      "loss": 1.3516,
      "theoretical_loss": 3.4181039200207377,
      "tokens_seen": 2313027584
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003023069207622869,
      "loss": 1.4178,
      "theoretical_loss": 3.418096139595206,
      "tokens_seen": 2313093120
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030228686058174527,
      "loss": 1.4479,
      "theoretical_loss": 3.418088359451832,
      "tokens_seen": 2313158656
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030226680040120363,
      "loss": 1.3881,
      "theoretical_loss": 3.418080579590597,
      "tokens_seen": 2313224192
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000302246740220662,
      "loss": 1.4307,
      "theoretical_loss": 3.418072800011484,
      "tokens_seen": 2313289728
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003022266800401204,
      "loss": 1.4377,
      "theoretical_loss": 3.4180650207144745,
      "tokens_seen": 2313355264
    },
    {
      "epoch": 0.7,
      "objective/train/advantage_avg": 0.49521371722221375,
      "objective/train/docs_used": 842373,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2390812635421753,
      "objective/train/original_loss": 1.2390811443328857,
      "objective/train/theoretical_loss": 3.41805724169955,
      "objective/train/tokens_used": 2333880800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24546091258525848,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050769329071045,
      "objective/train/weighted_lm_loss": 1.3019020557403564,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.41805724169955,
      "tokens_seen": 2313420800
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030220661985957877,
      "loss": 1.3697,
      "theoretical_loss": 3.41805724169955,
      "tokens_seen": 2313420800
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030218655967903713,
      "loss": 1.4236,
      "theoretical_loss": 3.418049462966692,
      "tokens_seen": 2313486336
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003021664994984955,
      "loss": 1.3971,
      "theoretical_loss": 3.4180416845158827,
      "tokens_seen": 2313551872
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030214643931795386,
      "loss": 1.3813,
      "theoretical_loss": 3.418033906347104,
      "tokens_seen": 2313617408
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003021263791374123,
      "loss": 1.4111,
      "theoretical_loss": 3.4180261284603373,
      "tokens_seen": 2313682944
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030210631895687064,
      "loss": 1.3943,
      "theoretical_loss": 3.4180183508555646,
      "tokens_seen": 2313748480
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000302086258776329,
      "loss": 1.481,
      "theoretical_loss": 3.4180105735327677,
      "tokens_seen": 2313814016
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030206619859578736,
      "loss": 1.4579,
      "theoretical_loss": 3.418002796491928,
      "tokens_seen": 2313879552
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003020461384152458,
      "loss": 1.4023,
      "theoretical_loss": 3.417995019733028,
      "tokens_seen": 2313945088
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030202607823470414,
      "loss": 1.2837,
      "theoretical_loss": 3.417987243256049,
      "tokens_seen": 2314010624
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003020060180541625,
      "loss": 1.3474,
      "theoretical_loss": 3.417979467060973,
      "tokens_seen": 2314076160
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030198595787362087,
      "loss": 1.4622,
      "theoretical_loss": 3.4179716911477818,
      "tokens_seen": 2314141696
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030196589769307923,
      "loss": 1.4107,
      "theoretical_loss": 3.4179639155164567,
      "tokens_seen": 2314207232
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030194583751253764,
      "loss": 1.3902,
      "theoretical_loss": 3.4179561401669805,
      "tokens_seen": 2314272768
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000301925777331996,
      "loss": 1.4099,
      "theoretical_loss": 3.417948365099334,
      "tokens_seen": 2314338304
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030190571715145437,
      "loss": 1.3417,
      "theoretical_loss": 3.417940590313499,
      "tokens_seen": 2314403840
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030188565697091273,
      "loss": 1.3901,
      "theoretical_loss": 3.4179328158094586,
      "tokens_seen": 2314469376
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030186559679037115,
      "loss": 1.4903,
      "theoretical_loss": 3.4179250415871936,
      "tokens_seen": 2314534912
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003018455366098295,
      "loss": 1.353,
      "theoretical_loss": 3.4179172676466854,
      "tokens_seen": 2314600448
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030182547642928787,
      "loss": 1.362,
      "theoretical_loss": 3.4179094939879167,
      "tokens_seen": 2314665984
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030180541624874623,
      "loss": 1.2929,
      "theoretical_loss": 3.417901720610869,
      "tokens_seen": 2314731520
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003017853560682046,
      "loss": 1.3433,
      "theoretical_loss": 3.417893947515524,
      "tokens_seen": 2314797056
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000301765295887663,
      "loss": 1.4274,
      "theoretical_loss": 3.417886174701864,
      "tokens_seen": 2314862592
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003017452357071214,
      "loss": 1.3367,
      "theoretical_loss": 3.41787840216987,
      "tokens_seen": 2314928128
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030172517552657974,
      "loss": 1.2132,
      "theoretical_loss": 3.417870629919524,
      "tokens_seen": 2314993664
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003017051153460381,
      "loss": 1.3696,
      "theoretical_loss": 3.4178628579508086,
      "tokens_seen": 2315059200
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003016850551654965,
      "loss": 1.4448,
      "theoretical_loss": 3.4178550862637045,
      "tokens_seen": 2315124736
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003016649949849549,
      "loss": 1.5263,
      "theoretical_loss": 3.4178473148581947,
      "tokens_seen": 2315190272
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030164493480441324,
      "loss": 1.2947,
      "theoretical_loss": 3.41783954373426,
      "tokens_seen": 2315255808
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003016248746238716,
      "loss": 1.4625,
      "theoretical_loss": 3.4178317728918826,
      "tokens_seen": 2315321344
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030160481444332996,
      "loss": 1.2648,
      "theoretical_loss": 3.417824002331045,
      "tokens_seen": 2315386880
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003015847542627884,
      "loss": 1.3203,
      "theoretical_loss": 3.417816232051728,
      "tokens_seen": 2315452416
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030156469408224674,
      "loss": 1.3841,
      "theoretical_loss": 3.4178084620539138,
      "tokens_seen": 2315517952
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003015446339017051,
      "loss": 1.3804,
      "theoretical_loss": 3.4178006923375843,
      "tokens_seen": 2315583488
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030152457372116347,
      "loss": 1.4287,
      "theoretical_loss": 3.4177929229027217,
      "tokens_seen": 2315649024
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003015045135406219,
      "loss": 1.3297,
      "theoretical_loss": 3.417785153749307,
      "tokens_seen": 2315714560
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030148445336008025,
      "loss": 1.3214,
      "theoretical_loss": 3.4177773848773225,
      "tokens_seen": 2315780096
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003014643931795386,
      "loss": 1.3716,
      "theoretical_loss": 3.4177696162867504,
      "tokens_seen": 2315845632
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030144433299899697,
      "loss": 1.4357,
      "theoretical_loss": 3.4177618479775718,
      "tokens_seen": 2315911168
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030142427281845533,
      "loss": 1.4576,
      "theoretical_loss": 3.4177540799497694,
      "tokens_seen": 2315976704
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030140421263791375,
      "loss": 1.3679,
      "theoretical_loss": 3.4177463122033243,
      "tokens_seen": 2316042240
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003013841524573721,
      "loss": 1.4155,
      "theoretical_loss": 3.4177385447382185,
      "tokens_seen": 2316107776
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003013640922768305,
      "loss": 1.4072,
      "theoretical_loss": 3.4177307775544343,
      "tokens_seen": 2316173312
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030134403209628884,
      "loss": 1.396,
      "theoretical_loss": 3.4177230106519527,
      "tokens_seen": 2316238848
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030132397191574725,
      "loss": 1.3533,
      "theoretical_loss": 3.4177152440307568,
      "tokens_seen": 2316304384
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003013039117352056,
      "loss": 1.4288,
      "theoretical_loss": 3.4177074776908274,
      "tokens_seen": 2316369920
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000301283851554664,
      "loss": 1.3367,
      "theoretical_loss": 3.417699711632147,
      "tokens_seen": 2316435456
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030126379137412234,
      "loss": 1.4412,
      "theoretical_loss": 3.4176919458546964,
      "tokens_seen": 2316500992
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003012437311935807,
      "loss": 1.3643,
      "theoretical_loss": 3.417684180358459,
      "tokens_seen": 2316566528
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003012236710130391,
      "loss": 1.4191,
      "theoretical_loss": 3.417676415143415,
      "tokens_seen": 2316632064
    },
    {
      "epoch": 0.7,
      "objective/train/advantage_avg": 0.49204257130622864,
      "objective/train/docs_used": 843608,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4274728298187256,
      "objective/train/original_loss": 1.4274728298187256,
      "objective/train/theoretical_loss": 3.417668650209548,
      "objective/train/tokens_used": 2337157600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24266274273395538,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504378080368042,
      "objective/train/weighted_lm_loss": 1.4994144439697266,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.417668650209548,
      "tokens_seen": 2316697600
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003012036108324975,
      "loss": 1.5093,
      "theoretical_loss": 3.417668650209548,
      "tokens_seen": 2316697600
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030118355065195584,
      "loss": 1.3052,
      "theoretical_loss": 3.4176608855568382,
      "tokens_seen": 2316763136
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003011634904714142,
      "loss": 1.4824,
      "theoretical_loss": 3.417653121185269,
      "tokens_seen": 2316828672
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003011434302908726,
      "loss": 1.4109,
      "theoretical_loss": 3.4176453570948215,
      "tokens_seen": 2316894208
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000301123370110331,
      "loss": 1.5477,
      "theoretical_loss": 3.4176375932854772,
      "tokens_seen": 2316959744
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030110330992978935,
      "loss": 1.4512,
      "theoretical_loss": 3.4176298297572183,
      "tokens_seen": 2317025280
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003010832497492477,
      "loss": 1.4323,
      "theoretical_loss": 3.4176220665100274,
      "tokens_seen": 2317090816
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030106318956870607,
      "loss": 1.5025,
      "theoretical_loss": 3.417614303543885,
      "tokens_seen": 2317156352
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030104312938816454,
      "loss": 1.3236,
      "theoretical_loss": 3.417606540858774,
      "tokens_seen": 2317221888
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003010230692076229,
      "loss": 1.4066,
      "theoretical_loss": 3.417598778454676,
      "tokens_seen": 2317287424
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030100300902708127,
      "loss": 1.4309,
      "theoretical_loss": 3.4175910163315733,
      "tokens_seen": 2317352960
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030098294884653963,
      "loss": 1.3868,
      "theoretical_loss": 3.417583254489447,
      "tokens_seen": 2317418496
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030096288866599804,
      "loss": 1.2915,
      "theoretical_loss": 3.417575492928279,
      "tokens_seen": 2317484032
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003009428284854564,
      "loss": 1.5186,
      "theoretical_loss": 3.4175677316480515,
      "tokens_seen": 2317549568
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030092276830491477,
      "loss": 1.4664,
      "theoretical_loss": 3.417559970648747,
      "tokens_seen": 2317615104
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030090270812437313,
      "loss": 1.3869,
      "theoretical_loss": 3.4175522099303466,
      "tokens_seen": 2317680640
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030088264794383155,
      "loss": 1.497,
      "theoretical_loss": 3.4175444494928318,
      "tokens_seen": 2317746176
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003008625877632899,
      "loss": 1.4449,
      "theoretical_loss": 3.4175366893361856,
      "tokens_seen": 2317811712
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030084252758274827,
      "loss": 1.3877,
      "theoretical_loss": 3.4175289294603894,
      "tokens_seen": 2317877248
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030082246740220663,
      "loss": 1.4647,
      "theoretical_loss": 3.417521169865425,
      "tokens_seen": 2317942784
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000300802407221665,
      "loss": 1.2753,
      "theoretical_loss": 3.4175134105512743,
      "tokens_seen": 2318008320
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003007823470411234,
      "loss": 1.4899,
      "theoretical_loss": 3.4175056515179194,
      "tokens_seen": 2318073856
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003007622868605818,
      "loss": 1.4616,
      "theoretical_loss": 3.4174978927653417,
      "tokens_seen": 2318139392
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030074222668004014,
      "loss": 1.3973,
      "theoretical_loss": 3.4174901342935238,
      "tokens_seen": 2318204928
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003007221664994985,
      "loss": 1.4956,
      "theoretical_loss": 3.417482376102447,
      "tokens_seen": 2318270464
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003007021063189569,
      "loss": 1.3843,
      "theoretical_loss": 3.417474618192094,
      "tokens_seen": 2318336000
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003006820461384153,
      "loss": 1.3878,
      "theoretical_loss": 3.4174668605624454,
      "tokens_seen": 2318401536
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030066198595787364,
      "loss": 1.3558,
      "theoretical_loss": 3.4174591032134845,
      "tokens_seen": 2318467072
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000300641925777332,
      "loss": 1.3243,
      "theoretical_loss": 3.4174513461451923,
      "tokens_seen": 2318532608
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030062186559679036,
      "loss": 1.3747,
      "theoretical_loss": 3.417443589357551,
      "tokens_seen": 2318598144
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003006018054162488,
      "loss": 1.4669,
      "theoretical_loss": 3.4174358328505425,
      "tokens_seen": 2318663680
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030058174523570714,
      "loss": 1.3951,
      "theoretical_loss": 3.417428076624149,
      "tokens_seen": 2318729216
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003005616850551655,
      "loss": 1.4042,
      "theoretical_loss": 3.417420320678352,
      "tokens_seen": 2318794752
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030054162487462387,
      "loss": 1.3389,
      "theoretical_loss": 3.4174125650131337,
      "tokens_seen": 2318860288
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003005215646940823,
      "loss": 1.3987,
      "theoretical_loss": 3.4174048096284757,
      "tokens_seen": 2318925824
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030050150451354065,
      "loss": 1.3578,
      "theoretical_loss": 3.4173970545243604,
      "tokens_seen": 2318991360
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000300481444332999,
      "loss": 1.4301,
      "theoretical_loss": 3.417389299700769,
      "tokens_seen": 2319056896
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030046138415245737,
      "loss": 1.3834,
      "theoretical_loss": 3.417381545157684,
      "tokens_seen": 2319122432
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030044132397191573,
      "loss": 1.4161,
      "theoretical_loss": 3.4173737908950876,
      "tokens_seen": 2319187968
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030042126379137415,
      "loss": 1.5378,
      "theoretical_loss": 3.417366036912961,
      "tokens_seen": 2319253504
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003004012036108325,
      "loss": 1.4179,
      "theoretical_loss": 3.417358283211286,
      "tokens_seen": 2319319040
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003003811434302909,
      "loss": 1.3591,
      "theoretical_loss": 3.4173505297900455,
      "tokens_seen": 2319384576
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030036108324974924,
      "loss": 1.2552,
      "theoretical_loss": 3.417342776649221,
      "tokens_seen": 2319450112
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030034102306920765,
      "loss": 1.3755,
      "theoretical_loss": 3.417335023788794,
      "tokens_seen": 2319515648
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000300320962888666,
      "loss": 1.4092,
      "theoretical_loss": 3.4173272712087472,
      "tokens_seen": 2319581184
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003003009027081244,
      "loss": 1.4512,
      "theoretical_loss": 3.417319518909062,
      "tokens_seen": 2319646720
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030028084252758274,
      "loss": 1.5277,
      "theoretical_loss": 3.4173117668897204,
      "tokens_seen": 2319712256
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003002607823470411,
      "loss": 1.2947,
      "theoretical_loss": 3.4173040151507044,
      "tokens_seen": 2319777792
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003002407221664995,
      "loss": 1.3472,
      "theoretical_loss": 3.4172962636919957,
      "tokens_seen": 2319843328
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003002206619859579,
      "loss": 1.4965,
      "theoretical_loss": 3.417288512513577,
      "tokens_seen": 2319908864
    },
    {
      "epoch": 0.7,
      "objective/train/advantage_avg": 0.4899619519710541,
      "objective/train/docs_used": 844817,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1608692407608032,
      "objective/train/original_loss": 1.1608688831329346,
      "objective/train/theoretical_loss": 3.4172807616154293,
      "objective/train/tokens_used": 2340434400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24098005890846252,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502210855484009,
      "objective/train/weighted_lm_loss": 1.2192366123199463,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.8607079386711121,
      "theoretical_loss": 3.4172807616154293,
      "tokens_seen": 2319974400
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030020060180541624,
      "loss": 1.2454,
      "theoretical_loss": 3.4172807616154293,
      "tokens_seen": 2319974400
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003001805416248746,
      "loss": 1.5541,
      "theoretical_loss": 3.417273010997535,
      "tokens_seen": 2320039936
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000300160481444333,
      "loss": 1.3994,
      "theoretical_loss": 3.4172652606598763,
      "tokens_seen": 2320105472
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003001404212637914,
      "loss": 1.2856,
      "theoretical_loss": 3.417257510602435,
      "tokens_seen": 2320171008
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030012036108324975,
      "loss": 1.3858,
      "theoretical_loss": 3.417249760825193,
      "tokens_seen": 2320236544
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003001003009027081,
      "loss": 1.2893,
      "theoretical_loss": 3.417242011328132,
      "tokens_seen": 2320302080
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030008024072216647,
      "loss": 1.324,
      "theoretical_loss": 3.417234262111234,
      "tokens_seen": 2320367616
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003000601805416249,
      "loss": 1.3969,
      "theoretical_loss": 3.4172265131744815,
      "tokens_seen": 2320433152
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00030004012036108325,
      "loss": 1.3492,
      "theoretical_loss": 3.417218764517856,
      "tokens_seen": 2320498688
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003000200601805416,
      "loss": 1.3949,
      "theoretical_loss": 3.4172110161413394,
      "tokens_seen": 2320564224
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0003,
      "loss": 1.5671,
      "theoretical_loss": 3.4172032680449136,
      "tokens_seen": 2320629760
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002999799398194584,
      "loss": 1.4069,
      "theoretical_loss": 3.417195520228561,
      "tokens_seen": 2320695296
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029995987963891675,
      "loss": 1.3529,
      "theoretical_loss": 3.4171877726922633,
      "tokens_seen": 2320760832
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002999398194583751,
      "loss": 1.3301,
      "theoretical_loss": 3.4171800254360027,
      "tokens_seen": 2320826368
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002999197592778335,
      "loss": 1.441,
      "theoretical_loss": 3.417172278459761,
      "tokens_seen": 2320891904
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002998996990972919,
      "loss": 1.3606,
      "theoretical_loss": 3.41716453176352,
      "tokens_seen": 2320957440
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029987963891675026,
      "loss": 1.4163,
      "theoretical_loss": 3.417156785347262,
      "tokens_seen": 2321022976
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002998595787362086,
      "loss": 1.3906,
      "theoretical_loss": 3.4171490392109685,
      "tokens_seen": 2321088512
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000299839518555667,
      "loss": 1.4302,
      "theoretical_loss": 3.4171412933546224,
      "tokens_seen": 2321154048
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029981945837512534,
      "loss": 1.4378,
      "theoretical_loss": 3.4171335477782043,
      "tokens_seen": 2321219584
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029979939819458376,
      "loss": 1.3013,
      "theoretical_loss": 3.4171258024816975,
      "tokens_seen": 2321285120
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002997793380140421,
      "loss": 1.3932,
      "theoretical_loss": 3.417118057465083,
      "tokens_seen": 2321350656
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002997592778335005,
      "loss": 1.3605,
      "theoretical_loss": 3.4171103127283433,
      "tokens_seen": 2321416192
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029973921765295885,
      "loss": 1.3563,
      "theoretical_loss": 3.4171025682714604,
      "tokens_seen": 2321481728
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029971915747241726,
      "loss": 1.5015,
      "theoretical_loss": 3.4170948240944163,
      "tokens_seen": 2321547264
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002996990972918756,
      "loss": 1.3669,
      "theoretical_loss": 3.417087080197193,
      "tokens_seen": 2321612800
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000299679037111334,
      "loss": 1.4195,
      "theoretical_loss": 3.417079336579772,
      "tokens_seen": 2321678336
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029965897693079235,
      "loss": 1.3142,
      "theoretical_loss": 3.4170715932421354,
      "tokens_seen": 2321743872
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002996389167502507,
      "loss": 1.4265,
      "theoretical_loss": 3.4170638501842663,
      "tokens_seen": 2321809408
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029961885656970913,
      "loss": 1.3105,
      "theoretical_loss": 3.417056107406145,
      "tokens_seen": 2321874944
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002995987963891675,
      "loss": 1.371,
      "theoretical_loss": 3.417048364907755,
      "tokens_seen": 2321940480
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029957873620862585,
      "loss": 1.4731,
      "theoretical_loss": 3.4170406226890773,
      "tokens_seen": 2322006016
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002995586760280842,
      "loss": 1.4471,
      "theoretical_loss": 3.417032880750094,
      "tokens_seen": 2322071552
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002995386158475427,
      "loss": 1.3434,
      "theoretical_loss": 3.417025139090788,
      "tokens_seen": 2322137088
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029951855566700105,
      "loss": 1.2583,
      "theoretical_loss": 3.4170173977111404,
      "tokens_seen": 2322202624
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002994984954864594,
      "loss": 1.4017,
      "theoretical_loss": 3.417009656611133,
      "tokens_seen": 2322268160
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029947843530591777,
      "loss": 1.4503,
      "theoretical_loss": 3.417001915790749,
      "tokens_seen": 2322333696
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029945837512537613,
      "loss": 1.4048,
      "theoretical_loss": 3.416994175249969,
      "tokens_seen": 2322399232
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029943831494483455,
      "loss": 1.3427,
      "theoretical_loss": 3.416986434988776,
      "tokens_seen": 2322464768
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002994182547642929,
      "loss": 1.312,
      "theoretical_loss": 3.4169786950071517,
      "tokens_seen": 2322530304
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002993981945837513,
      "loss": 1.301,
      "theoretical_loss": 3.416970955305078,
      "tokens_seen": 2322595840
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029937813440320964,
      "loss": 1.5134,
      "theoretical_loss": 3.416963215882537,
      "tokens_seen": 2322661376
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029935807422266805,
      "loss": 1.3316,
      "theoretical_loss": 3.416955476739511,
      "tokens_seen": 2322726912
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002993380140421264,
      "loss": 1.3684,
      "theoretical_loss": 3.4169477378759816,
      "tokens_seen": 2322792448
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002993179538615848,
      "loss": 1.3934,
      "theoretical_loss": 3.416939999291931,
      "tokens_seen": 2322857984
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029929789368104314,
      "loss": 1.4936,
      "theoretical_loss": 3.416932260987341,
      "tokens_seen": 2322923520
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002992778335005015,
      "loss": 1.3913,
      "theoretical_loss": 3.4169245229621943,
      "tokens_seen": 2322989056
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002992577733199599,
      "loss": 1.4097,
      "theoretical_loss": 3.4169167852164724,
      "tokens_seen": 2323054592
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002992377131394183,
      "loss": 1.3682,
      "theoretical_loss": 3.416909047750157,
      "tokens_seen": 2323120128
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029921765295887664,
      "loss": 1.3338,
      "theoretical_loss": 3.4169013105632304,
      "tokens_seen": 2323185664
    },
    {
      "epoch": 0.7,
      "objective/train/advantage_avg": 0.4909175932407379,
      "objective/train/docs_used": 845757,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2433336973190308,
      "objective/train/original_loss": 1.2433335781097412,
      "objective/train/theoretical_loss": 3.4168935736556754,
      "objective/train/tokens_used": 2343711200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24156539142131805,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503196716308594,
      "objective/train/weighted_lm_loss": 1.3057271242141724,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.4168935736556754,
      "tokens_seen": 2323251200
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000299197592778335,
      "loss": 1.4005,
      "theoretical_loss": 3.4168935736556754,
      "tokens_seen": 2323251200
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002991775325977934,
      "loss": 1.3658,
      "theoretical_loss": 3.416885837027473,
      "tokens_seen": 2323316736
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002991574724172518,
      "loss": 1.488,
      "theoretical_loss": 3.4168781006786055,
      "tokens_seen": 2323382272
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029913741223671015,
      "loss": 1.4166,
      "theoretical_loss": 3.4168703646090552,
      "tokens_seen": 2323447808
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002991173520561685,
      "loss": 1.4599,
      "theoretical_loss": 3.4168626288188038,
      "tokens_seen": 2323513344
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029909729187562687,
      "loss": 1.3593,
      "theoretical_loss": 3.4168548933078338,
      "tokens_seen": 2323578880
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002990772316950853,
      "loss": 1.4095,
      "theoretical_loss": 3.4168471580761266,
      "tokens_seen": 2323644416
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029905717151454365,
      "loss": 1.3453,
      "theoretical_loss": 3.4168394231236645,
      "tokens_seen": 2323709952
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000299037111334002,
      "loss": 1.4596,
      "theoretical_loss": 3.41683168845043,
      "tokens_seen": 2323775488
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002990170511534604,
      "loss": 1.292,
      "theoretical_loss": 3.416823954056405,
      "tokens_seen": 2323841024
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002989969909729188,
      "loss": 1.4743,
      "theoretical_loss": 3.416816219941571,
      "tokens_seen": 2323906560
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029897693079237715,
      "loss": 1.3486,
      "theoretical_loss": 3.4168084861059107,
      "tokens_seen": 2323972096
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002989568706118355,
      "loss": 1.3492,
      "theoretical_loss": 3.4168007525494053,
      "tokens_seen": 2324037632
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002989368104312939,
      "loss": 1.3661,
      "theoretical_loss": 3.4167930192720375,
      "tokens_seen": 2324103168
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002989167502507523,
      "loss": 1.4032,
      "theoretical_loss": 3.4167852862737895,
      "tokens_seen": 2324168704
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029889669007021066,
      "loss": 1.3981,
      "theoretical_loss": 3.416777553554643,
      "tokens_seen": 2324234240
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000298876629889669,
      "loss": 1.3772,
      "theoretical_loss": 3.4167698211145803,
      "tokens_seen": 2324299776
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002988565697091274,
      "loss": 1.4037,
      "theoretical_loss": 3.4167620889535835,
      "tokens_seen": 2324365312
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029883650952858574,
      "loss": 1.4478,
      "theoretical_loss": 3.4167543570716337,
      "tokens_seen": 2324430848
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029881644934804416,
      "loss": 1.3927,
      "theoretical_loss": 3.4167466254687144,
      "tokens_seen": 2324496384
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002987963891675025,
      "loss": 1.4033,
      "theoretical_loss": 3.4167388941448067,
      "tokens_seen": 2324561920
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002987763289869609,
      "loss": 1.2876,
      "theoretical_loss": 3.416731163099893,
      "tokens_seen": 2324627456
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029875626880641925,
      "loss": 1.3825,
      "theoretical_loss": 3.416723432333956,
      "tokens_seen": 2324692992
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029873620862587766,
      "loss": 1.3683,
      "theoretical_loss": 3.416715701846976,
      "tokens_seen": 2324758528
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000298716148445336,
      "loss": 1.403,
      "theoretical_loss": 3.416707971638937,
      "tokens_seen": 2324824064
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002986960882647944,
      "loss": 1.3104,
      "theoretical_loss": 3.41670024170982,
      "tokens_seen": 2324889600
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029867602808425275,
      "loss": 1.4087,
      "theoretical_loss": 3.416692512059607,
      "tokens_seen": 2324955136
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002986559679037111,
      "loss": 1.2915,
      "theoretical_loss": 3.4166847826882805,
      "tokens_seen": 2325020672
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029863590772316953,
      "loss": 1.4087,
      "theoretical_loss": 3.416677053595823,
      "tokens_seen": 2325086208
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002986158475426279,
      "loss": 1.2594,
      "theoretical_loss": 3.4166693247822155,
      "tokens_seen": 2325151744
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029859578736208625,
      "loss": 1.446,
      "theoretical_loss": 3.416661596247441,
      "tokens_seen": 2325217280
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002985757271815446,
      "loss": 1.3538,
      "theoretical_loss": 3.416653867991481,
      "tokens_seen": 2325282816
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029855566700100303,
      "loss": 1.4706,
      "theoretical_loss": 3.4166461400143175,
      "tokens_seen": 2325348352
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002985356068204614,
      "loss": 1.5014,
      "theoretical_loss": 3.4166384123159332,
      "tokens_seen": 2325413888
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029851554663991976,
      "loss": 1.4201,
      "theoretical_loss": 3.41663068489631,
      "tokens_seen": 2325479424
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002984954864593781,
      "loss": 1.2742,
      "theoretical_loss": 3.41662295775543,
      "tokens_seen": 2325544960
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002984754262788365,
      "loss": 1.5014,
      "theoretical_loss": 3.416615230893275,
      "tokens_seen": 2325610496
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002984553660982949,
      "loss": 1.2972,
      "theoretical_loss": 3.416607504309827,
      "tokens_seen": 2325676032
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029843530591775326,
      "loss": 1.3562,
      "theoretical_loss": 3.4165997780050685,
      "tokens_seen": 2325741568
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002984152457372116,
      "loss": 1.3749,
      "theoretical_loss": 3.416592051978981,
      "tokens_seen": 2325807104
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029839518555667,
      "loss": 1.3446,
      "theoretical_loss": 3.416584326231548,
      "tokens_seen": 2325872640
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002983751253761284,
      "loss": 1.4097,
      "theoretical_loss": 3.41657660076275,
      "tokens_seen": 2325938176
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029835506519558676,
      "loss": 1.4584,
      "theoretical_loss": 3.41656887557257,
      "tokens_seen": 2326003712
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002983350050150451,
      "loss": 1.4134,
      "theoretical_loss": 3.416561150660989,
      "tokens_seen": 2326069248
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002983149448345035,
      "loss": 1.3469,
      "theoretical_loss": 3.416553426027991,
      "tokens_seen": 2326134784
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029829488465396185,
      "loss": 1.4809,
      "theoretical_loss": 3.4165457016735563,
      "tokens_seen": 2326200320
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029827482447342026,
      "loss": 1.4343,
      "theoretical_loss": 3.416537977597668,
      "tokens_seen": 2326265856
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002982547642928786,
      "loss": 1.4259,
      "theoretical_loss": 3.416530253800308,
      "tokens_seen": 2326331392
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.000298234704112337,
      "loss": 1.3631,
      "theoretical_loss": 3.4165225302814584,
      "tokens_seen": 2326396928
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00029821464393179535,
      "loss": 1.3682,
      "theoretical_loss": 3.4165148070411013,
      "tokens_seen": 2326462464
    },
    {
      "epoch": 0.7,
      "objective/train/advantage_avg": 0.4948805868625641,
      "objective/train/docs_used": 847012,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3427120447158813,
      "objective/train/original_loss": 1.342712163925171,
      "objective/train/theoretical_loss": 3.4165070840792184,
      "objective/train/tokens_used": 2346988000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24513240158557892,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050734281539917,
      "objective/train/weighted_lm_loss": 1.410875678062439,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4165070840792184,
      "tokens_seen": 2326528000
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029819458375125377,
      "loss": 1.4239,
      "theoretical_loss": 3.4165070840792184,
      "tokens_seen": 2326528000
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029817452357071213,
      "loss": 1.3888,
      "theoretical_loss": 3.416499361395793,
      "tokens_seen": 2326593536
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002981544633901705,
      "loss": 1.3825,
      "theoretical_loss": 3.4164916389908058,
      "tokens_seen": 2326659072
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029813440320962885,
      "loss": 1.3307,
      "theoretical_loss": 3.4164839168642396,
      "tokens_seen": 2326724608
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002981143430290872,
      "loss": 1.3926,
      "theoretical_loss": 3.4164761950160765,
      "tokens_seen": 2326790144
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029809428284854563,
      "loss": 1.4471,
      "theoretical_loss": 3.4164684734462987,
      "tokens_seen": 2326855680
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000298074222668004,
      "loss": 1.2961,
      "theoretical_loss": 3.416460752154888,
      "tokens_seen": 2326921216
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029805416248746236,
      "loss": 1.3614,
      "theoretical_loss": 3.4164530311418266,
      "tokens_seen": 2326986752
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002980341023069208,
      "loss": 1.3456,
      "theoretical_loss": 3.4164453104070973,
      "tokens_seen": 2327052288
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002980140421263792,
      "loss": 1.3336,
      "theoretical_loss": 3.4164375899506814,
      "tokens_seen": 2327117824
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029799398194583755,
      "loss": 1.2872,
      "theoretical_loss": 3.4164298697725615,
      "tokens_seen": 2327183360
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002979739217652959,
      "loss": 1.3895,
      "theoretical_loss": 3.4164221498727194,
      "tokens_seen": 2327248896
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002979538615847543,
      "loss": 1.3267,
      "theoretical_loss": 3.416414430251137,
      "tokens_seen": 2327314432
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002979338014042127,
      "loss": 1.4254,
      "theoretical_loss": 3.4164067109077974,
      "tokens_seen": 2327379968
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029791374122367106,
      "loss": 1.3666,
      "theoretical_loss": 3.416398991842682,
      "tokens_seen": 2327445504
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002978936810431294,
      "loss": 1.4144,
      "theoretical_loss": 3.4163912730557726,
      "tokens_seen": 2327511040
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002978736208625878,
      "loss": 1.3826,
      "theoretical_loss": 3.4163835545470524,
      "tokens_seen": 2327576576
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029785356068204614,
      "loss": 1.2123,
      "theoretical_loss": 3.4163758363165027,
      "tokens_seen": 2327642112
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029783350050150456,
      "loss": 1.3047,
      "theoretical_loss": 3.4163681183641064,
      "tokens_seen": 2327707648
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002978134403209629,
      "loss": 1.3338,
      "theoretical_loss": 3.4163604006898445,
      "tokens_seen": 2327773184
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002977933801404213,
      "loss": 1.2278,
      "theoretical_loss": 3.4163526832937,
      "tokens_seen": 2327838720
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029777331995987965,
      "loss": 1.4212,
      "theoretical_loss": 3.416344966175655,
      "tokens_seen": 2327904256
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029775325977933806,
      "loss": 1.2537,
      "theoretical_loss": 3.416337249335691,
      "tokens_seen": 2327969792
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002977331995987964,
      "loss": 1.5032,
      "theoretical_loss": 3.4163295327737915,
      "tokens_seen": 2328035328
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002977131394182548,
      "loss": 1.3741,
      "theoretical_loss": 3.4163218164899374,
      "tokens_seen": 2328100864
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029769307923771315,
      "loss": 1.3313,
      "theoretical_loss": 3.416314100484111,
      "tokens_seen": 2328166400
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002976730190571715,
      "loss": 1.4013,
      "theoretical_loss": 3.4163063847562953,
      "tokens_seen": 2328231936
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029765295887662993,
      "loss": 1.392,
      "theoretical_loss": 3.4162986693064714,
      "tokens_seen": 2328297472
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002976328986960883,
      "loss": 1.3647,
      "theoretical_loss": 3.416290954134622,
      "tokens_seen": 2328363008
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029761283851554665,
      "loss": 1.3979,
      "theoretical_loss": 3.4162832392407294,
      "tokens_seen": 2328428544
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000297592778335005,
      "loss": 1.4231,
      "theoretical_loss": 3.416275524624776,
      "tokens_seen": 2328494080
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029757271815446343,
      "loss": 1.3301,
      "theoretical_loss": 3.4162678102867425,
      "tokens_seen": 2328559616
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002975526579739218,
      "loss": 1.4637,
      "theoretical_loss": 3.4162600962266128,
      "tokens_seen": 2328625152
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029753259779338016,
      "loss": 1.3721,
      "theoretical_loss": 3.416252382444368,
      "tokens_seen": 2328690688
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002975125376128385,
      "loss": 1.3626,
      "theoretical_loss": 3.416244668939991,
      "tokens_seen": 2328756224
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002974924774322969,
      "loss": 1.4055,
      "theoretical_loss": 3.416236955713463,
      "tokens_seen": 2328821760
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002974724172517553,
      "loss": 1.4246,
      "theoretical_loss": 3.4162292427647674,
      "tokens_seen": 2328887296
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029745235707121366,
      "loss": 1.3005,
      "theoretical_loss": 3.4162215300938854,
      "tokens_seen": 2328952832
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000297432296890672,
      "loss": 1.3846,
      "theoretical_loss": 3.4162138177007995,
      "tokens_seen": 2329018368
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002974122367101304,
      "loss": 1.5556,
      "theoretical_loss": 3.416206105585492,
      "tokens_seen": 2329083904
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002973921765295888,
      "loss": 1.2795,
      "theoretical_loss": 3.4161983937479454,
      "tokens_seen": 2329149440
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029737211634904716,
      "loss": 1.3106,
      "theoretical_loss": 3.416190682188141,
      "tokens_seen": 2329214976
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002973520561685055,
      "loss": 1.3128,
      "theoretical_loss": 3.4161829709060614,
      "tokens_seen": 2329280512
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002973319959879639,
      "loss": 1.4261,
      "theoretical_loss": 3.416175259901689,
      "tokens_seen": 2329346048
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029731193580742225,
      "loss": 1.3964,
      "theoretical_loss": 3.4161675491750056,
      "tokens_seen": 2329411584
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029729187562688067,
      "loss": 1.4029,
      "theoretical_loss": 3.416159838725994,
      "tokens_seen": 2329477120
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029727181544633903,
      "loss": 1.4027,
      "theoretical_loss": 3.4161521285546357,
      "tokens_seen": 2329542656
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002972517552657974,
      "loss": 1.4296,
      "theoretical_loss": 3.416144418660913,
      "tokens_seen": 2329608192
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029723169508525575,
      "loss": 1.4248,
      "theoretical_loss": 3.416136709044809,
      "tokens_seen": 2329673728
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029721163490471417,
      "loss": 1.3351,
      "theoretical_loss": 3.416128999706305,
      "tokens_seen": 2329739264
    },
    {
      "epoch": 0.71,
      "objective/train/advantage_avg": 0.49287140369415283,
      "objective/train/docs_used": 848291,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2761571407318115,
      "objective/train/original_loss": 1.2761571407318115,
      "objective/train/theoretical_loss": 3.416121290645383,
      "objective/train/tokens_used": 2350264800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24327786266803741,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505239963531494,
      "objective/train/weighted_lm_loss": 1.3405871391296387,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.416121290645383,
      "tokens_seen": 2329804800
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029719157472417253,
      "loss": 1.3664,
      "theoretical_loss": 3.416121290645383,
      "tokens_seen": 2329804800
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002971715145436309,
      "loss": 1.3065,
      "theoretical_loss": 3.4161135818620254,
      "tokens_seen": 2329870336
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029715145436308926,
      "loss": 1.3959,
      "theoretical_loss": 3.416105873356215,
      "tokens_seen": 2329935872
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002971313941825476,
      "loss": 1.355,
      "theoretical_loss": 3.4160981651279334,
      "tokens_seen": 2330001408
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029711133400200603,
      "loss": 1.4345,
      "theoretical_loss": 3.416090457177163,
      "tokens_seen": 2330066944
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002970912738214644,
      "loss": 1.2944,
      "theoretical_loss": 3.416082749503886,
      "tokens_seen": 2330132480
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029707121364092276,
      "loss": 1.4077,
      "theoretical_loss": 3.4160750421080843,
      "tokens_seen": 2330198016
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002970511534603811,
      "loss": 1.3187,
      "theoretical_loss": 3.416067334989741,
      "tokens_seen": 2330263552
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029703109327983954,
      "loss": 1.3345,
      "theoretical_loss": 3.4160596281488367,
      "tokens_seen": 2330329088
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002970110330992979,
      "loss": 1.3346,
      "theoretical_loss": 3.416051921585355,
      "tokens_seen": 2330394624
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029699097291875626,
      "loss": 1.466,
      "theoretical_loss": 3.4160442152992783,
      "tokens_seen": 2330460160
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002969709127382146,
      "loss": 1.4764,
      "theoretical_loss": 3.4160365092905876,
      "tokens_seen": 2330525696
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000296950852557673,
      "loss": 1.3107,
      "theoretical_loss": 3.416028803559266,
      "tokens_seen": 2330591232
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002969307923771314,
      "loss": 1.3666,
      "theoretical_loss": 3.4160210981052956,
      "tokens_seen": 2330656768
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029691073219658976,
      "loss": 1.4092,
      "theoretical_loss": 3.4160133929286585,
      "tokens_seen": 2330722304
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002968906720160481,
      "loss": 1.4409,
      "theoretical_loss": 3.416005688029337,
      "tokens_seen": 2330787840
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002968706118355065,
      "loss": 1.4321,
      "theoretical_loss": 3.415997983407313,
      "tokens_seen": 2330853376
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002968505516549649,
      "loss": 1.2887,
      "theoretical_loss": 3.4159902790625685,
      "tokens_seen": 2330918912
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029683049147442327,
      "loss": 1.387,
      "theoretical_loss": 3.4159825749950867,
      "tokens_seen": 2330984448
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029681043129388163,
      "loss": 1.4107,
      "theoretical_loss": 3.415974871204849,
      "tokens_seen": 2331049984
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029679037111334,
      "loss": 1.4775,
      "theoretical_loss": 3.4159671676918384,
      "tokens_seen": 2331115520
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002967703109327984,
      "loss": 1.3716,
      "theoretical_loss": 3.4159594644560363,
      "tokens_seen": 2331181056
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029675025075225677,
      "loss": 1.4092,
      "theoretical_loss": 3.4159517614974257,
      "tokens_seen": 2331246592
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029673019057171513,
      "loss": 1.4419,
      "theoretical_loss": 3.4159440588159877,
      "tokens_seen": 2331312128
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002967101303911735,
      "loss": 1.3933,
      "theoretical_loss": 3.415936356411706,
      "tokens_seen": 2331377664
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029669007021063186,
      "loss": 1.2937,
      "theoretical_loss": 3.4159286542845617,
      "tokens_seen": 2331443200
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002966700100300903,
      "loss": 1.3888,
      "theoretical_loss": 3.4159209524345373,
      "tokens_seen": 2331508736
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029664994984954864,
      "loss": 1.3479,
      "theoretical_loss": 3.4159132508616157,
      "tokens_seen": 2331574272
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000296629889669007,
      "loss": 1.4361,
      "theoretical_loss": 3.415905549565778,
      "tokens_seen": 2331639808
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029660982948846536,
      "loss": 1.4162,
      "theoretical_loss": 3.4158978485470075,
      "tokens_seen": 2331705344
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002965897693079238,
      "loss": 1.3423,
      "theoretical_loss": 3.4158901478052863,
      "tokens_seen": 2331770880
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029656970912738214,
      "loss": 1.4074,
      "theoretical_loss": 3.4158824473405955,
      "tokens_seen": 2331836416
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002965496489468405,
      "loss": 1.4053,
      "theoretical_loss": 3.4158747471529187,
      "tokens_seen": 2331901952
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002965295887662989,
      "loss": 1.4054,
      "theoretical_loss": 3.4158670472422377,
      "tokens_seen": 2331967488
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002965095285857573,
      "loss": 1.5285,
      "theoretical_loss": 3.4158593476085346,
      "tokens_seen": 2332033024
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002964894684052157,
      "loss": 1.406,
      "theoretical_loss": 3.4158516482517918,
      "tokens_seen": 2332098560
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029646940822467406,
      "loss": 1.3639,
      "theoretical_loss": 3.4158439491719914,
      "tokens_seen": 2332164096
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002964493480441324,
      "loss": 1.3163,
      "theoretical_loss": 3.415836250369116,
      "tokens_seen": 2332229632
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002964292878635908,
      "loss": 1.4165,
      "theoretical_loss": 3.4158285518431475,
      "tokens_seen": 2332295168
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002964092276830492,
      "loss": 1.4299,
      "theoretical_loss": 3.4158208535940684,
      "tokens_seen": 2332360704
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029638916750250756,
      "loss": 1.4551,
      "theoretical_loss": 3.4158131556218607,
      "tokens_seen": 2332426240
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002963691073219659,
      "loss": 1.3987,
      "theoretical_loss": 3.415805457926507,
      "tokens_seen": 2332491776
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002963490471414243,
      "loss": 1.4089,
      "theoretical_loss": 3.4157977605079894,
      "tokens_seen": 2332557312
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029632898696088265,
      "loss": 1.3313,
      "theoretical_loss": 3.41579006336629,
      "tokens_seen": 2332622848
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029630892678034107,
      "loss": 1.2496,
      "theoretical_loss": 3.415782366501391,
      "tokens_seen": 2332688384
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029628886659979943,
      "loss": 1.3005,
      "theoretical_loss": 3.415774669913275,
      "tokens_seen": 2332753920
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002962688064192578,
      "loss": 1.3489,
      "theoretical_loss": 3.415766973601924,
      "tokens_seen": 2332819456
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029624874623871615,
      "loss": 1.3154,
      "theoretical_loss": 3.4157592775673207,
      "tokens_seen": 2332884992
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029622868605817457,
      "loss": 1.3717,
      "theoretical_loss": 3.415751581809447,
      "tokens_seen": 2332950528
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029620862587763293,
      "loss": 1.412,
      "theoretical_loss": 3.4157438863282854,
      "tokens_seen": 2333016064
    },
    {
      "epoch": 0.71,
      "objective/train/advantage_avg": 0.4939133822917938,
      "objective/train/docs_used": 849464,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3251676559448242,
      "objective/train/original_loss": 1.3251676559448242,
      "objective/train/theoretical_loss": 3.415736191123818,
      "objective/train/tokens_used": 2353541600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2443237602710724,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506335496902466,
      "objective/train/weighted_lm_loss": 1.3922621011734009,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.415736191123818,
      "tokens_seen": 2333081600
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002961885656970913,
      "loss": 1.4274,
      "theoretical_loss": 3.415736191123818,
      "tokens_seen": 2333081600
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029616850551654966,
      "loss": 1.4842,
      "theoretical_loss": 3.4157284961960275,
      "tokens_seen": 2333147136
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000296148445336008,
      "loss": 1.4234,
      "theoretical_loss": 3.415720801544895,
      "tokens_seen": 2333212672
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029612838515546643,
      "loss": 1.3687,
      "theoretical_loss": 3.4157131071704043,
      "tokens_seen": 2333278208
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002961083249749248,
      "loss": 1.3123,
      "theoretical_loss": 3.4157054130725366,
      "tokens_seen": 2333343744
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029608826479438316,
      "loss": 1.429,
      "theoretical_loss": 3.4156977192512747,
      "tokens_seen": 2333409280
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002960682046138415,
      "loss": 1.4555,
      "theoretical_loss": 3.415690025706601,
      "tokens_seen": 2333474816
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029604814443329994,
      "loss": 1.4469,
      "theoretical_loss": 3.4156823324384975,
      "tokens_seen": 2333540352
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002960280842527583,
      "loss": 1.3562,
      "theoretical_loss": 3.415674639446946,
      "tokens_seen": 2333605888
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029600802407221666,
      "loss": 1.4085,
      "theoretical_loss": 3.4156669467319296,
      "tokens_seen": 2333671424
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000295987963891675,
      "loss": 1.4169,
      "theoretical_loss": 3.4156592542934305,
      "tokens_seen": 2333736960
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002959679037111334,
      "loss": 1.3723,
      "theoretical_loss": 3.4156515621314307,
      "tokens_seen": 2333802496
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002959478435305918,
      "loss": 1.3908,
      "theoretical_loss": 3.415643870245913,
      "tokens_seen": 2333868032
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029592778335005016,
      "loss": 1.477,
      "theoretical_loss": 3.4156361786368583,
      "tokens_seen": 2333933568
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029590772316950853,
      "loss": 1.4239,
      "theoretical_loss": 3.415628487304251,
      "tokens_seen": 2333999104
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002958876629889669,
      "loss": 1.3312,
      "theoretical_loss": 3.4156207962480716,
      "tokens_seen": 2334064640
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002958676028084253,
      "loss": 1.4248,
      "theoretical_loss": 3.4156131054683034,
      "tokens_seen": 2334130176
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029584754262788367,
      "loss": 1.4813,
      "theoretical_loss": 3.4156054149649284,
      "tokens_seen": 2334195712
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029582748244734203,
      "loss": 1.3009,
      "theoretical_loss": 3.415597724737929,
      "tokens_seen": 2334261248
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002958074222668004,
      "loss": 1.4058,
      "theoretical_loss": 3.4155900347872876,
      "tokens_seen": 2334326784
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002957873620862588,
      "loss": 1.4346,
      "theoretical_loss": 3.415582345112986,
      "tokens_seen": 2334392320
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029576730190571717,
      "loss": 1.477,
      "theoretical_loss": 3.415574655715007,
      "tokens_seen": 2334457856
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029574724172517553,
      "loss": 1.4928,
      "theoretical_loss": 3.415566966593333,
      "tokens_seen": 2334523392
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002957271815446339,
      "loss": 1.565,
      "theoretical_loss": 3.415559277747946,
      "tokens_seen": 2334588928
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029570712136409226,
      "loss": 1.3951,
      "theoretical_loss": 3.415551589178828,
      "tokens_seen": 2334654464
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002956870611835507,
      "loss": 1.3469,
      "theoretical_loss": 3.4155439008859623,
      "tokens_seen": 2334720000
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029566700100300904,
      "loss": 1.3959,
      "theoretical_loss": 3.4155362128693305,
      "tokens_seen": 2334785536
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002956469408224674,
      "loss": 1.4517,
      "theoretical_loss": 3.4155285251289147,
      "tokens_seen": 2334851072
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029562688064192576,
      "loss": 1.4226,
      "theoretical_loss": 3.415520837664698,
      "tokens_seen": 2334916608
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002956068204613842,
      "loss": 1.4173,
      "theoretical_loss": 3.415513150476662,
      "tokens_seen": 2334982144
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029558676028084254,
      "loss": 1.2561,
      "theoretical_loss": 3.4155054635647897,
      "tokens_seen": 2335047680
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002955667001003009,
      "loss": 1.409,
      "theoretical_loss": 3.4154977769290626,
      "tokens_seen": 2335113216
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029554663991975926,
      "loss": 1.3728,
      "theoretical_loss": 3.415490090569464,
      "tokens_seen": 2335178752
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002955265797392176,
      "loss": 1.3741,
      "theoretical_loss": 3.4154824044859757,
      "tokens_seen": 2335244288
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029550651955867604,
      "loss": 1.4113,
      "theoretical_loss": 3.41547471867858,
      "tokens_seen": 2335309824
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002954864593781344,
      "loss": 1.376,
      "theoretical_loss": 3.415467033147259,
      "tokens_seen": 2335375360
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029546639919759277,
      "loss": 1.5199,
      "theoretical_loss": 3.415459347891996,
      "tokens_seen": 2335440896
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029544633901705113,
      "loss": 1.3284,
      "theoretical_loss": 3.415451662912772,
      "tokens_seen": 2335506432
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029542627883650955,
      "loss": 1.2676,
      "theoretical_loss": 3.4154439782095705,
      "tokens_seen": 2335571968
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002954062186559679,
      "loss": 1.4166,
      "theoretical_loss": 3.4154362937823732,
      "tokens_seen": 2335637504
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029538615847542627,
      "loss": 1.4362,
      "theoretical_loss": 3.4154286096311623,
      "tokens_seen": 2335703040
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029536609829488463,
      "loss": 1.3847,
      "theoretical_loss": 3.415420925755921,
      "tokens_seen": 2335768576
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000295346038114343,
      "loss": 1.303,
      "theoretical_loss": 3.415413242156631,
      "tokens_seen": 2335834112
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002953259779338014,
      "loss": 1.4236,
      "theoretical_loss": 3.415405558833274,
      "tokens_seen": 2335899648
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002953059177532598,
      "loss": 1.2199,
      "theoretical_loss": 3.4153978757858336,
      "tokens_seen": 2335965184
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029528585757271814,
      "loss": 1.3601,
      "theoretical_loss": 3.4153901930142916,
      "tokens_seen": 2336030720
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002952657973921765,
      "loss": 1.4001,
      "theoretical_loss": 3.4153825105186306,
      "tokens_seen": 2336096256
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002952457372116349,
      "loss": 1.3305,
      "theoretical_loss": 3.415374828298833,
      "tokens_seen": 2336161792
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002952256770310933,
      "loss": 1.3699,
      "theoretical_loss": 3.41536714635488,
      "tokens_seen": 2336227328
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029520561685055164,
      "loss": 1.4949,
      "theoretical_loss": 3.4153594646867558,
      "tokens_seen": 2336292864
    },
    {
      "epoch": 0.71,
      "objective/train/advantage_avg": 0.4961789548397064,
      "objective/train/docs_used": 850768,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.266499400138855,
      "objective/train/original_loss": 1.266499400138855,
      "objective/train/theoretical_loss": 3.415351783294441,
      "objective/train/tokens_used": 2356818400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24644286930561066,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508708953857422,
      "objective/train/weighted_lm_loss": 1.3309162855148315,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.415351783294441,
      "tokens_seen": 2336358400
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029518555667001,
      "loss": 1.4341,
      "theoretical_loss": 3.415351783294441,
      "tokens_seen": 2336358400
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029516549648946836,
      "loss": 1.3854,
      "theoretical_loss": 3.4153441021779196,
      "tokens_seen": 2336423936
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002951454363089268,
      "loss": 1.3447,
      "theoretical_loss": 3.4153364213371726,
      "tokens_seen": 2336489472
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029512537612838514,
      "loss": 1.2917,
      "theoretical_loss": 3.415328740772183,
      "tokens_seen": 2336555008
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002951053159478435,
      "loss": 1.4101,
      "theoretical_loss": 3.415321060482933,
      "tokens_seen": 2336620544
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029508525576730187,
      "loss": 1.3453,
      "theoretical_loss": 3.4153133804694056,
      "tokens_seen": 2336686080
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002950651955867603,
      "loss": 1.3699,
      "theoretical_loss": 3.415305700731582,
      "tokens_seen": 2336751616
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029504513540621865,
      "loss": 1.4336,
      "theoretical_loss": 3.4152980212694453,
      "tokens_seen": 2336817152
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029502507522567706,
      "loss": 1.388,
      "theoretical_loss": 3.415290342082978,
      "tokens_seen": 2336882688
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002950050150451354,
      "loss": 1.4713,
      "theoretical_loss": 3.415282663172162,
      "tokens_seen": 2336948224
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002949849548645938,
      "loss": 1.3686,
      "theoretical_loss": 3.41527498453698,
      "tokens_seen": 2337013760
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002949648946840522,
      "loss": 1.3857,
      "theoretical_loss": 3.415267306177414,
      "tokens_seen": 2337079296
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029494483450351057,
      "loss": 1.4352,
      "theoretical_loss": 3.4152596280934473,
      "tokens_seen": 2337144832
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029492477432296893,
      "loss": 1.449,
      "theoretical_loss": 3.415251950285061,
      "tokens_seen": 2337210368
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002949047141424273,
      "loss": 1.3993,
      "theoretical_loss": 3.4152442727522385,
      "tokens_seen": 2337275904
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002948846539618857,
      "loss": 1.4564,
      "theoretical_loss": 3.415236595494962,
      "tokens_seen": 2337341440
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029486459378134407,
      "loss": 1.4095,
      "theoretical_loss": 3.415228918513213,
      "tokens_seen": 2337406976
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029484453360080243,
      "loss": 1.3644,
      "theoretical_loss": 3.415221241806975,
      "tokens_seen": 2337472512
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002948244734202608,
      "loss": 1.5359,
      "theoretical_loss": 3.41521356537623,
      "tokens_seen": 2337538048
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002948044132397192,
      "loss": 1.3672,
      "theoretical_loss": 3.4152058892209602,
      "tokens_seen": 2337603584
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029478435305917757,
      "loss": 1.4126,
      "theoretical_loss": 3.4151982133411485,
      "tokens_seen": 2337669120
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029476429287863593,
      "loss": 1.3932,
      "theoretical_loss": 3.4151905377367764,
      "tokens_seen": 2337734656
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002947442326980943,
      "loss": 1.3692,
      "theoretical_loss": 3.415182862407827,
      "tokens_seen": 2337800192
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029472417251755266,
      "loss": 1.2625,
      "theoretical_loss": 3.415175187354283,
      "tokens_seen": 2337865728
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002947041123370111,
      "loss": 1.5202,
      "theoretical_loss": 3.415167512576126,
      "tokens_seen": 2337931264
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029468405215646944,
      "loss": 1.3899,
      "theoretical_loss": 3.415159838073339,
      "tokens_seen": 2337996800
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002946639919759278,
      "loss": 1.2998,
      "theoretical_loss": 3.4151521638459035,
      "tokens_seen": 2338062336
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029464393179538616,
      "loss": 1.5447,
      "theoretical_loss": 3.4151444898938026,
      "tokens_seen": 2338127872
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002946238716148446,
      "loss": 1.5168,
      "theoretical_loss": 3.4151368162170193,
      "tokens_seen": 2338193408
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029460381143430294,
      "loss": 1.3654,
      "theoretical_loss": 3.415129142815535,
      "tokens_seen": 2338258944
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002945837512537613,
      "loss": 1.352,
      "theoretical_loss": 3.4151214696893324,
      "tokens_seen": 2338324480
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029456369107321966,
      "loss": 1.4816,
      "theoretical_loss": 3.415113796838394,
      "tokens_seen": 2338390016
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000294543630892678,
      "loss": 1.3357,
      "theoretical_loss": 3.415106124262702,
      "tokens_seen": 2338455552
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029452357071213644,
      "loss": 1.3332,
      "theoretical_loss": 3.415098451962239,
      "tokens_seen": 2338521088
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002945035105315948,
      "loss": 1.4162,
      "theoretical_loss": 3.4150907799369876,
      "tokens_seen": 2338586624
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029448345035105317,
      "loss": 1.3518,
      "theoretical_loss": 3.41508310818693,
      "tokens_seen": 2338652160
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029446339017051153,
      "loss": 1.3822,
      "theoretical_loss": 3.4150754367120486,
      "tokens_seen": 2338717696
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029444332998996995,
      "loss": 1.4056,
      "theoretical_loss": 3.415067765512326,
      "tokens_seen": 2338783232
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002944232698094283,
      "loss": 1.4607,
      "theoretical_loss": 3.4150600945877443,
      "tokens_seen": 2338848768
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029440320962888667,
      "loss": 1.3962,
      "theoretical_loss": 3.4150524239382865,
      "tokens_seen": 2338914304
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029438314944834503,
      "loss": 1.3861,
      "theoretical_loss": 3.415044753563934,
      "tokens_seen": 2338979840
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002943630892678034,
      "loss": 1.4576,
      "theoretical_loss": 3.41503708346467,
      "tokens_seen": 2339045376
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002943430290872618,
      "loss": 1.4702,
      "theoretical_loss": 3.415029413640477,
      "tokens_seen": 2339110912
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002943229689067202,
      "loss": 1.4302,
      "theoretical_loss": 3.4150217440913373,
      "tokens_seen": 2339176448
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029430290872617854,
      "loss": 1.3684,
      "theoretical_loss": 3.4150140748172326,
      "tokens_seen": 2339241984
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002942828485456369,
      "loss": 1.3773,
      "theoretical_loss": 3.4150064058181466,
      "tokens_seen": 2339307520
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002942627883650953,
      "loss": 1.3821,
      "theoretical_loss": 3.414998737094061,
      "tokens_seen": 2339373056
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002942427281845537,
      "loss": 1.3894,
      "theoretical_loss": 3.414991068644958,
      "tokens_seen": 2339438592
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029422266800401204,
      "loss": 1.4014,
      "theoretical_loss": 3.4149834004708204,
      "tokens_seen": 2339504128
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002942026078234704,
      "loss": 1.3994,
      "theoretical_loss": 3.414975732571631,
      "tokens_seen": 2339569664
    },
    {
      "epoch": 0.71,
      "objective/train/advantage_avg": 0.4898923337459564,
      "objective/train/docs_used": 851815,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3977073431015015,
      "objective/train/original_loss": 1.3977073431015015,
      "objective/train/theoretical_loss": 3.414968064947372,
      "objective/train/tokens_used": 2360095200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24073128402233124,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502128601074219,
      "objective/train/weighted_lm_loss": 1.4678014516830444,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.414968064947372,
      "tokens_seen": 2339635200
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029418254764292876,
      "loss": 1.3665,
      "theoretical_loss": 3.414968064947372,
      "tokens_seen": 2339635200
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002941624874623872,
      "loss": 1.3714,
      "theoretical_loss": 3.414960397598025,
      "tokens_seen": 2339700736
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029414242728184554,
      "loss": 1.3375,
      "theoretical_loss": 3.4149527305235736,
      "tokens_seen": 2339766272
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002941223671013039,
      "loss": 1.3869,
      "theoretical_loss": 3.4149450637239998,
      "tokens_seen": 2339831808
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029410230692076227,
      "loss": 1.3744,
      "theoretical_loss": 3.4149373971992856,
      "tokens_seen": 2339897344
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002940822467402207,
      "loss": 1.4747,
      "theoretical_loss": 3.414929730949414,
      "tokens_seen": 2339962880
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029406218655967905,
      "loss": 1.4208,
      "theoretical_loss": 3.414922064974368,
      "tokens_seen": 2340028416
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002940421263791374,
      "loss": 1.376,
      "theoretical_loss": 3.4149143992741284,
      "tokens_seen": 2340093952
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029402206619859577,
      "loss": 1.3363,
      "theoretical_loss": 3.4149067338486794,
      "tokens_seen": 2340159488
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029400200601805413,
      "loss": 1.4162,
      "theoretical_loss": 3.414899068698002,
      "tokens_seen": 2340225024
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029398194583751255,
      "loss": 1.3606,
      "theoretical_loss": 3.41489140382208,
      "tokens_seen": 2340290560
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002939618856569709,
      "loss": 1.3629,
      "theoretical_loss": 3.414883739220895,
      "tokens_seen": 2340356096
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002939418254764293,
      "loss": 1.3665,
      "theoretical_loss": 3.4148760748944293,
      "tokens_seen": 2340421632
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029392176529588764,
      "loss": 1.3563,
      "theoretical_loss": 3.414868410842666,
      "tokens_seen": 2340487168
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029390170511534605,
      "loss": 1.4443,
      "theoretical_loss": 3.414860747065587,
      "tokens_seen": 2340552704
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002938816449348044,
      "loss": 1.265,
      "theoretical_loss": 3.414853083563176,
      "tokens_seen": 2340618240
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002938615847542628,
      "loss": 1.4096,
      "theoretical_loss": 3.4148454203354133,
      "tokens_seen": 2340683776
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029384152457372114,
      "loss": 1.4036,
      "theoretical_loss": 3.4148377573822835,
      "tokens_seen": 2340749312
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002938214643931795,
      "loss": 1.4202,
      "theoretical_loss": 3.4148300947037677,
      "tokens_seen": 2340814848
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002938014042126379,
      "loss": 1.4343,
      "theoretical_loss": 3.414822432299849,
      "tokens_seen": 2340880384
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002937813440320963,
      "loss": 1.2972,
      "theoretical_loss": 3.4148147701705094,
      "tokens_seen": 2340945920
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029376128385155464,
      "loss": 1.3795,
      "theoretical_loss": 3.414807108315732,
      "tokens_seen": 2341011456
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000293741223671013,
      "loss": 1.4035,
      "theoretical_loss": 3.4147994467354987,
      "tokens_seen": 2341076992
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002937211634904714,
      "loss": 1.3684,
      "theoretical_loss": 3.414791785429792,
      "tokens_seen": 2341142528
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002937011033099298,
      "loss": 1.3834,
      "theoretical_loss": 3.414784124398595,
      "tokens_seen": 2341208064
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029368104312938815,
      "loss": 1.4005,
      "theoretical_loss": 3.41477646364189,
      "tokens_seen": 2341273600
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002936609829488465,
      "loss": 1.456,
      "theoretical_loss": 3.414768803159659,
      "tokens_seen": 2341339136
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029364092276830487,
      "loss": 1.4121,
      "theoretical_loss": 3.4147611429518845,
      "tokens_seen": 2341404672
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002936208625877633,
      "loss": 1.4305,
      "theoretical_loss": 3.4147534830185493,
      "tokens_seen": 2341470208
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029360080240722165,
      "loss": 1.4091,
      "theoretical_loss": 3.414745823359636,
      "tokens_seen": 2341535744
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029358074222668,
      "loss": 1.3981,
      "theoretical_loss": 3.4147381639751266,
      "tokens_seen": 2341601280
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002935606820461384,
      "loss": 1.3972,
      "theoretical_loss": 3.4147305048650045,
      "tokens_seen": 2341666816
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029354062186559684,
      "loss": 1.3758,
      "theoretical_loss": 3.414722846029251,
      "tokens_seen": 2341732352
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002935205616850552,
      "loss": 1.4694,
      "theoretical_loss": 3.4147151874678494,
      "tokens_seen": 2341797888
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029350050150451357,
      "loss": 1.4254,
      "theoretical_loss": 3.414707529180782,
      "tokens_seen": 2341863424
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029348044132397193,
      "loss": 1.3444,
      "theoretical_loss": 3.4146998711680308,
      "tokens_seen": 2341928960
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029346038114343035,
      "loss": 1.5072,
      "theoretical_loss": 3.4146922134295794,
      "tokens_seen": 2341994496
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002934403209628887,
      "loss": 1.5036,
      "theoretical_loss": 3.4146845559654095,
      "tokens_seen": 2342060032
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029342026078234707,
      "loss": 1.3654,
      "theoretical_loss": 3.414676898775504,
      "tokens_seen": 2342125568
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029340020060180543,
      "loss": 1.3912,
      "theoretical_loss": 3.4146692418598446,
      "tokens_seen": 2342191104
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002933801404212638,
      "loss": 1.4163,
      "theoretical_loss": 3.414661585218415,
      "tokens_seen": 2342256640
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002933600802407222,
      "loss": 1.3314,
      "theoretical_loss": 3.4146539288511963,
      "tokens_seen": 2342322176
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002933400200601806,
      "loss": 1.3907,
      "theoretical_loss": 3.4146462727581723,
      "tokens_seen": 2342387712
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029331995987963894,
      "loss": 1.4617,
      "theoretical_loss": 3.414638616939325,
      "tokens_seen": 2342453248
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002932998996990973,
      "loss": 1.4639,
      "theoretical_loss": 3.4146309613946366,
      "tokens_seen": 2342518784
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002932798395185557,
      "loss": 1.3681,
      "theoretical_loss": 3.4146233061240903,
      "tokens_seen": 2342584320
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002932597793380141,
      "loss": 1.3964,
      "theoretical_loss": 3.414615651127668,
      "tokens_seen": 2342649856
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029323971915747244,
      "loss": 1.4392,
      "theoretical_loss": 3.414607996405353,
      "tokens_seen": 2342715392
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002932196589769308,
      "loss": 1.4072,
      "theoretical_loss": 3.4146003419571267,
      "tokens_seen": 2342780928
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029319959879638916,
      "loss": 1.368,
      "theoretical_loss": 3.414592687782972,
      "tokens_seen": 2342846464
    },
    {
      "epoch": 0.71,
      "objective/train/advantage_avg": 0.49274343252182007,
      "objective/train/docs_used": 852909,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1868455410003662,
      "objective/train/original_loss": 1.1868455410003662,
      "objective/train/theoretical_loss": 3.4145850338828723,
      "objective/train/tokens_used": 2363372000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24326954782009125,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050511121749878,
      "objective/train/weighted_lm_loss": 1.2467751502990723,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4145850338828723,
      "tokens_seen": 2342912000
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002931795386158476,
      "loss": 1.3321,
      "theoretical_loss": 3.4145850338828723,
      "tokens_seen": 2342912000
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029315947843530594,
      "loss": 1.4193,
      "theoretical_loss": 3.414577380256809,
      "tokens_seen": 2342977536
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002931394182547643,
      "loss": 1.541,
      "theoretical_loss": 3.414569726904765,
      "tokens_seen": 2343043072
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029311935807422267,
      "loss": 1.3883,
      "theoretical_loss": 3.414562073826723,
      "tokens_seen": 2343108608
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002930992978936811,
      "loss": 1.3647,
      "theoretical_loss": 3.4145544210226655,
      "tokens_seen": 2343174144
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029307923771313945,
      "loss": 1.3994,
      "theoretical_loss": 3.414546768492575,
      "tokens_seen": 2343239680
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002930591775325978,
      "loss": 1.4082,
      "theoretical_loss": 3.414539116236434,
      "tokens_seen": 2343305216
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029303911735205617,
      "loss": 1.4453,
      "theoretical_loss": 3.4145314642542246,
      "tokens_seen": 2343370752
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029301905717151453,
      "loss": 1.4486,
      "theoretical_loss": 3.41452381254593,
      "tokens_seen": 2343436288
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029299899699097295,
      "loss": 1.2807,
      "theoretical_loss": 3.4145161611115324,
      "tokens_seen": 2343501824
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002929789368104313,
      "loss": 1.3074,
      "theoretical_loss": 3.4145085099510144,
      "tokens_seen": 2343567360
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002929588766298897,
      "loss": 1.372,
      "theoretical_loss": 3.4145008590643586,
      "tokens_seen": 2343632896
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029293881644934804,
      "loss": 1.3966,
      "theoretical_loss": 3.4144932084515474,
      "tokens_seen": 2343698432
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029291875626880645,
      "loss": 1.4515,
      "theoretical_loss": 3.4144855581125633,
      "tokens_seen": 2343763968
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002928986960882648,
      "loss": 1.3695,
      "theoretical_loss": 3.4144779080473895,
      "tokens_seen": 2343829504
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002928786359077232,
      "loss": 1.3215,
      "theoretical_loss": 3.4144702582560074,
      "tokens_seen": 2343895040
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029285857572718154,
      "loss": 1.3451,
      "theoretical_loss": 3.4144626087384005,
      "tokens_seen": 2343960576
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002928385155466399,
      "loss": 1.3591,
      "theoretical_loss": 3.414454959494551,
      "tokens_seen": 2344026112
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002928184553660983,
      "loss": 1.3521,
      "theoretical_loss": 3.4144473105244413,
      "tokens_seen": 2344091648
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002927983951855567,
      "loss": 1.2344,
      "theoretical_loss": 3.4144396618280544,
      "tokens_seen": 2344157184
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029277833500501504,
      "loss": 1.4558,
      "theoretical_loss": 3.414432013405372,
      "tokens_seen": 2344222720
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002927582748244734,
      "loss": 1.4028,
      "theoretical_loss": 3.4144243652563775,
      "tokens_seen": 2344288256
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002927382146439318,
      "loss": 1.4233,
      "theoretical_loss": 3.4144167173810533,
      "tokens_seen": 2344353792
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002927181544633902,
      "loss": 1.325,
      "theoretical_loss": 3.4144090697793814,
      "tokens_seen": 2344419328
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029269809428284855,
      "loss": 1.3999,
      "theoretical_loss": 3.4144014224513457,
      "tokens_seen": 2344484864
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002926780341023069,
      "loss": 1.4102,
      "theoretical_loss": 3.4143937753969267,
      "tokens_seen": 2344550400
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029265797392176527,
      "loss": 1.4705,
      "theoretical_loss": 3.414386128616109,
      "tokens_seen": 2344615936
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002926379137412237,
      "loss": 1.4542,
      "theoretical_loss": 3.414378482108874,
      "tokens_seen": 2344681472
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029261785356068205,
      "loss": 1.4459,
      "theoretical_loss": 3.4143708358752045,
      "tokens_seen": 2344747008
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002925977933801404,
      "loss": 1.4045,
      "theoretical_loss": 3.414363189915083,
      "tokens_seen": 2344812544
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002925777331995988,
      "loss": 1.3738,
      "theoretical_loss": 3.414355544228492,
      "tokens_seen": 2344878080
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002925576730190572,
      "loss": 1.4936,
      "theoretical_loss": 3.4143478988154143,
      "tokens_seen": 2344943616
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029253761283851555,
      "loss": 1.4649,
      "theoretical_loss": 3.414340253675833,
      "tokens_seen": 2345009152
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002925175526579739,
      "loss": 1.3958,
      "theoretical_loss": 3.4143326088097297,
      "tokens_seen": 2345074688
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002924974924774323,
      "loss": 1.4319,
      "theoretical_loss": 3.414324964217087,
      "tokens_seen": 2345140224
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002924774322968907,
      "loss": 1.4344,
      "theoretical_loss": 3.4143173198978882,
      "tokens_seen": 2345205760
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029245737211634906,
      "loss": 1.4405,
      "theoretical_loss": 3.414309675852116,
      "tokens_seen": 2345271296
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002924373119358074,
      "loss": 1.3013,
      "theoretical_loss": 3.4143020320797515,
      "tokens_seen": 2345336832
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002924172517552658,
      "loss": 1.2646,
      "theoretical_loss": 3.414294388580779,
      "tokens_seen": 2345402368
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029239719157472414,
      "loss": 1.4186,
      "theoretical_loss": 3.41428674535518,
      "tokens_seen": 2345467904
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029237713139418256,
      "loss": 1.3774,
      "theoretical_loss": 3.4142791024029377,
      "tokens_seen": 2345533440
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002923570712136409,
      "loss": 1.385,
      "theoretical_loss": 3.4142714597240342,
      "tokens_seen": 2345598976
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002923370110330993,
      "loss": 1.2938,
      "theoretical_loss": 3.4142638173184525,
      "tokens_seen": 2345664512
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029231695085255765,
      "loss": 1.3587,
      "theoretical_loss": 3.414256175186175,
      "tokens_seen": 2345730048
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029229689067201606,
      "loss": 1.3382,
      "theoretical_loss": 3.4142485333271844,
      "tokens_seen": 2345795584
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002922768304914744,
      "loss": 1.3288,
      "theoretical_loss": 3.414240891741463,
      "tokens_seen": 2345861120
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002922567703109328,
      "loss": 1.4444,
      "theoretical_loss": 3.4142332504289934,
      "tokens_seen": 2345926656
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029223671013039115,
      "loss": 1.3972,
      "theoretical_loss": 3.4142256093897587,
      "tokens_seen": 2345992192
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002922166499498495,
      "loss": 1.383,
      "theoretical_loss": 3.414217968623741,
      "tokens_seen": 2346057728
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002921965897693079,
      "loss": 1.3714,
      "theoretical_loss": 3.414210328130923,
      "tokens_seen": 2346123264
    },
    {
      "epoch": 0.71,
      "objective/train/advantage_avg": 0.48892834782600403,
      "objective/train/docs_used": 854185,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4054796695709229,
      "objective/train/original_loss": 1.4054795503616333,
      "objective/train/theoretical_loss": 3.414202687911288,
      "objective/train/tokens_used": 2366648800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24034816026687622,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050114631652832,
      "objective/train/weighted_lm_loss": 1.4757565259933472,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.414202687911288,
      "tokens_seen": 2346188800
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002921765295887663,
      "loss": 1.4862,
      "theoretical_loss": 3.414202687911288,
      "tokens_seen": 2346188800
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029215646940822465,
      "loss": 1.4191,
      "theoretical_loss": 3.4141950479648173,
      "tokens_seen": 2346254336
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000292136409227683,
      "loss": 1.433,
      "theoretical_loss": 3.4141874082914945,
      "tokens_seen": 2346319872
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029211634904714143,
      "loss": 1.3311,
      "theoretical_loss": 3.4141797688913016,
      "tokens_seen": 2346385408
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002920962888665998,
      "loss": 1.3722,
      "theoretical_loss": 3.414172129764222,
      "tokens_seen": 2346450944
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029207622868605815,
      "loss": 1.3673,
      "theoretical_loss": 3.4141644909102373,
      "tokens_seen": 2346516480
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002920561685055165,
      "loss": 1.3301,
      "theoretical_loss": 3.414156852329331,
      "tokens_seen": 2346582016
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029203610832497493,
      "loss": 1.5029,
      "theoretical_loss": 3.414149214021485,
      "tokens_seen": 2346647552
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029201604814443335,
      "loss": 1.4316,
      "theoretical_loss": 3.414141575986682,
      "tokens_seen": 2346713088
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002919959879638917,
      "loss": 1.3673,
      "theoretical_loss": 3.4141339382249054,
      "tokens_seen": 2346778624
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002919759277833501,
      "loss": 1.2917,
      "theoretical_loss": 3.4141263007361373,
      "tokens_seen": 2346844160
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029195586760280844,
      "loss": 1.4477,
      "theoretical_loss": 3.4141186635203598,
      "tokens_seen": 2346909696
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029193580742226685,
      "loss": 1.4363,
      "theoretical_loss": 3.4141110265775563,
      "tokens_seen": 2346975232
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002919157472417252,
      "loss": 1.3328,
      "theoretical_loss": 3.4141033899077096,
      "tokens_seen": 2347040768
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002918956870611836,
      "loss": 1.3282,
      "theoretical_loss": 3.414095753510801,
      "tokens_seen": 2347106304
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029187562688064194,
      "loss": 1.4236,
      "theoretical_loss": 3.4140881173868145,
      "tokens_seen": 2347171840
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002918555667001003,
      "loss": 1.4115,
      "theoretical_loss": 3.414080481535732,
      "tokens_seen": 2347237376
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002918355065195587,
      "loss": 1.2932,
      "theoretical_loss": 3.4140728459575365,
      "tokens_seen": 2347302912
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002918154463390171,
      "loss": 1.2945,
      "theoretical_loss": 3.4140652106522102,
      "tokens_seen": 2347368448
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029179538615847544,
      "loss": 1.3485,
      "theoretical_loss": 3.4140575756197364,
      "tokens_seen": 2347433984
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002917753259779338,
      "loss": 1.3844,
      "theoretical_loss": 3.414049940860097,
      "tokens_seen": 2347499520
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002917552657973922,
      "loss": 1.434,
      "theoretical_loss": 3.414042306373275,
      "tokens_seen": 2347565056
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002917352056168506,
      "loss": 1.3932,
      "theoretical_loss": 3.4140346721592527,
      "tokens_seen": 2347630592
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029171514543630895,
      "loss": 1.3966,
      "theoretical_loss": 3.4140270382180136,
      "tokens_seen": 2347696128
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002916950852557673,
      "loss": 1.2959,
      "theoretical_loss": 3.4140194045495393,
      "tokens_seen": 2347761664
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029167502507522567,
      "loss": 1.3725,
      "theoretical_loss": 3.414011771153813,
      "tokens_seen": 2347827200
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002916549648946841,
      "loss": 1.4128,
      "theoretical_loss": 3.4140041380308177,
      "tokens_seen": 2347892736
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029163490471414245,
      "loss": 1.3242,
      "theoretical_loss": 3.413996505180535,
      "tokens_seen": 2347958272
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002916148445336008,
      "loss": 1.3612,
      "theoretical_loss": 3.413988872602949,
      "tokens_seen": 2348023808
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002915947843530592,
      "loss": 1.388,
      "theoretical_loss": 3.4139812402980407,
      "tokens_seen": 2348089344
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002915747241725176,
      "loss": 1.393,
      "theoretical_loss": 3.4139736082657937,
      "tokens_seen": 2348154880
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029155466399197595,
      "loss": 1.3316,
      "theoretical_loss": 3.4139659765061907,
      "tokens_seen": 2348220416
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002915346038114343,
      "loss": 1.354,
      "theoretical_loss": 3.4139583450192137,
      "tokens_seen": 2348285952
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002915145436308927,
      "loss": 1.3195,
      "theoretical_loss": 3.413950713804846,
      "tokens_seen": 2348351488
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002914944834503511,
      "loss": 1.3509,
      "theoretical_loss": 3.4139430828630704,
      "tokens_seen": 2348417024
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029147442326980946,
      "loss": 1.334,
      "theoretical_loss": 3.4139354521938685,
      "tokens_seen": 2348482560
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002914543630892678,
      "loss": 1.471,
      "theoretical_loss": 3.413927821797224,
      "tokens_seen": 2348548096
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002914343029087262,
      "loss": 1.3431,
      "theoretical_loss": 3.413920191673119,
      "tokens_seen": 2348613632
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029141424272818454,
      "loss": 1.4522,
      "theoretical_loss": 3.413912561821537,
      "tokens_seen": 2348679168
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029139418254764296,
      "loss": 1.433,
      "theoretical_loss": 3.4139049322424597,
      "tokens_seen": 2348744704
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002913741223671013,
      "loss": 1.394,
      "theoretical_loss": 3.41389730293587,
      "tokens_seen": 2348810240
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002913540621865597,
      "loss": 1.2434,
      "theoretical_loss": 3.4138896739017506,
      "tokens_seen": 2348875776
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029133400200601805,
      "loss": 1.3597,
      "theoretical_loss": 3.4138820451400846,
      "tokens_seen": 2348941312
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029131394182547646,
      "loss": 1.4567,
      "theoretical_loss": 3.413874416650854,
      "tokens_seen": 2349006848
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002912938816449348,
      "loss": 1.3318,
      "theoretical_loss": 3.413866788434042,
      "tokens_seen": 2349072384
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002912738214643932,
      "loss": 1.3928,
      "theoretical_loss": 3.413859160489631,
      "tokens_seen": 2349137920
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029125376128385155,
      "loss": 1.2929,
      "theoretical_loss": 3.4138515328176036,
      "tokens_seen": 2349203456
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002912337011033099,
      "loss": 1.4437,
      "theoretical_loss": 3.4138439054179424,
      "tokens_seen": 2349268992
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029121364092276833,
      "loss": 1.5006,
      "theoretical_loss": 3.413836278290631,
      "tokens_seen": 2349334528
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002911935807422267,
      "loss": 1.2875,
      "theoretical_loss": 3.4138286514356504,
      "tokens_seen": 2349400064
    },
    {
      "epoch": 0.71,
      "objective/train/advantage_avg": 0.4913260340690613,
      "objective/train/docs_used": 855374,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4019585847854614,
      "objective/train/original_loss": 1.401958703994751,
      "objective/train/theoretical_loss": 3.413821024852985,
      "objective/train/tokens_used": 2369925600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24233514070510864,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05036461353302,
      "objective/train/weighted_lm_loss": 1.4724071025848389,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.413821024852985,
      "tokens_seen": 2349465600
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029117352056168505,
      "loss": 1.4421,
      "theoretical_loss": 3.413821024852985,
      "tokens_seen": 2349465600
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002911534603811434,
      "loss": 1.3918,
      "theoretical_loss": 3.4138133985426165,
      "tokens_seen": 2349531136
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029113340020060183,
      "loss": 1.4846,
      "theoretical_loss": 3.4138057725045274,
      "tokens_seen": 2349596672
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002911133400200602,
      "loss": 1.4004,
      "theoretical_loss": 3.413798146738701,
      "tokens_seen": 2349662208
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029109327983951855,
      "loss": 1.4741,
      "theoretical_loss": 3.41379052124512,
      "tokens_seen": 2349727744
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002910732196589769,
      "loss": 1.4357,
      "theoretical_loss": 3.4137828960237666,
      "tokens_seen": 2349793280
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002910531594784353,
      "loss": 1.474,
      "theoretical_loss": 3.413775271074624,
      "tokens_seen": 2349858816
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002910330992978937,
      "loss": 1.4338,
      "theoretical_loss": 3.4137676463976745,
      "tokens_seen": 2349924352
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029101303911735206,
      "loss": 1.4132,
      "theoretical_loss": 3.4137600219929007,
      "tokens_seen": 2349989888
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002909929789368104,
      "loss": 1.3828,
      "theoretical_loss": 3.413752397860286,
      "tokens_seen": 2350055424
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002909729187562688,
      "loss": 1.3443,
      "theoretical_loss": 3.413744773999812,
      "tokens_seen": 2350120960
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002909528585757272,
      "loss": 1.3763,
      "theoretical_loss": 3.4137371504114626,
      "tokens_seen": 2350186496
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029093279839518556,
      "loss": 1.3962,
      "theoretical_loss": 3.41372952709522,
      "tokens_seen": 2350252032
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002909127382146439,
      "loss": 1.3868,
      "theoretical_loss": 3.4137219040510667,
      "tokens_seen": 2350317568
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002908926780341023,
      "loss": 1.2614,
      "theoretical_loss": 3.413714281278985,
      "tokens_seen": 2350383104
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029087261785356065,
      "loss": 1.408,
      "theoretical_loss": 3.4137066587789584,
      "tokens_seen": 2350448640
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029085255767301906,
      "loss": 1.3117,
      "theoretical_loss": 3.4136990365509696,
      "tokens_seen": 2350514176
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002908324974924774,
      "loss": 1.4366,
      "theoretical_loss": 3.4136914145950006,
      "tokens_seen": 2350579712
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002908124373119358,
      "loss": 1.4226,
      "theoretical_loss": 3.413683792911035,
      "tokens_seen": 2350645248
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029079237713139415,
      "loss": 1.3311,
      "theoretical_loss": 3.4136761714990547,
      "tokens_seen": 2350710784
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029077231695085257,
      "loss": 1.4078,
      "theoretical_loss": 3.413668550359043,
      "tokens_seen": 2350776320
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029075225677031093,
      "loss": 1.3661,
      "theoretical_loss": 3.413660929490982,
      "tokens_seen": 2350841856
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002907321965897693,
      "loss": 1.4555,
      "theoretical_loss": 3.413653308894855,
      "tokens_seen": 2350907392
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029071213640922765,
      "loss": 1.49,
      "theoretical_loss": 3.4136456885706448,
      "tokens_seen": 2350972928
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000290692076228686,
      "loss": 1.2469,
      "theoretical_loss": 3.4136380685183334,
      "tokens_seen": 2351038464
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029067201604814443,
      "loss": 1.3867,
      "theoretical_loss": 3.413630448737904,
      "tokens_seen": 2351104000
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002906519558676028,
      "loss": 1.3138,
      "theoretical_loss": 3.4136228292293396,
      "tokens_seen": 2351169536
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029063189568706116,
      "loss": 1.4084,
      "theoretical_loss": 3.4136152099926225,
      "tokens_seen": 2351235072
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002906118355065195,
      "loss": 1.4123,
      "theoretical_loss": 3.413607591027735,
      "tokens_seen": 2351300608
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029059177532597794,
      "loss": 1.3259,
      "theoretical_loss": 3.4135999723346613,
      "tokens_seen": 2351366144
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002905717151454363,
      "loss": 1.3953,
      "theoretical_loss": 3.4135923539133826,
      "tokens_seen": 2351431680
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029055165496489466,
      "loss": 1.5073,
      "theoretical_loss": 3.413584735763882,
      "tokens_seen": 2351497216
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002905315947843531,
      "loss": 1.401,
      "theoretical_loss": 3.4135771178861427,
      "tokens_seen": 2351562752
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002905115346038115,
      "loss": 1.2584,
      "theoretical_loss": 3.413569500280147,
      "tokens_seen": 2351628288
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029049147442326986,
      "loss": 1.3738,
      "theoretical_loss": 3.413561882945878,
      "tokens_seen": 2351693824
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002904714142427282,
      "loss": 1.4384,
      "theoretical_loss": 3.413554265883318,
      "tokens_seen": 2351759360
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002904513540621866,
      "loss": 1.3615,
      "theoretical_loss": 3.41354664909245,
      "tokens_seen": 2351824896
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029043129388164494,
      "loss": 1.3652,
      "theoretical_loss": 3.413539032573257,
      "tokens_seen": 2351890432
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029041123370110336,
      "loss": 1.2973,
      "theoretical_loss": 3.4135314163257213,
      "tokens_seen": 2351955968
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002903911735205617,
      "loss": 1.3999,
      "theoretical_loss": 3.4135238003498256,
      "tokens_seen": 2352021504
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002903711133400201,
      "loss": 1.4244,
      "theoretical_loss": 3.413516184645553,
      "tokens_seen": 2352087040
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029035105315947845,
      "loss": 1.416,
      "theoretical_loss": 3.413508569212886,
      "tokens_seen": 2352152576
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029033099297893686,
      "loss": 1.3524,
      "theoretical_loss": 3.4135009540518073,
      "tokens_seen": 2352218112
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002903109327983952,
      "loss": 1.2907,
      "theoretical_loss": 3.4134933391623,
      "tokens_seen": 2352283648
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002902908726178536,
      "loss": 1.3588,
      "theoretical_loss": 3.413485724544347,
      "tokens_seen": 2352349184
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029027081243731195,
      "loss": 1.3314,
      "theoretical_loss": 3.41347811019793,
      "tokens_seen": 2352414720
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002902507522567703,
      "loss": 1.4542,
      "theoretical_loss": 3.4134704961230327,
      "tokens_seen": 2352480256
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029023069207622873,
      "loss": 1.3822,
      "theoretical_loss": 3.4134628823196373,
      "tokens_seen": 2352545792
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002902106318956871,
      "loss": 1.4183,
      "theoretical_loss": 3.413455268787727,
      "tokens_seen": 2352611328
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029019057171514545,
      "loss": 1.3419,
      "theoretical_loss": 3.4134476555272846,
      "tokens_seen": 2352676864
    },
    {
      "epoch": 0.71,
      "objective/train/advantage_avg": 0.49002406001091003,
      "objective/train/docs_used": 856640,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4419714212417603,
      "objective/train/original_loss": 1.4419713020324707,
      "objective/train/theoretical_loss": 3.4134400425382925,
      "objective/train/tokens_used": 2373202400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24065379798412323,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502256155014038,
      "objective/train/weighted_lm_loss": 1.5143520832061768,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4134400425382925,
      "tokens_seen": 2352742400
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002901705115346038,
      "loss": 1.413,
      "theoretical_loss": 3.4134400425382925,
      "tokens_seen": 2352742400
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029015045135406223,
      "loss": 1.3929,
      "theoretical_loss": 3.4134324298207335,
      "tokens_seen": 2352807936
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002901303911735206,
      "loss": 1.3694,
      "theoretical_loss": 3.4134248173745902,
      "tokens_seen": 2352873472
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029011033099297896,
      "loss": 1.3472,
      "theoretical_loss": 3.4134172051998464,
      "tokens_seen": 2352939008
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002900902708124373,
      "loss": 1.4112,
      "theoretical_loss": 3.4134095932964836,
      "tokens_seen": 2353004544
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002900702106318957,
      "loss": 1.379,
      "theoretical_loss": 3.413401981664485,
      "tokens_seen": 2353070080
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002900501504513541,
      "loss": 1.448,
      "theoretical_loss": 3.4133943703038336,
      "tokens_seen": 2353135616
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00029003009027081246,
      "loss": 1.4544,
      "theoretical_loss": 3.4133867592145117,
      "tokens_seen": 2353201152
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002900100300902708,
      "loss": 1.3789,
      "theoretical_loss": 3.413379148396503,
      "tokens_seen": 2353266688
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002899899699097292,
      "loss": 1.3177,
      "theoretical_loss": 3.4133715378497893,
      "tokens_seen": 2353332224
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002899699097291876,
      "loss": 1.4174,
      "theoretical_loss": 3.4133639275743537,
      "tokens_seen": 2353397760
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028994984954864596,
      "loss": 1.3138,
      "theoretical_loss": 3.4133563175701793,
      "tokens_seen": 2353463296
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002899297893681043,
      "loss": 1.3773,
      "theoretical_loss": 3.413348707837248,
      "tokens_seen": 2353528832
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002899097291875627,
      "loss": 1.3826,
      "theoretical_loss": 3.4133410983755437,
      "tokens_seen": 2353594368
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028988966900702105,
      "loss": 1.3554,
      "theoretical_loss": 3.4133334891850486,
      "tokens_seen": 2353659904
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028986960882647946,
      "loss": 1.4183,
      "theoretical_loss": 3.413325880265745,
      "tokens_seen": 2353725440
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002898495486459378,
      "loss": 1.3623,
      "theoretical_loss": 3.4133182716176167,
      "tokens_seen": 2353790976
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002898294884653962,
      "loss": 1.4959,
      "theoretical_loss": 3.4133106632406456,
      "tokens_seen": 2353856512
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028980942828485455,
      "loss": 1.3187,
      "theoretical_loss": 3.4133030551348154,
      "tokens_seen": 2353922048
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028978936810431297,
      "loss": 1.2849,
      "theoretical_loss": 3.413295447300108,
      "tokens_seen": 2353987584
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028976930792377133,
      "loss": 1.2783,
      "theoretical_loss": 3.4132878397365065,
      "tokens_seen": 2354053120
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002897492477432297,
      "loss": 1.4263,
      "theoretical_loss": 3.413280232443994,
      "tokens_seen": 2354118656
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028972918756268805,
      "loss": 1.3687,
      "theoretical_loss": 3.413272625422553,
      "tokens_seen": 2354184192
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002897091273821464,
      "loss": 1.2869,
      "theoretical_loss": 3.413265018672166,
      "tokens_seen": 2354249728
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028968906720160483,
      "loss": 1.4152,
      "theoretical_loss": 3.4132574121928165,
      "tokens_seen": 2354315264
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002896690070210632,
      "loss": 1.4719,
      "theoretical_loss": 3.413249805984487,
      "tokens_seen": 2354380800
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028964894684052156,
      "loss": 1.3307,
      "theoretical_loss": 3.41324220004716,
      "tokens_seen": 2354446336
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002896288866599799,
      "loss": 1.2669,
      "theoretical_loss": 3.4132345943808184,
      "tokens_seen": 2354511872
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028960882647943834,
      "loss": 1.3986,
      "theoretical_loss": 3.413226988985446,
      "tokens_seen": 2354577408
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002895887662988967,
      "loss": 1.3163,
      "theoretical_loss": 3.4132193838610236,
      "tokens_seen": 2354642944
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028956870611835506,
      "loss": 1.2988,
      "theoretical_loss": 3.413211779007536,
      "tokens_seen": 2354708480
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002895486459378134,
      "loss": 1.3919,
      "theoretical_loss": 3.4132041744249646,
      "tokens_seen": 2354774016
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002895285857572718,
      "loss": 1.3878,
      "theoretical_loss": 3.4131965701132927,
      "tokens_seen": 2354839552
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002895085255767302,
      "loss": 1.3392,
      "theoretical_loss": 3.4131889660725037,
      "tokens_seen": 2354905088
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028948846539618856,
      "loss": 1.3485,
      "theoretical_loss": 3.4131813623025797,
      "tokens_seen": 2354970624
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002894684052156469,
      "loss": 1.257,
      "theoretical_loss": 3.4131737588035036,
      "tokens_seen": 2355036160
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002894483450351053,
      "loss": 1.3717,
      "theoretical_loss": 3.4131661555752584,
      "tokens_seen": 2355101696
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002894282848545637,
      "loss": 1.3901,
      "theoretical_loss": 3.4131585526178267,
      "tokens_seen": 2355167232
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028940822467402207,
      "loss": 1.3414,
      "theoretical_loss": 3.4131509499311914,
      "tokens_seen": 2355232768
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028938816449348043,
      "loss": 1.3771,
      "theoretical_loss": 3.413143347515336,
      "tokens_seen": 2355298304
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002893681043129388,
      "loss": 1.3988,
      "theoretical_loss": 3.413135745370242,
      "tokens_seen": 2355363840
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002893480441323972,
      "loss": 1.3882,
      "theoretical_loss": 3.4131281434958933,
      "tokens_seen": 2355429376
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028932798395185557,
      "loss": 1.3251,
      "theoretical_loss": 3.413120541892272,
      "tokens_seen": 2355494912
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028930792377131393,
      "loss": 1.4723,
      "theoretical_loss": 3.4131129405593614,
      "tokens_seen": 2355560448
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002892878635907723,
      "loss": 1.3995,
      "theoretical_loss": 3.4131053394971445,
      "tokens_seen": 2355625984
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028926780341023066,
      "loss": 1.3342,
      "theoretical_loss": 3.4130977387056034,
      "tokens_seen": 2355691520
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002892477432296891,
      "loss": 1.3925,
      "theoretical_loss": 3.413090138184722,
      "tokens_seen": 2355757056
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028922768304914744,
      "loss": 1.4451,
      "theoretical_loss": 3.4130825379344816,
      "tokens_seen": 2355822592
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002892076228686058,
      "loss": 1.3767,
      "theoretical_loss": 3.4130749379548666,
      "tokens_seen": 2355888128
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028918756268806416,
      "loss": 1.3591,
      "theoretical_loss": 3.4130673382458587,
      "tokens_seen": 2355953664
    },
    {
      "epoch": 0.71,
      "objective/train/advantage_avg": 0.48490390181541443,
      "objective/train/docs_used": 857700,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3055931329727173,
      "objective/train/original_loss": 1.3055932521820068,
      "objective/train/theoretical_loss": 3.4130597388074415,
      "objective/train/tokens_used": 2376479200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2365303486585617,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0496927499771118,
      "objective/train/weighted_lm_loss": 1.3705573081970215,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4130597388074415,
      "tokens_seen": 2356019200
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002891675025075226,
      "loss": 1.4183,
      "theoretical_loss": 3.4130597388074415,
      "tokens_seen": 2356019200
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028914744232698094,
      "loss": 1.3565,
      "theoretical_loss": 3.4130521396395976,
      "tokens_seen": 2356084736
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002891273821464393,
      "loss": 1.418,
      "theoretical_loss": 3.41304454074231,
      "tokens_seen": 2356150272
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028910732196589766,
      "loss": 1.2638,
      "theoretical_loss": 3.4130369421155606,
      "tokens_seen": 2356215808
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000289087261785356,
      "loss": 1.3945,
      "theoretical_loss": 3.4130293437593333,
      "tokens_seen": 2356281344
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028906720160481444,
      "loss": 1.42,
      "theoretical_loss": 3.4130217456736105,
      "tokens_seen": 2356346880
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002890471414242728,
      "loss": 1.407,
      "theoretical_loss": 3.4130141478583758,
      "tokens_seen": 2356412416
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002890270812437312,
      "loss": 1.3643,
      "theoretical_loss": 3.413006550313611,
      "tokens_seen": 2356477952
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002890070210631896,
      "loss": 1.4187,
      "theoretical_loss": 3.412998953039299,
      "tokens_seen": 2356543488
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000288986960882648,
      "loss": 1.3319,
      "theoretical_loss": 3.4129913560354233,
      "tokens_seen": 2356609024
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028896690070210636,
      "loss": 1.368,
      "theoretical_loss": 3.4129837593019667,
      "tokens_seen": 2356674560
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002889468405215647,
      "loss": 1.3355,
      "theoretical_loss": 3.4129761628389113,
      "tokens_seen": 2356740096
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002889267803410231,
      "loss": 1.4302,
      "theoretical_loss": 3.412968566646241,
      "tokens_seen": 2356805632
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028890672016048145,
      "loss": 1.4558,
      "theoretical_loss": 3.4129609707239377,
      "tokens_seen": 2356871168
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028888665997993986,
      "loss": 1.3406,
      "theoretical_loss": 3.412953375071985,
      "tokens_seen": 2356936704
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028886659979939823,
      "loss": 1.2958,
      "theoretical_loss": 3.4129457796903653,
      "tokens_seen": 2357002240
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002888465396188566,
      "loss": 1.4567,
      "theoretical_loss": 3.412938184579062,
      "tokens_seen": 2357067776
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028882647943831495,
      "loss": 1.3508,
      "theoretical_loss": 3.412930589738057,
      "tokens_seen": 2357133312
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028880641925777337,
      "loss": 1.3766,
      "theoretical_loss": 3.412922995167334,
      "tokens_seen": 2357198848
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028878635907723173,
      "loss": 1.3972,
      "theoretical_loss": 3.4129154008668756,
      "tokens_seen": 2357264384
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002887662988966901,
      "loss": 1.3119,
      "theoretical_loss": 3.4129078068366647,
      "tokens_seen": 2357329920
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028874623871614845,
      "loss": 1.3143,
      "theoretical_loss": 3.412900213076684,
      "tokens_seen": 2357395456
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002887261785356068,
      "loss": 1.462,
      "theoretical_loss": 3.4128926195869163,
      "tokens_seen": 2357460992
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028870611835506523,
      "loss": 1.3359,
      "theoretical_loss": 3.4128850263673454,
      "tokens_seen": 2357526528
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002886860581745236,
      "loss": 1.3531,
      "theoretical_loss": 3.412877433417953,
      "tokens_seen": 2357592064
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028866599799398196,
      "loss": 1.268,
      "theoretical_loss": 3.412869840738723,
      "tokens_seen": 2357657600
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002886459378134403,
      "loss": 1.3133,
      "theoretical_loss": 3.4128622483296365,
      "tokens_seen": 2357723136
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028862587763289874,
      "loss": 1.4848,
      "theoretical_loss": 3.412854656190679,
      "tokens_seen": 2357788672
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002886058174523571,
      "loss": 1.3459,
      "theoretical_loss": 3.412847064321831,
      "tokens_seen": 2357854208
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028858575727181546,
      "loss": 1.4045,
      "theoretical_loss": 3.4128394727230766,
      "tokens_seen": 2357919744
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002885656970912738,
      "loss": 1.3477,
      "theoretical_loss": 3.412831881394399,
      "tokens_seen": 2357985280
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002885456369107322,
      "loss": 1.3664,
      "theoretical_loss": 3.41282429033578,
      "tokens_seen": 2358050816
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002885255767301906,
      "loss": 1.3513,
      "theoretical_loss": 3.412816699547203,
      "tokens_seen": 2358116352
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028850551654964896,
      "loss": 1.5216,
      "theoretical_loss": 3.412809109028651,
      "tokens_seen": 2358181888
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002884854563691073,
      "loss": 1.2942,
      "theoretical_loss": 3.4128015187801073,
      "tokens_seen": 2358247424
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002884653961885657,
      "loss": 1.422,
      "theoretical_loss": 3.4127939288015536,
      "tokens_seen": 2358312960
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002884453360080241,
      "loss": 1.4217,
      "theoretical_loss": 3.412786339092974,
      "tokens_seen": 2358378496
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028842527582748247,
      "loss": 1.2808,
      "theoretical_loss": 3.4127787496543505,
      "tokens_seen": 2358444032
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028840521564694083,
      "loss": 1.2502,
      "theoretical_loss": 3.4127711604856668,
      "tokens_seen": 2358509568
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002883851554663992,
      "loss": 1.3825,
      "theoretical_loss": 3.412763571586905,
      "tokens_seen": 2358575104
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002883650952858576,
      "loss": 1.3998,
      "theoretical_loss": 3.4127559829580485,
      "tokens_seen": 2358640640
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028834503510531597,
      "loss": 1.3265,
      "theoretical_loss": 3.41274839459908,
      "tokens_seen": 2358706176
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028832497492477433,
      "loss": 1.3933,
      "theoretical_loss": 3.4127408065099827,
      "tokens_seen": 2358771712
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002883049147442327,
      "loss": 1.4284,
      "theoretical_loss": 3.412733218690739,
      "tokens_seen": 2358837248
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028828485456369106,
      "loss": 1.4396,
      "theoretical_loss": 3.4127256311413325,
      "tokens_seen": 2358902784
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002882647943831495,
      "loss": 1.395,
      "theoretical_loss": 3.4127180438617457,
      "tokens_seen": 2358968320
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028824473420260784,
      "loss": 1.2626,
      "theoretical_loss": 3.4127104568519613,
      "tokens_seen": 2359033856
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002882246740220662,
      "loss": 1.3569,
      "theoretical_loss": 3.4127028701119624,
      "tokens_seen": 2359099392
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028820461384152456,
      "loss": 1.4633,
      "theoretical_loss": 3.4126952836417317,
      "tokens_seen": 2359164928
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.000288184553660983,
      "loss": 1.206,
      "theoretical_loss": 3.4126876974412528,
      "tokens_seen": 2359230464
    },
    {
      "epoch": 0.71,
      "objective/train/advantage_avg": 0.4935346841812134,
      "objective/train/docs_used": 858941,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2413978576660156,
      "objective/train/original_loss": 1.2413978576660156,
      "objective/train/theoretical_loss": 3.412680111510508,
      "objective/train/tokens_used": 2379756000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24394705891609192,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050593614578247,
      "objective/train/weighted_lm_loss": 1.3039952516555786,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.412680111510508,
      "tokens_seen": 2359296000
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028816449348044134,
      "loss": 1.3065,
      "theoretical_loss": 3.412680111510508,
      "tokens_seen": 2359296000
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002881444332998997,
      "loss": 1.4402,
      "theoretical_loss": 3.4126725258494806,
      "tokens_seen": 2359361536
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.00028812437311935806,
      "loss": 1.4587,
      "theoretical_loss": 3.412664940458153,
      "tokens_seen": 2359427072
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002881043129388164,
      "loss": 1.2967,
      "theoretical_loss": 3.4126573553365085,
      "tokens_seen": 2359492608
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028808425275827484,
      "loss": 1.4269,
      "theoretical_loss": 3.41264977048453,
      "tokens_seen": 2359558144
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002880641925777332,
      "loss": 1.4274,
      "theoretical_loss": 3.4126421859022003,
      "tokens_seen": 2359623680
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028804413239719157,
      "loss": 1.3486,
      "theoretical_loss": 3.4126346015895024,
      "tokens_seen": 2359689216
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028802407221664993,
      "loss": 1.5244,
      "theoretical_loss": 3.412627017546419,
      "tokens_seen": 2359754752
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028800401203610835,
      "loss": 1.4275,
      "theoretical_loss": 3.412619433772934,
      "tokens_seen": 2359820288
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002879839518555667,
      "loss": 1.4237,
      "theoretical_loss": 3.412611850269029,
      "tokens_seen": 2359885824
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028796389167502507,
      "loss": 1.4197,
      "theoretical_loss": 3.4126042670346877,
      "tokens_seen": 2359951360
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028794383149448343,
      "loss": 1.3392,
      "theoretical_loss": 3.412596684069893,
      "tokens_seen": 2360016896
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002879237713139418,
      "loss": 1.4158,
      "theoretical_loss": 3.412589101374627,
      "tokens_seen": 2360082432
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002879037111334002,
      "loss": 1.4151,
      "theoretical_loss": 3.4125815189488735,
      "tokens_seen": 2360147968
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002878836509528586,
      "loss": 1.2271,
      "theoretical_loss": 3.412573936792616,
      "tokens_seen": 2360213504
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028786359077231694,
      "loss": 1.4626,
      "theoretical_loss": 3.4125663549058363,
      "tokens_seen": 2360279040
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002878435305917753,
      "loss": 1.4082,
      "theoretical_loss": 3.4125587732885174,
      "tokens_seen": 2360344576
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002878234704112337,
      "loss": 1.374,
      "theoretical_loss": 3.4125511919406426,
      "tokens_seen": 2360410112
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002878034102306921,
      "loss": 1.3666,
      "theoretical_loss": 3.412543610862195,
      "tokens_seen": 2360475648
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028778335005015044,
      "loss": 1.3218,
      "theoretical_loss": 3.4125360300531575,
      "tokens_seen": 2360541184
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002877632898696088,
      "loss": 1.4264,
      "theoretical_loss": 3.4125284495135126,
      "tokens_seen": 2360606720
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028774322968906716,
      "loss": 1.3768,
      "theoretical_loss": 3.412520869243244,
      "tokens_seen": 2360672256
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002877231695085256,
      "loss": 1.3664,
      "theoretical_loss": 3.4125132892423338,
      "tokens_seen": 2360737792
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028770310932798394,
      "loss": 1.3167,
      "theoretical_loss": 3.4125057095107656,
      "tokens_seen": 2360803328
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002876830491474423,
      "loss": 1.4651,
      "theoretical_loss": 3.412498130048522,
      "tokens_seen": 2360868864
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028766298896690067,
      "loss": 1.3685,
      "theoretical_loss": 3.412490550855586,
      "tokens_seen": 2360934400
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002876429287863591,
      "loss": 1.3205,
      "theoretical_loss": 3.412482971931941,
      "tokens_seen": 2360999936
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028762286860581745,
      "loss": 1.4782,
      "theoretical_loss": 3.412475393277569,
      "tokens_seen": 2361065472
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002876028084252758,
      "loss": 1.479,
      "theoretical_loss": 3.412467814892454,
      "tokens_seen": 2361131008
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028758274824473417,
      "loss": 1.3586,
      "theoretical_loss": 3.412460236776578,
      "tokens_seen": 2361196544
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028756268806419253,
      "loss": 1.3893,
      "theoretical_loss": 3.412452658929925,
      "tokens_seen": 2361262080
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028754262788365095,
      "loss": 1.315,
      "theoretical_loss": 3.412445081352477,
      "tokens_seen": 2361327616
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028752256770310936,
      "loss": 1.3957,
      "theoretical_loss": 3.4124375040442176,
      "tokens_seen": 2361393152
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002875025075225677,
      "loss": 1.4489,
      "theoretical_loss": 3.4124299270051295,
      "tokens_seen": 2361458688
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002874824473420261,
      "loss": 1.3612,
      "theoretical_loss": 3.412422350235196,
      "tokens_seen": 2361524224
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002874623871614845,
      "loss": 1.4339,
      "theoretical_loss": 3.4124147737343993,
      "tokens_seen": 2361589760
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028744232698094287,
      "loss": 1.3345,
      "theoretical_loss": 3.4124071975027235,
      "tokens_seen": 2361655296
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028742226680040123,
      "loss": 1.4308,
      "theoretical_loss": 3.4123996215401506,
      "tokens_seen": 2361720832
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002874022066198596,
      "loss": 1.416,
      "theoretical_loss": 3.412392045846664,
      "tokens_seen": 2361786368
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000287382146439318,
      "loss": 1.3405,
      "theoretical_loss": 3.4123844704222464,
      "tokens_seen": 2361851904
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028736208625877637,
      "loss": 1.3416,
      "theoretical_loss": 3.4123768952668807,
      "tokens_seen": 2361917440
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028734202607823473,
      "loss": 1.3992,
      "theoretical_loss": 3.4123693203805505,
      "tokens_seen": 2361982976
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002873219658976931,
      "loss": 1.3881,
      "theoretical_loss": 3.4123617457632385,
      "tokens_seen": 2362048512
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028730190571715146,
      "loss": 1.3841,
      "theoretical_loss": 3.4123541714149273,
      "tokens_seen": 2362114048
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002872818455366099,
      "loss": 1.4378,
      "theoretical_loss": 3.412346597335601,
      "tokens_seen": 2362179584
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028726178535606824,
      "loss": 1.4505,
      "theoretical_loss": 3.4123390235252407,
      "tokens_seen": 2362245120
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002872417251755266,
      "loss": 1.3813,
      "theoretical_loss": 3.412331449983831,
      "tokens_seen": 2362310656
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028722166499498496,
      "loss": 1.4595,
      "theoretical_loss": 3.4123238767113544,
      "tokens_seen": 2362376192
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002872016048144434,
      "loss": 1.3724,
      "theoretical_loss": 3.4123163037077937,
      "tokens_seen": 2362441728
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028718154463390174,
      "loss": 1.3135,
      "theoretical_loss": 3.412308730973132,
      "tokens_seen": 2362507264
    },
    {
      "epoch": 0.72,
      "objective/train/advantage_avg": 0.49517935514450073,
      "objective/train/docs_used": 860261,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3953934907913208,
      "objective/train/original_loss": 1.3953936100006104,
      "objective/train/theoretical_loss": 3.4123011585073524,
      "objective/train/tokens_used": 2383032800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24537505209445953,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507653951644897,
      "objective/train/weighted_lm_loss": 1.4661709070205688,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 3.4123011585073524,
      "tokens_seen": 2362572800
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002871614844533601,
      "loss": 1.4096,
      "theoretical_loss": 3.4123011585073524,
      "tokens_seen": 2362572800
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028714142427281846,
      "loss": 1.3973,
      "theoretical_loss": 3.412293586310438,
      "tokens_seen": 2362638336
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002871213640922768,
      "loss": 1.3411,
      "theoretical_loss": 3.4122860143823717,
      "tokens_seen": 2362703872
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028710130391173524,
      "loss": 1.3037,
      "theoretical_loss": 3.412278442723136,
      "tokens_seen": 2362769408
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002870812437311936,
      "loss": 1.3667,
      "theoretical_loss": 3.4122708713327143,
      "tokens_seen": 2362834944
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028706118355065197,
      "loss": 1.3476,
      "theoretical_loss": 3.41226330021109,
      "tokens_seen": 2362900480
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028704112337011033,
      "loss": 1.3564,
      "theoretical_loss": 3.4122557293582454,
      "tokens_seen": 2362966016
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028702106318956875,
      "loss": 1.3164,
      "theoretical_loss": 3.412248158774164,
      "tokens_seen": 2363031552
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002870010030090271,
      "loss": 1.3331,
      "theoretical_loss": 3.412240588458829,
      "tokens_seen": 2363097088
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028698094282848547,
      "loss": 1.3504,
      "theoretical_loss": 3.4122330184122225,
      "tokens_seen": 2363162624
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028696088264794383,
      "loss": 1.4894,
      "theoretical_loss": 3.412225448634328,
      "tokens_seen": 2363228160
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002869408224674022,
      "loss": 1.4215,
      "theoretical_loss": 3.4122178791251287,
      "tokens_seen": 2363293696
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002869207622868606,
      "loss": 1.367,
      "theoretical_loss": 3.4122103098846077,
      "tokens_seen": 2363359232
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000286900702106319,
      "loss": 1.3313,
      "theoretical_loss": 3.4122027409127473,
      "tokens_seen": 2363424768
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028688064192577734,
      "loss": 1.3425,
      "theoretical_loss": 3.4121951722095316,
      "tokens_seen": 2363490304
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002868605817452357,
      "loss": 1.3619,
      "theoretical_loss": 3.4121876037749423,
      "tokens_seen": 2363555840
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002868405215646941,
      "loss": 1.4526,
      "theoretical_loss": 3.4121800356089635,
      "tokens_seen": 2363621376
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002868204613841525,
      "loss": 1.441,
      "theoretical_loss": 3.412172467711578,
      "tokens_seen": 2363686912
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028680040120361084,
      "loss": 1.4237,
      "theoretical_loss": 3.412164900082769,
      "tokens_seen": 2363752448
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002867803410230692,
      "loss": 1.352,
      "theoretical_loss": 3.4121573327225185,
      "tokens_seen": 2363817984
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028676028084252756,
      "loss": 1.3535,
      "theoretical_loss": 3.4121497656308106,
      "tokens_seen": 2363883520
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000286740220661986,
      "loss": 1.3871,
      "theoretical_loss": 3.4121421988076275,
      "tokens_seen": 2363949056
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028672016048144434,
      "loss": 1.2955,
      "theoretical_loss": 3.412134632252953,
      "tokens_seen": 2364014592
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002867001003009027,
      "loss": 1.4163,
      "theoretical_loss": 3.4121270659667697,
      "tokens_seen": 2364080128
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028668004012036107,
      "loss": 1.1814,
      "theoretical_loss": 3.412119499949061,
      "tokens_seen": 2364145664
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002866599799398195,
      "loss": 1.3029,
      "theoretical_loss": 3.4121119341998094,
      "tokens_seen": 2364211200
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028663991975927785,
      "loss": 1.4221,
      "theoretical_loss": 3.4121043687189982,
      "tokens_seen": 2364276736
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002866198595787362,
      "loss": 1.488,
      "theoretical_loss": 3.4120968035066106,
      "tokens_seen": 2364342272
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028659979939819457,
      "loss": 1.3999,
      "theoretical_loss": 3.412089238562629,
      "tokens_seen": 2364407808
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028657973921765293,
      "loss": 1.4091,
      "theoretical_loss": 3.4120816738870374,
      "tokens_seen": 2364473344
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028655967903711135,
      "loss": 1.4392,
      "theoretical_loss": 3.4120741094798186,
      "tokens_seen": 2364538880
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002865396188565697,
      "loss": 1.3811,
      "theoretical_loss": 3.412066545340955,
      "tokens_seen": 2364604416
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002865195586760281,
      "loss": 1.3293,
      "theoretical_loss": 3.41205898147043,
      "tokens_seen": 2364669952
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028649949849548644,
      "loss": 1.3572,
      "theoretical_loss": 3.4120514178682266,
      "tokens_seen": 2364735488
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028647943831494485,
      "loss": 1.3665,
      "theoretical_loss": 3.412043854534328,
      "tokens_seen": 2364801024
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002864593781344032,
      "loss": 1.3854,
      "theoretical_loss": 3.4120362914687172,
      "tokens_seen": 2364866560
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002864393179538616,
      "loss": 1.3827,
      "theoretical_loss": 3.412028728671377,
      "tokens_seen": 2364932096
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028641925777331994,
      "loss": 1.3358,
      "theoretical_loss": 3.412021166142291,
      "tokens_seen": 2364997632
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002863991975927783,
      "loss": 1.3453,
      "theoretical_loss": 3.4120136038814417,
      "tokens_seen": 2365063168
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002863791374122367,
      "loss": 1.4026,
      "theoretical_loss": 3.4120060418888123,
      "tokens_seen": 2365128704
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002863590772316951,
      "loss": 1.4174,
      "theoretical_loss": 3.411998480164386,
      "tokens_seen": 2365194240
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028633901705115344,
      "loss": 1.4174,
      "theoretical_loss": 3.411990918708146,
      "tokens_seen": 2365259776
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002863189568706118,
      "loss": 1.3573,
      "theoretical_loss": 3.411983357520075,
      "tokens_seen": 2365325312
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002862988966900702,
      "loss": 1.2974,
      "theoretical_loss": 3.411975796600156,
      "tokens_seen": 2365390848
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002862788365095286,
      "loss": 1.3196,
      "theoretical_loss": 3.4119682359483727,
      "tokens_seen": 2365456384
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028625877632898694,
      "loss": 1.4088,
      "theoretical_loss": 3.4119606755647074,
      "tokens_seen": 2365521920
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002862387161484453,
      "loss": 1.4309,
      "theoretical_loss": 3.4119531154491436,
      "tokens_seen": 2365587456
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028621865596790367,
      "loss": 1.3625,
      "theoretical_loss": 3.411945555601664,
      "tokens_seen": 2365652992
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002861985957873621,
      "loss": 1.3657,
      "theoretical_loss": 3.4119379960222522,
      "tokens_seen": 2365718528
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028617853560682045,
      "loss": 1.428,
      "theoretical_loss": 3.411930436710891,
      "tokens_seen": 2365784064
    },
    {
      "epoch": 0.72,
      "objective/train/advantage_avg": 0.49338480830192566,
      "objective/train/docs_used": 861421,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1170531511306763,
      "objective/train/original_loss": 1.1170530319213867,
      "objective/train/theoretical_loss": 3.411922877667563,
      "objective/train/tokens_used": 2386309600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24374377727508545,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505775213241577,
      "objective/train/weighted_lm_loss": 1.1735343933105469,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.411922877667563,
      "tokens_seen": 2365849600
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002861584754262788,
      "loss": 1.2403,
      "theoretical_loss": 3.411922877667563,
      "tokens_seen": 2365849600
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028613841524573717,
      "loss": 1.4482,
      "theoretical_loss": 3.411915318892252,
      "tokens_seen": 2365915136
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002861183550651956,
      "loss": 1.4337,
      "theoretical_loss": 3.411907760384941,
      "tokens_seen": 2365980672
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028609829488465395,
      "loss": 1.3923,
      "theoretical_loss": 3.411900202145613,
      "tokens_seen": 2366046208
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002860782347041123,
      "loss": 1.212,
      "theoretical_loss": 3.4118926441742508,
      "tokens_seen": 2366111744
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002860581745235707,
      "loss": 1.4373,
      "theoretical_loss": 3.4118850864708374,
      "tokens_seen": 2366177280
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002860381143430291,
      "loss": 1.271,
      "theoretical_loss": 3.4118775290353565,
      "tokens_seen": 2366242816
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002860180541624875,
      "loss": 1.3657,
      "theoretical_loss": 3.41186997186779,
      "tokens_seen": 2366308352
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028599799398194587,
      "loss": 1.3447,
      "theoretical_loss": 3.4118624149681227,
      "tokens_seen": 2366373888
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028597793380140423,
      "loss": 1.3061,
      "theoretical_loss": 3.411854858336336,
      "tokens_seen": 2366439424
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002859578736208626,
      "loss": 1.458,
      "theoretical_loss": 3.4118473019724145,
      "tokens_seen": 2366504960
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000285937813440321,
      "loss": 1.3637,
      "theoretical_loss": 3.41183974587634,
      "tokens_seen": 2366570496
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002859177532597794,
      "loss": 1.2614,
      "theoretical_loss": 3.4118321900480963,
      "tokens_seen": 2366636032
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028589769307923774,
      "loss": 1.3377,
      "theoretical_loss": 3.411824634487666,
      "tokens_seen": 2366701568
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002858776328986961,
      "loss": 1.4183,
      "theoretical_loss": 3.4118170791950333,
      "tokens_seen": 2366767104
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002858575727181545,
      "loss": 1.2931,
      "theoretical_loss": 3.41180952417018,
      "tokens_seen": 2366832640
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002858375125376129,
      "loss": 1.411,
      "theoretical_loss": 3.4118019694130894,
      "tokens_seen": 2366898176
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028581745235707124,
      "loss": 1.5327,
      "theoretical_loss": 3.411794414923745,
      "tokens_seen": 2366963712
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002857973921765296,
      "loss": 1.5123,
      "theoretical_loss": 3.41178686070213,
      "tokens_seen": 2367029248
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028577733199598796,
      "loss": 1.299,
      "theoretical_loss": 3.4117793067482274,
      "tokens_seen": 2367094784
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002857572718154464,
      "loss": 1.3863,
      "theoretical_loss": 3.41177175306202,
      "tokens_seen": 2367160320
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028573721163490474,
      "loss": 1.3732,
      "theoretical_loss": 3.4117641996434913,
      "tokens_seen": 2367225856
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002857171514543631,
      "loss": 1.338,
      "theoretical_loss": 3.411756646492624,
      "tokens_seen": 2367291392
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028569709127382147,
      "loss": 1.426,
      "theoretical_loss": 3.4117490936094015,
      "tokens_seen": 2367356928
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002856770310932799,
      "loss": 1.4043,
      "theoretical_loss": 3.4117415409938068,
      "tokens_seen": 2367422464
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028565697091273825,
      "loss": 1.3252,
      "theoretical_loss": 3.411733988645823,
      "tokens_seen": 2367488000
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002856369107321966,
      "loss": 1.351,
      "theoretical_loss": 3.411726436565433,
      "tokens_seen": 2367553536
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028561685055165497,
      "loss": 1.356,
      "theoretical_loss": 3.41171888475262,
      "tokens_seen": 2367619072
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028559679037111333,
      "loss": 1.4944,
      "theoretical_loss": 3.411711333207368,
      "tokens_seen": 2367684608
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028557673019057175,
      "loss": 1.3685,
      "theoretical_loss": 3.411703781929659,
      "tokens_seen": 2367750144
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002855566700100301,
      "loss": 1.3699,
      "theoretical_loss": 3.411696230919476,
      "tokens_seen": 2367815680
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002855366098294885,
      "loss": 1.415,
      "theoretical_loss": 3.411688680176803,
      "tokens_seen": 2367881216
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028551654964894684,
      "loss": 1.2988,
      "theoretical_loss": 3.411681129701623,
      "tokens_seen": 2367946752
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028549648946840525,
      "loss": 1.4258,
      "theoretical_loss": 3.411673579493918,
      "tokens_seen": 2368012288
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002854764292878636,
      "loss": 1.32,
      "theoretical_loss": 3.4116660295536727,
      "tokens_seen": 2368077824
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000285456369107322,
      "loss": 1.3647,
      "theoretical_loss": 3.4116584798808693,
      "tokens_seen": 2368143360
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028543630892678034,
      "loss": 1.2672,
      "theoretical_loss": 3.411650930475491,
      "tokens_seen": 2368208896
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002854162487462387,
      "loss": 1.422,
      "theoretical_loss": 3.4116433813375213,
      "tokens_seen": 2368274432
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002853961885656971,
      "loss": 1.2738,
      "theoretical_loss": 3.411635832466943,
      "tokens_seen": 2368339968
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002853761283851555,
      "loss": 1.4202,
      "theoretical_loss": 3.411628283863739,
      "tokens_seen": 2368405504
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028535606820461384,
      "loss": 1.3476,
      "theoretical_loss": 3.4116207355278934,
      "tokens_seen": 2368471040
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002853360080240722,
      "loss": 1.3975,
      "theoretical_loss": 3.411613187459388,
      "tokens_seen": 2368536576
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002853159478435306,
      "loss": 1.3343,
      "theoretical_loss": 3.411605639658207,
      "tokens_seen": 2368602112
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000285295887662989,
      "loss": 1.3679,
      "theoretical_loss": 3.4115980921243327,
      "tokens_seen": 2368667648
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028527582748244735,
      "loss": 1.4282,
      "theoretical_loss": 3.411590544857749,
      "tokens_seen": 2368733184
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002852557673019057,
      "loss": 1.3133,
      "theoretical_loss": 3.4115829978584387,
      "tokens_seen": 2368798720
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028523570712136407,
      "loss": 1.3949,
      "theoretical_loss": 3.4115754511263847,
      "tokens_seen": 2368864256
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002852156469408225,
      "loss": 1.4856,
      "theoretical_loss": 3.411567904661571,
      "tokens_seen": 2368929792
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028519558676028085,
      "loss": 1.3758,
      "theoretical_loss": 3.4115603584639795,
      "tokens_seen": 2368995328
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002851755265797392,
      "loss": 1.4183,
      "theoretical_loss": 3.411552812533594,
      "tokens_seen": 2369060864
    },
    {
      "epoch": 0.72,
      "objective/train/advantage_avg": 0.49455055594444275,
      "objective/train/docs_used": 862685,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2091267108917236,
      "objective/train/original_loss": 1.209126591682434,
      "objective/train/theoretical_loss": 3.411545266870398,
      "objective/train/tokens_used": 2389586400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24479496479034424,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506995916366577,
      "objective/train/weighted_lm_loss": 1.2703179121017456,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.411545266870398,
      "tokens_seen": 2369126400
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028515546639919757,
      "loss": 1.323,
      "theoretical_loss": 3.411545266870398,
      "tokens_seen": 2369126400
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000285135406218656,
      "loss": 1.2753,
      "theoretical_loss": 3.411537721474374,
      "tokens_seen": 2369191936
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028511534603811435,
      "loss": 1.3736,
      "theoretical_loss": 3.4115301763455057,
      "tokens_seen": 2369257472
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002850952858575727,
      "loss": 1.3813,
      "theoretical_loss": 3.4115226314837757,
      "tokens_seen": 2369323008
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002850752256770311,
      "loss": 1.3315,
      "theoretical_loss": 3.4115150868891675,
      "tokens_seen": 2369388544
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002850551654964895,
      "loss": 1.4673,
      "theoretical_loss": 3.4115075425616643,
      "tokens_seen": 2369454080
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028503510531594785,
      "loss": 1.3416,
      "theoretical_loss": 3.411499998501249,
      "tokens_seen": 2369519616
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002850150451354062,
      "loss": 1.3866,
      "theoretical_loss": 3.411492454707905,
      "tokens_seen": 2369585152
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002849949849548646,
      "loss": 1.3872,
      "theoretical_loss": 3.4114849111816152,
      "tokens_seen": 2369650688
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028497492477432294,
      "loss": 1.4255,
      "theoretical_loss": 3.4114773679223633,
      "tokens_seen": 2369716224
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028495486459378136,
      "loss": 1.3359,
      "theoretical_loss": 3.4114698249301316,
      "tokens_seen": 2369781760
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002849348044132397,
      "loss": 1.3725,
      "theoretical_loss": 3.411462282204904,
      "tokens_seen": 2369847296
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002849147442326981,
      "loss": 1.4143,
      "theoretical_loss": 3.411454739746664,
      "tokens_seen": 2369912832
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028489468405215644,
      "loss": 1.3451,
      "theoretical_loss": 3.4114471975553933,
      "tokens_seen": 2369978368
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028487462387161486,
      "loss": 1.3051,
      "theoretical_loss": 3.411439655631076,
      "tokens_seen": 2370043904
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002848545636910732,
      "loss": 1.4046,
      "theoretical_loss": 3.411432113973696,
      "tokens_seen": 2370109440
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002848345035105316,
      "loss": 1.3703,
      "theoretical_loss": 3.411424572583235,
      "tokens_seen": 2370174976
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028481444332998995,
      "loss": 1.3628,
      "theoretical_loss": 3.411417031459677,
      "tokens_seen": 2370240512
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002847943831494483,
      "loss": 1.3671,
      "theoretical_loss": 3.4114094906030052,
      "tokens_seen": 2370306048
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002847743229689067,
      "loss": 1.3937,
      "theoretical_loss": 3.4114019500132025,
      "tokens_seen": 2370371584
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002847542627883651,
      "loss": 1.4171,
      "theoretical_loss": 3.411394409690252,
      "tokens_seen": 2370437120
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028473420260782345,
      "loss": 1.2947,
      "theoretical_loss": 3.4113868696341374,
      "tokens_seen": 2370502656
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002847141424272818,
      "loss": 1.4485,
      "theoretical_loss": 3.4113793298448414,
      "tokens_seen": 2370568192
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028469408224674023,
      "loss": 1.3994,
      "theoretical_loss": 3.411371790322347,
      "tokens_seen": 2370633728
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002846740220661986,
      "loss": 1.4403,
      "theoretical_loss": 3.411364251066638,
      "tokens_seen": 2370699264
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028465396188565695,
      "loss": 1.4168,
      "theoretical_loss": 3.4113567120776978,
      "tokens_seen": 2370764800
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002846339017051153,
      "loss": 1.3877,
      "theoretical_loss": 3.411349173355508,
      "tokens_seen": 2370830336
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002846138415245737,
      "loss": 1.3868,
      "theoretical_loss": 3.4113416349000536,
      "tokens_seen": 2370895872
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002845937813440321,
      "loss": 1.4396,
      "theoretical_loss": 3.4113340967113173,
      "tokens_seen": 2370961408
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028457372116349046,
      "loss": 1.4328,
      "theoretical_loss": 3.411326558789282,
      "tokens_seen": 2371026944
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002845536609829488,
      "loss": 1.471,
      "theoretical_loss": 3.4113190211339304,
      "tokens_seen": 2371092480
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028453360080240724,
      "loss": 1.4427,
      "theoretical_loss": 3.4113114837452465,
      "tokens_seen": 2371158016
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028451354062186565,
      "loss": 1.3042,
      "theoretical_loss": 3.4113039466232133,
      "tokens_seen": 2371223552
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000284493480441324,
      "loss": 1.3929,
      "theoretical_loss": 3.411296409767814,
      "tokens_seen": 2371289088
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002844734202607824,
      "loss": 1.4847,
      "theoretical_loss": 3.4112888731790316,
      "tokens_seen": 2371354624
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028445336008024074,
      "loss": 1.3281,
      "theoretical_loss": 3.4112813368568498,
      "tokens_seen": 2371420160
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002844332998996991,
      "loss": 1.3983,
      "theoretical_loss": 3.4112738008012506,
      "tokens_seen": 2371485696
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002844132397191575,
      "loss": 1.3353,
      "theoretical_loss": 3.4112662650122187,
      "tokens_seen": 2371551232
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002843931795386159,
      "loss": 1.4209,
      "theoretical_loss": 3.4112587294897367,
      "tokens_seen": 2371616768
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028437311935807424,
      "loss": 1.3369,
      "theoretical_loss": 3.4112511942337873,
      "tokens_seen": 2371682304
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002843530591775326,
      "loss": 1.41,
      "theoretical_loss": 3.4112436592443545,
      "tokens_seen": 2371747840
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000284332998996991,
      "loss": 1.3617,
      "theoretical_loss": 3.4112361245214213,
      "tokens_seen": 2371813376
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002843129388164494,
      "loss": 1.4142,
      "theoretical_loss": 3.4112285900649706,
      "tokens_seen": 2371878912
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028429287863590775,
      "loss": 1.3484,
      "theoretical_loss": 3.411221055874986,
      "tokens_seen": 2371944448
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002842728184553661,
      "loss": 1.3267,
      "theoretical_loss": 3.41121352195145,
      "tokens_seen": 2372009984
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028425275827482447,
      "loss": 1.3266,
      "theoretical_loss": 3.4112059882943466,
      "tokens_seen": 2372075520
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002842326980942829,
      "loss": 1.3621,
      "theoretical_loss": 3.411198454903659,
      "tokens_seen": 2372141056
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028421263791374125,
      "loss": 1.4345,
      "theoretical_loss": 3.4111909217793697,
      "tokens_seen": 2372206592
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002841925777331996,
      "loss": 1.3736,
      "theoretical_loss": 3.4111833889214624,
      "tokens_seen": 2372272128
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000284172517552658,
      "loss": 1.3711,
      "theoretical_loss": 3.411175856329921,
      "tokens_seen": 2372337664
    },
    {
      "epoch": 0.72,
      "objective/train/advantage_avg": 0.4933544397354126,
      "objective/train/docs_used": 863878,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2882392406463623,
      "objective/train/original_loss": 1.2882393598556519,
      "objective/train/theoretical_loss": 3.4111683240047275,
      "objective/train/tokens_used": 2392863200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24381835758686066,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505748987197876,
      "objective/train/weighted_lm_loss": 1.3533143997192383,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4111683240047275,
      "tokens_seen": 2372403200
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002841524573721164,
      "loss": 1.3541,
      "theoretical_loss": 3.4111683240047275,
      "tokens_seen": 2372403200
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028413239719157475,
      "loss": 1.2996,
      "theoretical_loss": 3.4111607919458655,
      "tokens_seen": 2372468736
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002841123370110331,
      "loss": 1.2552,
      "theoretical_loss": 3.4111532601533185,
      "tokens_seen": 2372534272
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002840922768304915,
      "loss": 1.2844,
      "theoretical_loss": 3.4111457286270697,
      "tokens_seen": 2372599808
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002840722166499499,
      "loss": 1.3451,
      "theoretical_loss": 3.4111381973671024,
      "tokens_seen": 2372665344
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028405215646940825,
      "loss": 1.3347,
      "theoretical_loss": 3.4111306663733996,
      "tokens_seen": 2372730880
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002840320962888666,
      "loss": 1.3674,
      "theoretical_loss": 3.411123135645944,
      "tokens_seen": 2372796416
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000284012036108325,
      "loss": 1.3597,
      "theoretical_loss": 3.4111156051847202,
      "tokens_seen": 2372861952
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028399197592778334,
      "loss": 1.3392,
      "theoretical_loss": 3.4111080749897105,
      "tokens_seen": 2372927488
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028397191574724176,
      "loss": 1.4288,
      "theoretical_loss": 3.411100545060898,
      "tokens_seen": 2372993024
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002839518555667001,
      "loss": 1.3071,
      "theoretical_loss": 3.411093015398267,
      "tokens_seen": 2373058560
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002839317953861585,
      "loss": 1.363,
      "theoretical_loss": 3.4110854860017987,
      "tokens_seen": 2373124096
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028391173520561684,
      "loss": 1.4554,
      "theoretical_loss": 3.4110779568714786,
      "tokens_seen": 2373189632
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028389167502507526,
      "loss": 1.406,
      "theoretical_loss": 3.4110704280072888,
      "tokens_seen": 2373255168
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002838716148445336,
      "loss": 1.3662,
      "theoretical_loss": 3.4110628994092127,
      "tokens_seen": 2373320704
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000283851554663992,
      "loss": 1.4123,
      "theoretical_loss": 3.411055371077233,
      "tokens_seen": 2373386240
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028383149448345035,
      "loss": 1.3779,
      "theoretical_loss": 3.4110478430113345,
      "tokens_seen": 2373451776
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002838114343029087,
      "loss": 1.2899,
      "theoretical_loss": 3.4110403152114985,
      "tokens_seen": 2373517312
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002837913741223671,
      "loss": 1.3795,
      "theoretical_loss": 3.41103278767771,
      "tokens_seen": 2373582848
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002837713139418255,
      "loss": 1.37,
      "theoretical_loss": 3.411025260409951,
      "tokens_seen": 2373648384
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028375125376128385,
      "loss": 1.2707,
      "theoretical_loss": 3.4110177334082055,
      "tokens_seen": 2373713920
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002837311935807422,
      "loss": 1.3647,
      "theoretical_loss": 3.411010206672456,
      "tokens_seen": 2373779456
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028371113340020063,
      "loss": 1.381,
      "theoretical_loss": 3.4110026802026865,
      "tokens_seen": 2373844992
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000283691073219659,
      "loss": 1.5372,
      "theoretical_loss": 3.41099515399888,
      "tokens_seen": 2373910528
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028367101303911735,
      "loss": 1.373,
      "theoretical_loss": 3.4109876280610196,
      "tokens_seen": 2373976064
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002836509528585757,
      "loss": 1.2556,
      "theoretical_loss": 3.410980102389089,
      "tokens_seen": 2374041600
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002836308926780341,
      "loss": 1.2285,
      "theoretical_loss": 3.410972576983071,
      "tokens_seen": 2374107136
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002836108324974925,
      "loss": 1.3233,
      "theoretical_loss": 3.410965051842949,
      "tokens_seen": 2374172672
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028359077231695086,
      "loss": 1.3765,
      "theoretical_loss": 3.4109575269687062,
      "tokens_seen": 2374238208
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002835707121364092,
      "loss": 1.3937,
      "theoretical_loss": 3.4109500023603263,
      "tokens_seen": 2374303744
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002835506519558676,
      "loss": 1.5089,
      "theoretical_loss": 3.410942478017792,
      "tokens_seen": 2374369280
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000283530591775326,
      "loss": 1.3983,
      "theoretical_loss": 3.4109349539410867,
      "tokens_seen": 2374434816
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028351053159478436,
      "loss": 1.3459,
      "theoretical_loss": 3.4109274301301937,
      "tokens_seen": 2374500352
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002834904714142427,
      "loss": 1.3619,
      "theoretical_loss": 3.4109199065850966,
      "tokens_seen": 2374565888
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002834704112337011,
      "loss": 1.3206,
      "theoretical_loss": 3.4109123833057784,
      "tokens_seen": 2374631424
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028345035105315945,
      "loss": 1.3738,
      "theoretical_loss": 3.4109048602922223,
      "tokens_seen": 2374696960
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028343029087261786,
      "loss": 1.2962,
      "theoretical_loss": 3.410897337544412,
      "tokens_seen": 2374762496
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002834102306920762,
      "loss": 1.2867,
      "theoretical_loss": 3.41088981506233,
      "tokens_seen": 2374828032
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002833901705115346,
      "loss": 1.3421,
      "theoretical_loss": 3.41088229284596,
      "tokens_seen": 2374893568
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028337011033099295,
      "loss": 1.4261,
      "theoretical_loss": 3.4108747708952856,
      "tokens_seen": 2374959104
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028335005015045137,
      "loss": 1.2937,
      "theoretical_loss": 3.41086724921029,
      "tokens_seen": 2375024640
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028332998996990973,
      "loss": 1.3353,
      "theoretical_loss": 3.410859727790956,
      "tokens_seen": 2375090176
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002833099297893681,
      "loss": 1.4124,
      "theoretical_loss": 3.410852206637267,
      "tokens_seen": 2375155712
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028328986960882645,
      "loss": 1.4078,
      "theoretical_loss": 3.4108446857492067,
      "tokens_seen": 2375221248
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002832698094282848,
      "loss": 1.3693,
      "theoretical_loss": 3.410837165126758,
      "tokens_seen": 2375286784
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028324974924774323,
      "loss": 1.4214,
      "theoretical_loss": 3.4108296447699047,
      "tokens_seen": 2375352320
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002832296890672016,
      "loss": 1.4179,
      "theoretical_loss": 3.410822124678629,
      "tokens_seen": 2375417856
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028320962888665996,
      "loss": 1.3838,
      "theoretical_loss": 3.4108146048529155,
      "tokens_seen": 2375483392
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002831895687061183,
      "loss": 1.3985,
      "theoretical_loss": 3.4108070852927472,
      "tokens_seen": 2375548928
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028316950852557674,
      "loss": 1.3064,
      "theoretical_loss": 3.4107995659981065,
      "tokens_seen": 2375614464
    },
    {
      "epoch": 0.72,
      "objective/train/advantage_avg": 0.491377055644989,
      "objective/train/docs_used": 865150,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3522849082946777,
      "objective/train/original_loss": 1.3522846698760986,
      "objective/train/theoretical_loss": 3.410792046968978,
      "objective/train/tokens_used": 2396140000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2419682741165161,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503677129745483,
      "objective/train/weighted_lm_loss": 1.4204065799713135,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.410792046968978,
      "tokens_seen": 2375680000
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002831494483450351,
      "loss": 1.367,
      "theoretical_loss": 3.410792046968978,
      "tokens_seen": 2375680000
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028312938816449346,
      "loss": 1.2566,
      "theoretical_loss": 3.4107845282053435,
      "tokens_seen": 2375745536
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002831093279839518,
      "loss": 1.4991,
      "theoretical_loss": 3.4107770097071874,
      "tokens_seen": 2375811072
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002830892678034102,
      "loss": 1.3281,
      "theoretical_loss": 3.4107694914744933,
      "tokens_seen": 2375876608
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002830692076228686,
      "loss": 1.3091,
      "theoretical_loss": 3.410761973507243,
      "tokens_seen": 2375942144
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028304914744232696,
      "loss": 1.3551,
      "theoretical_loss": 3.4107544558054217,
      "tokens_seen": 2376007680
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002830290872617854,
      "loss": 1.3969,
      "theoretical_loss": 3.410746938369011,
      "tokens_seen": 2376073216
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028300902708124374,
      "loss": 1.4709,
      "theoretical_loss": 3.410739421197995,
      "tokens_seen": 2376138752
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028298896690070216,
      "loss": 1.3439,
      "theoretical_loss": 3.4107319042923576,
      "tokens_seen": 2376204288
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002829689067201605,
      "loss": 1.3548,
      "theoretical_loss": 3.410724387652081,
      "tokens_seen": 2376269824
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002829488465396189,
      "loss": 1.4649,
      "theoretical_loss": 3.410716871277149,
      "tokens_seen": 2376335360
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028292878635907725,
      "loss": 1.3756,
      "theoretical_loss": 3.410709355167545,
      "tokens_seen": 2376400896
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028290872617853566,
      "loss": 1.2976,
      "theoretical_loss": 3.410701839323252,
      "tokens_seen": 2376466432
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000282888665997994,
      "loss": 1.3501,
      "theoretical_loss": 3.410694323744254,
      "tokens_seen": 2376531968
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002828686058174524,
      "loss": 1.362,
      "theoretical_loss": 3.4106868084305337,
      "tokens_seen": 2376597504
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028284854563691075,
      "loss": 1.3372,
      "theoretical_loss": 3.4106792933820747,
      "tokens_seen": 2376663040
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002828284854563691,
      "loss": 1.287,
      "theoretical_loss": 3.41067177859886,
      "tokens_seen": 2376728576
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002828084252758275,
      "loss": 1.4681,
      "theoretical_loss": 3.4106642640808733,
      "tokens_seen": 2376794112
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002827883650952859,
      "loss": 1.4946,
      "theoretical_loss": 3.4106567498280973,
      "tokens_seen": 2376859648
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028276830491474425,
      "loss": 1.3338,
      "theoretical_loss": 3.4106492358405163,
      "tokens_seen": 2376925184
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002827482447342026,
      "loss": 1.4537,
      "theoretical_loss": 3.410641722118113,
      "tokens_seen": 2376990720
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028272818455366103,
      "loss": 1.4823,
      "theoretical_loss": 3.4106342086608707,
      "tokens_seen": 2377056256
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002827081243731194,
      "loss": 1.3074,
      "theoretical_loss": 3.410626695468773,
      "tokens_seen": 2377121792
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028268806419257775,
      "loss": 1.4324,
      "theoretical_loss": 3.4106191825418035,
      "tokens_seen": 2377187328
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002826680040120361,
      "loss": 1.2721,
      "theoretical_loss": 3.4106116698799447,
      "tokens_seen": 2377252864
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002826479438314945,
      "loss": 1.339,
      "theoretical_loss": 3.4106041574831805,
      "tokens_seen": 2377318400
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002826278836509529,
      "loss": 1.3673,
      "theoretical_loss": 3.410596645351494,
      "tokens_seen": 2377383936
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028260782347041126,
      "loss": 1.4386,
      "theoretical_loss": 3.410589133484869,
      "tokens_seen": 2377449472
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002825877632898696,
      "loss": 1.3197,
      "theoretical_loss": 3.410581621883288,
      "tokens_seen": 2377515008
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000282567703109328,
      "loss": 1.3045,
      "theoretical_loss": 3.410574110546735,
      "tokens_seen": 2377580544
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002825476429287864,
      "loss": 1.3153,
      "theoretical_loss": 3.4105665994751933,
      "tokens_seen": 2377646080
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028252758274824476,
      "loss": 1.3824,
      "theoretical_loss": 3.4105590886686463,
      "tokens_seen": 2377711616
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002825075225677031,
      "loss": 1.3438,
      "theoretical_loss": 3.4105515781270768,
      "tokens_seen": 2377777152
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002824874623871615,
      "loss": 1.3245,
      "theoretical_loss": 3.410544067850469,
      "tokens_seen": 2377842688
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028246740220661985,
      "loss": 1.2807,
      "theoretical_loss": 3.4105365578388054,
      "tokens_seen": 2377908224
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028244734202607826,
      "loss": 1.4237,
      "theoretical_loss": 3.4105290480920702,
      "tokens_seen": 2377973760
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002824272818455366,
      "loss": 1.3975,
      "theoretical_loss": 3.410521538610246,
      "tokens_seen": 2378039296
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000282407221664995,
      "loss": 1.3736,
      "theoretical_loss": 3.4105140293933163,
      "tokens_seen": 2378104832
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028238716148445335,
      "loss": 1.3945,
      "theoretical_loss": 3.410506520441265,
      "tokens_seen": 2378170368
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028236710130391177,
      "loss": 1.4458,
      "theoretical_loss": 3.4104990117540748,
      "tokens_seen": 2378235904
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028234704112337013,
      "loss": 1.3661,
      "theoretical_loss": 3.410491503331729,
      "tokens_seen": 2378301440
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002823269809428285,
      "loss": 1.3876,
      "theoretical_loss": 3.4104839951742116,
      "tokens_seen": 2378366976
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028230692076228685,
      "loss": 1.2836,
      "theoretical_loss": 3.410476487281506,
      "tokens_seen": 2378432512
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002822868605817452,
      "loss": 1.3863,
      "theoretical_loss": 3.410468979653595,
      "tokens_seen": 2378498048
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028226680040120363,
      "loss": 1.3352,
      "theoretical_loss": 3.410461472290462,
      "tokens_seen": 2378563584
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000282246740220662,
      "loss": 1.3819,
      "theoretical_loss": 3.41045396519209,
      "tokens_seen": 2378629120
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028222668004012036,
      "loss": 1.2808,
      "theoretical_loss": 3.410446458358464,
      "tokens_seen": 2378694656
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002822066198595787,
      "loss": 1.3684,
      "theoretical_loss": 3.4104389517895655,
      "tokens_seen": 2378760192
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028218655967903714,
      "loss": 1.3542,
      "theoretical_loss": 3.4104314454853792,
      "tokens_seen": 2378825728
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002821664994984955,
      "loss": 1.4057,
      "theoretical_loss": 3.4104239394458875,
      "tokens_seen": 2378891264
    },
    {
      "epoch": 0.72,
      "objective/train/advantage_avg": 0.4911629259586334,
      "objective/train/docs_used": 866449,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1117297410964966,
      "objective/train/original_loss": 1.1117297410964966,
      "objective/train/theoretical_loss": 3.4104164336710747,
      "objective/train/tokens_used": 2399416800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24161279201507568,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050344467163086,
      "objective/train/weighted_lm_loss": 1.1677311658859253,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4104164336710747,
      "tokens_seen": 2378956800
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028214643931795386,
      "loss": 1.2754,
      "theoretical_loss": 3.4104164336710747,
      "tokens_seen": 2378956800
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002821263791374122,
      "loss": 1.338,
      "theoretical_loss": 3.410408928160923,
      "tokens_seen": 2379022336
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002821063189568706,
      "loss": 1.3769,
      "theoretical_loss": 3.4104014229154167,
      "tokens_seen": 2379087872
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000282086258776329,
      "loss": 1.3442,
      "theoretical_loss": 3.410393917934539,
      "tokens_seen": 2379153408
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028206619859578736,
      "loss": 1.3137,
      "theoretical_loss": 3.4103864132182733,
      "tokens_seen": 2379218944
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002820461384152457,
      "loss": 1.2679,
      "theoretical_loss": 3.410378908766603,
      "tokens_seen": 2379284480
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002820260782347041,
      "loss": 1.5057,
      "theoretical_loss": 3.4103714045795113,
      "tokens_seen": 2379350016
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002820060180541625,
      "loss": 1.3965,
      "theoretical_loss": 3.4103639006569813,
      "tokens_seen": 2379415552
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028198595787362087,
      "loss": 1.3676,
      "theoretical_loss": 3.410356396998997,
      "tokens_seen": 2379481088
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028196589769307923,
      "loss": 1.4133,
      "theoretical_loss": 3.4103488936055415,
      "tokens_seen": 2379546624
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002819458375125376,
      "loss": 1.5189,
      "theoretical_loss": 3.4103413904765985,
      "tokens_seen": 2379612160
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000281925777331996,
      "loss": 1.4296,
      "theoretical_loss": 3.4103338876121505,
      "tokens_seen": 2379677696
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028190571715145437,
      "loss": 1.4757,
      "theoretical_loss": 3.410326385012182,
      "tokens_seen": 2379743232
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028188565697091273,
      "loss": 1.3792,
      "theoretical_loss": 3.410318882676676,
      "tokens_seen": 2379808768
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002818655967903711,
      "loss": 1.2884,
      "theoretical_loss": 3.4103113806056156,
      "tokens_seen": 2379874304
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028184553660982946,
      "loss": 1.4593,
      "theoretical_loss": 3.410303878798984,
      "tokens_seen": 2379939840
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002818254764292879,
      "loss": 1.4088,
      "theoretical_loss": 3.4102963772567656,
      "tokens_seen": 2380005376
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028180541624874624,
      "loss": 1.3692,
      "theoretical_loss": 3.410288875978943,
      "tokens_seen": 2380070912
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002817853560682046,
      "loss": 1.4899,
      "theoretical_loss": 3.4102813749655,
      "tokens_seen": 2380136448
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028176529588766296,
      "loss": 1.345,
      "theoretical_loss": 3.4102738742164194,
      "tokens_seen": 2380201984
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002817452357071214,
      "loss": 1.3,
      "theoretical_loss": 3.4102663737316847,
      "tokens_seen": 2380267520
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028172517552657974,
      "loss": 1.2935,
      "theoretical_loss": 3.4102588735112804,
      "tokens_seen": 2380333056
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002817051153460381,
      "loss": 1.3475,
      "theoretical_loss": 3.4102513735551887,
      "tokens_seen": 2380398592
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028168505516549646,
      "loss": 1.2804,
      "theoretical_loss": 3.4102438738633936,
      "tokens_seen": 2380464128
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002816649949849548,
      "loss": 1.4615,
      "theoretical_loss": 3.410236374435878,
      "tokens_seen": 2380529664
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028164493480441324,
      "loss": 1.3655,
      "theoretical_loss": 3.410228875272626,
      "tokens_seen": 2380595200
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002816248746238716,
      "loss": 1.3973,
      "theoretical_loss": 3.41022137637362,
      "tokens_seen": 2380660736
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028160481444332997,
      "loss": 1.284,
      "theoretical_loss": 3.410213877738845,
      "tokens_seen": 2380726272
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028158475426278833,
      "loss": 1.3491,
      "theoretical_loss": 3.410206379368283,
      "tokens_seen": 2380791808
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028156469408224674,
      "loss": 1.2795,
      "theoretical_loss": 3.410198881261918,
      "tokens_seen": 2380857344
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002815446339017051,
      "loss": 1.3707,
      "theoretical_loss": 3.4101913834197335,
      "tokens_seen": 2380922880
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002815245737211635,
      "loss": 1.4159,
      "theoretical_loss": 3.4101838858417124,
      "tokens_seen": 2380988416
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002815045135406219,
      "loss": 1.4045,
      "theoretical_loss": 3.4101763885278387,
      "tokens_seen": 2381053952
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028148445336008025,
      "loss": 1.3424,
      "theoretical_loss": 3.4101688914780954,
      "tokens_seen": 2381119488
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028146439317953866,
      "loss": 1.4552,
      "theoretical_loss": 3.4101613946924663,
      "tokens_seen": 2381185024
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000281444332998997,
      "loss": 1.4502,
      "theoretical_loss": 3.4101538981709347,
      "tokens_seen": 2381250560
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002814242728184554,
      "loss": 1.3414,
      "theoretical_loss": 3.410146401913484,
      "tokens_seen": 2381316096
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028140421263791375,
      "loss": 1.4518,
      "theoretical_loss": 3.410138905920097,
      "tokens_seen": 2381381632
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028138415245737217,
      "loss": 1.2978,
      "theoretical_loss": 3.4101314101907585,
      "tokens_seen": 2381447168
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028136409227683053,
      "loss": 1.3014,
      "theoretical_loss": 3.4101239147254505,
      "tokens_seen": 2381512704
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002813440320962889,
      "loss": 1.3514,
      "theoretical_loss": 3.410116419524158,
      "tokens_seen": 2381578240
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028132397191574725,
      "loss": 1.316,
      "theoretical_loss": 3.4101089245868628,
      "tokens_seen": 2381643776
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002813039117352056,
      "loss": 1.4536,
      "theoretical_loss": 3.4101014299135493,
      "tokens_seen": 2381709312
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028128385155466403,
      "loss": 1.2803,
      "theoretical_loss": 3.4100939355042006,
      "tokens_seen": 2381774848
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002812637913741224,
      "loss": 1.3968,
      "theoretical_loss": 3.4100864413588003,
      "tokens_seen": 2381840384
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028124373119358076,
      "loss": 1.3289,
      "theoretical_loss": 3.4100789474773316,
      "tokens_seen": 2381905920
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002812236710130391,
      "loss": 1.4101,
      "theoretical_loss": 3.4100714538597785,
      "tokens_seen": 2381971456
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028120361083249754,
      "loss": 1.3063,
      "theoretical_loss": 3.410063960506124,
      "tokens_seen": 2382036992
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002811835506519559,
      "loss": 1.3526,
      "theoretical_loss": 3.410056467416352,
      "tokens_seen": 2382102528
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028116349047141426,
      "loss": 1.3527,
      "theoretical_loss": 3.410048974590445,
      "tokens_seen": 2382168064
    },
    {
      "epoch": 0.72,
      "objective/train/advantage_avg": 0.4892081618309021,
      "objective/train/docs_used": 867590,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3218554258346558,
      "objective/train/original_loss": 1.3218555450439453,
      "objective/train/theoretical_loss": 3.410041482028387,
      "objective/train/tokens_used": 2402693600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2400255799293518,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501409769058228,
      "objective/train/weighted_lm_loss": 1.3880363702774048,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.410041482028387,
      "tokens_seen": 2382233600
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002811434302908726,
      "loss": 1.3616,
      "theoretical_loss": 3.410041482028387,
      "tokens_seen": 2382233600
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000281123370110331,
      "loss": 1.3266,
      "theoretical_loss": 3.4100339897301617,
      "tokens_seen": 2382299136
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002811033099297894,
      "loss": 1.5085,
      "theoretical_loss": 3.4100264976957524,
      "tokens_seen": 2382364672
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028108324974924776,
      "loss": 1.4979,
      "theoretical_loss": 3.4100190059251423,
      "tokens_seen": 2382430208
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002810631895687061,
      "loss": 1.4288,
      "theoretical_loss": 3.4100115144183154,
      "tokens_seen": 2382495744
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002810431293881645,
      "loss": 1.2653,
      "theoretical_loss": 3.4100040231752544,
      "tokens_seen": 2382561280
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002810230692076229,
      "loss": 1.321,
      "theoretical_loss": 3.4099965321959433,
      "tokens_seen": 2382626816
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028100300902708127,
      "loss": 1.3868,
      "theoretical_loss": 3.4099890414803653,
      "tokens_seen": 2382692352
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028098294884653963,
      "loss": 1.4178,
      "theoretical_loss": 3.409981551028504,
      "tokens_seen": 2382757888
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000280962888665998,
      "loss": 1.345,
      "theoretical_loss": 3.409974060840343,
      "tokens_seen": 2382823424
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028094282848545635,
      "loss": 1.2879,
      "theoretical_loss": 3.4099665709158655,
      "tokens_seen": 2382888960
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028092276830491477,
      "loss": 1.3769,
      "theoretical_loss": 3.409959081255055,
      "tokens_seen": 2382954496
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028090270812437313,
      "loss": 1.3527,
      "theoretical_loss": 3.409951591857895,
      "tokens_seen": 2383020032
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002808826479438315,
      "loss": 1.3729,
      "theoretical_loss": 3.4099441027243693,
      "tokens_seen": 2383085568
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028086258776328986,
      "loss": 1.3542,
      "theoretical_loss": 3.409936613854461,
      "tokens_seen": 2383151104
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002808425275827483,
      "loss": 1.4483,
      "theoretical_loss": 3.4099291252481536,
      "tokens_seen": 2383216640
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028082246740220664,
      "loss": 1.4306,
      "theoretical_loss": 3.4099216369054304,
      "tokens_seen": 2383282176
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000280802407221665,
      "loss": 1.3321,
      "theoretical_loss": 3.4099141488262754,
      "tokens_seen": 2383347712
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028078234704112336,
      "loss": 1.373,
      "theoretical_loss": 3.4099066610106714,
      "tokens_seen": 2383413248
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002807622868605818,
      "loss": 1.2806,
      "theoretical_loss": 3.4098991734586024,
      "tokens_seen": 2383478784
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028074222668004014,
      "loss": 1.3457,
      "theoretical_loss": 3.409891686170052,
      "tokens_seen": 2383544320
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002807221664994985,
      "loss": 1.4646,
      "theoretical_loss": 3.409884199145003,
      "tokens_seen": 2383609856
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028070210631895686,
      "loss": 1.3708,
      "theoretical_loss": 3.40987671238344,
      "tokens_seen": 2383675392
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002806820461384152,
      "loss": 1.3144,
      "theoretical_loss": 3.4098692258853447,
      "tokens_seen": 2383740928
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028066198595787364,
      "loss": 1.3496,
      "theoretical_loss": 3.4098617396507023,
      "tokens_seen": 2383806464
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000280641925777332,
      "loss": 1.4212,
      "theoretical_loss": 3.409854253679496,
      "tokens_seen": 2383872000
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028062186559679037,
      "loss": 1.3911,
      "theoretical_loss": 3.4098467679717084,
      "tokens_seen": 2383937536
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028060180541624873,
      "loss": 1.3483,
      "theoretical_loss": 3.409839282527323,
      "tokens_seen": 2384003072
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028058174523570715,
      "loss": 1.4317,
      "theoretical_loss": 3.4098317973463246,
      "tokens_seen": 2384068608
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002805616850551655,
      "loss": 1.3497,
      "theoretical_loss": 3.4098243124286958,
      "tokens_seen": 2384134144
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028054162487462387,
      "loss": 1.3906,
      "theoretical_loss": 3.4098168277744203,
      "tokens_seen": 2384199680
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028052156469408223,
      "loss": 1.4146,
      "theoretical_loss": 3.4098093433834813,
      "tokens_seen": 2384265216
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002805015045135406,
      "loss": 1.2968,
      "theoretical_loss": 3.4098018592558628,
      "tokens_seen": 2384330752
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000280481444332999,
      "loss": 1.4277,
      "theoretical_loss": 3.4097943753915474,
      "tokens_seen": 2384396288
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028046138415245737,
      "loss": 1.3876,
      "theoretical_loss": 3.4097868917905196,
      "tokens_seen": 2384461824
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028044132397191574,
      "loss": 1.3161,
      "theoretical_loss": 3.4097794084527626,
      "tokens_seen": 2384527360
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002804212637913741,
      "loss": 1.3115,
      "theoretical_loss": 3.4097719253782595,
      "tokens_seen": 2384592896
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002804012036108325,
      "loss": 1.3514,
      "theoretical_loss": 3.409764442566994,
      "tokens_seen": 2384658432
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002803811434302909,
      "loss": 1.4292,
      "theoretical_loss": 3.40975696001895,
      "tokens_seen": 2384723968
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028036108324974924,
      "loss": 1.3851,
      "theoretical_loss": 3.4097494777341106,
      "tokens_seen": 2384789504
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002803410230692076,
      "loss": 1.386,
      "theoretical_loss": 3.4097419957124595,
      "tokens_seen": 2384855040
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028032096288866596,
      "loss": 1.5324,
      "theoretical_loss": 3.40973451395398,
      "tokens_seen": 2384920576
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002803009027081244,
      "loss": 1.3586,
      "theoretical_loss": 3.4097270324586555,
      "tokens_seen": 2384986112
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028028084252758274,
      "loss": 1.4185,
      "theoretical_loss": 3.40971955122647,
      "tokens_seen": 2385051648
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002802607823470411,
      "loss": 1.3044,
      "theoretical_loss": 3.409712070257407,
      "tokens_seen": 2385117184
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028024072216649947,
      "loss": 1.3276,
      "theoretical_loss": 3.4097045895514495,
      "tokens_seen": 2385182720
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002802206619859579,
      "loss": 1.375,
      "theoretical_loss": 3.4096971091085817,
      "tokens_seen": 2385248256
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028020060180541624,
      "loss": 1.3595,
      "theoretical_loss": 3.409689628928786,
      "tokens_seen": 2385313792
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002801805416248746,
      "loss": 1.5323,
      "theoretical_loss": 3.4096821490120472,
      "tokens_seen": 2385379328
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028016048144433297,
      "loss": 1.3995,
      "theoretical_loss": 3.409674669358348,
      "tokens_seen": 2385444864
    },
    {
      "epoch": 0.72,
      "objective/train/advantage_avg": 0.4905834496021271,
      "objective/train/docs_used": 868828,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2843033075332642,
      "objective/train/original_loss": 1.2843031883239746,
      "objective/train/theoretical_loss": 3.4096671899676725,
      "objective/train/tokens_used": 2405970400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24114829301834106,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502841472625732,
      "objective/train/weighted_lm_loss": 1.3487316370010376,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4096671899676725,
      "tokens_seen": 2385510400
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028014042126379133,
      "loss": 1.3948,
      "theoretical_loss": 3.4096671899676725,
      "tokens_seen": 2385510400
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028012036108324975,
      "loss": 1.3329,
      "theoretical_loss": 3.4096597108400037,
      "tokens_seen": 2385575936
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002801003009027081,
      "loss": 1.4729,
      "theoretical_loss": 3.4096522319753255,
      "tokens_seen": 2385641472
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028008024072216647,
      "loss": 1.3632,
      "theoretical_loss": 3.4096447533736214,
      "tokens_seen": 2385707008
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028006018054162483,
      "loss": 1.3147,
      "theoretical_loss": 3.4096372750348745,
      "tokens_seen": 2385772544
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028004012036108325,
      "loss": 1.403,
      "theoretical_loss": 3.4096297969590688,
      "tokens_seen": 2385838080
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028002006018054167,
      "loss": 1.3412,
      "theoretical_loss": 3.4096223191461874,
      "tokens_seen": 2385903616
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00028000000000000003,
      "loss": 1.3975,
      "theoretical_loss": 3.4096148415962144,
      "tokens_seen": 2385969152
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002799799398194584,
      "loss": 1.3942,
      "theoretical_loss": 3.409607364309133,
      "tokens_seen": 2386034688
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027995987963891675,
      "loss": 1.4491,
      "theoretical_loss": 3.4095998872849265,
      "tokens_seen": 2386100224
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027993981945837517,
      "loss": 1.4067,
      "theoretical_loss": 3.4095924105235786,
      "tokens_seen": 2386165760
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027991975927783353,
      "loss": 1.517,
      "theoretical_loss": 3.4095849340250735,
      "tokens_seen": 2386231296
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002798996990972919,
      "loss": 1.4199,
      "theoretical_loss": 3.4095774577893936,
      "tokens_seen": 2386296832
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027987963891675026,
      "loss": 1.3236,
      "theoretical_loss": 3.4095699818165235,
      "tokens_seen": 2386362368
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002798595787362087,
      "loss": 1.3312,
      "theoretical_loss": 3.4095625061064463,
      "tokens_seen": 2386427904
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027983951855566704,
      "loss": 1.3805,
      "theoretical_loss": 3.409555030659145,
      "tokens_seen": 2386493440
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002798194583751254,
      "loss": 1.3913,
      "theoretical_loss": 3.4095475554746044,
      "tokens_seen": 2386558976
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027979939819458376,
      "loss": 1.3859,
      "theoretical_loss": 3.409540080552807,
      "tokens_seen": 2386624512
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002797793380140422,
      "loss": 1.346,
      "theoretical_loss": 3.4095326058937365,
      "tokens_seen": 2386690048
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027975927783350054,
      "loss": 1.4115,
      "theoretical_loss": 3.4095251314973765,
      "tokens_seen": 2386755584
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002797392176529589,
      "loss": 1.4681,
      "theoretical_loss": 3.409517657363711,
      "tokens_seen": 2386821120
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027971915747241726,
      "loss": 1.4044,
      "theoretical_loss": 3.4095101834927233,
      "tokens_seen": 2386886656
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002796990972918756,
      "loss": 1.3531,
      "theoretical_loss": 3.4095027098843964,
      "tokens_seen": 2386952192
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027967903711133404,
      "loss": 1.3755,
      "theoretical_loss": 3.409495236538715,
      "tokens_seen": 2387017728
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002796589769307924,
      "loss": 1.464,
      "theoretical_loss": 3.4094877634556617,
      "tokens_seen": 2387083264
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027963891675025077,
      "loss": 1.3926,
      "theoretical_loss": 3.40948029063522,
      "tokens_seen": 2387148800
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027961885656970913,
      "loss": 1.4067,
      "theoretical_loss": 3.4094728180773743,
      "tokens_seen": 2387214336
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027959879638916755,
      "loss": 1.428,
      "theoretical_loss": 3.409465345782108,
      "tokens_seen": 2387279872
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002795787362086259,
      "loss": 1.407,
      "theoretical_loss": 3.4094578737494037,
      "tokens_seen": 2387345408
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027955867602808427,
      "loss": 1.3977,
      "theoretical_loss": 3.409450401979246,
      "tokens_seen": 2387410944
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027953861584754263,
      "loss": 1.4472,
      "theoretical_loss": 3.409442930471618,
      "tokens_seen": 2387476480
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000279518555667001,
      "loss": 1.4008,
      "theoretical_loss": 3.4094354592265033,
      "tokens_seen": 2387542016
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002794984954864594,
      "loss": 1.3914,
      "theoretical_loss": 3.4094279882438854,
      "tokens_seen": 2387607552
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002794784353059178,
      "loss": 1.2669,
      "theoretical_loss": 3.409420517523748,
      "tokens_seen": 2387673088
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027945837512537614,
      "loss": 1.3368,
      "theoretical_loss": 3.409413047066075,
      "tokens_seen": 2387738624
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002794383149448345,
      "loss": 1.3659,
      "theoretical_loss": 3.4094055768708493,
      "tokens_seen": 2387804160
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002794182547642929,
      "loss": 1.2961,
      "theoretical_loss": 3.409398106938055,
      "tokens_seen": 2387869696
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002793981945837513,
      "loss": 1.4343,
      "theoretical_loss": 3.4093906372676757,
      "tokens_seen": 2387935232
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027937813440320964,
      "loss": 1.2658,
      "theoretical_loss": 3.409383167859694,
      "tokens_seen": 2388000768
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000279358074222668,
      "loss": 1.4267,
      "theoretical_loss": 3.4093756987140953,
      "tokens_seen": 2388066304
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027933801404212636,
      "loss": 1.5043,
      "theoretical_loss": 3.4093682298308616,
      "tokens_seen": 2388131840
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002793179538615848,
      "loss": 1.3136,
      "theoretical_loss": 3.409360761209977,
      "tokens_seen": 2388197376
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027929789368104314,
      "loss": 1.3333,
      "theoretical_loss": 3.4093532928514256,
      "tokens_seen": 2388262912
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002792778335005015,
      "loss": 1.4093,
      "theoretical_loss": 3.4093458247551904,
      "tokens_seen": 2388328448
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027925777331995987,
      "loss": 1.4406,
      "theoretical_loss": 3.4093383569212548,
      "tokens_seen": 2388393984
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002792377131394183,
      "loss": 1.3351,
      "theoretical_loss": 3.4093308893496026,
      "tokens_seen": 2388459520
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027921765295887664,
      "loss": 1.3925,
      "theoretical_loss": 3.409323422040218,
      "tokens_seen": 2388525056
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000279197592778335,
      "loss": 1.3183,
      "theoretical_loss": 3.4093159549930836,
      "tokens_seen": 2388590592
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027917753259779337,
      "loss": 1.4713,
      "theoretical_loss": 3.4093084882081834,
      "tokens_seen": 2388656128
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027915747241725173,
      "loss": 1.4413,
      "theoretical_loss": 3.4093010216855015,
      "tokens_seen": 2388721664
    },
    {
      "epoch": 0.72,
      "objective/train/advantage_avg": 0.4933336079120636,
      "objective/train/docs_used": 870071,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3479670286178589,
      "objective/train/original_loss": 1.3479667901992798,
      "objective/train/theoretical_loss": 3.409293555425021,
      "objective/train/tokens_used": 2409247200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24408937990665436,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505741834640503,
      "objective/train/weighted_lm_loss": 1.4160805940628052,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.409293555425021,
      "tokens_seen": 2388787200
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027913741223671015,
      "loss": 1.5155,
      "theoretical_loss": 3.409293555425021,
      "tokens_seen": 2388787200
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002791173520561685,
      "loss": 1.3454,
      "theoretical_loss": 3.4092860894267254,
      "tokens_seen": 2388852736
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027909729187562687,
      "loss": 1.6466,
      "theoretical_loss": 3.4092786236905983,
      "tokens_seen": 2388918272
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027907723169508523,
      "loss": 1.3003,
      "theoretical_loss": 3.4092711582166237,
      "tokens_seen": 2388983808
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027905717151454365,
      "loss": 1.3868,
      "theoretical_loss": 3.4092636930047853,
      "tokens_seen": 2389049344
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000279037111334002,
      "loss": 1.3814,
      "theoretical_loss": 3.4092562280550656,
      "tokens_seen": 2389114880
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002790170511534604,
      "loss": 1.3104,
      "theoretical_loss": 3.4092487633674495,
      "tokens_seen": 2389180416
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027899699097291874,
      "loss": 1.4137,
      "theoretical_loss": 3.4092412989419203,
      "tokens_seen": 2389245952
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002789769307923771,
      "loss": 1.3536,
      "theoretical_loss": 3.409233834778461,
      "tokens_seen": 2389311488
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002789568706118355,
      "loss": 1.2926,
      "theoretical_loss": 3.4092263708770556,
      "tokens_seen": 2389377024
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002789368104312939,
      "loss": 1.4629,
      "theoretical_loss": 3.4092189072376877,
      "tokens_seen": 2389442560
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027891675025075224,
      "loss": 1.3369,
      "theoretical_loss": 3.4092114438603414,
      "tokens_seen": 2389508096
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002788966900702106,
      "loss": 1.417,
      "theoretical_loss": 3.409203980744999,
      "tokens_seen": 2389573632
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000278876629889669,
      "loss": 1.4836,
      "theoretical_loss": 3.4091965178916457,
      "tokens_seen": 2389639168
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002788565697091274,
      "loss": 1.4034,
      "theoretical_loss": 3.4091890553002644,
      "tokens_seen": 2389704704
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027883650952858574,
      "loss": 1.4485,
      "theoretical_loss": 3.4091815929708384,
      "tokens_seen": 2389770240
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002788164493480441,
      "loss": 1.3791,
      "theoretical_loss": 3.409174130903352,
      "tokens_seen": 2389835776
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027879638916750247,
      "loss": 1.4163,
      "theoretical_loss": 3.4091666690977878,
      "tokens_seen": 2389901312
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002787763289869609,
      "loss": 1.3895,
      "theoretical_loss": 3.4091592075541306,
      "tokens_seen": 2389966848
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027875626880641925,
      "loss": 1.3336,
      "theoretical_loss": 3.4091517462723635,
      "tokens_seen": 2390032384
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002787362086258776,
      "loss": 1.4127,
      "theoretical_loss": 3.40914428525247,
      "tokens_seen": 2390097920
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027871614844533597,
      "loss": 1.3519,
      "theoretical_loss": 3.409136824494434,
      "tokens_seen": 2390163456
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002786960882647944,
      "loss": 1.3314,
      "theoretical_loss": 3.4091293639982387,
      "tokens_seen": 2390228992
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027867602808425275,
      "loss": 1.3807,
      "theoretical_loss": 3.409121903763868,
      "tokens_seen": 2390294528
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002786559679037111,
      "loss": 1.3558,
      "theoretical_loss": 3.4091144437913057,
      "tokens_seen": 2390360064
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002786359077231695,
      "loss": 1.3615,
      "theoretical_loss": 3.4091069840805353,
      "tokens_seen": 2390425600
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002786158475426279,
      "loss": 1.2941,
      "theoretical_loss": 3.4090995246315403,
      "tokens_seen": 2390491136
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027859578736208625,
      "loss": 1.3135,
      "theoretical_loss": 3.4090920654443044,
      "tokens_seen": 2390556672
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002785757271815446,
      "loss": 1.3201,
      "theoretical_loss": 3.4090846065188116,
      "tokens_seen": 2390622208
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000278555667001003,
      "loss": 1.4071,
      "theoretical_loss": 3.409077147855045,
      "tokens_seen": 2390687744
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027853560682046134,
      "loss": 1.3438,
      "theoretical_loss": 3.4090696894529886,
      "tokens_seen": 2390753280
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002785155466399198,
      "loss": 1.2999,
      "theoretical_loss": 3.409062231312626,
      "tokens_seen": 2390818816
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002784954864593782,
      "loss": 1.3929,
      "theoretical_loss": 3.409054773433941,
      "tokens_seen": 2390884352
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027847542627883654,
      "loss": 1.3761,
      "theoretical_loss": 3.409047315816917,
      "tokens_seen": 2390949888
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002784553660982949,
      "loss": 1.4683,
      "theoretical_loss": 3.409039858461537,
      "tokens_seen": 2391015424
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002784353059177533,
      "loss": 1.5177,
      "theoretical_loss": 3.4090324013677855,
      "tokens_seen": 2391080960
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002784152457372117,
      "loss": 1.4328,
      "theoretical_loss": 3.409024944535646,
      "tokens_seen": 2391146496
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027839518555667004,
      "loss": 1.3899,
      "theoretical_loss": 3.4090174879651025,
      "tokens_seen": 2391212032
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002783751253761284,
      "loss": 1.2706,
      "theoretical_loss": 3.4090100316561385,
      "tokens_seen": 2391277568
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027835506519558676,
      "loss": 1.3393,
      "theoretical_loss": 3.4090025756087368,
      "tokens_seen": 2391343104
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002783350050150452,
      "loss": 1.447,
      "theoretical_loss": 3.4089951198228814,
      "tokens_seen": 2391408640
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027831494483450354,
      "loss": 1.4453,
      "theoretical_loss": 3.4089876642985573,
      "tokens_seen": 2391474176
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002782948846539619,
      "loss": 1.3446,
      "theoretical_loss": 3.408980209035746,
      "tokens_seen": 2391539712
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027827482447342027,
      "loss": 1.4468,
      "theoretical_loss": 3.408972754034433,
      "tokens_seen": 2391605248
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002782547642928787,
      "loss": 1.4642,
      "theoretical_loss": 3.408965299294601,
      "tokens_seen": 2391670784
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027823470411233705,
      "loss": 1.323,
      "theoretical_loss": 3.408957844816234,
      "tokens_seen": 2391736320
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002782146439317954,
      "loss": 1.3402,
      "theoretical_loss": 3.4089503905993155,
      "tokens_seen": 2391801856
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027819458375125377,
      "loss": 1.3848,
      "theoretical_loss": 3.408942936643829,
      "tokens_seen": 2391867392
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027817452357071213,
      "loss": 1.3698,
      "theoretical_loss": 3.4089354829497585,
      "tokens_seen": 2391932928
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027815446339017055,
      "loss": 1.3473,
      "theoretical_loss": 3.408928029517088,
      "tokens_seen": 2391998464
    },
    {
      "epoch": 0.72,
      "objective/train/advantage_avg": 0.4913603961467743,
      "objective/train/docs_used": 871244,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4427021741867065,
      "objective/train/original_loss": 1.442702054977417,
      "objective/train/theoretical_loss": 3.4089205763458,
      "objective/train/tokens_used": 2412524000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2418006807565689,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503653287887573,
      "objective/train/weighted_lm_loss": 1.515334129333496,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.4089205763458,
      "tokens_seen": 2392064000
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002781344032096289,
      "loss": 1.4657,
      "theoretical_loss": 3.4089205763458,
      "tokens_seen": 2392064000
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002781143430290873,
      "loss": 1.3831,
      "theoretical_loss": 3.4089131234358794,
      "tokens_seen": 2392129536
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027809428284854564,
      "loss": 1.4064,
      "theoretical_loss": 3.408905670787309,
      "tokens_seen": 2392195072
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027807422266800405,
      "loss": 1.4649,
      "theoretical_loss": 3.408898218400073,
      "tokens_seen": 2392260608
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002780541624874624,
      "loss": 1.3333,
      "theoretical_loss": 3.408890766274155,
      "tokens_seen": 2392326144
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002780341023069208,
      "loss": 1.4838,
      "theoretical_loss": 3.4088833144095387,
      "tokens_seen": 2392391680
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.00027801404212637914,
      "loss": 1.3752,
      "theoretical_loss": 3.408875862806208,
      "tokens_seen": 2392457216
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002779939819458375,
      "loss": 1.3393,
      "theoretical_loss": 3.4088684114641454,
      "tokens_seen": 2392522752
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002779739217652959,
      "loss": 1.3639,
      "theoretical_loss": 3.408860960383336,
      "tokens_seen": 2392588288
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002779538615847543,
      "loss": 1.3684,
      "theoretical_loss": 3.408853509563763,
      "tokens_seen": 2392653824
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027793380140421264,
      "loss": 1.3694,
      "theoretical_loss": 3.40884605900541,
      "tokens_seen": 2392719360
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000277913741223671,
      "loss": 1.4169,
      "theoretical_loss": 3.4088386087082605,
      "tokens_seen": 2392784896
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002778936810431294,
      "loss": 1.3295,
      "theoretical_loss": 3.4088311586722986,
      "tokens_seen": 2392850432
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002778736208625878,
      "loss": 1.3954,
      "theoretical_loss": 3.4088237088975073,
      "tokens_seen": 2392915968
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027785356068204614,
      "loss": 1.4213,
      "theoretical_loss": 3.4088162593838716,
      "tokens_seen": 2392981504
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002778335005015045,
      "loss": 1.3003,
      "theoretical_loss": 3.4088088101313736,
      "tokens_seen": 2393047040
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027781344032096287,
      "loss": 1.3861,
      "theoretical_loss": 3.4088013611399983,
      "tokens_seen": 2393112576
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002777933801404213,
      "loss": 1.4077,
      "theoretical_loss": 3.408793912409729,
      "tokens_seen": 2393178112
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027777331995987965,
      "loss": 1.3255,
      "theoretical_loss": 3.4087864639405487,
      "tokens_seen": 2393243648
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000277753259779338,
      "loss": 1.4718,
      "theoretical_loss": 3.408779015732442,
      "tokens_seen": 2393309184
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027773319959879637,
      "loss": 1.3515,
      "theoretical_loss": 3.408771567785392,
      "tokens_seen": 2393374720
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002777131394182548,
      "loss": 1.4081,
      "theoretical_loss": 3.408764120099383,
      "tokens_seen": 2393440256
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027769307923771315,
      "loss": 1.3341,
      "theoretical_loss": 3.4087566726743983,
      "tokens_seen": 2393505792
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002776730190571715,
      "loss": 1.4242,
      "theoretical_loss": 3.4087492255104213,
      "tokens_seen": 2393571328
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002776529588766299,
      "loss": 1.3583,
      "theoretical_loss": 3.4087417786074363,
      "tokens_seen": 2393636864
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002776328986960883,
      "loss": 1.394,
      "theoretical_loss": 3.408734331965427,
      "tokens_seen": 2393702400
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027761283851554665,
      "loss": 1.4666,
      "theoretical_loss": 3.4087268855843766,
      "tokens_seen": 2393767936
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000277592778335005,
      "loss": 1.4444,
      "theoretical_loss": 3.4087194394642695,
      "tokens_seen": 2393833472
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002775727181544634,
      "loss": 1.3068,
      "theoretical_loss": 3.4087119936050887,
      "tokens_seen": 2393899008
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027755265797392174,
      "loss": 1.4855,
      "theoretical_loss": 3.4087045480068183,
      "tokens_seen": 2393964544
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027753259779338016,
      "loss": 1.4209,
      "theoretical_loss": 3.408697102669442,
      "tokens_seen": 2394030080
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002775125376128385,
      "loss": 1.3811,
      "theoretical_loss": 3.4086896575929435,
      "tokens_seen": 2394095616
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002774924774322969,
      "loss": 1.3069,
      "theoretical_loss": 3.4086822127773067,
      "tokens_seen": 2394161152
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027747241725175524,
      "loss": 1.2873,
      "theoretical_loss": 3.408674768222515,
      "tokens_seen": 2394226688
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027745235707121366,
      "loss": 1.3696,
      "theoretical_loss": 3.408667323928552,
      "tokens_seen": 2394292224
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000277432296890672,
      "loss": 1.302,
      "theoretical_loss": 3.4086598798954015,
      "tokens_seen": 2394357760
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002774122367101304,
      "loss": 1.4001,
      "theoretical_loss": 3.4086524361230475,
      "tokens_seen": 2394423296
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027739217652958875,
      "loss": 1.2984,
      "theoretical_loss": 3.4086449926114737,
      "tokens_seen": 2394488832
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002773721163490471,
      "loss": 1.3578,
      "theoretical_loss": 3.4086375493606638,
      "tokens_seen": 2394554368
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002773520561685055,
      "loss": 1.3155,
      "theoretical_loss": 3.4086301063706017,
      "tokens_seen": 2394619904
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002773319959879639,
      "loss": 1.5127,
      "theoretical_loss": 3.40862266364127,
      "tokens_seen": 2394685440
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027731193580742225,
      "loss": 1.2866,
      "theoretical_loss": 3.408615221172654,
      "tokens_seen": 2394750976
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002772918756268806,
      "loss": 1.3712,
      "theoretical_loss": 3.408607778964737,
      "tokens_seen": 2394816512
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027727181544633903,
      "loss": 1.3067,
      "theoretical_loss": 3.408600337017502,
      "tokens_seen": 2394882048
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002772517552657974,
      "loss": 1.3723,
      "theoretical_loss": 3.4085928953309335,
      "tokens_seen": 2394947584
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027723169508525575,
      "loss": 1.3337,
      "theoretical_loss": 3.4085854539050144,
      "tokens_seen": 2395013120
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002772116349047141,
      "loss": 1.4321,
      "theoretical_loss": 3.4085780127397296,
      "tokens_seen": 2395078656
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002771915747241725,
      "loss": 1.4042,
      "theoretical_loss": 3.408570571835062,
      "tokens_seen": 2395144192
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002771715145436309,
      "loss": 1.4289,
      "theoretical_loss": 3.408563131190996,
      "tokens_seen": 2395209728
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027715145436308926,
      "loss": 1.4009,
      "theoretical_loss": 3.408555690807514,
      "tokens_seen": 2395275264
    },
    {
      "epoch": 0.73,
      "objective/train/advantage_avg": 0.494269996881485,
      "objective/train/docs_used": 872384,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2380151748657227,
      "objective/train/original_loss": 1.2380151748657227,
      "objective/train/theoretical_loss": 3.408548250684601,
      "objective/train/tokens_used": 2415800800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24452869594097137,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506701469421387,
      "objective/train/weighted_lm_loss": 1.3007664680480957,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.408548250684601,
      "tokens_seen": 2395340800
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002771313941825476,
      "loss": 1.3399,
      "theoretical_loss": 3.408548250684601,
      "tokens_seen": 2395340800
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000277111334002006,
      "loss": 1.4471,
      "theoretical_loss": 3.408540810822241,
      "tokens_seen": 2395406336
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002770912738214644,
      "loss": 1.2952,
      "theoretical_loss": 3.408533371220417,
      "tokens_seen": 2395471872
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027707121364092276,
      "loss": 1.3488,
      "theoretical_loss": 3.4085259318791126,
      "tokens_seen": 2395537408
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002770511534603811,
      "loss": 1.302,
      "theoretical_loss": 3.4085184927983114,
      "tokens_seen": 2395602944
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002770310932798395,
      "loss": 1.3956,
      "theoretical_loss": 3.4085110539779984,
      "tokens_seen": 2395668480
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002770110330992979,
      "loss": 1.4536,
      "theoretical_loss": 3.408503615418156,
      "tokens_seen": 2395734016
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002769909729187563,
      "loss": 1.4421,
      "theoretical_loss": 3.408496177118769,
      "tokens_seen": 2395799552
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002769709127382147,
      "loss": 1.3,
      "theoretical_loss": 3.4084887390798206,
      "tokens_seen": 2395865088
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027695085255767304,
      "loss": 1.4446,
      "theoretical_loss": 3.4084813013012942,
      "tokens_seen": 2395930624
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002769307923771314,
      "loss": 1.2566,
      "theoretical_loss": 3.4084738637831746,
      "tokens_seen": 2395996160
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002769107321965898,
      "loss": 1.4133,
      "theoretical_loss": 3.4084664265254445,
      "tokens_seen": 2396061696
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002768906720160482,
      "loss": 1.2286,
      "theoretical_loss": 3.4084589895280883,
      "tokens_seen": 2396127232
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027687061183550655,
      "loss": 1.2769,
      "theoretical_loss": 3.4084515527910897,
      "tokens_seen": 2396192768
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002768505516549649,
      "loss": 1.4317,
      "theoretical_loss": 3.408444116314432,
      "tokens_seen": 2396258304
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027683049147442327,
      "loss": 1.4338,
      "theoretical_loss": 3.4084366800980996,
      "tokens_seen": 2396323840
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002768104312938817,
      "loss": 1.4377,
      "theoretical_loss": 3.408429244142076,
      "tokens_seen": 2396389376
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027679037111334005,
      "loss": 1.4405,
      "theoretical_loss": 3.4084218084463447,
      "tokens_seen": 2396454912
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002767703109327984,
      "loss": 1.2922,
      "theoretical_loss": 3.4084143730108893,
      "tokens_seen": 2396520448
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027675025075225677,
      "loss": 1.5183,
      "theoretical_loss": 3.4084069378356947,
      "tokens_seen": 2396585984
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002767301905717152,
      "loss": 1.3149,
      "theoretical_loss": 3.408399502920744,
      "tokens_seen": 2396651520
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027671013039117355,
      "loss": 1.3354,
      "theoretical_loss": 3.4083920682660205,
      "tokens_seen": 2396717056
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002766900702106319,
      "loss": 1.3887,
      "theoretical_loss": 3.408384633871509,
      "tokens_seen": 2396782592
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002766700100300903,
      "loss": 1.4733,
      "theoretical_loss": 3.408377199737192,
      "tokens_seen": 2396848128
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002766499498495487,
      "loss": 1.3928,
      "theoretical_loss": 3.4083697658630543,
      "tokens_seen": 2396913664
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027662988966900705,
      "loss": 1.4135,
      "theoretical_loss": 3.4083623322490793,
      "tokens_seen": 2396979200
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002766098294884654,
      "loss": 1.3773,
      "theoretical_loss": 3.408354898895251,
      "tokens_seen": 2397044736
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002765897693079238,
      "loss": 1.3781,
      "theoretical_loss": 3.4083474658015525,
      "tokens_seen": 2397110272
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027656970912738214,
      "loss": 1.4103,
      "theoretical_loss": 3.4083400329679687,
      "tokens_seen": 2397175808
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027654964894684056,
      "loss": 1.3621,
      "theoretical_loss": 3.4083326003944823,
      "tokens_seen": 2397241344
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002765295887662989,
      "loss": 1.3853,
      "theoretical_loss": 3.408325168081078,
      "tokens_seen": 2397306880
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002765095285857573,
      "loss": 1.3558,
      "theoretical_loss": 3.4083177360277386,
      "tokens_seen": 2397372416
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027648946840521564,
      "loss": 1.3991,
      "theoretical_loss": 3.408310304234449,
      "tokens_seen": 2397437952
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027646940822467406,
      "loss": 1.3696,
      "theoretical_loss": 3.408302872701192,
      "tokens_seen": 2397503488
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002764493480441324,
      "loss": 1.3981,
      "theoretical_loss": 3.4082954414279523,
      "tokens_seen": 2397569024
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002764292878635908,
      "loss": 1.3382,
      "theoretical_loss": 3.4082880104147133,
      "tokens_seen": 2397634560
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027640922768304915,
      "loss": 1.4302,
      "theoretical_loss": 3.4082805796614584,
      "tokens_seen": 2397700096
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002763891675025075,
      "loss": 1.3145,
      "theoretical_loss": 3.4082731491681715,
      "tokens_seen": 2397765632
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002763691073219659,
      "loss": 1.3896,
      "theoretical_loss": 3.408265718934837,
      "tokens_seen": 2397831168
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002763490471414243,
      "loss": 1.4258,
      "theoretical_loss": 3.408258288961439,
      "tokens_seen": 2397896704
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027632898696088265,
      "loss": 1.425,
      "theoretical_loss": 3.4082508592479597,
      "tokens_seen": 2397962240
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000276308926780341,
      "loss": 1.3971,
      "theoretical_loss": 3.4082434297943838,
      "tokens_seen": 2398027776
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027628886659979943,
      "loss": 1.45,
      "theoretical_loss": 3.4082360006006955,
      "tokens_seen": 2398093312
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002762688064192578,
      "loss": 1.333,
      "theoretical_loss": 3.408228571666878,
      "tokens_seen": 2398158848
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027624874623871615,
      "loss": 1.3961,
      "theoretical_loss": 3.4082211429929155,
      "tokens_seen": 2398224384
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002762286860581745,
      "loss": 1.3169,
      "theoretical_loss": 3.4082137145787916,
      "tokens_seen": 2398289920
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002762086258776329,
      "loss": 1.4354,
      "theoretical_loss": 3.4082062864244906,
      "tokens_seen": 2398355456
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002761885656970913,
      "loss": 1.3413,
      "theoretical_loss": 3.4081988585299956,
      "tokens_seen": 2398420992
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027616850551654966,
      "loss": 1.3634,
      "theoretical_loss": 3.4081914308952905,
      "tokens_seen": 2398486528
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000276148445336008,
      "loss": 1.3407,
      "theoretical_loss": 3.4081840035203594,
      "tokens_seen": 2398552064
    },
    {
      "epoch": 0.73,
      "objective/train/advantage_avg": 0.4919467866420746,
      "objective/train/docs_used": 873551,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2826989889144897,
      "objective/train/original_loss": 1.2826988697052002,
      "objective/train/theoretical_loss": 3.4081765764051863,
      "objective/train/tokens_used": 2419077600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2425064742565155,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504274368286133,
      "objective/train/weighted_lm_loss": 1.347258448600769,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4081765764051863,
      "tokens_seen": 2398617600
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002761283851554664,
      "loss": 1.348,
      "theoretical_loss": 3.4081765764051863,
      "tokens_seen": 2398617600
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002761083249749248,
      "loss": 1.3802,
      "theoretical_loss": 3.408169149549755,
      "tokens_seen": 2398683136
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027608826479438316,
      "loss": 1.3776,
      "theoretical_loss": 3.4081617229540484,
      "tokens_seen": 2398748672
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002760682046138415,
      "loss": 1.3335,
      "theoretical_loss": 3.4081542966180516,
      "tokens_seen": 2398814208
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002760481444332999,
      "loss": 1.4929,
      "theoretical_loss": 3.4081468705417475,
      "tokens_seen": 2398879744
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027602808425275825,
      "loss": 1.3604,
      "theoretical_loss": 3.40813944472512,
      "tokens_seen": 2398945280
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027600802407221666,
      "loss": 1.3857,
      "theoretical_loss": 3.4081320191681534,
      "tokens_seen": 2399010816
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000275987963891675,
      "loss": 1.2695,
      "theoretical_loss": 3.4081245938708316,
      "tokens_seen": 2399076352
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002759679037111334,
      "loss": 1.2857,
      "theoretical_loss": 3.408117168833138,
      "tokens_seen": 2399141888
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027594784353059175,
      "loss": 1.322,
      "theoretical_loss": 3.4081097440550563,
      "tokens_seen": 2399207424
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027592778335005017,
      "loss": 1.3534,
      "theoretical_loss": 3.408102319536571,
      "tokens_seen": 2399272960
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027590772316950853,
      "loss": 1.3536,
      "theoretical_loss": 3.408094895277665,
      "tokens_seen": 2399338496
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002758876629889669,
      "loss": 1.3426,
      "theoretical_loss": 3.408087471278323,
      "tokens_seen": 2399404032
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027586760280842525,
      "loss": 1.1925,
      "theoretical_loss": 3.4080800475385287,
      "tokens_seen": 2399469568
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002758475426278836,
      "loss": 1.3632,
      "theoretical_loss": 3.4080726240582653,
      "tokens_seen": 2399535104
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027582748244734203,
      "loss": 1.465,
      "theoretical_loss": 3.4080652008375174,
      "tokens_seen": 2399600640
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002758074222668004,
      "loss": 1.4142,
      "theoretical_loss": 3.4080577778762686,
      "tokens_seen": 2399666176
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027578736208625876,
      "loss": 1.5178,
      "theoretical_loss": 3.4080503551745025,
      "tokens_seen": 2399731712
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002757673019057171,
      "loss": 1.3587,
      "theoretical_loss": 3.4080429327322026,
      "tokens_seen": 2399797248
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027574724172517554,
      "loss": 1.3798,
      "theoretical_loss": 3.408035510549354,
      "tokens_seen": 2399862784
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002757271815446339,
      "loss": 1.4844,
      "theoretical_loss": 3.4080280886259393,
      "tokens_seen": 2399928320
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027570712136409226,
      "loss": 1.3046,
      "theoretical_loss": 3.408020666961943,
      "tokens_seen": 2399993856
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002756870611835506,
      "loss": 1.3871,
      "theoretical_loss": 3.408013245557349,
      "tokens_seen": 2400059392
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000275667001003009,
      "loss": 1.3423,
      "theoretical_loss": 3.408005824412141,
      "tokens_seen": 2400124928
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002756469408224674,
      "loss": 1.4734,
      "theoretical_loss": 3.4079984035263022,
      "tokens_seen": 2400190464
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027562688064192576,
      "loss": 1.3354,
      "theoretical_loss": 3.4079909828998174,
      "tokens_seen": 2400256000
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002756068204613841,
      "loss": 1.3268,
      "theoretical_loss": 3.40798356253267,
      "tokens_seen": 2400321536
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002755867602808425,
      "loss": 1.3117,
      "theoretical_loss": 3.407976142424844,
      "tokens_seen": 2400387072
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002755667001003009,
      "loss": 1.3995,
      "theoretical_loss": 3.4079687225763236,
      "tokens_seen": 2400452608
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027554663991975927,
      "loss": 1.3966,
      "theoretical_loss": 3.4079613029870917,
      "tokens_seen": 2400518144
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027552657973921763,
      "loss": 1.4065,
      "theoretical_loss": 3.4079538836571333,
      "tokens_seen": 2400583680
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027550651955867604,
      "loss": 1.3694,
      "theoretical_loss": 3.407946464586431,
      "tokens_seen": 2400649216
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027548645937813446,
      "loss": 1.4367,
      "theoretical_loss": 3.40793904577497,
      "tokens_seen": 2400714752
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002754663991975928,
      "loss": 1.4363,
      "theoretical_loss": 3.4079316272227334,
      "tokens_seen": 2400780288
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002754463390170512,
      "loss": 1.454,
      "theoretical_loss": 3.407924208929705,
      "tokens_seen": 2400845824
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027542627883650955,
      "loss": 1.4198,
      "theoretical_loss": 3.407916790895869,
      "tokens_seen": 2400911360
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002754062186559679,
      "loss": 1.3776,
      "theoretical_loss": 3.4079093731212087,
      "tokens_seen": 2400976896
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002753861584754263,
      "loss": 1.3594,
      "theoretical_loss": 3.4079019556057086,
      "tokens_seen": 2401042432
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002753660982948847,
      "loss": 1.2442,
      "theoretical_loss": 3.4078945383493524,
      "tokens_seen": 2401107968
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027534603811434305,
      "loss": 1.3316,
      "theoretical_loss": 3.4078871213521245,
      "tokens_seen": 2401173504
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002753259779338014,
      "loss": 1.4382,
      "theoretical_loss": 3.4078797046140075,
      "tokens_seen": 2401239040
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027530591775325983,
      "loss": 1.368,
      "theoretical_loss": 3.4078722881349863,
      "tokens_seen": 2401304576
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002752858575727182,
      "loss": 1.3632,
      "theoretical_loss": 3.407864871915044,
      "tokens_seen": 2401370112
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027526579739217655,
      "loss": 1.404,
      "theoretical_loss": 3.407857455954166,
      "tokens_seen": 2401435648
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002752457372116349,
      "loss": 1.4283,
      "theoretical_loss": 3.4078500402523346,
      "tokens_seen": 2401501184
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002752457372116349,
      "loss": 1.409,
      "theoretical_loss": 3.407842624809534,
      "tokens_seen": 2401566720
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002752256770310933,
      "loss": 1.2854,
      "theoretical_loss": 3.4078352096257483,
      "tokens_seen": 2401632256
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002752056168505517,
      "loss": 1.3404,
      "theoretical_loss": 3.4078277947009616,
      "tokens_seen": 2401697792
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027518555667001006,
      "loss": 1.3927,
      "theoretical_loss": 3.407820380035157,
      "tokens_seen": 2401763328
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002751654964894684,
      "loss": 1.3722,
      "theoretical_loss": 3.40781296562832,
      "tokens_seen": 2401828864
    },
    {
      "epoch": 0.73,
      "objective/train/advantage_avg": 0.49251076579093933,
      "objective/train/docs_used": 874334,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2582896947860718,
      "objective/train/original_loss": 1.2582894563674927,
      "objective/train/theoretical_loss": 3.4078055514804326,
      "objective/train/tokens_used": 2422354400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24296115338802338,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504862070083618,
      "objective/train/weighted_lm_loss": 1.321784496307373,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4078055514804326,
      "tokens_seen": 2401894400
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002751454363089268,
      "loss": 1.3181,
      "theoretical_loss": 3.4078055514804326,
      "tokens_seen": 2401894400
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002751253761283852,
      "loss": 1.4669,
      "theoretical_loss": 3.40779813759148,
      "tokens_seen": 2401959936
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027510531594784356,
      "loss": 1.3958,
      "theoretical_loss": 3.4077907239614453,
      "tokens_seen": 2402025472
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002750852557673019,
      "loss": 1.489,
      "theoretical_loss": 3.4077833105903133,
      "tokens_seen": 2402091008
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002750651955867603,
      "loss": 1.4557,
      "theoretical_loss": 3.407775897478067,
      "tokens_seen": 2402156544
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027504513540621865,
      "loss": 1.31,
      "theoretical_loss": 3.4077684846246905,
      "tokens_seen": 2402222080
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027502507522567706,
      "loss": 1.4025,
      "theoretical_loss": 3.4077610720301674,
      "tokens_seen": 2402287616
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002750050150451354,
      "loss": 1.4083,
      "theoretical_loss": 3.4077536596944826,
      "tokens_seen": 2402353152
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002749849548645938,
      "loss": 1.3979,
      "theoretical_loss": 3.4077462476176192,
      "tokens_seen": 2402418688
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027496489468405215,
      "loss": 1.39,
      "theoretical_loss": 3.4077388357995617,
      "tokens_seen": 2402484224
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027494483450351057,
      "loss": 1.5401,
      "theoretical_loss": 3.407731424240293,
      "tokens_seen": 2402549760
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027492477432296893,
      "loss": 1.3421,
      "theoretical_loss": 3.407724012939798,
      "tokens_seen": 2402615296
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027492477432296893,
      "loss": 1.4431,
      "theoretical_loss": 3.4077166018980605,
      "tokens_seen": 2402680832
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002749047141424273,
      "loss": 1.3816,
      "theoretical_loss": 3.407709191115064,
      "tokens_seen": 2402746368
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027488465396188565,
      "loss": 1.3011,
      "theoretical_loss": 3.4077017805907923,
      "tokens_seen": 2402811904
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000274864593781344,
      "loss": 1.4425,
      "theoretical_loss": 3.4076943703252294,
      "tokens_seen": 2402877440
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027484453360080243,
      "loss": 1.6009,
      "theoretical_loss": 3.40768696031836,
      "tokens_seen": 2402942976
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002748244734202608,
      "loss": 1.3645,
      "theoretical_loss": 3.4076795505701667,
      "tokens_seen": 2403008512
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027480441323971916,
      "loss": 1.4569,
      "theoretical_loss": 3.4076721410806345,
      "tokens_seen": 2403074048
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002747843530591775,
      "loss": 1.3804,
      "theoretical_loss": 3.4076647318497466,
      "tokens_seen": 2403139584
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027476429287863594,
      "loss": 1.5031,
      "theoretical_loss": 3.407657322877488,
      "tokens_seen": 2403205120
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002747442326980943,
      "loss": 1.3054,
      "theoretical_loss": 3.407649914163841,
      "tokens_seen": 2403270656
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027472417251755266,
      "loss": 1.2504,
      "theoretical_loss": 3.407642505708791,
      "tokens_seen": 2403336192
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000274704112337011,
      "loss": 1.3692,
      "theoretical_loss": 3.407635097512321,
      "tokens_seen": 2403401728
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002746840521564694,
      "loss": 1.3899,
      "theoretical_loss": 3.407627689574415,
      "tokens_seen": 2403467264
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002746639919759278,
      "loss": 1.5402,
      "theoretical_loss": 3.4076202818950576,
      "tokens_seen": 2403532800
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027464393179538616,
      "loss": 1.2691,
      "theoretical_loss": 3.4076128744742324,
      "tokens_seen": 2403598336
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002746238716148445,
      "loss": 1.3329,
      "theoretical_loss": 3.4076054673119227,
      "tokens_seen": 2403663872
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002746038114343029,
      "loss": 1.3548,
      "theoretical_loss": 3.4075980604081133,
      "tokens_seen": 2403729408
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002745837512537613,
      "loss": 1.4748,
      "theoretical_loss": 3.407590653762788,
      "tokens_seen": 2403794944
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027456369107321967,
      "loss": 1.332,
      "theoretical_loss": 3.4075832473759298,
      "tokens_seen": 2403860480
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027454363089267803,
      "loss": 1.487,
      "theoretical_loss": 3.4075758412475237,
      "tokens_seen": 2403926016
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002745235707121364,
      "loss": 1.5392,
      "theoretical_loss": 3.407568435377553,
      "tokens_seen": 2403991552
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027450351053159475,
      "loss": 1.3797,
      "theoretical_loss": 3.4075610297660024,
      "tokens_seen": 2404057088
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027448345035105317,
      "loss": 1.5073,
      "theoretical_loss": 3.4075536244128553,
      "tokens_seen": 2404122624
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027446339017051153,
      "loss": 1.438,
      "theoretical_loss": 3.407546219318095,
      "tokens_seen": 2404188160
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002744433299899699,
      "loss": 1.3342,
      "theoretical_loss": 3.407538814481707,
      "tokens_seen": 2404253696
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027442326980942826,
      "loss": 1.5498,
      "theoretical_loss": 3.407531409903674,
      "tokens_seen": 2404319232
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027440320962888667,
      "loss": 1.4886,
      "theoretical_loss": 3.4075240055839804,
      "tokens_seen": 2404384768
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027438314944834503,
      "loss": 1.523,
      "theoretical_loss": 3.40751660152261,
      "tokens_seen": 2404450304
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002743630892678034,
      "loss": 1.3497,
      "theoretical_loss": 3.4075091977195466,
      "tokens_seen": 2404515840
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027434302908726176,
      "loss": 1.2843,
      "theoretical_loss": 3.4075017941747747,
      "tokens_seen": 2404581376
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002743229689067202,
      "loss": 1.3705,
      "theoretical_loss": 3.407494390888278,
      "tokens_seen": 2404646912
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027430290872617854,
      "loss": 1.3631,
      "theoretical_loss": 3.40748698786004,
      "tokens_seen": 2404712448
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002742828485456369,
      "loss": 1.3932,
      "theoretical_loss": 3.4074795850900452,
      "tokens_seen": 2404777984
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027426278836509526,
      "loss": 1.311,
      "theoretical_loss": 3.407472182578277,
      "tokens_seen": 2404843520
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002742427281845536,
      "loss": 1.6323,
      "theoretical_loss": 3.4074647803247204,
      "tokens_seen": 2404909056
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027422266800401204,
      "loss": 1.2463,
      "theoretical_loss": 3.407457378329358,
      "tokens_seen": 2404974592
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002742026078234704,
      "loss": 1.294,
      "theoretical_loss": 3.407449976592175,
      "tokens_seen": 2405040128
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027418254764292877,
      "loss": 1.4245,
      "theoretical_loss": 3.4074425751131545,
      "tokens_seen": 2405105664
    },
    {
      "epoch": 0.73,
      "objective/train/advantage_avg": 0.4963158071041107,
      "objective/train/docs_used": 875512,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2482362985610962,
      "objective/train/original_loss": 1.2482361793518066,
      "objective/train/theoretical_loss": 3.407435173892281,
      "objective/train/tokens_used": 2425631200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2464946061372757,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508848428726196,
      "objective/train/weighted_lm_loss": 1.3116883039474487,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.407435173892281,
      "tokens_seen": 2405171200
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027416248746238713,
      "loss": 1.3687,
      "theoretical_loss": 3.407435173892281,
      "tokens_seen": 2405171200
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027414242728184554,
      "loss": 1.4181,
      "theoretical_loss": 3.407427772929538,
      "tokens_seen": 2405236736
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002741223671013039,
      "loss": 1.472,
      "theoretical_loss": 3.4074203722249097,
      "tokens_seen": 2405302272
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027410230692076227,
      "loss": 1.4404,
      "theoretical_loss": 3.40741297177838,
      "tokens_seen": 2405367808
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027408224674022063,
      "loss": 1.4416,
      "theoretical_loss": 3.407405571589933,
      "tokens_seen": 2405433344
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000274062186559679,
      "loss": 1.3929,
      "theoretical_loss": 3.4073981716595525,
      "tokens_seen": 2405498880
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002740421263791374,
      "loss": 1.5238,
      "theoretical_loss": 3.407390771987223,
      "tokens_seen": 2405564416
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002740220661985958,
      "loss": 1.3049,
      "theoretical_loss": 3.4073833725729274,
      "tokens_seen": 2405629952
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002740020060180542,
      "loss": 1.4586,
      "theoretical_loss": 3.4073759734166504,
      "tokens_seen": 2405695488
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027398194583751255,
      "loss": 1.3599,
      "theoretical_loss": 3.407368574518376,
      "tokens_seen": 2405761024
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027396188565697097,
      "loss": 1.2615,
      "theoretical_loss": 3.4073611758780884,
      "tokens_seen": 2405826560
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027394182547642933,
      "loss": 1.4972,
      "theoretical_loss": 3.407353777495771,
      "tokens_seen": 2405892096
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002739217652958877,
      "loss": 1.4166,
      "theoretical_loss": 3.4073463793714076,
      "tokens_seen": 2405957632
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027390170511534605,
      "loss": 1.4782,
      "theoretical_loss": 3.4073389815049833,
      "tokens_seen": 2406023168
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002738816449348044,
      "loss": 1.4165,
      "theoretical_loss": 3.4073315838964806,
      "tokens_seen": 2406088704
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027386158475426283,
      "loss": 1.3975,
      "theoretical_loss": 3.407324186545885,
      "tokens_seen": 2406154240
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002738415245737212,
      "loss": 1.47,
      "theoretical_loss": 3.407316789453179,
      "tokens_seen": 2406219776
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027382146439317956,
      "loss": 1.4386,
      "theoretical_loss": 3.407309392618348,
      "tokens_seen": 2406285312
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002738014042126379,
      "loss": 1.3644,
      "theoretical_loss": 3.407301996041375,
      "tokens_seen": 2406350848
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027378134403209634,
      "loss": 1.3659,
      "theoretical_loss": 3.4072945997222446,
      "tokens_seen": 2406416384
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002737612838515547,
      "loss": 1.3361,
      "theoretical_loss": 3.40728720366094,
      "tokens_seen": 2406481920
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027374122367101306,
      "loss": 1.4545,
      "theoretical_loss": 3.4072798078574458,
      "tokens_seen": 2406547456
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002737211634904714,
      "loss": 1.3562,
      "theoretical_loss": 3.407272412311746,
      "tokens_seen": 2406612992
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002737011033099298,
      "loss": 1.4541,
      "theoretical_loss": 3.4072650170238243,
      "tokens_seen": 2406678528
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002736810431293882,
      "loss": 1.4274,
      "theoretical_loss": 3.407257621993665,
      "tokens_seen": 2406744064
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027366098294884656,
      "loss": 1.3916,
      "theoretical_loss": 3.407250227221252,
      "tokens_seen": 2406809600
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002736409227683049,
      "loss": 1.3536,
      "theoretical_loss": 3.4072428327065696,
      "tokens_seen": 2406875136
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002736208625877633,
      "loss": 1.3957,
      "theoretical_loss": 3.407235438449601,
      "tokens_seen": 2406940672
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002736008024072217,
      "loss": 1.3299,
      "theoretical_loss": 3.4072280444503305,
      "tokens_seen": 2407006208
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027358074222668007,
      "loss": 1.4163,
      "theoretical_loss": 3.4072206507087426,
      "tokens_seen": 2407071744
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027356068204613843,
      "loss": 1.3775,
      "theoretical_loss": 3.4072132572248206,
      "tokens_seen": 2407137280
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002735406218655968,
      "loss": 1.3614,
      "theoretical_loss": 3.4072058639985494,
      "tokens_seen": 2407202816
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027352056168505515,
      "loss": 1.4765,
      "theoretical_loss": 3.407198471029912,
      "tokens_seen": 2407268352
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027350050150451357,
      "loss": 1.5233,
      "theoretical_loss": 3.4071910783188932,
      "tokens_seen": 2407333888
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027348044132397193,
      "loss": 1.4003,
      "theoretical_loss": 3.4071836858654767,
      "tokens_seen": 2407399424
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002734603811434303,
      "loss": 1.3244,
      "theoretical_loss": 3.407176293669646,
      "tokens_seen": 2407464960
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027344032096288866,
      "loss": 1.3745,
      "theoretical_loss": 3.4071689017313864,
      "tokens_seen": 2407530496
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002734202607823471,
      "loss": 1.339,
      "theoretical_loss": 3.4071615100506807,
      "tokens_seen": 2407596032
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027340020060180544,
      "loss": 1.4092,
      "theoretical_loss": 3.4071541186275134,
      "tokens_seen": 2407661568
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002733801404212638,
      "loss": 1.3946,
      "theoretical_loss": 3.4071467274618685,
      "tokens_seen": 2407727104
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027336008024072216,
      "loss": 1.5277,
      "theoretical_loss": 3.40713933655373,
      "tokens_seen": 2407792640
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002733400200601806,
      "loss": 1.5064,
      "theoretical_loss": 3.4071319459030818,
      "tokens_seen": 2407858176
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027331995987963894,
      "loss": 1.448,
      "theoretical_loss": 3.4071245555099083,
      "tokens_seen": 2407923712
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002732998996990973,
      "loss": 1.5021,
      "theoretical_loss": 3.407117165374193,
      "tokens_seen": 2407989248
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027327983951855566,
      "loss": 1.45,
      "theoretical_loss": 3.40710977549592,
      "tokens_seen": 2408054784
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000273259779338014,
      "loss": 1.4872,
      "theoretical_loss": 3.407102385875074,
      "tokens_seen": 2408120320
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027323971915747244,
      "loss": 1.4425,
      "theoretical_loss": 3.4070949965116384,
      "tokens_seen": 2408185856
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002732196589769308,
      "loss": 1.3957,
      "theoretical_loss": 3.407087607405597,
      "tokens_seen": 2408251392
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027319959879638917,
      "loss": 1.3022,
      "theoretical_loss": 3.4070802185569344,
      "tokens_seen": 2408316928
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027317953861584753,
      "loss": 1.4629,
      "theoretical_loss": 3.4070728299656348,
      "tokens_seen": 2408382464
    },
    {
      "epoch": 0.73,
      "objective/train/advantage_avg": 0.4950404465198517,
      "objective/train/docs_used": 876655,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.356238842010498,
      "objective/train/original_loss": 1.356238842010498,
      "objective/train/theoretical_loss": 3.4070654416316812,
      "objective/train/tokens_used": 2428908000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24524113535881042,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050750970840454,
      "objective/train/weighted_lm_loss": 1.4250178337097168,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4070654416316812,
      "tokens_seen": 2408448000
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027315947843530594,
      "loss": 1.4135,
      "theoretical_loss": 3.4070654416316812,
      "tokens_seen": 2408448000
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002731394182547643,
      "loss": 1.3863,
      "theoretical_loss": 3.4070580535550588,
      "tokens_seen": 2408513536
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027311935807422267,
      "loss": 1.4117,
      "theoretical_loss": 3.407050665735751,
      "tokens_seen": 2408579072
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027309929789368103,
      "loss": 1.4984,
      "theoretical_loss": 3.4070432781737416,
      "tokens_seen": 2408644608
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002730792377131394,
      "loss": 1.3279,
      "theoretical_loss": 3.4070358908690155,
      "tokens_seen": 2408710144
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002730591775325978,
      "loss": 1.3615,
      "theoretical_loss": 3.407028503821556,
      "tokens_seen": 2408775680
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027303911735205617,
      "loss": 1.3698,
      "theoretical_loss": 3.407021117031347,
      "tokens_seen": 2408841216
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027301905717151453,
      "loss": 1.4209,
      "theoretical_loss": 3.407013730498374,
      "tokens_seen": 2408906752
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002729989969909729,
      "loss": 1.3809,
      "theoretical_loss": 3.407006344222619,
      "tokens_seen": 2408972288
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002729789368104313,
      "loss": 1.3248,
      "theoretical_loss": 3.406998958204067,
      "tokens_seen": 2409037824
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002729588766298897,
      "loss": 1.34,
      "theoretical_loss": 3.4069915724427027,
      "tokens_seen": 2409103360
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027293881644934804,
      "loss": 1.4347,
      "theoretical_loss": 3.406984186938509,
      "tokens_seen": 2409168896
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002729187562688064,
      "loss": 1.3651,
      "theoretical_loss": 3.4069768016914708,
      "tokens_seen": 2409234432
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027289869608826476,
      "loss": 1.3357,
      "theoretical_loss": 3.4069694167015716,
      "tokens_seen": 2409299968
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002728786359077232,
      "loss": 1.3484,
      "theoretical_loss": 3.4069620319687957,
      "tokens_seen": 2409365504
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027285857572718154,
      "loss": 1.4469,
      "theoretical_loss": 3.406954647493127,
      "tokens_seen": 2409431040
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002728385155466399,
      "loss": 1.4892,
      "theoretical_loss": 3.40694726327455,
      "tokens_seen": 2409496576
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027281845536609827,
      "loss": 1.4092,
      "theoretical_loss": 3.406939879313048,
      "tokens_seen": 2409562112
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002727983951855567,
      "loss": 1.3862,
      "theoretical_loss": 3.406932495608606,
      "tokens_seen": 2409627648
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027277833500501504,
      "loss": 1.4928,
      "theoretical_loss": 3.406925112161207,
      "tokens_seen": 2409693184
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002727582748244734,
      "loss": 1.2641,
      "theoretical_loss": 3.4069177289708357,
      "tokens_seen": 2409758720
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027273821464393177,
      "loss": 1.4144,
      "theoretical_loss": 3.4069103460374763,
      "tokens_seen": 2409824256
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027271815446339013,
      "loss": 1.3919,
      "theoretical_loss": 3.4069029633611128,
      "tokens_seen": 2409889792
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027269809428284855,
      "loss": 1.3197,
      "theoretical_loss": 3.406895580941729,
      "tokens_seen": 2409955328
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002726780341023069,
      "loss": 1.485,
      "theoretical_loss": 3.406888198779309,
      "tokens_seen": 2410020864
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027265797392176527,
      "loss": 1.4323,
      "theoretical_loss": 3.4068808168738367,
      "tokens_seen": 2410086400
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027263791374122363,
      "loss": 1.3897,
      "theoretical_loss": 3.4068734352252967,
      "tokens_seen": 2410151936
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027261785356068205,
      "loss": 1.3545,
      "theoretical_loss": 3.4068660538336726,
      "tokens_seen": 2410217472
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002725977933801404,
      "loss": 1.3328,
      "theoretical_loss": 3.4068586726989487,
      "tokens_seen": 2410283008
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002725777331995988,
      "loss": 1.37,
      "theoretical_loss": 3.406851291821109,
      "tokens_seen": 2410348544
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027255767301905714,
      "loss": 1.4659,
      "theoretical_loss": 3.406843911200138,
      "tokens_seen": 2410414080
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002725376128385155,
      "loss": 1.4472,
      "theoretical_loss": 3.406836530836019,
      "tokens_seen": 2410479616
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027251755265797397,
      "loss": 1.3925,
      "theoretical_loss": 3.4068291507287363,
      "tokens_seen": 2410545152
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027249749247743233,
      "loss": 1.3926,
      "theoretical_loss": 3.406821770878274,
      "tokens_seen": 2410610688
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002724774322968907,
      "loss": 1.4523,
      "theoretical_loss": 3.406814391284617,
      "tokens_seen": 2410676224
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027245737211634906,
      "loss": 1.346,
      "theoretical_loss": 3.4068070119477483,
      "tokens_seen": 2410741760
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002724373119358075,
      "loss": 1.3643,
      "theoretical_loss": 3.406799632867652,
      "tokens_seen": 2410807296
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027241725175526584,
      "loss": 1.2805,
      "theoretical_loss": 3.406792254044313,
      "tokens_seen": 2410872832
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002723971915747242,
      "loss": 1.4401,
      "theoretical_loss": 3.4067848754777152,
      "tokens_seen": 2410938368
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027237713139418256,
      "loss": 1.394,
      "theoretical_loss": 3.406777497167842,
      "tokens_seen": 2411003904
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000272357071213641,
      "loss": 1.3866,
      "theoretical_loss": 3.406770119114678,
      "tokens_seen": 2411069440
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027233701103309934,
      "loss": 1.3142,
      "theoretical_loss": 3.4067627413182078,
      "tokens_seen": 2411134976
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002723169508525577,
      "loss": 1.5169,
      "theoretical_loss": 3.4067553637784145,
      "tokens_seen": 2411200512
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027229689067201606,
      "loss": 1.3725,
      "theoretical_loss": 3.4067479864952825,
      "tokens_seen": 2411266048
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002722768304914744,
      "loss": 1.3382,
      "theoretical_loss": 3.406740609468796,
      "tokens_seen": 2411331584
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027225677031093284,
      "loss": 1.4348,
      "theoretical_loss": 3.4067332326989392,
      "tokens_seen": 2411397120
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002722367101303912,
      "loss": 1.5057,
      "theoretical_loss": 3.406725856185696,
      "tokens_seen": 2411462656
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027221664994984957,
      "loss": 1.3791,
      "theoretical_loss": 3.406718479929051,
      "tokens_seen": 2411528192
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027219658976930793,
      "loss": 1.3986,
      "theoretical_loss": 3.4067111039289877,
      "tokens_seen": 2411593728
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027217652958876635,
      "loss": 1.3651,
      "theoretical_loss": 3.40670372818549,
      "tokens_seen": 2411659264
    },
    {
      "epoch": 0.73,
      "objective/train/advantage_avg": 0.4916989803314209,
      "objective/train/docs_used": 877919,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3934146165847778,
      "objective/train/original_loss": 1.3934146165847778,
      "objective/train/theoretical_loss": 3.4066963526985425,
      "objective/train/tokens_used": 2432184800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24208833277225494,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504006147384644,
      "objective/train/weighted_lm_loss": 1.4634616374969482,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4066963526985425,
      "tokens_seen": 2411724800
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002721564694082247,
      "loss": 1.4305,
      "theoretical_loss": 3.4066963526985425,
      "tokens_seen": 2411724800
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027213640922768307,
      "loss": 1.388,
      "theoretical_loss": 3.4066889774681295,
      "tokens_seen": 2411790336
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027211634904714143,
      "loss": 1.4366,
      "theoretical_loss": 3.406681602494235,
      "tokens_seen": 2411855872
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002720962888665998,
      "loss": 1.3681,
      "theoretical_loss": 3.4066742277768425,
      "tokens_seen": 2411921408
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002720762286860582,
      "loss": 1.3836,
      "theoretical_loss": 3.4066668533159365,
      "tokens_seen": 2411986944
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027205616850551657,
      "loss": 1.3646,
      "theoretical_loss": 3.4066594791115015,
      "tokens_seen": 2412052480
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027203610832497493,
      "loss": 1.4471,
      "theoretical_loss": 3.4066521051635212,
      "tokens_seen": 2412118016
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002720160481444333,
      "loss": 1.4715,
      "theoretical_loss": 3.40664473147198,
      "tokens_seen": 2412183552
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002719959879638917,
      "loss": 1.4116,
      "theoretical_loss": 3.4066373580368614,
      "tokens_seen": 2412249088
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002719759277833501,
      "loss": 1.3616,
      "theoretical_loss": 3.40662998485815,
      "tokens_seen": 2412314624
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027195586760280844,
      "loss": 1.486,
      "theoretical_loss": 3.4066226119358296,
      "tokens_seen": 2412380160
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002719358074222668,
      "loss": 1.4544,
      "theoretical_loss": 3.4066152392698847,
      "tokens_seen": 2412445696
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027191574724172516,
      "loss": 1.4177,
      "theoretical_loss": 3.4066078668602997,
      "tokens_seen": 2412511232
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002718956870611836,
      "loss": 1.3936,
      "theoretical_loss": 3.4066004947070576,
      "tokens_seen": 2412576768
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027187562688064194,
      "loss": 1.3463,
      "theoretical_loss": 3.406593122810144,
      "tokens_seen": 2412642304
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002718555667001003,
      "loss": 1.3734,
      "theoretical_loss": 3.4065857511695414,
      "tokens_seen": 2412707840
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027183550651955867,
      "loss": 1.4524,
      "theoretical_loss": 3.4065783797852354,
      "tokens_seen": 2412773376
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002718154463390171,
      "loss": 1.398,
      "theoretical_loss": 3.406571008657209,
      "tokens_seen": 2412838912
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027179538615847544,
      "loss": 1.4109,
      "theoretical_loss": 3.4065636377854474,
      "tokens_seen": 2412904448
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002717753259779338,
      "loss": 1.3966,
      "theoretical_loss": 3.4065562671699334,
      "tokens_seen": 2412969984
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027175526579739217,
      "loss": 1.3326,
      "theoretical_loss": 3.4065488968106528,
      "tokens_seen": 2413035520
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027173520561685053,
      "loss": 1.491,
      "theoretical_loss": 3.406541526707588,
      "tokens_seen": 2413101056
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027171514543630895,
      "loss": 1.3166,
      "theoretical_loss": 3.406534156860724,
      "tokens_seen": 2413166592
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002716950852557673,
      "loss": 1.4332,
      "theoretical_loss": 3.4065267872700455,
      "tokens_seen": 2413232128
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027167502507522567,
      "loss": 1.4722,
      "theoretical_loss": 3.406519417935536,
      "tokens_seen": 2413297664
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027165496489468403,
      "loss": 1.3839,
      "theoretical_loss": 3.406512048857179,
      "tokens_seen": 2413363200
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027163490471414245,
      "loss": 1.4486,
      "theoretical_loss": 3.40650468003496,
      "tokens_seen": 2413428736
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002716148445336008,
      "loss": 1.433,
      "theoretical_loss": 3.406497311468862,
      "tokens_seen": 2413494272
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002715947843530592,
      "loss": 1.4256,
      "theoretical_loss": 3.4064899431588698,
      "tokens_seen": 2413559808
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027157472417251754,
      "loss": 1.4317,
      "theoretical_loss": 3.406482575104967,
      "tokens_seen": 2413625344
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002715546639919759,
      "loss": 1.4684,
      "theoretical_loss": 3.4064752073071385,
      "tokens_seen": 2413690880
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002715346038114343,
      "loss": 1.3201,
      "theoretical_loss": 3.4064678397653676,
      "tokens_seen": 2413756416
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002715145436308927,
      "loss": 1.3137,
      "theoretical_loss": 3.4064604724796395,
      "tokens_seen": 2413821952
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027149448345035104,
      "loss": 1.4454,
      "theoretical_loss": 3.406453105449937,
      "tokens_seen": 2413887488
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002714744232698094,
      "loss": 1.477,
      "theoretical_loss": 3.4064457386762457,
      "tokens_seen": 2413953024
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002714543630892678,
      "loss": 1.2854,
      "theoretical_loss": 3.4064383721585485,
      "tokens_seen": 2414018560
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002714343029087262,
      "loss": 1.4424,
      "theoretical_loss": 3.4064310058968306,
      "tokens_seen": 2414084096
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027141424272818454,
      "loss": 1.3616,
      "theoretical_loss": 3.4064236398910754,
      "tokens_seen": 2414149632
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002713941825476429,
      "loss": 1.2485,
      "theoretical_loss": 3.406416274141267,
      "tokens_seen": 2414215168
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027137412236710127,
      "loss": 1.3594,
      "theoretical_loss": 3.40640890864739,
      "tokens_seen": 2414280704
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002713540621865597,
      "loss": 1.3213,
      "theoretical_loss": 3.406401543409429,
      "tokens_seen": 2414346240
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027133400200601805,
      "loss": 1.4303,
      "theoretical_loss": 3.406394178427367,
      "tokens_seen": 2414411776
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002713139418254764,
      "loss": 1.4688,
      "theoretical_loss": 3.4063868137011886,
      "tokens_seen": 2414477312
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027129388164493477,
      "loss": 1.4372,
      "theoretical_loss": 3.4063794492308785,
      "tokens_seen": 2414542848
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002712738214643932,
      "loss": 1.3836,
      "theoretical_loss": 3.4063720850164203,
      "tokens_seen": 2414608384
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027125376128385155,
      "loss": 1.4252,
      "theoretical_loss": 3.4063647210577988,
      "tokens_seen": 2414673920
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002712337011033099,
      "loss": 1.3975,
      "theoretical_loss": 3.406357357354997,
      "tokens_seen": 2414739456
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002712136409227683,
      "loss": 1.4091,
      "theoretical_loss": 3.4063499939080004,
      "tokens_seen": 2414804992
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002711935807422267,
      "loss": 1.3581,
      "theoretical_loss": 3.406342630716792,
      "tokens_seen": 2414870528
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027117352056168505,
      "loss": 1.2689,
      "theoretical_loss": 3.406335267781357,
      "tokens_seen": 2414936064
    },
    {
      "epoch": 0.73,
      "objective/train/advantage_avg": 0.4938090145587921,
      "objective/train/docs_used": 879191,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.21418035030365,
      "objective/train/original_loss": 1.2141801118850708,
      "objective/train/theoretical_loss": 3.406327905101679,
      "objective/train/tokens_used": 2435461600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24418877065181732,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506223440170288,
      "objective/train/weighted_lm_loss": 1.2754276990890503,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.406327905101679,
      "tokens_seen": 2415001600
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002711534603811434,
      "loss": 1.4868,
      "theoretical_loss": 3.406327905101679,
      "tokens_seen": 2415001600
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002711334002006018,
      "loss": 1.4142,
      "theoretical_loss": 3.4063205426777423,
      "tokens_seen": 2415067136
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027111334002006014,
      "loss": 1.4756,
      "theoretical_loss": 3.406313180509531,
      "tokens_seen": 2415132672
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027109327983951856,
      "loss": 1.4618,
      "theoretical_loss": 3.4063058185970294,
      "tokens_seen": 2415198208
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002710732196589769,
      "loss": 1.4166,
      "theoretical_loss": 3.4062984569402213,
      "tokens_seen": 2415263744
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002710531594784353,
      "loss": 1.2962,
      "theoretical_loss": 3.4062910955390917,
      "tokens_seen": 2415329280
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027103309929789364,
      "loss": 1.4436,
      "theoretical_loss": 3.406283734393624,
      "tokens_seen": 2415394816
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002710130391173521,
      "loss": 1.4101,
      "theoretical_loss": 3.4062763735038026,
      "tokens_seen": 2415460352
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002709929789368105,
      "loss": 1.3445,
      "theoretical_loss": 3.406269012869612,
      "tokens_seen": 2415525888
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027097291875626884,
      "loss": 1.3562,
      "theoretical_loss": 3.406261652491036,
      "tokens_seen": 2415591424
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002709528585757272,
      "loss": 1.3539,
      "theoretical_loss": 3.4062542923680588,
      "tokens_seen": 2415656960
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027093279839518556,
      "loss": 1.3603,
      "theoretical_loss": 3.406246932500665,
      "tokens_seen": 2415722496
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000270912738214644,
      "loss": 1.3438,
      "theoretical_loss": 3.4062395728888384,
      "tokens_seen": 2415788032
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027089267803410234,
      "loss": 1.4744,
      "theoretical_loss": 3.4062322135325633,
      "tokens_seen": 2415853568
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002708726178535607,
      "loss": 1.4682,
      "theoretical_loss": 3.406224854431824,
      "tokens_seen": 2415919104
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027085255767301907,
      "loss": 1.4239,
      "theoretical_loss": 3.4062174955866045,
      "tokens_seen": 2415984640
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002708324974924775,
      "loss": 1.5097,
      "theoretical_loss": 3.4062101369968887,
      "tokens_seen": 2416050176
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027081243731193584,
      "loss": 1.2625,
      "theoretical_loss": 3.406202778662662,
      "tokens_seen": 2416115712
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002707923771313942,
      "loss": 1.456,
      "theoretical_loss": 3.406195420583907,
      "tokens_seen": 2416181248
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027077231695085257,
      "loss": 1.4416,
      "theoretical_loss": 3.406188062760609,
      "tokens_seen": 2416246784
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027075225677031093,
      "loss": 1.4013,
      "theoretical_loss": 3.406180705192752,
      "tokens_seen": 2416312320
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027073219658976935,
      "loss": 1.3828,
      "theoretical_loss": 3.40617334788032,
      "tokens_seen": 2416377856
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002707121364092277,
      "loss": 1.2586,
      "theoretical_loss": 3.4061659908232973,
      "tokens_seen": 2416443392
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027069207622868607,
      "loss": 1.4307,
      "theoretical_loss": 3.406158634021668,
      "tokens_seen": 2416508928
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027067201604814443,
      "loss": 1.3357,
      "theoretical_loss": 3.406151277475417,
      "tokens_seen": 2416574464
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027065195586760285,
      "loss": 1.3412,
      "theoretical_loss": 3.406143921184527,
      "tokens_seen": 2416640000
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002706318956870612,
      "loss": 1.4217,
      "theoretical_loss": 3.4061365651489837,
      "tokens_seen": 2416705536
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002706118355065196,
      "loss": 1.37,
      "theoretical_loss": 3.4061292093687707,
      "tokens_seen": 2416771072
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027059177532597794,
      "loss": 1.487,
      "theoretical_loss": 3.4061218538438722,
      "tokens_seen": 2416836608
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002705717151454363,
      "loss": 1.4373,
      "theoretical_loss": 3.4061144985742726,
      "tokens_seen": 2416902144
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002705516549648947,
      "loss": 1.3805,
      "theoretical_loss": 3.406107143559956,
      "tokens_seen": 2416967680
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002705315947843531,
      "loss": 1.2759,
      "theoretical_loss": 3.4060997888009066,
      "tokens_seen": 2417033216
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027051153460381144,
      "loss": 1.4551,
      "theoretical_loss": 3.4060924342971086,
      "tokens_seen": 2417098752
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002704914744232698,
      "loss": 1.3696,
      "theoretical_loss": 3.406085080048546,
      "tokens_seen": 2417164288
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002704714142427282,
      "loss": 1.5157,
      "theoretical_loss": 3.4060777260552038,
      "tokens_seen": 2417229824
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002704513540621866,
      "loss": 1.3386,
      "theoretical_loss": 3.4060703723170653,
      "tokens_seen": 2417295360
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027043129388164494,
      "loss": 1.5317,
      "theoretical_loss": 3.406063018834115,
      "tokens_seen": 2417360896
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002704112337011033,
      "loss": 1.4527,
      "theoretical_loss": 3.406055665606338,
      "tokens_seen": 2417426432
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027039117352056167,
      "loss": 1.366,
      "theoretical_loss": 3.4060483126337173,
      "tokens_seen": 2417491968
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002703711133400201,
      "loss": 1.4725,
      "theoretical_loss": 3.4060409599162376,
      "tokens_seen": 2417557504
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027035105315947845,
      "loss": 1.3074,
      "theoretical_loss": 3.406033607453883,
      "tokens_seen": 2417623040
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002703309929789368,
      "loss": 1.452,
      "theoretical_loss": 3.406026255246638,
      "tokens_seen": 2417688576
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027031093279839517,
      "loss": 1.5374,
      "theoretical_loss": 3.4060189032944868,
      "tokens_seen": 2417754112
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002702908726178536,
      "loss": 1.2758,
      "theoretical_loss": 3.4060115515974134,
      "tokens_seen": 2417819648
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027027081243731195,
      "loss": 1.4189,
      "theoretical_loss": 3.4060042001554023,
      "tokens_seen": 2417885184
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002702507522567703,
      "loss": 1.3611,
      "theoretical_loss": 3.4059968489684374,
      "tokens_seen": 2417950720
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002702306920762287,
      "loss": 1.4008,
      "theoretical_loss": 3.4059894980365035,
      "tokens_seen": 2418016256
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002702106318956871,
      "loss": 1.43,
      "theoretical_loss": 3.4059821473595844,
      "tokens_seen": 2418081792
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027019057171514545,
      "loss": 1.4282,
      "theoretical_loss": 3.4059747969376644,
      "tokens_seen": 2418147328
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002701705115346038,
      "loss": 1.3117,
      "theoretical_loss": 3.4059674467707275,
      "tokens_seen": 2418212864
    },
    {
      "epoch": 0.73,
      "objective/train/advantage_avg": 0.49238163232803345,
      "objective/train/docs_used": 880852,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3269131183624268,
      "objective/train/original_loss": 1.3269132375717163,
      "objective/train/theoretical_loss": 3.4059600968587587,
      "objective/train/tokens_used": 2438738400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2434282749891281,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504757165908813,
      "objective/train/weighted_lm_loss": 1.3939543962478638,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4059600968587587,
      "tokens_seen": 2418278400
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002701504513540622,
      "loss": 1.4229,
      "theoretical_loss": 3.4059600968587587,
      "tokens_seen": 2418278400
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027013039117352054,
      "loss": 1.4918,
      "theoretical_loss": 3.4059527472017415,
      "tokens_seen": 2418343936
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027011033099297896,
      "loss": 1.3014,
      "theoretical_loss": 3.4059453977996603,
      "tokens_seen": 2418409472
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002700902708124373,
      "loss": 1.368,
      "theoretical_loss": 3.4059380486524997,
      "tokens_seen": 2418475008
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002700702106318957,
      "loss": 1.3406,
      "theoretical_loss": 3.4059306997602437,
      "tokens_seen": 2418540544
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027005015045135404,
      "loss": 1.3165,
      "theoretical_loss": 3.4059233511228766,
      "tokens_seen": 2418606080
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00027003009027081246,
      "loss": 1.4074,
      "theoretical_loss": 3.4059160027403825,
      "tokens_seen": 2418671616
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002700100300902708,
      "loss": 1.3767,
      "theoretical_loss": 3.405908654612746,
      "tokens_seen": 2418737152
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002699899699097292,
      "loss": 1.3615,
      "theoretical_loss": 3.405901306739951,
      "tokens_seen": 2418802688
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026996990972918755,
      "loss": 1.4322,
      "theoretical_loss": 3.4058939591219817,
      "tokens_seen": 2418868224
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002699498495486459,
      "loss": 1.3842,
      "theoretical_loss": 3.405886611758823,
      "tokens_seen": 2418933760
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002699297893681043,
      "loss": 1.2933,
      "theoretical_loss": 3.405879264650458,
      "tokens_seen": 2418999296
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002699097291875627,
      "loss": 1.3397,
      "theoretical_loss": 3.405871917796872,
      "tokens_seen": 2419064832
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026988966900702105,
      "loss": 1.4379,
      "theoretical_loss": 3.4058645711980495,
      "tokens_seen": 2419130368
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002698696088264794,
      "loss": 1.31,
      "theoretical_loss": 3.405857224853974,
      "tokens_seen": 2419195904
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026984954864593783,
      "loss": 1.5153,
      "theoretical_loss": 3.4058498787646294,
      "tokens_seen": 2419261440
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002698294884653962,
      "loss": 1.514,
      "theoretical_loss": 3.405842532930001,
      "tokens_seen": 2419326976
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026980942828485455,
      "loss": 1.2973,
      "theoretical_loss": 3.405835187350072,
      "tokens_seen": 2419392512
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002697893681043129,
      "loss": 1.294,
      "theoretical_loss": 3.405827842024828,
      "tokens_seen": 2419458048
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002697693079237713,
      "loss": 1.3433,
      "theoretical_loss": 3.4058204969542523,
      "tokens_seen": 2419523584
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002697492477432297,
      "loss": 1.4672,
      "theoretical_loss": 3.4058131521383292,
      "tokens_seen": 2419589120
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026972918756268806,
      "loss": 1.3173,
      "theoretical_loss": 3.4058058075770434,
      "tokens_seen": 2419654656
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002697091273821464,
      "loss": 1.3611,
      "theoretical_loss": 3.4057984632703793,
      "tokens_seen": 2419720192
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002696890672016048,
      "loss": 1.3925,
      "theoretical_loss": 3.4057911192183203,
      "tokens_seen": 2419785728
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002696690070210632,
      "loss": 1.3471,
      "theoretical_loss": 3.4057837754208515,
      "tokens_seen": 2419851264
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026964894684052156,
      "loss": 1.3978,
      "theoretical_loss": 3.405776431877957,
      "tokens_seen": 2419916800
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002696288866599799,
      "loss": 1.3089,
      "theoretical_loss": 3.4057690885896204,
      "tokens_seen": 2419982336
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002696088264794383,
      "loss": 1.3859,
      "theoretical_loss": 3.405761745555827,
      "tokens_seen": 2420047872
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026958876629889665,
      "loss": 1.5214,
      "theoretical_loss": 3.4057544027765605,
      "tokens_seen": 2420113408
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026956870611835506,
      "loss": 1.5091,
      "theoretical_loss": 3.4057470602518056,
      "tokens_seen": 2420178944
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002695486459378134,
      "loss": 1.4617,
      "theoretical_loss": 3.4057397179815463,
      "tokens_seen": 2420244480
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002695285857572718,
      "loss": 1.459,
      "theoretical_loss": 3.405732375965767,
      "tokens_seen": 2420310016
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002695085255767302,
      "loss": 1.3674,
      "theoretical_loss": 3.4057250342044516,
      "tokens_seen": 2420375552
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002694884653961886,
      "loss": 1.3738,
      "theoretical_loss": 3.4057176926975847,
      "tokens_seen": 2420441088
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000269468405215647,
      "loss": 1.2369,
      "theoretical_loss": 3.4057103514451503,
      "tokens_seen": 2420506624
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026944834503510534,
      "loss": 1.3899,
      "theoretical_loss": 3.4057030104471337,
      "tokens_seen": 2420572160
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002694282848545637,
      "loss": 1.3786,
      "theoretical_loss": 3.405695669703518,
      "tokens_seen": 2420637696
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026940822467402207,
      "loss": 1.3668,
      "theoretical_loss": 3.4056883292142883,
      "tokens_seen": 2420703232
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002693881644934805,
      "loss": 1.3388,
      "theoretical_loss": 3.4056809889794284,
      "tokens_seen": 2420768768
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026936810431293885,
      "loss": 1.4391,
      "theoretical_loss": 3.4056736489989223,
      "tokens_seen": 2420834304
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002693480441323972,
      "loss": 1.3244,
      "theoretical_loss": 3.405666309272755,
      "tokens_seen": 2420899840
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026932798395185557,
      "loss": 1.383,
      "theoretical_loss": 3.405658969800911,
      "tokens_seen": 2420965376
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.000269307923771314,
      "loss": 1.4155,
      "theoretical_loss": 3.4056516305833737,
      "tokens_seen": 2421030912
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026928786359077235,
      "loss": 1.452,
      "theoretical_loss": 3.4056442916201286,
      "tokens_seen": 2421096448
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002692678034102307,
      "loss": 1.4006,
      "theoretical_loss": 3.4056369529111583,
      "tokens_seen": 2421161984
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002692477432296891,
      "loss": 1.3943,
      "theoretical_loss": 3.4056296144564486,
      "tokens_seen": 2421227520
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002692276830491475,
      "loss": 1.4184,
      "theoretical_loss": 3.4056222762559836,
      "tokens_seen": 2421293056
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026920762286860585,
      "loss": 1.5109,
      "theoretical_loss": 3.4056149383097467,
      "tokens_seen": 2421358592
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002691875626880642,
      "loss": 1.299,
      "theoretical_loss": 3.4056076006177234,
      "tokens_seen": 2421424128
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002691675025075226,
      "loss": 1.4462,
      "theoretical_loss": 3.4056002631798967,
      "tokens_seen": 2421489664
    },
    {
      "epoch": 0.73,
      "objective/train/advantage_avg": 0.49602895975112915,
      "objective/train/docs_used": 881885,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1949819326400757,
      "objective/train/original_loss": 1.1949820518493652,
      "objective/train/theoretical_loss": 3.4055929259962525,
      "objective/train/tokens_used": 2442015200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24621139466762543,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508546829223633,
      "objective/train/weighted_lm_loss": 1.2557467222213745,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 3.4055929259962525,
      "tokens_seen": 2421555200
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026914744232698094,
      "loss": 1.3355,
      "theoretical_loss": 3.4055929259962525,
      "tokens_seen": 2421555200
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026912738214643936,
      "loss": 1.2843,
      "theoretical_loss": 3.4055855890667734,
      "tokens_seen": 2421620736
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002691073219658977,
      "loss": 1.4107,
      "theoretical_loss": 3.4055782523914453,
      "tokens_seen": 2421686272
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002690872617853561,
      "loss": 1.4325,
      "theoretical_loss": 3.4055709159702516,
      "tokens_seen": 2421751808
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026906720160481444,
      "loss": 1.3672,
      "theoretical_loss": 3.4055635798031765,
      "tokens_seen": 2421817344
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026904714142427286,
      "loss": 1.5063,
      "theoretical_loss": 3.4055562438902047,
      "tokens_seen": 2421882880
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002690270812437312,
      "loss": 1.3084,
      "theoretical_loss": 3.405548908231321,
      "tokens_seen": 2421948416
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002690070210631896,
      "loss": 1.3314,
      "theoretical_loss": 3.405541572826509,
      "tokens_seen": 2422013952
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026898696088264795,
      "loss": 1.4614,
      "theoretical_loss": 3.4055342376757527,
      "tokens_seen": 2422079488
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002689669007021063,
      "loss": 1.2526,
      "theoretical_loss": 3.4055269027790374,
      "tokens_seen": 2422145024
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002689468405215647,
      "loss": 1.4743,
      "theoretical_loss": 3.405519568136347,
      "tokens_seen": 2422210560
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002689267803410231,
      "loss": 1.3941,
      "theoretical_loss": 3.4055122337476655,
      "tokens_seen": 2422276096
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026890672016048145,
      "loss": 1.473,
      "theoretical_loss": 3.4055048996129775,
      "tokens_seen": 2422341632
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002688866599799398,
      "loss": 1.5385,
      "theoretical_loss": 3.4054975657322677,
      "tokens_seen": 2422407168
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026886659979939823,
      "loss": 1.504,
      "theoretical_loss": 3.40549023210552,
      "tokens_seen": 2422472704
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002688465396188566,
      "loss": 1.4371,
      "theoretical_loss": 3.4054828987327186,
      "tokens_seen": 2422538240
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026882647943831495,
      "loss": 1.518,
      "theoretical_loss": 3.4054755656138487,
      "tokens_seen": 2422603776
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002688064192577733,
      "loss": 1.5205,
      "theoretical_loss": 3.4054682327488934,
      "tokens_seen": 2422669312
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002687863590772317,
      "loss": 1.3667,
      "theoretical_loss": 3.4054609001378378,
      "tokens_seen": 2422734848
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002687662988966901,
      "loss": 1.3133,
      "theoretical_loss": 3.405453567780666,
      "tokens_seen": 2422800384
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026874623871614846,
      "loss": 1.3761,
      "theoretical_loss": 3.4054462356773625,
      "tokens_seen": 2422865920
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002687261785356068,
      "loss": 1.5033,
      "theoretical_loss": 3.405438903827912,
      "tokens_seen": 2422931456
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002687061183550652,
      "loss": 1.4638,
      "theoretical_loss": 3.4054315722322976,
      "tokens_seen": 2422996992
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002686860581745236,
      "loss": 1.2912,
      "theoretical_loss": 3.4054242408905053,
      "tokens_seen": 2423062528
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026866599799398196,
      "loss": 1.3828,
      "theoretical_loss": 3.405416909802518,
      "tokens_seen": 2423128064
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002686459378134403,
      "loss": 1.5001,
      "theoretical_loss": 3.405409578968321,
      "tokens_seen": 2423193600
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002686258776328987,
      "loss": 1.3365,
      "theoretical_loss": 3.4054022483878983,
      "tokens_seen": 2423259136
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026860581745235705,
      "loss": 1.3558,
      "theoretical_loss": 3.405394918061234,
      "tokens_seen": 2423324672
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026858575727181546,
      "loss": 1.4352,
      "theoretical_loss": 3.405387587988313,
      "tokens_seen": 2423390208
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002685656970912738,
      "loss": 1.5216,
      "theoretical_loss": 3.405380258169119,
      "tokens_seen": 2423455744
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002685456369107322,
      "loss": 1.336,
      "theoretical_loss": 3.4053729286036374,
      "tokens_seen": 2423521280
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026852557673019055,
      "loss": 1.4067,
      "theoretical_loss": 3.4053655992918515,
      "tokens_seen": 2423586816
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026850551654964897,
      "loss": 1.4403,
      "theoretical_loss": 3.405358270233746,
      "tokens_seen": 2423652352
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026848545636910733,
      "loss": 1.3097,
      "theoretical_loss": 3.4053509414293055,
      "tokens_seen": 2423717888
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002684653961885657,
      "loss": 1.3365,
      "theoretical_loss": 3.405343612878514,
      "tokens_seen": 2423783424
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026844533600802405,
      "loss": 1.2672,
      "theoretical_loss": 3.405336284581356,
      "tokens_seen": 2423848960
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002684252758274824,
      "loss": 1.3132,
      "theoretical_loss": 3.405328956537816,
      "tokens_seen": 2423914496
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026840521564694083,
      "loss": 1.4668,
      "theoretical_loss": 3.4053216287478785,
      "tokens_seen": 2423980032
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002683851554663992,
      "loss": 1.3527,
      "theoretical_loss": 3.405314301211527,
      "tokens_seen": 2424045568
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026836509528585756,
      "loss": 1.32,
      "theoretical_loss": 3.405306973928747,
      "tokens_seen": 2424111104
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002683450351053159,
      "loss": 1.2838,
      "theoretical_loss": 3.4052996468995222,
      "tokens_seen": 2424176640
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026832497492477433,
      "loss": 1.3656,
      "theoretical_loss": 3.4052923201238374,
      "tokens_seen": 2424242176
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002683049147442327,
      "loss": 1.4321,
      "theoretical_loss": 3.4052849936016765,
      "tokens_seen": 2424307712
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026828485456369106,
      "loss": 1.383,
      "theoretical_loss": 3.405277667333024,
      "tokens_seen": 2424373248
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002682647943831494,
      "loss": 1.4362,
      "theoretical_loss": 3.4052703413178644,
      "tokens_seen": 2424438784
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002682447342026078,
      "loss": 1.3063,
      "theoretical_loss": 3.4052630155561823,
      "tokens_seen": 2424504320
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002682246740220662,
      "loss": 1.423,
      "theoretical_loss": 3.4052556900479614,
      "tokens_seen": 2424569856
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026820461384152456,
      "loss": 1.3755,
      "theoretical_loss": 3.405248364793187,
      "tokens_seen": 2424635392
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002681845536609829,
      "loss": 1.3261,
      "theoretical_loss": 3.405241039791843,
      "tokens_seen": 2424700928
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002681644934804413,
      "loss": 1.4648,
      "theoretical_loss": 3.405233715043913,
      "tokens_seen": 2424766464
    },
    {
      "epoch": 0.73,
      "objective/train/advantage_avg": 0.49374017119407654,
      "objective/train/docs_used": 882990,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4434094429016113,
      "objective/train/original_loss": 1.4434094429016113,
      "objective/train/theoretical_loss": 3.405226390549383,
      "objective/train/tokens_used": 2445292000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24409286677837372,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506149530410767,
      "objective/train/weighted_lm_loss": 1.5164159536361694,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.405226390549383,
      "tokens_seen": 2424832000
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002681444332998997,
      "loss": 1.4335,
      "theoretical_loss": 3.405226390549383,
      "tokens_seen": 2424832000
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026812437311935807,
      "loss": 1.3473,
      "theoretical_loss": 3.405219066308236,
      "tokens_seen": 2424897536
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026810431293881643,
      "loss": 1.534,
      "theoretical_loss": 3.405211742320457,
      "tokens_seen": 2424963072
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002680842527582748,
      "loss": 1.403,
      "theoretical_loss": 3.4052044185860306,
      "tokens_seen": 2425028608
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026806419257773315,
      "loss": 1.3787,
      "theoretical_loss": 3.4051970951049406,
      "tokens_seen": 2425094144
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026804413239719157,
      "loss": 1.4989,
      "theoretical_loss": 3.405189771877172,
      "tokens_seen": 2425159680
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026802407221664993,
      "loss": 1.4318,
      "theoretical_loss": 3.4051824489027087,
      "tokens_seen": 2425225216
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00026800401203610835,
      "loss": 1.3948,
      "theoretical_loss": 3.4051751261815353,
      "tokens_seen": 2425290752
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002679839518555667,
      "loss": 1.442,
      "theoretical_loss": 3.4051678037136357,
      "tokens_seen": 2425356288
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002679638916750251,
      "loss": 1.446,
      "theoretical_loss": 3.4051604814989953,
      "tokens_seen": 2425421824
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002679438314944835,
      "loss": 1.374,
      "theoretical_loss": 3.405153159537598,
      "tokens_seen": 2425487360
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026792377131394185,
      "loss": 1.4332,
      "theoretical_loss": 3.405145837829428,
      "tokens_seen": 2425552896
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002679037111334002,
      "loss": 1.3292,
      "theoretical_loss": 3.40513851637447,
      "tokens_seen": 2425618432
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026788365095285863,
      "loss": 1.5134,
      "theoretical_loss": 3.405131195172708,
      "tokens_seen": 2425683968
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000267863590772317,
      "loss": 1.4563,
      "theoretical_loss": 3.405123874224127,
      "tokens_seen": 2425749504
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026784353059177535,
      "loss": 1.2588,
      "theoretical_loss": 3.405116553528711,
      "tokens_seen": 2425815040
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002678234704112337,
      "loss": 1.3895,
      "theoretical_loss": 3.4051092330864443,
      "tokens_seen": 2425880576
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002678034102306921,
      "loss": 1.2796,
      "theoretical_loss": 3.4051019128973117,
      "tokens_seen": 2425946112
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002677833500501505,
      "loss": 1.5088,
      "theoretical_loss": 3.405094592961297,
      "tokens_seen": 2426011648
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026776328986960886,
      "loss": 1.4472,
      "theoretical_loss": 3.4050872732783857,
      "tokens_seen": 2426077184
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002677432296890672,
      "loss": 1.3462,
      "theoretical_loss": 3.405079953848561,
      "tokens_seen": 2426142720
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002677231695085256,
      "loss": 1.3992,
      "theoretical_loss": 3.405072634671808,
      "tokens_seen": 2426208256
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000267703109327984,
      "loss": 1.3659,
      "theoretical_loss": 3.4050653157481108,
      "tokens_seen": 2426273792
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026768304914744236,
      "loss": 1.3606,
      "theoretical_loss": 3.405057997077454,
      "tokens_seen": 2426339328
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002676629889669007,
      "loss": 1.5083,
      "theoretical_loss": 3.4050506786598222,
      "tokens_seen": 2426404864
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002676429287863591,
      "loss": 1.4378,
      "theoretical_loss": 3.4050433604951995,
      "tokens_seen": 2426470400
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026762286860581745,
      "loss": 1.3994,
      "theoretical_loss": 3.4050360425835704,
      "tokens_seen": 2426535936
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026760280842527586,
      "loss": 1.4406,
      "theoretical_loss": 3.4050287249249194,
      "tokens_seen": 2426601472
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002675827482447342,
      "loss": 1.4009,
      "theoretical_loss": 3.405021407519231,
      "tokens_seen": 2426667008
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002675626880641926,
      "loss": 1.3844,
      "theoretical_loss": 3.405014090366489,
      "tokens_seen": 2426732544
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026754262788365095,
      "loss": 1.3753,
      "theoretical_loss": 3.4050067734666785,
      "tokens_seen": 2426798080
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026752256770310937,
      "loss": 1.356,
      "theoretical_loss": 3.404999456819784,
      "tokens_seen": 2426863616
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026750250752256773,
      "loss": 1.4931,
      "theoretical_loss": 3.4049921404257897,
      "tokens_seen": 2426929152
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002674824473420261,
      "loss": 1.2523,
      "theoretical_loss": 3.4049848242846794,
      "tokens_seen": 2426994688
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026746238716148445,
      "loss": 1.3244,
      "theoretical_loss": 3.404977508396439,
      "tokens_seen": 2427060224
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002674423269809428,
      "loss": 1.4286,
      "theoretical_loss": 3.404970192761051,
      "tokens_seen": 2427125760
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026742226680040123,
      "loss": 1.3919,
      "theoretical_loss": 3.4049628773785017,
      "tokens_seen": 2427191296
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002674022066198596,
      "loss": 1.3122,
      "theoretical_loss": 3.404955562248775,
      "tokens_seen": 2427256832
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026738214643931796,
      "loss": 1.3585,
      "theoretical_loss": 3.404948247371854,
      "tokens_seen": 2427322368
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002673620862587763,
      "loss": 1.4194,
      "theoretical_loss": 3.404940932747725,
      "tokens_seen": 2427387904
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026734202607823474,
      "loss": 1.2782,
      "theoretical_loss": 3.4049336183763717,
      "tokens_seen": 2427453440
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002673219658976931,
      "loss": 1.4428,
      "theoretical_loss": 3.404926304257778,
      "tokens_seen": 2427518976
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026730190571715146,
      "loss": 1.3477,
      "theoretical_loss": 3.404918990391929,
      "tokens_seen": 2427584512
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002672818455366098,
      "loss": 1.4296,
      "theoretical_loss": 3.4049116767788092,
      "tokens_seen": 2427650048
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002672617853560682,
      "loss": 1.375,
      "theoretical_loss": 3.4049043634184026,
      "tokens_seen": 2427715584
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002672417251755266,
      "loss": 1.2886,
      "theoretical_loss": 3.404897050310694,
      "tokens_seen": 2427781120
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026722166499498496,
      "loss": 1.433,
      "theoretical_loss": 3.4048897374556675,
      "tokens_seen": 2427846656
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002672016048144433,
      "loss": 1.4079,
      "theoretical_loss": 3.404882424853308,
      "tokens_seen": 2427912192
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002671815446339017,
      "loss": 1.4907,
      "theoretical_loss": 3.4048751125035994,
      "tokens_seen": 2427977728
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002671614844533601,
      "loss": 1.4479,
      "theoretical_loss": 3.4048678004065267,
      "tokens_seen": 2428043264
    },
    {
      "epoch": 0.74,
      "objective/train/advantage_avg": 0.49451345205307007,
      "objective/train/docs_used": 884091,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.207484483718872,
      "objective/train/original_loss": 1.2074843645095825,
      "objective/train/theoretical_loss": 3.404860488562074,
      "objective/train/tokens_used": 2448568800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2447201907634735,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506954193115234,
      "objective/train/weighted_lm_loss": 1.2686408758163452,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.404860488562074,
      "tokens_seen": 2428108800
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026714142427281847,
      "loss": 1.3966,
      "theoretical_loss": 3.404860488562074,
      "tokens_seen": 2428108800
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026712136409227683,
      "loss": 1.3262,
      "theoretical_loss": 3.404853176970226,
      "tokens_seen": 2428174336
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002671013039117352,
      "loss": 1.3745,
      "theoretical_loss": 3.404845865630967,
      "tokens_seen": 2428239872
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026708124373119355,
      "loss": 1.4232,
      "theoretical_loss": 3.4048385545442814,
      "tokens_seen": 2428305408
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026706118355065197,
      "loss": 1.5509,
      "theoretical_loss": 3.404831243710154,
      "tokens_seen": 2428370944
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026704112337011033,
      "loss": 1.3149,
      "theoretical_loss": 3.4048239331285686,
      "tokens_seen": 2428436480
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002670210631895687,
      "loss": 1.3011,
      "theoretical_loss": 3.40481662279951,
      "tokens_seen": 2428502016
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026700100300902706,
      "loss": 1.4375,
      "theoretical_loss": 3.4048093127229633,
      "tokens_seen": 2428567552
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026698094282848547,
      "loss": 1.3755,
      "theoretical_loss": 3.404802002898912,
      "tokens_seen": 2428633088
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026696088264794383,
      "loss": 1.3673,
      "theoretical_loss": 3.4047946933273407,
      "tokens_seen": 2428698624
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002669408224674022,
      "loss": 1.3865,
      "theoretical_loss": 3.4047873840082348,
      "tokens_seen": 2428764160
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026692076228686056,
      "loss": 1.3209,
      "theoretical_loss": 3.4047800749415775,
      "tokens_seen": 2428829696
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000266900702106319,
      "loss": 1.5067,
      "theoretical_loss": 3.404772766127354,
      "tokens_seen": 2428895232
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026688064192577734,
      "loss": 1.3723,
      "theoretical_loss": 3.404765457565549,
      "tokens_seen": 2428960768
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002668605817452357,
      "loss": 1.3724,
      "theoretical_loss": 3.404758149256146,
      "tokens_seen": 2429026304
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026684052156469406,
      "loss": 1.4204,
      "theoretical_loss": 3.4047508411991303,
      "tokens_seen": 2429091840
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002668204613841524,
      "loss": 1.3214,
      "theoretical_loss": 3.404743533394486,
      "tokens_seen": 2429157376
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026680040120361084,
      "loss": 1.4017,
      "theoretical_loss": 3.404736225842198,
      "tokens_seen": 2429222912
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002667803410230692,
      "loss": 1.4021,
      "theoretical_loss": 3.4047289185422507,
      "tokens_seen": 2429288448
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026676028084252757,
      "loss": 1.3938,
      "theoretical_loss": 3.404721611494628,
      "tokens_seen": 2429353984
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026674022066198593,
      "loss": 1.3536,
      "theoretical_loss": 3.404714304699315,
      "tokens_seen": 2429419520
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026672016048144434,
      "loss": 1.325,
      "theoretical_loss": 3.4047069981562954,
      "tokens_seen": 2429485056
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002667001003009027,
      "loss": 1.326,
      "theoretical_loss": 3.404699691865555,
      "tokens_seen": 2429550592
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026668004012036107,
      "loss": 1.5077,
      "theoretical_loss": 3.4046923858270772,
      "tokens_seen": 2429616128
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026665997993981943,
      "loss": 1.3491,
      "theoretical_loss": 3.4046850800408466,
      "tokens_seen": 2429681664
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002666399197592778,
      "loss": 1.3646,
      "theoretical_loss": 3.4046777745068484,
      "tokens_seen": 2429747200
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002666198595787362,
      "loss": 1.4175,
      "theoretical_loss": 3.4046704692250658,
      "tokens_seen": 2429812736
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026659979939819457,
      "loss": 1.3241,
      "theoretical_loss": 3.4046631641954845,
      "tokens_seen": 2429878272
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026657973921765293,
      "loss": 1.3223,
      "theoretical_loss": 3.4046558594180887,
      "tokens_seen": 2429943808
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002665596790371113,
      "loss": 1.3162,
      "theoretical_loss": 3.4046485548928627,
      "tokens_seen": 2430009344
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002665396188565697,
      "loss": 1.545,
      "theoretical_loss": 3.404641250619791,
      "tokens_seen": 2430074880
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002665195586760281,
      "loss": 1.4652,
      "theoretical_loss": 3.404633946598858,
      "tokens_seen": 2430140416
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002664994984954865,
      "loss": 1.4806,
      "theoretical_loss": 3.4046266428300482,
      "tokens_seen": 2430205952
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026647943831494485,
      "loss": 1.3945,
      "theoretical_loss": 3.4046193393133466,
      "tokens_seen": 2430271488
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002664593781344032,
      "loss": 1.4203,
      "theoretical_loss": 3.404612036048737,
      "tokens_seen": 2430337024
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026643931795386163,
      "loss": 1.4171,
      "theoretical_loss": 3.4046047330362046,
      "tokens_seen": 2430402560
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026641925777332,
      "loss": 1.4153,
      "theoretical_loss": 3.404597430275733,
      "tokens_seen": 2430468096
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026639919759277836,
      "loss": 1.4549,
      "theoretical_loss": 3.4045901277673076,
      "tokens_seen": 2430533632
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002663791374122367,
      "loss": 1.3738,
      "theoretical_loss": 3.404582825510913,
      "tokens_seen": 2430599168
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026635907723169514,
      "loss": 1.3101,
      "theoretical_loss": 3.4045755235065327,
      "tokens_seen": 2430664704
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002663390170511535,
      "loss": 1.4377,
      "theoretical_loss": 3.4045682217541517,
      "tokens_seen": 2430730240
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026631895687061186,
      "loss": 1.428,
      "theoretical_loss": 3.404560920253755,
      "tokens_seen": 2430795776
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002662988966900702,
      "loss": 1.3512,
      "theoretical_loss": 3.404553619005326,
      "tokens_seen": 2430861312
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002662788365095286,
      "loss": 1.4131,
      "theoretical_loss": 3.4045463180088507,
      "tokens_seen": 2430926848
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000266258776328987,
      "loss": 1.4449,
      "theoretical_loss": 3.4045390172643124,
      "tokens_seen": 2430992384
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026623871614844536,
      "loss": 1.3796,
      "theoretical_loss": 3.4045317167716957,
      "tokens_seen": 2431057920
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002662186559679037,
      "loss": 1.3318,
      "theoretical_loss": 3.404524416530986,
      "tokens_seen": 2431123456
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002661985957873621,
      "loss": 1.3648,
      "theoretical_loss": 3.4045171165421673,
      "tokens_seen": 2431188992
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002661785356068205,
      "loss": 1.4181,
      "theoretical_loss": 3.4045098168052235,
      "tokens_seen": 2431254528
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026615847542627887,
      "loss": 1.3543,
      "theoretical_loss": 3.40450251732014,
      "tokens_seen": 2431320064
    },
    {
      "epoch": 0.74,
      "objective/train/advantage_avg": 0.4945901036262512,
      "objective/train/docs_used": 885315,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3446402549743652,
      "objective/train/original_loss": 1.3446402549743652,
      "objective/train/theoretical_loss": 3.4044952180869013,
      "objective/train/tokens_used": 2451845600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24508242309093475,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507049560546875,
      "objective/train/weighted_lm_loss": 1.412768006324768,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4044952180869013,
      "tokens_seen": 2431385600
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026613841524573723,
      "loss": 1.4305,
      "theoretical_loss": 3.4044952180869013,
      "tokens_seen": 2431385600
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002661183550651956,
      "loss": 1.4738,
      "theoretical_loss": 3.404487919105491,
      "tokens_seen": 2431451136
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026609829488465395,
      "loss": 1.3542,
      "theoretical_loss": 3.404480620375895,
      "tokens_seen": 2431516672
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026607823470411237,
      "loss": 1.4316,
      "theoretical_loss": 3.4044733218980965,
      "tokens_seen": 2431582208
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026605817452357073,
      "loss": 1.4506,
      "theoretical_loss": 3.4044660236720805,
      "tokens_seen": 2431647744
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002660381143430291,
      "loss": 1.3536,
      "theoretical_loss": 3.4044587256978325,
      "tokens_seen": 2431713280
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026601805416248746,
      "loss": 1.4244,
      "theoretical_loss": 3.4044514279753355,
      "tokens_seen": 2431778816
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026599799398194587,
      "loss": 1.4284,
      "theoretical_loss": 3.404444130504575,
      "tokens_seen": 2431844352
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026597793380140423,
      "loss": 1.2243,
      "theoretical_loss": 3.404436833285535,
      "tokens_seen": 2431909888
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002659578736208626,
      "loss": 1.4133,
      "theoretical_loss": 3.4044295363182,
      "tokens_seen": 2431975424
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026593781344032096,
      "loss": 1.4152,
      "theoretical_loss": 3.4044222396025554,
      "tokens_seen": 2432040960
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002659177532597794,
      "loss": 1.3918,
      "theoretical_loss": 3.404414943138585,
      "tokens_seen": 2432106496
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026589769307923774,
      "loss": 1.3368,
      "theoretical_loss": 3.4044076469262734,
      "tokens_seen": 2432172032
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002658776328986961,
      "loss": 1.3747,
      "theoretical_loss": 3.4044003509656053,
      "tokens_seen": 2432237568
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026585757271815446,
      "loss": 1.2834,
      "theoretical_loss": 3.404393055256565,
      "tokens_seen": 2432303104
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002658375125376128,
      "loss": 1.3219,
      "theoretical_loss": 3.4043857597991374,
      "tokens_seen": 2432368640
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026581745235707124,
      "loss": 1.4185,
      "theoretical_loss": 3.4043784645933064,
      "tokens_seen": 2432434176
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002657973921765296,
      "loss": 1.3412,
      "theoretical_loss": 3.4043711696390577,
      "tokens_seen": 2432499712
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026577733199598797,
      "loss": 1.4549,
      "theoretical_loss": 3.4043638749363745,
      "tokens_seen": 2432565248
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026575727181544633,
      "loss": 1.3193,
      "theoretical_loss": 3.4043565804852425,
      "tokens_seen": 2432630784
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026573721163490474,
      "loss": 1.4424,
      "theoretical_loss": 3.4043492862856457,
      "tokens_seen": 2432696320
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002657171514543631,
      "loss": 1.4179,
      "theoretical_loss": 3.404341992337568,
      "tokens_seen": 2432761856
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026569709127382147,
      "loss": 1.3836,
      "theoretical_loss": 3.4043346986409952,
      "tokens_seen": 2432827392
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026567703109327983,
      "loss": 1.3829,
      "theoretical_loss": 3.4043274051959114,
      "tokens_seen": 2432892928
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002656569709127382,
      "loss": 1.4907,
      "theoretical_loss": 3.404320112002301,
      "tokens_seen": 2432958464
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002656369107321966,
      "loss": 1.5053,
      "theoretical_loss": 3.4043128190601486,
      "tokens_seen": 2433024000
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026561685055165497,
      "loss": 1.4029,
      "theoretical_loss": 3.4043055263694386,
      "tokens_seen": 2433089536
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026559679037111333,
      "loss": 1.2968,
      "theoretical_loss": 3.4042982339301555,
      "tokens_seen": 2433155072
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002655767301905717,
      "loss": 1.4061,
      "theoretical_loss": 3.4042909417422846,
      "tokens_seen": 2433220608
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002655566700100301,
      "loss": 1.3697,
      "theoretical_loss": 3.4042836498058096,
      "tokens_seen": 2433286144
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002655366098294885,
      "loss": 1.467,
      "theoretical_loss": 3.4042763581207156,
      "tokens_seen": 2433351680
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026551654964894684,
      "loss": 1.4566,
      "theoretical_loss": 3.404269066686987,
      "tokens_seen": 2433417216
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002654964894684052,
      "loss": 1.3202,
      "theoretical_loss": 3.404261775504608,
      "tokens_seen": 2433482752
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026547642928786356,
      "loss": 1.3484,
      "theoretical_loss": 3.4042544845735634,
      "tokens_seen": 2433548288
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000265456369107322,
      "loss": 1.3804,
      "theoretical_loss": 3.404247193893838,
      "tokens_seen": 2433613824
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026543630892678034,
      "loss": 1.2621,
      "theoretical_loss": 3.4042399034654167,
      "tokens_seen": 2433679360
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002654162487462387,
      "loss": 1.4533,
      "theoretical_loss": 3.404232613288283,
      "tokens_seen": 2433744896
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026539618856569707,
      "loss": 1.4781,
      "theoretical_loss": 3.4042253233624225,
      "tokens_seen": 2433810432
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002653761283851555,
      "loss": 1.387,
      "theoretical_loss": 3.4042180336878194,
      "tokens_seen": 2433875968
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026535606820461384,
      "loss": 1.3869,
      "theoretical_loss": 3.404210744264458,
      "tokens_seen": 2433941504
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002653360080240722,
      "loss": 1.3245,
      "theoretical_loss": 3.404203455092323,
      "tokens_seen": 2434007040
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026531594784353057,
      "loss": 1.3057,
      "theoretical_loss": 3.404196166171399,
      "tokens_seen": 2434072576
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026529588766298893,
      "loss": 1.306,
      "theoretical_loss": 3.404188877501671,
      "tokens_seen": 2434138112
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026527582748244735,
      "loss": 1.3726,
      "theoretical_loss": 3.404181589083123,
      "tokens_seen": 2434203648
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002652557673019057,
      "loss": 1.3881,
      "theoretical_loss": 3.40417430091574,
      "tokens_seen": 2434269184
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026523570712136407,
      "loss": 1.5127,
      "theoretical_loss": 3.4041670129995065,
      "tokens_seen": 2434334720
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026521564694082243,
      "loss": 1.2576,
      "theoretical_loss": 3.4041597253344067,
      "tokens_seen": 2434400256
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026519558676028085,
      "loss": 1.3542,
      "theoretical_loss": 3.404152437920426,
      "tokens_seen": 2434465792
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002651755265797392,
      "loss": 1.4498,
      "theoretical_loss": 3.404145150757548,
      "tokens_seen": 2434531328
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002651554663991976,
      "loss": 1.4932,
      "theoretical_loss": 3.404137863845758,
      "tokens_seen": 2434596864
    },
    {
      "epoch": 0.74,
      "objective/train/advantage_avg": 0.4945903718471527,
      "objective/train/docs_used": 886508,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3504252433776855,
      "objective/train/original_loss": 1.350425124168396,
      "objective/train/theoretical_loss": 3.40413057718504,
      "objective/train/tokens_used": 2455122400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24498647451400757,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507044792175293,
      "objective/train/weighted_lm_loss": 1.4188997745513916,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.40413057718504,
      "tokens_seen": 2434662400
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026513540621865594,
      "loss": 1.3603,
      "theoretical_loss": 3.40413057718504,
      "tokens_seen": 2434662400
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002651153460381143,
      "loss": 1.4186,
      "theoretical_loss": 3.404123290775379,
      "tokens_seen": 2434727936
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002650952858575727,
      "loss": 1.3896,
      "theoretical_loss": 3.4041160046167596,
      "tokens_seen": 2434793472
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002650752256770311,
      "loss": 1.435,
      "theoretical_loss": 3.4041087187091663,
      "tokens_seen": 2434859008
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026505516549648944,
      "loss": 1.4761,
      "theoretical_loss": 3.404101433052584,
      "tokens_seen": 2434924544
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002650351053159478,
      "loss": 1.3639,
      "theoretical_loss": 3.4040941476469966,
      "tokens_seen": 2434990080
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002650150451354062,
      "loss": 1.2265,
      "theoretical_loss": 3.4040868624923895,
      "tokens_seen": 2435055616
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026499498495486464,
      "loss": 1.5446,
      "theoretical_loss": 3.4040795775887465,
      "tokens_seen": 2435121152
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000264974924774323,
      "loss": 1.483,
      "theoretical_loss": 3.404072292936053,
      "tokens_seen": 2435186688
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026495486459378136,
      "loss": 1.2824,
      "theoretical_loss": 3.404065008534293,
      "tokens_seen": 2435252224
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002649348044132398,
      "loss": 1.3606,
      "theoretical_loss": 3.4040577243834513,
      "tokens_seen": 2435317760
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026491474423269814,
      "loss": 1.2604,
      "theoretical_loss": 3.4040504404835126,
      "tokens_seen": 2435383296
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002648946840521565,
      "loss": 1.3487,
      "theoretical_loss": 3.404043156834461,
      "tokens_seen": 2435448832
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026487462387161486,
      "loss": 1.3726,
      "theoretical_loss": 3.404035873436282,
      "tokens_seen": 2435514368
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002648545636910732,
      "loss": 1.3941,
      "theoretical_loss": 3.4040285902889598,
      "tokens_seen": 2435579904
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026483450351053164,
      "loss": 1.4502,
      "theoretical_loss": 3.4040213073924788,
      "tokens_seen": 2435645440
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026481444332999,
      "loss": 1.5324,
      "theoretical_loss": 3.404014024746824,
      "tokens_seen": 2435710976
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026479438314944837,
      "loss": 1.3318,
      "theoretical_loss": 3.4040067423519793,
      "tokens_seen": 2435776512
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026477432296890673,
      "loss": 1.3465,
      "theoretical_loss": 3.40399946020793,
      "tokens_seen": 2435842048
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026475426278836514,
      "loss": 1.4496,
      "theoretical_loss": 3.4039921783146605,
      "tokens_seen": 2435907584
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002647342026078235,
      "loss": 1.3646,
      "theoretical_loss": 3.4039848966721555,
      "tokens_seen": 2435973120
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026471414242728187,
      "loss": 1.2993,
      "theoretical_loss": 3.4039776152803993,
      "tokens_seen": 2436038656
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026469408224674023,
      "loss": 1.3971,
      "theoretical_loss": 3.403970334139377,
      "tokens_seen": 2436104192
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002646740220661986,
      "loss": 1.3594,
      "theoretical_loss": 3.4039630532490732,
      "tokens_seen": 2436169728
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000264653961885657,
      "loss": 1.4705,
      "theoretical_loss": 3.4039557726094722,
      "tokens_seen": 2436235264
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026463390170511537,
      "loss": 1.2666,
      "theoretical_loss": 3.4039484922205583,
      "tokens_seen": 2436300800
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026461384152457373,
      "loss": 1.3349,
      "theoretical_loss": 3.403941212082317,
      "tokens_seen": 2436366336
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002645937813440321,
      "loss": 1.3308,
      "theoretical_loss": 3.4039339321947324,
      "tokens_seen": 2436431872
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002645737211634905,
      "loss": 1.3171,
      "theoretical_loss": 3.4039266525577894,
      "tokens_seen": 2436497408
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002645536609829489,
      "loss": 1.3633,
      "theoretical_loss": 3.4039193731714716,
      "tokens_seen": 2436562944
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026453360080240724,
      "loss": 1.5984,
      "theoretical_loss": 3.4039120940357654,
      "tokens_seen": 2436628480
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002645135406218656,
      "loss": 1.3462,
      "theoretical_loss": 3.4039048151506544,
      "tokens_seen": 2436694016
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026449348044132396,
      "loss": 1.3776,
      "theoretical_loss": 3.403897536516123,
      "tokens_seen": 2436759552
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002644734202607824,
      "loss": 1.4008,
      "theoretical_loss": 3.403890258132156,
      "tokens_seen": 2436825088
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026445336008024074,
      "loss": 1.357,
      "theoretical_loss": 3.4038829799987385,
      "tokens_seen": 2436890624
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002644332998996991,
      "loss": 1.3646,
      "theoretical_loss": 3.403875702115855,
      "tokens_seen": 2436956160
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026441323971915747,
      "loss": 1.4294,
      "theoretical_loss": 3.40386842448349,
      "tokens_seen": 2437021696
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002643931795386159,
      "loss": 1.3474,
      "theoretical_loss": 3.403861147101628,
      "tokens_seen": 2437087232
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026437311935807424,
      "loss": 1.4246,
      "theoretical_loss": 3.4038538699702534,
      "tokens_seen": 2437152768
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002643530591775326,
      "loss": 1.387,
      "theoretical_loss": 3.4038465930893516,
      "tokens_seen": 2437218304
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026433299899699097,
      "loss": 1.3677,
      "theoretical_loss": 3.4038393164589067,
      "tokens_seen": 2437283840
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026431293881644933,
      "loss": 1.3136,
      "theoretical_loss": 3.4038320400789033,
      "tokens_seen": 2437349376
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026429287863590775,
      "loss": 1.4419,
      "theoretical_loss": 3.403824763949326,
      "tokens_seen": 2437414912
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002642728184553661,
      "loss": 1.3528,
      "theoretical_loss": 3.403817488070161,
      "tokens_seen": 2437480448
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026425275827482447,
      "loss": 1.3433,
      "theoretical_loss": 3.4038102124413907,
      "tokens_seen": 2437545984
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026423269809428283,
      "loss": 1.4642,
      "theoretical_loss": 3.4038029370630007,
      "tokens_seen": 2437611520
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026421263791374125,
      "loss": 1.4069,
      "theoretical_loss": 3.403795661934976,
      "tokens_seen": 2437677056
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002641925777331996,
      "loss": 1.3707,
      "theoretical_loss": 3.4037883870573005,
      "tokens_seen": 2437742592
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000264172517552658,
      "loss": 1.3977,
      "theoretical_loss": 3.403781112429959,
      "tokens_seen": 2437808128
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026415245737211634,
      "loss": 1.2358,
      "theoretical_loss": 3.403773838052937,
      "tokens_seen": 2437873664
    },
    {
      "epoch": 0.74,
      "objective/train/advantage_avg": 0.4919215142726898,
      "objective/train/docs_used": 887686,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3000240325927734,
      "objective/train/original_loss": 1.3000237941741943,
      "objective/train/theoretical_loss": 3.4037665639262182,
      "objective/train/tokens_used": 2458399200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24289992451667786,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050426959991455,
      "objective/train/weighted_lm_loss": 1.3654282093048096,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4037665639262182,
      "tokens_seen": 2437939200
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002641323971915747,
      "loss": 1.3234,
      "theoretical_loss": 3.4037665639262182,
      "tokens_seen": 2437939200
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002641123370110331,
      "loss": 1.3288,
      "theoretical_loss": 3.4037592900497877,
      "tokens_seen": 2438004736
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002640922768304915,
      "loss": 1.3435,
      "theoretical_loss": 3.4037520164236303,
      "tokens_seen": 2438070272
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026407221664994984,
      "loss": 1.3918,
      "theoretical_loss": 3.4037447430477306,
      "tokens_seen": 2438135808
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002640521564694082,
      "loss": 1.3661,
      "theoretical_loss": 3.403737469922073,
      "tokens_seen": 2438201344
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002640320962888666,
      "loss": 1.5205,
      "theoretical_loss": 3.4037301970466416,
      "tokens_seen": 2438266880
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000264012036108325,
      "loss": 1.3661,
      "theoretical_loss": 3.4037229244214227,
      "tokens_seen": 2438332416
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026399197592778334,
      "loss": 1.297,
      "theoretical_loss": 3.403715652046399,
      "tokens_seen": 2438397952
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002639719157472417,
      "loss": 1.4241,
      "theoretical_loss": 3.4037083799215573,
      "tokens_seen": 2438463488
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026395185556670007,
      "loss": 1.4096,
      "theoretical_loss": 3.4037011080468806,
      "tokens_seen": 2438529024
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002639317953861585,
      "loss": 1.3514,
      "theoretical_loss": 3.403693836422354,
      "tokens_seen": 2438594560
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026391173520561685,
      "loss": 1.2611,
      "theoretical_loss": 3.403686565047962,
      "tokens_seen": 2438660096
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002638916750250752,
      "loss": 1.3516,
      "theoretical_loss": 3.40367929392369,
      "tokens_seen": 2438725632
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026387161484453357,
      "loss": 1.379,
      "theoretical_loss": 3.4036720230495225,
      "tokens_seen": 2438791168
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000263851554663992,
      "loss": 1.2943,
      "theoretical_loss": 3.4036647524254438,
      "tokens_seen": 2438856704
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026383149448345035,
      "loss": 1.3903,
      "theoretical_loss": 3.4036574820514387,
      "tokens_seen": 2438922240
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002638114343029087,
      "loss": 1.3943,
      "theoretical_loss": 3.4036502119274914,
      "tokens_seen": 2438987776
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002637913741223671,
      "loss": 1.373,
      "theoretical_loss": 3.403642942053587,
      "tokens_seen": 2439053312
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002637713139418255,
      "loss": 1.2119,
      "theoretical_loss": 3.403635672429711,
      "tokens_seen": 2439118848
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026375125376128385,
      "loss": 1.3451,
      "theoretical_loss": 3.403628403055847,
      "tokens_seen": 2439184384
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002637311935807422,
      "loss": 1.3159,
      "theoretical_loss": 3.4036211339319795,
      "tokens_seen": 2439249920
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002637111334002006,
      "loss": 1.3094,
      "theoretical_loss": 3.4036138650580945,
      "tokens_seen": 2439315456
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026369107321965894,
      "loss": 1.2918,
      "theoretical_loss": 3.4036065964341753,
      "tokens_seen": 2439380992
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026367101303911736,
      "loss": 1.4747,
      "theoretical_loss": 3.4035993280602073,
      "tokens_seen": 2439446528
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002636509528585757,
      "loss": 1.3697,
      "theoretical_loss": 3.4035920599361753,
      "tokens_seen": 2439512064
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002636308926780341,
      "loss": 1.4758,
      "theoretical_loss": 3.4035847920620634,
      "tokens_seen": 2439577600
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026361083249749244,
      "loss": 1.4531,
      "theoretical_loss": 3.403577524437857,
      "tokens_seen": 2439643136
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026359077231695086,
      "loss": 1.416,
      "theoretical_loss": 3.4035702570635404,
      "tokens_seen": 2439708672
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002635707121364092,
      "loss": 1.3332,
      "theoretical_loss": 3.403562989939098,
      "tokens_seen": 2439774208
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002635506519558676,
      "loss": 1.3915,
      "theoretical_loss": 3.403555723064515,
      "tokens_seen": 2439839744
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026353059177532595,
      "loss": 1.4767,
      "theoretical_loss": 3.403548456439776,
      "tokens_seen": 2439905280
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026351053159478436,
      "loss": 1.4475,
      "theoretical_loss": 3.4035411900648653,
      "tokens_seen": 2439970816
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002634904714142428,
      "loss": 1.302,
      "theoretical_loss": 3.4035339239397686,
      "tokens_seen": 2440036352
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026347041123370114,
      "loss": 1.4058,
      "theoretical_loss": 3.4035266580644694,
      "tokens_seen": 2440101888
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002634503510531595,
      "loss": 1.4143,
      "theoretical_loss": 3.403519392438953,
      "tokens_seen": 2440167424
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026343029087261787,
      "loss": 1.4671,
      "theoretical_loss": 3.403512127063204,
      "tokens_seen": 2440232960
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002634102306920763,
      "loss": 1.4262,
      "theoretical_loss": 3.4035048619372072,
      "tokens_seen": 2440298496
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026339017051153464,
      "loss": 1.3052,
      "theoretical_loss": 3.403497597060947,
      "tokens_seen": 2440364032
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000263370110330993,
      "loss": 1.4256,
      "theoretical_loss": 3.403490332434409,
      "tokens_seen": 2440429568
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026335005015045137,
      "loss": 1.3991,
      "theoretical_loss": 3.403483068057577,
      "tokens_seen": 2440495104
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026332998996990973,
      "loss": 1.3722,
      "theoretical_loss": 3.403475803930436,
      "tokens_seen": 2440560640
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026330992978936815,
      "loss": 1.321,
      "theoretical_loss": 3.4034685400529705,
      "tokens_seen": 2440626176
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002632898696088265,
      "loss": 1.4218,
      "theoretical_loss": 3.4034612764251655,
      "tokens_seen": 2440691712
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026326980942828487,
      "loss": 1.4571,
      "theoretical_loss": 3.4034540130470057,
      "tokens_seen": 2440757248
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026324974924774323,
      "loss": 1.2847,
      "theoretical_loss": 3.4034467499184755,
      "tokens_seen": 2440822784
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026322968906720165,
      "loss": 1.3929,
      "theoretical_loss": 3.4034394870395603,
      "tokens_seen": 2440888320
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026320962888666,
      "loss": 1.3662,
      "theoretical_loss": 3.4034322244102437,
      "tokens_seen": 2440953856
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002631895687061184,
      "loss": 1.3079,
      "theoretical_loss": 3.403424962030512,
      "tokens_seen": 2441019392
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026316950852557674,
      "loss": 1.4189,
      "theoretical_loss": 3.4034176999003485,
      "tokens_seen": 2441084928
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002631494483450351,
      "loss": 1.305,
      "theoretical_loss": 3.4034104380197383,
      "tokens_seen": 2441150464
    },
    {
      "epoch": 0.74,
      "objective/train/advantage_avg": 0.49078458547592163,
      "objective/train/docs_used": 888977,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1415976285934448,
      "objective/train/original_loss": 1.1415977478027344,
      "objective/train/theoretical_loss": 3.4034031763886667,
      "objective/train/tokens_used": 2461676000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24142290651798248,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050305724143982,
      "objective/train/weighted_lm_loss": 1.1990419626235962,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4034031763886667,
      "tokens_seen": 2441216000
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002631293881644935,
      "loss": 1.2664,
      "theoretical_loss": 3.4034031763886667,
      "tokens_seen": 2441216000
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002631093279839519,
      "loss": 1.354,
      "theoretical_loss": 3.403395915007118,
      "tokens_seen": 2441281536
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026308926780341024,
      "loss": 1.4432,
      "theoretical_loss": 3.4033886538750764,
      "tokens_seen": 2441347072
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002630692076228686,
      "loss": 1.3918,
      "theoretical_loss": 3.4033813929925274,
      "tokens_seen": 2441412608
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000263049147442327,
      "loss": 1.311,
      "theoretical_loss": 3.4033741323594557,
      "tokens_seen": 2441478144
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002630290872617854,
      "loss": 1.4175,
      "theoretical_loss": 3.4033668719758454,
      "tokens_seen": 2441543680
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026300902708124374,
      "loss": 1.4062,
      "theoretical_loss": 3.4033596118416822,
      "tokens_seen": 2441609216
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002629889669007021,
      "loss": 1.3603,
      "theoretical_loss": 3.4033523519569497,
      "tokens_seen": 2441674752
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026296890672016047,
      "loss": 1.2512,
      "theoretical_loss": 3.4033450923216337,
      "tokens_seen": 2441740288
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002629488465396189,
      "loss": 1.3234,
      "theoretical_loss": 3.4033378329357182,
      "tokens_seen": 2441805824
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026292878635907725,
      "loss": 1.2949,
      "theoretical_loss": 3.4033305737991886,
      "tokens_seen": 2441871360
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002629087261785356,
      "loss": 1.3915,
      "theoretical_loss": 3.4033233149120288,
      "tokens_seen": 2441936896
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026288866599799397,
      "loss": 1.44,
      "theoretical_loss": 3.403316056274224,
      "tokens_seen": 2442002432
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002628686058174524,
      "loss": 1.4693,
      "theoretical_loss": 3.4033087978857592,
      "tokens_seen": 2442067968
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026284854563691075,
      "loss": 1.3318,
      "theoretical_loss": 3.403301539746619,
      "tokens_seen": 2442133504
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002628284854563691,
      "loss": 1.387,
      "theoretical_loss": 3.4032942818567875,
      "tokens_seen": 2442199040
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002628084252758275,
      "loss": 1.3016,
      "theoretical_loss": 3.40328702421625,
      "tokens_seen": 2442264576
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002627883650952859,
      "loss": 1.4415,
      "theoretical_loss": 3.4032797668249914,
      "tokens_seen": 2442330112
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026276830491474425,
      "loss": 1.3942,
      "theoretical_loss": 3.4032725096829965,
      "tokens_seen": 2442395648
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002627482447342026,
      "loss": 1.433,
      "theoretical_loss": 3.4032652527902494,
      "tokens_seen": 2442461184
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000262728184553661,
      "loss": 1.4054,
      "theoretical_loss": 3.4032579961467357,
      "tokens_seen": 2442526720
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026270812437311934,
      "loss": 1.3551,
      "theoretical_loss": 3.403250739752439,
      "tokens_seen": 2442592256
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026268806419257776,
      "loss": 1.3574,
      "theoretical_loss": 3.4032434836073455,
      "tokens_seen": 2442657792
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002626680040120361,
      "loss": 1.4374,
      "theoretical_loss": 3.403236227711439,
      "tokens_seen": 2442723328
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002626479438314945,
      "loss": 1.3756,
      "theoretical_loss": 3.4032289720647046,
      "tokens_seen": 2442788864
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026262788365095284,
      "loss": 1.4949,
      "theoretical_loss": 3.403221716667127,
      "tokens_seen": 2442854400
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026260782347041126,
      "loss": 1.3872,
      "theoretical_loss": 3.40321446151869,
      "tokens_seen": 2442919936
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002625877632898696,
      "loss": 1.3521,
      "theoretical_loss": 3.4032072066193804,
      "tokens_seen": 2442985472
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000262567703109328,
      "loss": 1.4214,
      "theoretical_loss": 3.403199951969181,
      "tokens_seen": 2443051008
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026254764292878635,
      "loss": 1.4585,
      "theoretical_loss": 3.4031926975680777,
      "tokens_seen": 2443116544
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002625275827482447,
      "loss": 1.3832,
      "theoretical_loss": 3.4031854434160556,
      "tokens_seen": 2443182080
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002625075225677031,
      "loss": 1.4164,
      "theoretical_loss": 3.403178189513098,
      "tokens_seen": 2443247616
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002624874623871615,
      "loss": 1.4486,
      "theoretical_loss": 3.4031709358591904,
      "tokens_seen": 2443313152
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026246740220661985,
      "loss": 1.331,
      "theoretical_loss": 3.403163682454318,
      "tokens_seen": 2443378688
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002624473420260782,
      "loss": 1.4572,
      "theoretical_loss": 3.4031564292984653,
      "tokens_seen": 2443444224
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026242728184553663,
      "loss": 1.3559,
      "theoretical_loss": 3.403149176391617,
      "tokens_seen": 2443509760
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000262407221664995,
      "loss": 1.4034,
      "theoretical_loss": 3.4031419237337577,
      "tokens_seen": 2443575296
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026238716148445335,
      "loss": 1.3409,
      "theoretical_loss": 3.4031346713248727,
      "tokens_seen": 2443640832
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002623671013039117,
      "loss": 1.4204,
      "theoretical_loss": 3.403127419164946,
      "tokens_seen": 2443706368
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002623470411233701,
      "loss": 1.3904,
      "theoretical_loss": 3.403120167253963,
      "tokens_seen": 2443771904
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002623269809428285,
      "loss": 1.4452,
      "theoretical_loss": 3.403112915591908,
      "tokens_seen": 2443837440
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026230692076228686,
      "loss": 1.4095,
      "theoretical_loss": 3.4031056641787663,
      "tokens_seen": 2443902976
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002622868605817452,
      "loss": 1.3928,
      "theoretical_loss": 3.4030984130145225,
      "tokens_seen": 2443968512
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002622668004012036,
      "loss": 1.4099,
      "theoretical_loss": 3.4030911620991615,
      "tokens_seen": 2444034048
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000262246740220662,
      "loss": 1.5027,
      "theoretical_loss": 3.4030839114326676,
      "tokens_seen": 2444099584
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026222668004012036,
      "loss": 1.398,
      "theoretical_loss": 3.4030766610150263,
      "tokens_seen": 2444165120
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002622066198595787,
      "loss": 1.3897,
      "theoretical_loss": 3.4030694108462214,
      "tokens_seen": 2444230656
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002621865596790371,
      "loss": 1.4365,
      "theoretical_loss": 3.4030621609262384,
      "tokens_seen": 2444296192
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026216649949849545,
      "loss": 1.4383,
      "theoretical_loss": 3.4030549112550625,
      "tokens_seen": 2444361728
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026214643931795386,
      "loss": 1.3169,
      "theoretical_loss": 3.4030476618326775,
      "tokens_seen": 2444427264
    },
    {
      "epoch": 0.74,
      "objective/train/advantage_avg": 0.4923287630081177,
      "objective/train/docs_used": 890203,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4758617877960205,
      "objective/train/original_loss": 1.4758615493774414,
      "objective/train/theoretical_loss": 3.4030404126590685,
      "objective/train/tokens_used": 2464952800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24277667701244354,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504668951034546,
      "objective/train/weighted_lm_loss": 1.550306797027588,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4030404126590685,
      "tokens_seen": 2444492800
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002621263791374122,
      "loss": 1.497,
      "theoretical_loss": 3.4030404126590685,
      "tokens_seen": 2444492800
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002621063189568706,
      "loss": 1.4096,
      "theoretical_loss": 3.4030331637342206,
      "tokens_seen": 2444558336
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026208625877632895,
      "loss": 1.4855,
      "theoretical_loss": 3.4030259150581186,
      "tokens_seen": 2444623872
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026206619859578737,
      "loss": 1.4273,
      "theoretical_loss": 3.403018666630747,
      "tokens_seen": 2444689408
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026204613841524573,
      "loss": 1.3665,
      "theoretical_loss": 3.403011418452091,
      "tokens_seen": 2444754944
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002620260782347041,
      "loss": 1.3691,
      "theoretical_loss": 3.4030041705221343,
      "tokens_seen": 2444820480
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002620060180541625,
      "loss": 1.3168,
      "theoretical_loss": 3.402996922840863,
      "tokens_seen": 2444886016
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026198595787362087,
      "loss": 1.534,
      "theoretical_loss": 3.4029896754082616,
      "tokens_seen": 2444951552
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002619658976930793,
      "loss": 1.4262,
      "theoretical_loss": 3.402982428224315,
      "tokens_seen": 2445017088
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026194583751253765,
      "loss": 1.4527,
      "theoretical_loss": 3.4029751812890074,
      "tokens_seen": 2445082624
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000261925777331996,
      "loss": 1.447,
      "theoretical_loss": 3.402967934602324,
      "tokens_seen": 2445148160
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026190571715145437,
      "loss": 1.4015,
      "theoretical_loss": 3.4029606881642493,
      "tokens_seen": 2445213696
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002618856569709128,
      "loss": 1.3862,
      "theoretical_loss": 3.4029534419747685,
      "tokens_seen": 2445279232
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026186559679037115,
      "loss": 1.3106,
      "theoretical_loss": 3.4029461960338665,
      "tokens_seen": 2445344768
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002618455366098295,
      "loss": 1.4301,
      "theoretical_loss": 3.4029389503415275,
      "tokens_seen": 2445410304
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002618254764292879,
      "loss": 1.4275,
      "theoretical_loss": 3.4029317048977368,
      "tokens_seen": 2445475840
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002618054162487463,
      "loss": 1.4415,
      "theoretical_loss": 3.4029244597024793,
      "tokens_seen": 2445541376
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026178535606820465,
      "loss": 1.4841,
      "theoretical_loss": 3.4029172147557394,
      "tokens_seen": 2445606912
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000261765295887663,
      "loss": 1.3046,
      "theoretical_loss": 3.402909970057502,
      "tokens_seen": 2445672448
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002617452357071214,
      "loss": 1.2988,
      "theoretical_loss": 3.4029027256077526,
      "tokens_seen": 2445737984
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026172517552657974,
      "loss": 1.3594,
      "theoretical_loss": 3.402895481406475,
      "tokens_seen": 2445803520
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026170511534603816,
      "loss": 1.3393,
      "theoretical_loss": 3.4028882374536544,
      "tokens_seen": 2445869056
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002616850551654965,
      "loss": 1.3447,
      "theoretical_loss": 3.4028809937492763,
      "tokens_seen": 2445934592
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002616649949849549,
      "loss": 1.4469,
      "theoretical_loss": 3.4028737502933244,
      "tokens_seen": 2446000128
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026164493480441324,
      "loss": 1.2713,
      "theoretical_loss": 3.4028665070857844,
      "tokens_seen": 2446065664
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026162487462387166,
      "loss": 1.3913,
      "theoretical_loss": 3.4028592641266404,
      "tokens_seen": 2446131200
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026160481444333,
      "loss": 1.5243,
      "theoretical_loss": 3.402852021415878,
      "tokens_seen": 2446196736
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002615847542627884,
      "loss": 1.2964,
      "theoretical_loss": 3.4028447789534813,
      "tokens_seen": 2446262272
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026156469408224675,
      "loss": 1.3468,
      "theoretical_loss": 3.402837536739436,
      "tokens_seen": 2446327808
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002615446339017051,
      "loss": 1.4048,
      "theoretical_loss": 3.4028302947737257,
      "tokens_seen": 2446393344
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002615245737211635,
      "loss": 1.3515,
      "theoretical_loss": 3.4028230530563364,
      "tokens_seen": 2446458880
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002615045135406219,
      "loss": 1.3604,
      "theoretical_loss": 3.402815811587252,
      "tokens_seen": 2446524416
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026148445336008025,
      "loss": 1.4386,
      "theoretical_loss": 3.4028085703664583,
      "tokens_seen": 2446589952
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002614643931795386,
      "loss": 1.4079,
      "theoretical_loss": 3.4028013293939394,
      "tokens_seen": 2446655488
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026144433299899703,
      "loss": 1.4282,
      "theoretical_loss": 3.4027940886696806,
      "tokens_seen": 2446721024
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002614242728184554,
      "loss": 1.3656,
      "theoretical_loss": 3.4027868481936663,
      "tokens_seen": 2446786560
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026140421263791375,
      "loss": 1.3597,
      "theoretical_loss": 3.402779607965882,
      "tokens_seen": 2446852096
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002613841524573721,
      "loss": 1.4943,
      "theoretical_loss": 3.4027723679863113,
      "tokens_seen": 2446917632
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002613640922768305,
      "loss": 1.3962,
      "theoretical_loss": 3.4027651282549405,
      "tokens_seen": 2446983168
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002613440320962889,
      "loss": 1.3132,
      "theoretical_loss": 3.4027578887717533,
      "tokens_seen": 2447048704
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026132397191574726,
      "loss": 1.4707,
      "theoretical_loss": 3.4027506495367357,
      "tokens_seen": 2447114240
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002613039117352056,
      "loss": 1.4143,
      "theoretical_loss": 3.402743410549871,
      "tokens_seen": 2447179776
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000261283851554664,
      "loss": 1.3832,
      "theoretical_loss": 3.4027361718111457,
      "tokens_seen": 2447245312
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002612637913741224,
      "loss": 1.3975,
      "theoretical_loss": 3.4027289333205433,
      "tokens_seen": 2447310848
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026124373119358076,
      "loss": 1.3276,
      "theoretical_loss": 3.4027216950780494,
      "tokens_seen": 2447376384
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002612236710130391,
      "loss": 1.4255,
      "theoretical_loss": 3.4027144570836487,
      "tokens_seen": 2447441920
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002612036108324975,
      "loss": 1.3319,
      "theoretical_loss": 3.4027072193373264,
      "tokens_seen": 2447507456
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026118355065195585,
      "loss": 1.2875,
      "theoretical_loss": 3.4026999818390666,
      "tokens_seen": 2447572992
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026116349047141426,
      "loss": 1.3453,
      "theoretical_loss": 3.4026927445888546,
      "tokens_seen": 2447638528
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002611434302908726,
      "loss": 1.3121,
      "theoretical_loss": 3.4026855075866753,
      "tokens_seen": 2447704064
    },
    {
      "epoch": 0.74,
      "objective/train/advantage_avg": 0.4945920407772064,
      "objective/train/docs_used": 890833,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1802425384521484,
      "objective/train/original_loss": 1.1802425384521484,
      "objective/train/theoretical_loss": 3.402678270832513,
      "objective/train/tokens_used": 2468229600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24488870799541473,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507042407989502,
      "objective/train/weighted_lm_loss": 1.2400504350662231,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.402678270832513,
      "tokens_seen": 2447769600
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000261123370110331,
      "loss": 1.2503,
      "theoretical_loss": 3.402678270832513,
      "tokens_seen": 2447769600
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026110330992978935,
      "loss": 1.1615,
      "theoretical_loss": 3.402671034326354,
      "tokens_seen": 2447835136
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026108324974924777,
      "loss": 1.3101,
      "theoretical_loss": 3.4026637980681813,
      "tokens_seen": 2447900672
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026106318956870613,
      "loss": 1.3411,
      "theoretical_loss": 3.402656562057981,
      "tokens_seen": 2447966208
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002610431293881645,
      "loss": 1.3467,
      "theoretical_loss": 3.4026493262957374,
      "tokens_seen": 2448031744
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026102306920762285,
      "loss": 1.4195,
      "theoretical_loss": 3.402642090781436,
      "tokens_seen": 2448097280
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002610030090270812,
      "loss": 1.3753,
      "theoretical_loss": 3.4026348555150605,
      "tokens_seen": 2448162816
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026098294884653963,
      "loss": 1.3954,
      "theoretical_loss": 3.402627620496597,
      "tokens_seen": 2448228352
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000260962888665998,
      "loss": 1.348,
      "theoretical_loss": 3.40262038572603,
      "tokens_seen": 2448293888
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026094282848545636,
      "loss": 1.4445,
      "theoretical_loss": 3.4026131512033437,
      "tokens_seen": 2448359424
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002609227683049147,
      "loss": 1.3441,
      "theoretical_loss": 3.402605916928524,
      "tokens_seen": 2448424960
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026090270812437313,
      "loss": 1.2841,
      "theoretical_loss": 3.402598682901555,
      "tokens_seen": 2448490496
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002608826479438315,
      "loss": 1.3954,
      "theoretical_loss": 3.4025914491224225,
      "tokens_seen": 2448556032
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026086258776328986,
      "loss": 1.4123,
      "theoretical_loss": 3.4025842155911104,
      "tokens_seen": 2448621568
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002608425275827482,
      "loss": 1.3616,
      "theoretical_loss": 3.4025769823076035,
      "tokens_seen": 2448687104
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002608224674022066,
      "loss": 1.3398,
      "theoretical_loss": 3.4025697492718874,
      "tokens_seen": 2448752640
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000260802407221665,
      "loss": 1.4187,
      "theoretical_loss": 3.402562516483947,
      "tokens_seen": 2448818176
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026078234704112336,
      "loss": 1.3192,
      "theoretical_loss": 3.4025552839437663,
      "tokens_seen": 2448883712
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002607622868605817,
      "loss": 1.4515,
      "theoretical_loss": 3.402548051651331,
      "tokens_seen": 2448949248
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002607422266800401,
      "loss": 1.5012,
      "theoretical_loss": 3.402540819606626,
      "tokens_seen": 2449014784
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002607221664994985,
      "loss": 1.5629,
      "theoretical_loss": 3.4025335878096357,
      "tokens_seen": 2449080320
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026070210631895687,
      "loss": 1.2825,
      "theoretical_loss": 3.4025263562603447,
      "tokens_seen": 2449145856
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026068204613841523,
      "loss": 1.3558,
      "theoretical_loss": 3.402519124958739,
      "tokens_seen": 2449211392
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002606619859578736,
      "loss": 1.321,
      "theoretical_loss": 3.4025118939048027,
      "tokens_seen": 2449276928
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026064192577733195,
      "loss": 1.4108,
      "theoretical_loss": 3.402504663098521,
      "tokens_seen": 2449342464
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026062186559679037,
      "loss": 1.3083,
      "theoretical_loss": 3.402497432539879,
      "tokens_seen": 2449408000
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026060180541624873,
      "loss": 1.2982,
      "theoretical_loss": 3.4024902022288606,
      "tokens_seen": 2449473536
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002605817452357071,
      "loss": 1.3373,
      "theoretical_loss": 3.4024829721654513,
      "tokens_seen": 2449539072
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026056168505516546,
      "loss": 1.2915,
      "theoretical_loss": 3.4024757423496363,
      "tokens_seen": 2449604608
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026054162487462387,
      "loss": 1.3639,
      "theoretical_loss": 3.4024685127814003,
      "tokens_seen": 2449670144
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026052156469408223,
      "loss": 1.3467,
      "theoretical_loss": 3.4024612834607284,
      "tokens_seen": 2449735680
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026050150451354065,
      "loss": 1.3163,
      "theoretical_loss": 3.4024540543876047,
      "tokens_seen": 2449801216
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000260481444332999,
      "loss": 1.3511,
      "theoretical_loss": 3.402446825562015,
      "tokens_seen": 2449866752
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026046138415245743,
      "loss": 1.2748,
      "theoretical_loss": 3.4024395969839434,
      "tokens_seen": 2449932288
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002604413239719158,
      "loss": 1.1739,
      "theoretical_loss": 3.402432368653376,
      "tokens_seen": 2449997824
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026042126379137415,
      "loss": 1.3325,
      "theoretical_loss": 3.4024251405702963,
      "tokens_seen": 2450063360
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002604012036108325,
      "loss": 1.3481,
      "theoretical_loss": 3.4024179127346903,
      "tokens_seen": 2450128896
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002603811434302909,
      "loss": 1.5204,
      "theoretical_loss": 3.4024106851465423,
      "tokens_seen": 2450194432
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002603610832497493,
      "loss": 1.3328,
      "theoretical_loss": 3.4024034578058373,
      "tokens_seen": 2450259968
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026034102306920766,
      "loss": 1.3119,
      "theoretical_loss": 3.4023962307125606,
      "tokens_seen": 2450325504
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000260320962888666,
      "loss": 1.1592,
      "theoretical_loss": 3.402389003866696,
      "tokens_seen": 2450391040
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002603009027081244,
      "loss": 1.4716,
      "theoretical_loss": 3.40238177726823,
      "tokens_seen": 2450456576
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002602808425275828,
      "loss": 1.4143,
      "theoretical_loss": 3.4023745509171466,
      "tokens_seen": 2450522112
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026026078234704116,
      "loss": 1.3377,
      "theoretical_loss": 3.4023673248134307,
      "tokens_seen": 2450587648
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002602407221664995,
      "loss": 1.3415,
      "theoretical_loss": 3.4023600989570673,
      "tokens_seen": 2450653184
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002602206619859579,
      "loss": 1.2798,
      "theoretical_loss": 3.4023528733480415,
      "tokens_seen": 2450718720
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026020060180541625,
      "loss": 1.3242,
      "theoretical_loss": 3.4023456479863383,
      "tokens_seen": 2450784256
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026018054162487466,
      "loss": 1.3711,
      "theoretical_loss": 3.402338422871942,
      "tokens_seen": 2450849792
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000260160481444333,
      "loss": 1.2613,
      "theoretical_loss": 3.4023311980048385,
      "tokens_seen": 2450915328
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002601404212637914,
      "loss": 1.3357,
      "theoretical_loss": 3.4023239733850117,
      "tokens_seen": 2450980864
    },
    {
      "epoch": 0.74,
      "objective/train/advantage_avg": 0.49414753913879395,
      "objective/train/docs_used": 891914,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.40836763381958,
      "objective/train/original_loss": 1.40836763381958,
      "objective/train/theoretical_loss": 3.402316749012447,
      "objective/train/tokens_used": 2471506400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24440991878509521,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506573915481567,
      "objective/train/weighted_lm_loss": 1.479498028755188,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.402316749012447,
      "tokens_seen": 2451046400
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026012036108324975,
      "loss": 1.4,
      "theoretical_loss": 3.402316749012447,
      "tokens_seen": 2451046400
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026010030090270817,
      "loss": 1.2378,
      "theoretical_loss": 3.40230952488713,
      "tokens_seen": 2451111936
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026008024072216653,
      "loss": 1.4209,
      "theoretical_loss": 3.402302301009044,
      "tokens_seen": 2451177472
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002600601805416249,
      "loss": 1.2653,
      "theoretical_loss": 3.4022950773781755,
      "tokens_seen": 2451243008
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026004012036108325,
      "loss": 1.2351,
      "theoretical_loss": 3.4022878539945087,
      "tokens_seen": 2451308544
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002600200601805416,
      "loss": 1.2914,
      "theoretical_loss": 3.4022806308580282,
      "tokens_seen": 2451374080
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00026000000000000003,
      "loss": 1.5026,
      "theoretical_loss": 3.40227340796872,
      "tokens_seen": 2451439616
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002599799398194584,
      "loss": 1.3784,
      "theoretical_loss": 3.4022661853265683,
      "tokens_seen": 2451505152
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025995987963891676,
      "loss": 1.3147,
      "theoretical_loss": 3.402258962931558,
      "tokens_seen": 2451570688
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002599398194583751,
      "loss": 1.3721,
      "theoretical_loss": 3.402251740783674,
      "tokens_seen": 2451636224
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025991975927783353,
      "loss": 1.3973,
      "theoretical_loss": 3.4022445188829016,
      "tokens_seen": 2451701760
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002598996990972919,
      "loss": 1.2354,
      "theoretical_loss": 3.4022372972292256,
      "tokens_seen": 2451767296
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025987963891675026,
      "loss": 1.3274,
      "theoretical_loss": 3.4022300758226307,
      "tokens_seen": 2451832832
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002598595787362086,
      "loss": 1.3718,
      "theoretical_loss": 3.402222854663102,
      "tokens_seen": 2451898368
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000259839518555667,
      "loss": 1.4404,
      "theoretical_loss": 3.402215633750625,
      "tokens_seen": 2451963904
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002598194583751254,
      "loss": 1.3104,
      "theoretical_loss": 3.4022084130851837,
      "tokens_seen": 2452029440
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025979939819458376,
      "loss": 1.312,
      "theoretical_loss": 3.4022011926667637,
      "tokens_seen": 2452094976
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002597793380140421,
      "loss": 1.3051,
      "theoretical_loss": 3.4021939724953496,
      "tokens_seen": 2452160512
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002597592778335005,
      "loss": 1.3202,
      "theoretical_loss": 3.4021867525709264,
      "tokens_seen": 2452226048
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002597392176529589,
      "loss": 1.2948,
      "theoretical_loss": 3.4021795328934794,
      "tokens_seen": 2452291584
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025971915747241727,
      "loss": 1.3602,
      "theoretical_loss": 3.402172313462993,
      "tokens_seen": 2452357120
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025969909729187563,
      "loss": 1.4173,
      "theoretical_loss": 3.4021650942794524,
      "tokens_seen": 2452422656
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000259679037111334,
      "loss": 1.35,
      "theoretical_loss": 3.4021578753428425,
      "tokens_seen": 2452488192
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025965897693079235,
      "loss": 1.2788,
      "theoretical_loss": 3.402150656653149,
      "tokens_seen": 2452553728
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025963891675025077,
      "loss": 1.3237,
      "theoretical_loss": 3.4021434382103557,
      "tokens_seen": 2452619264
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025961885656970913,
      "loss": 1.3388,
      "theoretical_loss": 3.402136220014448,
      "tokens_seen": 2452684800
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002595987963891675,
      "loss": 1.332,
      "theoretical_loss": 3.402129002065411,
      "tokens_seen": 2452750336
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025957873620862586,
      "loss": 1.3561,
      "theoretical_loss": 3.4021217843632297,
      "tokens_seen": 2452815872
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025955867602808427,
      "loss": 1.2965,
      "theoretical_loss": 3.4021145669078887,
      "tokens_seen": 2452881408
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025953861584754263,
      "loss": 1.3925,
      "theoretical_loss": 3.4021073496993735,
      "tokens_seen": 2452946944
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000259518555667001,
      "loss": 1.2059,
      "theoretical_loss": 3.4021001327376688,
      "tokens_seen": 2453012480
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025949849548645936,
      "loss": 1.2857,
      "theoretical_loss": 3.402092916022759,
      "tokens_seen": 2453078016
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002594784353059178,
      "loss": 1.5237,
      "theoretical_loss": 3.40208569955463,
      "tokens_seen": 2453143552
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025945837512537614,
      "loss": 1.266,
      "theoretical_loss": 3.4020784833332662,
      "tokens_seen": 2453209088
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002594383149448345,
      "loss": 1.4285,
      "theoretical_loss": 3.402071267358653,
      "tokens_seen": 2453274624
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025941825476429286,
      "loss": 1.3796,
      "theoretical_loss": 3.402064051630775,
      "tokens_seen": 2453340160
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002593981945837512,
      "loss": 1.3176,
      "theoretical_loss": 3.402056836149617,
      "tokens_seen": 2453405696
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025937813440320964,
      "loss": 1.4305,
      "theoretical_loss": 3.4020496209151645,
      "tokens_seen": 2453471232
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000259358074222668,
      "loss": 1.2628,
      "theoretical_loss": 3.402042405927402,
      "tokens_seen": 2453536768
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025933801404212636,
      "loss": 1.2964,
      "theoretical_loss": 3.4020351911863154,
      "tokens_seen": 2453602304
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002593179538615847,
      "loss": 1.2706,
      "theoretical_loss": 3.4020279766918886,
      "tokens_seen": 2453667840
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025929789368104314,
      "loss": 1.3526,
      "theoretical_loss": 3.4020207624441063,
      "tokens_seen": 2453733376
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002592778335005015,
      "loss": 1.3797,
      "theoretical_loss": 3.4020135484429552,
      "tokens_seen": 2453798912
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025925777331995987,
      "loss": 1.3908,
      "theoretical_loss": 3.4020063346884184,
      "tokens_seen": 2453864448
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025923771313941823,
      "loss": 1.3486,
      "theoretical_loss": 3.401999121180482,
      "tokens_seen": 2453929984
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002592176529588766,
      "loss": 1.2081,
      "theoretical_loss": 3.401991907919131,
      "tokens_seen": 2453995520
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000259197592778335,
      "loss": 1.1292,
      "theoretical_loss": 3.4019846949043497,
      "tokens_seen": 2454061056
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025917753259779337,
      "loss": 1.2867,
      "theoretical_loss": 3.401977482136124,
      "tokens_seen": 2454126592
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025915747241725173,
      "loss": 1.3705,
      "theoretical_loss": 3.401970269614438,
      "tokens_seen": 2454192128
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002591374122367101,
      "loss": 1.4184,
      "theoretical_loss": 3.4019630573392767,
      "tokens_seen": 2454257664
    },
    {
      "epoch": 0.74,
      "objective/train/advantage_avg": 0.4939999580383301,
      "objective/train/docs_used": 893100,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3085309267044067,
      "objective/train/original_loss": 1.3085310459136963,
      "objective/train/theoretical_loss": 3.401955845310626,
      "objective/train/tokens_used": 2474783200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24433110654354095,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506421327590942,
      "objective/train/weighted_lm_loss": 1.3746453523635864,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0168063640594482,
      "theoretical_loss": 3.401955845310626,
      "tokens_seen": 2454323200
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002591173520561685,
      "loss": 1.3035,
      "theoretical_loss": 3.401955845310626,
      "tokens_seen": 2454323200
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002590972918756269,
      "loss": 1.3076,
      "theoretical_loss": 3.40194863352847,
      "tokens_seen": 2454388736
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025907723169508524,
      "loss": 1.3329,
      "theoretical_loss": 3.4019414219927944,
      "tokens_seen": 2454454272
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002590571715145436,
      "loss": 1.3796,
      "theoretical_loss": 3.4019342107035833,
      "tokens_seen": 2454519808
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025903711133400196,
      "loss": 1.2165,
      "theoretical_loss": 3.4019269996608226,
      "tokens_seen": 2454585344
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002590170511534604,
      "loss": 1.2821,
      "theoretical_loss": 3.4019197888644968,
      "tokens_seen": 2454650880
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002589969909729188,
      "loss": 1.2855,
      "theoretical_loss": 3.4019125783145907,
      "tokens_seen": 2454716416
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025897693079237716,
      "loss": 1.1614,
      "theoretical_loss": 3.40190536801109,
      "tokens_seen": 2454781952
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002589568706118355,
      "loss": 1.3843,
      "theoretical_loss": 3.4018981579539793,
      "tokens_seen": 2454847488
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025893681043129393,
      "loss": 1.3264,
      "theoretical_loss": 3.4018909481432433,
      "tokens_seen": 2454913024
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002589167502507523,
      "loss": 1.337,
      "theoretical_loss": 3.4018837385788676,
      "tokens_seen": 2454978560
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025889669007021066,
      "loss": 1.3849,
      "theoretical_loss": 3.401876529260837,
      "tokens_seen": 2455044096
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000258876629889669,
      "loss": 1.2025,
      "theoretical_loss": 3.401869320189136,
      "tokens_seen": 2455109632
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002588565697091274,
      "loss": 1.4661,
      "theoretical_loss": 3.4018621113637506,
      "tokens_seen": 2455175168
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002588365095285858,
      "loss": 1.455,
      "theoretical_loss": 3.401854902784665,
      "tokens_seen": 2455240704
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025881644934804416,
      "loss": 1.4124,
      "theoretical_loss": 3.4018476944518645,
      "tokens_seen": 2455306240
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002587963891675025,
      "loss": 1.4596,
      "theoretical_loss": 3.401840486365334,
      "tokens_seen": 2455371776
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002587763289869609,
      "loss": 1.3175,
      "theoretical_loss": 3.401833278525059,
      "tokens_seen": 2455437312
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002587562688064193,
      "loss": 1.2167,
      "theoretical_loss": 3.4018260709310235,
      "tokens_seen": 2455502848
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025873620862587767,
      "loss": 1.284,
      "theoretical_loss": 3.4018188635832134,
      "tokens_seen": 2455568384
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025871614844533603,
      "loss": 1.2654,
      "theoretical_loss": 3.4018116564816134,
      "tokens_seen": 2455633920
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002586960882647944,
      "loss": 1.4107,
      "theoretical_loss": 3.4018044496262085,
      "tokens_seen": 2455699456
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025867602808425275,
      "loss": 1.2912,
      "theoretical_loss": 3.4017972430169836,
      "tokens_seen": 2455764992
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025865596790371117,
      "loss": 1.2562,
      "theoretical_loss": 3.4017900366539244,
      "tokens_seen": 2455830528
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025863590772316953,
      "loss": 1.4499,
      "theoretical_loss": 3.401782830537015,
      "tokens_seen": 2455896064
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002586158475426279,
      "loss": 1.3248,
      "theoretical_loss": 3.401775624666241,
      "tokens_seen": 2455961600
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025859578736208626,
      "loss": 1.4181,
      "theoretical_loss": 3.4017684190415873,
      "tokens_seen": 2456027136
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025857572718154467,
      "loss": 1.4035,
      "theoretical_loss": 3.401761213663039,
      "tokens_seen": 2456092672
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025855566700100303,
      "loss": 1.3495,
      "theoretical_loss": 3.4017540085305806,
      "tokens_seen": 2456158208
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002585356068204614,
      "loss": 1.3418,
      "theoretical_loss": 3.401746803644198,
      "tokens_seen": 2456223744
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025851554663991976,
      "loss": 1.2319,
      "theoretical_loss": 3.4017395990038755,
      "tokens_seen": 2456289280
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002584954864593782,
      "loss": 1.4569,
      "theoretical_loss": 3.4017323946095988,
      "tokens_seen": 2456354816
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025847542627883654,
      "loss": 1.2668,
      "theoretical_loss": 3.401725190461352,
      "tokens_seen": 2456420352
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002584553660982949,
      "loss": 1.5,
      "theoretical_loss": 3.401717986559121,
      "tokens_seen": 2456485888
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025843530591775326,
      "loss": 1.2077,
      "theoretical_loss": 3.4017107829028905,
      "tokens_seen": 2456551424
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002584152457372116,
      "loss": 1.2503,
      "theoretical_loss": 3.4017035794926453,
      "tokens_seen": 2456616960
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025839518555667004,
      "loss": 1.3735,
      "theoretical_loss": 3.4016963763283714,
      "tokens_seen": 2456682496
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002583751253761284,
      "loss": 1.2589,
      "theoretical_loss": 3.4016891734100523,
      "tokens_seen": 2456748032
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025835506519558677,
      "loss": 1.345,
      "theoretical_loss": 3.401681970737674,
      "tokens_seen": 2456813568
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025833500501504513,
      "loss": 1.3832,
      "theoretical_loss": 3.401674768311222,
      "tokens_seen": 2456879104
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025831494483450354,
      "loss": 1.2825,
      "theoretical_loss": 3.4016675661306803,
      "tokens_seen": 2456944640
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002582948846539619,
      "loss": 1.422,
      "theoretical_loss": 3.401660364196035,
      "tokens_seen": 2457010176
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025827482447342027,
      "loss": 1.3122,
      "theoretical_loss": 3.4016531625072695,
      "tokens_seen": 2457075712
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025825476429287863,
      "loss": 1.3852,
      "theoretical_loss": 3.4016459610643706,
      "tokens_seen": 2457141248
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000258234704112337,
      "loss": 1.4064,
      "theoretical_loss": 3.4016387598673226,
      "tokens_seen": 2457206784
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002582146439317954,
      "loss": 1.4962,
      "theoretical_loss": 3.4016315589161104,
      "tokens_seen": 2457272320
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025819458375125377,
      "loss": 1.2795,
      "theoretical_loss": 3.4016243582107193,
      "tokens_seen": 2457337856
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025817452357071213,
      "loss": 1.34,
      "theoretical_loss": 3.4016171577511343,
      "tokens_seen": 2457403392
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002581544633901705,
      "loss": 1.2904,
      "theoretical_loss": 3.4016099575373406,
      "tokens_seen": 2457468928
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002581344032096289,
      "loss": 1.2825,
      "theoretical_loss": 3.4016027575693233,
      "tokens_seen": 2457534464
    },
    {
      "epoch": 0.74,
      "objective/train/advantage_avg": 0.49502822756767273,
      "objective/train/docs_used": 894282,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2373509407043457,
      "objective/train/original_loss": 1.2373509407043457,
      "objective/train/theoretical_loss": 3.4015955578470667,
      "objective/train/tokens_used": 2478060000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2453298270702362,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507501363754272,
      "objective/train/weighted_lm_loss": 1.3001329898834229,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4015955578470667,
      "tokens_seen": 2457600000
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002581143430290873,
      "loss": 1.31,
      "theoretical_loss": 3.4015955578470667,
      "tokens_seen": 2457600000
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025809428284854564,
      "loss": 1.3188,
      "theoretical_loss": 3.4015883583705566,
      "tokens_seen": 2457665536
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000258074222668004,
      "loss": 1.31,
      "theoretical_loss": 3.401581159139778,
      "tokens_seen": 2457731072
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025805416248746236,
      "loss": 1.3992,
      "theoretical_loss": 3.401573960154716,
      "tokens_seen": 2457796608
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002580341023069208,
      "loss": 1.3003,
      "theoretical_loss": 3.4015667614153555,
      "tokens_seen": 2457862144
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025801404212637914,
      "loss": 1.3432,
      "theoretical_loss": 3.4015595629216815,
      "tokens_seen": 2457927680
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002579939819458375,
      "loss": 1.2839,
      "theoretical_loss": 3.401552364673679,
      "tokens_seen": 2457993216
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025797392176529586,
      "loss": 1.3609,
      "theoretical_loss": 3.401545166671333,
      "tokens_seen": 2458058752
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002579538615847543,
      "loss": 1.2471,
      "theoretical_loss": 3.4015379689146292,
      "tokens_seen": 2458124288
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025793380140421264,
      "loss": 1.2933,
      "theoretical_loss": 3.401530771403552,
      "tokens_seen": 2458189824
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.000257913741223671,
      "loss": 1.2508,
      "theoretical_loss": 3.401523574138087,
      "tokens_seen": 2458255360
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025789368104312937,
      "loss": 1.5449,
      "theoretical_loss": 3.4015163771182184,
      "tokens_seen": 2458320896
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025787362086258773,
      "loss": 1.386,
      "theoretical_loss": 3.4015091803439326,
      "tokens_seen": 2458386432
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.00025785356068204615,
      "loss": 1.4345,
      "theoretical_loss": 3.4015019838152134,
      "tokens_seen": 2458451968
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002578335005015045,
      "loss": 1.3723,
      "theoretical_loss": 3.4014947875320463,
      "tokens_seen": 2458517504
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025781344032096287,
      "loss": 1.414,
      "theoretical_loss": 3.401487591494417,
      "tokens_seen": 2458583040
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025779338014042123,
      "loss": 1.3169,
      "theoretical_loss": 3.4014803957023094,
      "tokens_seen": 2458648576
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025777331995987965,
      "loss": 1.4082,
      "theoretical_loss": 3.40147320015571,
      "tokens_seen": 2458714112
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000257753259779338,
      "loss": 1.3314,
      "theoretical_loss": 3.401466004854602,
      "tokens_seen": 2458779648
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002577331995987964,
      "loss": 1.2407,
      "theoretical_loss": 3.4014588097989726,
      "tokens_seen": 2458845184
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025771313941825474,
      "loss": 1.3623,
      "theoretical_loss": 3.4014516149888054,
      "tokens_seen": 2458910720
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002576930792377131,
      "loss": 1.3868,
      "theoretical_loss": 3.4014444204240863,
      "tokens_seen": 2458976256
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002576730190571715,
      "loss": 1.3668,
      "theoretical_loss": 3.4014372261048,
      "tokens_seen": 2459041792
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002576529588766299,
      "loss": 1.1813,
      "theoretical_loss": 3.401430032030931,
      "tokens_seen": 2459107328
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025763289869608824,
      "loss": 1.2412,
      "theoretical_loss": 3.401422838202466,
      "tokens_seen": 2459172864
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002576128385155466,
      "loss": 1.2419,
      "theoretical_loss": 3.4014156446193886,
      "tokens_seen": 2459238400
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000257592778335005,
      "loss": 1.2859,
      "theoretical_loss": 3.401408451281684,
      "tokens_seen": 2459303936
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002575727181544634,
      "loss": 1.3009,
      "theoretical_loss": 3.4014012581893383,
      "tokens_seen": 2459369472
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025755265797392174,
      "loss": 1.2806,
      "theoretical_loss": 3.401394065342336,
      "tokens_seen": 2459435008
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002575325977933801,
      "loss": 1.3978,
      "theoretical_loss": 3.401386872740662,
      "tokens_seen": 2459500544
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025751253761283847,
      "loss": 1.3914,
      "theoretical_loss": 3.4013796803843013,
      "tokens_seen": 2459566080
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025749247743229694,
      "loss": 1.254,
      "theoretical_loss": 3.40137248827324,
      "tokens_seen": 2459631616
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002574724172517553,
      "loss": 1.4254,
      "theoretical_loss": 3.4013652964074614,
      "tokens_seen": 2459697152
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025745235707121366,
      "loss": 1.397,
      "theoretical_loss": 3.4013581047869526,
      "tokens_seen": 2459762688
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000257432296890672,
      "loss": 1.2478,
      "theoretical_loss": 3.4013509134116973,
      "tokens_seen": 2459828224
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025741223671013044,
      "loss": 1.36,
      "theoretical_loss": 3.4013437222816814,
      "tokens_seen": 2459893760
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002573921765295888,
      "loss": 1.4222,
      "theoretical_loss": 3.4013365313968893,
      "tokens_seen": 2459959296
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025737211634904717,
      "loss": 1.3819,
      "theoretical_loss": 3.401329340757307,
      "tokens_seen": 2460024832
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025735205616850553,
      "loss": 1.3071,
      "theoretical_loss": 3.4013221503629185,
      "tokens_seen": 2460090368
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025733199598796394,
      "loss": 1.4059,
      "theoretical_loss": 3.4013149602137096,
      "tokens_seen": 2460155904
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002573119358074223,
      "loss": 1.3259,
      "theoretical_loss": 3.4013077703096655,
      "tokens_seen": 2460221440
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025729187562688067,
      "loss": 1.4831,
      "theoretical_loss": 3.401300580650771,
      "tokens_seen": 2460286976
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025727181544633903,
      "loss": 1.258,
      "theoretical_loss": 3.4012933912370116,
      "tokens_seen": 2460352512
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002572517552657974,
      "loss": 1.3732,
      "theoretical_loss": 3.4012862020683716,
      "tokens_seen": 2460418048
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002572316950852558,
      "loss": 1.254,
      "theoretical_loss": 3.4012790131448374,
      "tokens_seen": 2460483584
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025721163490471417,
      "loss": 1.273,
      "theoretical_loss": 3.4012718244663924,
      "tokens_seen": 2460549120
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025719157472417253,
      "loss": 1.214,
      "theoretical_loss": 3.401264636033024,
      "tokens_seen": 2460614656
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002571715145436309,
      "loss": 1.2292,
      "theoretical_loss": 3.401257447844715,
      "tokens_seen": 2460680192
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002571514543630893,
      "loss": 1.3342,
      "theoretical_loss": 3.4012502599014516,
      "tokens_seen": 2460745728
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002571313941825477,
      "loss": 1.2262,
      "theoretical_loss": 3.401243072203219,
      "tokens_seen": 2460811264
    },
    {
      "epoch": 0.75,
      "objective/train/advantage_avg": 0.49362707138061523,
      "objective/train/docs_used": 895477,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.343868374824524,
      "objective/train/original_loss": 1.343868374824524,
      "objective/train/theoretical_loss": 3.401235884750002,
      "objective/train/tokens_used": 2481336800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24398308992385864,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506030321121216,
      "objective/train/weighted_lm_loss": 1.4120122194290161,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.401235884750002,
      "tokens_seen": 2460876800
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025711133400200604,
      "loss": 1.3685,
      "theoretical_loss": 3.401235884750002,
      "tokens_seen": 2460876800
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002570912738214644,
      "loss": 1.3813,
      "theoretical_loss": 3.4012286975417863,
      "tokens_seen": 2460942336
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025707121364092276,
      "loss": 1.3466,
      "theoretical_loss": 3.4012215105785564,
      "tokens_seen": 2461007872
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002570511534603812,
      "loss": 1.3152,
      "theoretical_loss": 3.4012143238602977,
      "tokens_seen": 2461073408
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025703109327983954,
      "loss": 1.308,
      "theoretical_loss": 3.401207137386995,
      "tokens_seen": 2461138944
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002570110330992979,
      "loss": 1.2721,
      "theoretical_loss": 3.4011999511586337,
      "tokens_seen": 2461204480
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025699097291875626,
      "loss": 1.3899,
      "theoretical_loss": 3.401192765175199,
      "tokens_seen": 2461270016
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002569709127382147,
      "loss": 1.3686,
      "theoretical_loss": 3.4011855794366763,
      "tokens_seen": 2461335552
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025695085255767304,
      "loss": 1.3639,
      "theoretical_loss": 3.40117839394305,
      "tokens_seen": 2461401088
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002569307923771314,
      "loss": 1.3552,
      "theoretical_loss": 3.4011712086943056,
      "tokens_seen": 2461466624
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025691073219658977,
      "loss": 1.3359,
      "theoretical_loss": 3.4011640236904284,
      "tokens_seen": 2461532160
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025689067201604813,
      "loss": 1.2785,
      "theoretical_loss": 3.4011568389314037,
      "tokens_seen": 2461597696
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025687061183550655,
      "loss": 1.3046,
      "theoretical_loss": 3.4011496544172157,
      "tokens_seen": 2461663232
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002568505516549649,
      "loss": 1.2352,
      "theoretical_loss": 3.4011424701478505,
      "tokens_seen": 2461728768
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025683049147442327,
      "loss": 1.3022,
      "theoretical_loss": 3.401135286123293,
      "tokens_seen": 2461794304
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025681043129388163,
      "loss": 1.3771,
      "theoretical_loss": 3.4011281023435282,
      "tokens_seen": 2461859840
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025679037111334005,
      "loss": 1.3887,
      "theoretical_loss": 3.401120918808541,
      "tokens_seen": 2461925376
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002567703109327984,
      "loss": 1.2789,
      "theoretical_loss": 3.401113735518317,
      "tokens_seen": 2461990912
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002567502507522568,
      "loss": 1.3101,
      "theoretical_loss": 3.401106552472841,
      "tokens_seen": 2462056448
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025673019057171514,
      "loss": 1.1907,
      "theoretical_loss": 3.401099369672099,
      "tokens_seen": 2462121984
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002567101303911735,
      "loss": 1.313,
      "theoretical_loss": 3.401092187116075,
      "tokens_seen": 2462187520
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002566900702106319,
      "loss": 1.4623,
      "theoretical_loss": 3.4010850048047545,
      "tokens_seen": 2462253056
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002566700100300903,
      "loss": 1.4135,
      "theoretical_loss": 3.4010778227381233,
      "tokens_seen": 2462318592
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025664994984954864,
      "loss": 1.3287,
      "theoretical_loss": 3.4010706409161653,
      "tokens_seen": 2462384128
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000256629889669007,
      "loss": 1.3523,
      "theoretical_loss": 3.401063459338867,
      "tokens_seen": 2462449664
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002566098294884654,
      "loss": 1.243,
      "theoretical_loss": 3.4010562780062124,
      "tokens_seen": 2462515200
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002565897693079238,
      "loss": 1.4564,
      "theoretical_loss": 3.4010490969181877,
      "tokens_seen": 2462580736
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025656970912738214,
      "loss": 1.3365,
      "theoretical_loss": 3.401041916074777,
      "tokens_seen": 2462646272
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002565496489468405,
      "loss": 1.4023,
      "theoretical_loss": 3.4010347354759665,
      "tokens_seen": 2462711808
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025652958876629887,
      "loss": 1.2891,
      "theoretical_loss": 3.4010275551217406,
      "tokens_seen": 2462777344
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002565095285857573,
      "loss": 1.4773,
      "theoretical_loss": 3.401020375012085,
      "tokens_seen": 2462842880
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025648946840521565,
      "loss": 1.518,
      "theoretical_loss": 3.401013195146984,
      "tokens_seen": 2462908416
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000256469408224674,
      "loss": 1.2208,
      "theoretical_loss": 3.401006015526424,
      "tokens_seen": 2462973952
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025644934804413237,
      "loss": 1.3844,
      "theoretical_loss": 3.4009988361503893,
      "tokens_seen": 2463039488
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002564292878635908,
      "loss": 1.3378,
      "theoretical_loss": 3.400991657018865,
      "tokens_seen": 2463105024
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025640922768304915,
      "loss": 1.2754,
      "theoretical_loss": 3.4009844781318366,
      "tokens_seen": 2463170560
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002563891675025075,
      "loss": 1.3887,
      "theoretical_loss": 3.4009772994892895,
      "tokens_seen": 2463236096
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002563691073219659,
      "loss": 1.3309,
      "theoretical_loss": 3.400970121091208,
      "tokens_seen": 2463301632
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025634904714142424,
      "loss": 1.4519,
      "theoretical_loss": 3.400962942937579,
      "tokens_seen": 2463367168
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025632898696088265,
      "loss": 1.3427,
      "theoretical_loss": 3.4009557650283857,
      "tokens_seen": 2463432704
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000256308926780341,
      "loss": 1.452,
      "theoretical_loss": 3.4009485873636143,
      "tokens_seen": 2463498240
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002562888665997994,
      "loss": 1.2855,
      "theoretical_loss": 3.4009414099432496,
      "tokens_seen": 2463563776
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025626880641925774,
      "loss": 1.2364,
      "theoretical_loss": 3.400934232767277,
      "tokens_seen": 2463629312
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025624874623871616,
      "loss": 1.3253,
      "theoretical_loss": 3.4009270558356817,
      "tokens_seen": 2463694848
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002562286860581745,
      "loss": 1.3565,
      "theoretical_loss": 3.4009198791484487,
      "tokens_seen": 2463760384
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002562086258776329,
      "loss": 1.2793,
      "theoretical_loss": 3.4009127027055635,
      "tokens_seen": 2463825920
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025618856569709124,
      "loss": 1.2404,
      "theoretical_loss": 3.400905526507011,
      "tokens_seen": 2463891456
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025616850551654966,
      "loss": 1.3549,
      "theoretical_loss": 3.4008983505527763,
      "tokens_seen": 2463956992
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000256148445336008,
      "loss": 1.3181,
      "theoretical_loss": 3.4008911748428448,
      "tokens_seen": 2464022528
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002561283851554664,
      "loss": 1.3429,
      "theoretical_loss": 3.4008839993772018,
      "tokens_seen": 2464088064
    },
    {
      "epoch": 0.75,
      "objective/train/advantage_avg": 0.4892596900463104,
      "objective/train/docs_used": 896700,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1192070245742798,
      "objective/train/original_loss": 1.1192069053649902,
      "objective/train/theoretical_loss": 3.4008768241558323,
      "objective/train/tokens_used": 2484613600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24045877158641815,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050148367881775,
      "objective/train/weighted_lm_loss": 1.1754449605941772,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4008768241558323,
      "tokens_seen": 2464153600
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025610832497492475,
      "loss": 1.326,
      "theoretical_loss": 3.4008768241558323,
      "tokens_seen": 2464153600
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002560882647943831,
      "loss": 1.3885,
      "theoretical_loss": 3.400869649178721,
      "tokens_seen": 2464219136
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002560682046138415,
      "loss": 1.4017,
      "theoretical_loss": 3.400862474445854,
      "tokens_seen": 2464284672
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002560481444332999,
      "loss": 1.2943,
      "theoretical_loss": 3.4008552999572164,
      "tokens_seen": 2464350208
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025602808425275825,
      "loss": 1.3635,
      "theoretical_loss": 3.4008481257127925,
      "tokens_seen": 2464415744
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002560080240722166,
      "loss": 1.2881,
      "theoretical_loss": 3.4008409517125684,
      "tokens_seen": 2464481280
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002559879638916751,
      "loss": 1.2883,
      "theoretical_loss": 3.400833777956529,
      "tokens_seen": 2464546816
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025596790371113344,
      "loss": 1.3582,
      "theoretical_loss": 3.400826604444659,
      "tokens_seen": 2464612352
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002559478435305918,
      "loss": 1.2922,
      "theoretical_loss": 3.400819431176944,
      "tokens_seen": 2464677888
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025592778335005017,
      "loss": 1.2583,
      "theoretical_loss": 3.40081225815337,
      "tokens_seen": 2464743424
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025590772316950853,
      "loss": 1.2349,
      "theoretical_loss": 3.4008050853739213,
      "tokens_seen": 2464808960
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025588766298896695,
      "loss": 1.189,
      "theoretical_loss": 3.400797912838583,
      "tokens_seen": 2464874496
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002558676028084253,
      "loss": 1.4325,
      "theoretical_loss": 3.4007907405473405,
      "tokens_seen": 2464940032
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025584754262788367,
      "loss": 1.337,
      "theoretical_loss": 3.400783568500179,
      "tokens_seen": 2465005568
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025582748244734203,
      "loss": 1.4259,
      "theoretical_loss": 3.4007763966970836,
      "tokens_seen": 2465071104
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025580742226680045,
      "loss": 1.3157,
      "theoretical_loss": 3.40076922513804,
      "tokens_seen": 2465136640
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002557873620862588,
      "loss": 1.4049,
      "theoretical_loss": 3.400762053823033,
      "tokens_seen": 2465202176
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002557673019057172,
      "loss": 1.2823,
      "theoretical_loss": 3.400754882752048,
      "tokens_seen": 2465267712
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025574724172517554,
      "loss": 1.3364,
      "theoretical_loss": 3.40074771192507,
      "tokens_seen": 2465333248
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002557271815446339,
      "loss": 1.2796,
      "theoretical_loss": 3.400740541342084,
      "tokens_seen": 2465398784
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002557071213640923,
      "loss": 1.4031,
      "theoretical_loss": 3.400733371003076,
      "tokens_seen": 2465464320
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002556870611835507,
      "loss": 1.3906,
      "theoretical_loss": 3.4007262009080303,
      "tokens_seen": 2465529856
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025566700100300904,
      "loss": 1.3738,
      "theoretical_loss": 3.4007190310569326,
      "tokens_seen": 2465595392
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002556469408224674,
      "loss": 1.3789,
      "theoretical_loss": 3.4007118614497682,
      "tokens_seen": 2465660928
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002556268806419258,
      "loss": 1.3523,
      "theoretical_loss": 3.4007046920865225,
      "tokens_seen": 2465726464
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002556068204613842,
      "loss": 1.3352,
      "theoretical_loss": 3.40069752296718,
      "tokens_seen": 2465792000
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025558676028084254,
      "loss": 1.3983,
      "theoretical_loss": 3.400690354091726,
      "tokens_seen": 2465857536
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002555667001003009,
      "loss": 1.151,
      "theoretical_loss": 3.4006831854601467,
      "tokens_seen": 2465923072
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025554663991975927,
      "loss": 1.3228,
      "theoretical_loss": 3.4006760170724264,
      "tokens_seen": 2465988608
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002555265797392177,
      "loss": 1.31,
      "theoretical_loss": 3.4006688489285506,
      "tokens_seen": 2466054144
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025550651955867605,
      "loss": 1.3319,
      "theoretical_loss": 3.400661681028504,
      "tokens_seen": 2466119680
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002554864593781344,
      "loss": 1.4189,
      "theoretical_loss": 3.400654513372273,
      "tokens_seen": 2466185216
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025546639919759277,
      "loss": 1.4358,
      "theoretical_loss": 3.4006473459598423,
      "tokens_seen": 2466250752
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002554463390170512,
      "loss": 1.3175,
      "theoretical_loss": 3.4006401787911966,
      "tokens_seen": 2466316288
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025542627883650955,
      "loss": 1.4242,
      "theoretical_loss": 3.4006330118663213,
      "tokens_seen": 2466381824
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002554062186559679,
      "loss": 1.2336,
      "theoretical_loss": 3.4006258451852025,
      "tokens_seen": 2466447360
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002553861584754263,
      "loss": 1.5104,
      "theoretical_loss": 3.4006186787478243,
      "tokens_seen": 2466512896
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025536609829488464,
      "loss": 1.4681,
      "theoretical_loss": 3.4006115125541725,
      "tokens_seen": 2466578432
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025534603811434305,
      "loss": 1.2553,
      "theoretical_loss": 3.4006043466042324,
      "tokens_seen": 2466643968
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002553259779338014,
      "loss": 1.2993,
      "theoretical_loss": 3.400597180897989,
      "tokens_seen": 2466709504
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002553059177532598,
      "loss": 1.4172,
      "theoretical_loss": 3.4005900154354274,
      "tokens_seen": 2466775040
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025528585757271814,
      "loss": 1.3613,
      "theoretical_loss": 3.4005828502165336,
      "tokens_seen": 2466840576
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025526579739217656,
      "loss": 1.2753,
      "theoretical_loss": 3.400575685241292,
      "tokens_seen": 2466906112
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002552457372116349,
      "loss": 1.3221,
      "theoretical_loss": 3.400568520509688,
      "tokens_seen": 2466971648
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002552256770310933,
      "loss": 1.4451,
      "theoretical_loss": 3.400561356021707,
      "tokens_seen": 2467037184
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025520561685055164,
      "loss": 1.3069,
      "theoretical_loss": 3.4005541917773345,
      "tokens_seen": 2467102720
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025518555667001006,
      "loss": 1.471,
      "theoretical_loss": 3.4005470277765557,
      "tokens_seen": 2467168256
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002551654964894684,
      "loss": 1.4229,
      "theoretical_loss": 3.400539864019355,
      "tokens_seen": 2467233792
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002551454363089268,
      "loss": 1.2809,
      "theoretical_loss": 3.400532700505719,
      "tokens_seen": 2467299328
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025512537612838515,
      "loss": 1.4737,
      "theoretical_loss": 3.4005255372356316,
      "tokens_seen": 2467364864
    },
    {
      "epoch": 0.75,
      "objective/train/advantage_avg": 0.49229684472084045,
      "objective/train/docs_used": 897881,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2669014930725098,
      "objective/train/original_loss": 1.2669013738632202,
      "objective/train/theoretical_loss": 3.4005183742090788,
      "objective/train/tokens_used": 2487890400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24292722344398499,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504645109176636,
      "objective/train/weighted_lm_loss": 1.33077073097229,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4005183742090788,
      "tokens_seen": 2467430400
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002551053159478435,
      "loss": 1.3571,
      "theoretical_loss": 3.4005183742090788,
      "tokens_seen": 2467430400
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002550852557673019,
      "loss": 1.4088,
      "theoretical_loss": 3.4005112114260463,
      "tokens_seen": 2467495936
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002550651955867603,
      "loss": 1.2904,
      "theoretical_loss": 3.400504048886518,
      "tokens_seen": 2467561472
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025504513540621865,
      "loss": 1.2511,
      "theoretical_loss": 3.4004968865904805,
      "tokens_seen": 2467627008
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000255025075225677,
      "loss": 1.4574,
      "theoretical_loss": 3.400489724537918,
      "tokens_seen": 2467692544
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025500501504513543,
      "loss": 1.2711,
      "theoretical_loss": 3.4004825627288167,
      "tokens_seen": 2467758080
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002549849548645938,
      "loss": 1.342,
      "theoretical_loss": 3.4004754011631615,
      "tokens_seen": 2467823616
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025496489468405215,
      "loss": 1.4064,
      "theoretical_loss": 3.4004682398409374,
      "tokens_seen": 2467889152
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002549448345035105,
      "loss": 1.3256,
      "theoretical_loss": 3.4004610787621297,
      "tokens_seen": 2467954688
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002549247743229689,
      "loss": 1.3873,
      "theoretical_loss": 3.400453917926724,
      "tokens_seen": 2468020224
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002549047141424273,
      "loss": 1.2139,
      "theoretical_loss": 3.4004467573347053,
      "tokens_seen": 2468085760
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025488465396188566,
      "loss": 1.2301,
      "theoretical_loss": 3.400439596986059,
      "tokens_seen": 2468151296
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000254864593781344,
      "loss": 1.4388,
      "theoretical_loss": 3.40043243688077,
      "tokens_seen": 2468216832
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002548445336008024,
      "loss": 1.4174,
      "theoretical_loss": 3.400425277018824,
      "tokens_seen": 2468282368
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002548244734202608,
      "loss": 1.3693,
      "theoretical_loss": 3.400418117400206,
      "tokens_seen": 2468347904
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025480441323971916,
      "loss": 1.3352,
      "theoretical_loss": 3.400410958024902,
      "tokens_seen": 2468413440
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002547843530591775,
      "loss": 1.3319,
      "theoretical_loss": 3.400403798892896,
      "tokens_seen": 2468478976
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002547642928786359,
      "loss": 1.3178,
      "theoretical_loss": 3.4003966400041747,
      "tokens_seen": 2468544512
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025474423269809425,
      "loss": 1.2921,
      "theoretical_loss": 3.4003894813587223,
      "tokens_seen": 2468610048
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025472417251755266,
      "loss": 1.3249,
      "theoretical_loss": 3.400382322956524,
      "tokens_seen": 2468675584
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000254704112337011,
      "loss": 1.2372,
      "theoretical_loss": 3.4003751647975657,
      "tokens_seen": 2468741120
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002546840521564694,
      "loss": 1.3175,
      "theoretical_loss": 3.4003680068818327,
      "tokens_seen": 2468806656
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025466399197592775,
      "loss": 1.4449,
      "theoretical_loss": 3.4003608492093096,
      "tokens_seen": 2468872192
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025464393179538616,
      "loss": 1.3068,
      "theoretical_loss": 3.4003536917799826,
      "tokens_seen": 2468937728
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025462387161484453,
      "loss": 1.3685,
      "theoretical_loss": 3.400346534593836,
      "tokens_seen": 2469003264
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002546038114343029,
      "loss": 1.3316,
      "theoretical_loss": 3.400339377650856,
      "tokens_seen": 2469068800
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025458375125376125,
      "loss": 1.4013,
      "theoretical_loss": 3.400332220951027,
      "tokens_seen": 2469134336
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002545636910732196,
      "loss": 1.3777,
      "theoretical_loss": 3.4003250644943352,
      "tokens_seen": 2469199872
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025454363089267803,
      "loss": 1.3104,
      "theoretical_loss": 3.4003179082807655,
      "tokens_seen": 2469265408
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002545235707121364,
      "loss": 1.3287,
      "theoretical_loss": 3.4003107523103027,
      "tokens_seen": 2469330944
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002545035105315948,
      "loss": 1.3503,
      "theoretical_loss": 3.400303596582933,
      "tokens_seen": 2469396480
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025448345035105317,
      "loss": 1.2825,
      "theoretical_loss": 3.4002964410986407,
      "tokens_seen": 2469462016
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002544633901705116,
      "loss": 1.2763,
      "theoretical_loss": 3.400289285857412,
      "tokens_seen": 2469527552
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025444332998996995,
      "loss": 1.3787,
      "theoretical_loss": 3.4002821308592317,
      "tokens_seen": 2469593088
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002544232698094283,
      "loss": 1.2865,
      "theoretical_loss": 3.4002749761040847,
      "tokens_seen": 2469658624
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002544032096288867,
      "loss": 1.3835,
      "theoretical_loss": 3.4002678215919575,
      "tokens_seen": 2469724160
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025438314944834504,
      "loss": 1.3892,
      "theoretical_loss": 3.400260667322834,
      "tokens_seen": 2469789696
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025436308926780345,
      "loss": 1.2897,
      "theoretical_loss": 3.4002535132967004,
      "tokens_seen": 2469855232
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002543430290872618,
      "loss": 1.2677,
      "theoretical_loss": 3.400246359513542,
      "tokens_seen": 2469920768
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002543229689067202,
      "loss": 1.3208,
      "theoretical_loss": 3.400239205973344,
      "tokens_seen": 2469986304
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025430290872617854,
      "loss": 1.4319,
      "theoretical_loss": 3.4002320526760914,
      "tokens_seen": 2470051840
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025428284854563696,
      "loss": 1.5093,
      "theoretical_loss": 3.4002248996217697,
      "tokens_seen": 2470117376
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002542627883650953,
      "loss": 1.3516,
      "theoretical_loss": 3.4002177468103643,
      "tokens_seen": 2470182912
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002542427281845537,
      "loss": 1.2956,
      "theoretical_loss": 3.40021059424186,
      "tokens_seen": 2470248448
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025422266800401204,
      "loss": 1.3898,
      "theoretical_loss": 3.4002034419162426,
      "tokens_seen": 2470313984
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025420260782347046,
      "loss": 1.3057,
      "theoretical_loss": 3.400196289833498,
      "tokens_seen": 2470379520
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002541825476429288,
      "loss": 1.345,
      "theoretical_loss": 3.40018913799361,
      "tokens_seen": 2470445056
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002541624874623872,
      "loss": 1.2188,
      "theoretical_loss": 3.400181986396565,
      "tokens_seen": 2470510592
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025414242728184555,
      "loss": 1.3434,
      "theoretical_loss": 3.400174835042348,
      "tokens_seen": 2470576128
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002541223671013039,
      "loss": 1.3535,
      "theoretical_loss": 3.4001676839309445,
      "tokens_seen": 2470641664
    },
    {
      "epoch": 0.75,
      "objective/train/advantage_avg": 0.4886086583137512,
      "objective/train/docs_used": 899059,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.303771734237671,
      "objective/train/original_loss": 1.3037716150283813,
      "objective/train/theoretical_loss": 3.4001605330623397,
      "objective/train/tokens_used": 2491167200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24004243314266205,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500810146331787,
      "objective/train/weighted_lm_loss": 1.3695186376571655,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.4001605330623397,
      "tokens_seen": 2470707200
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002541023069207623,
      "loss": 1.3356,
      "theoretical_loss": 3.4001605330623397,
      "tokens_seen": 2470707200
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002540822467402207,
      "loss": 1.245,
      "theoretical_loss": 3.4001533824365184,
      "tokens_seen": 2470772736
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025406218655967905,
      "loss": 1.3114,
      "theoretical_loss": 3.400146232053467,
      "tokens_seen": 2470838272
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002540421263791374,
      "loss": 1.3768,
      "theoretical_loss": 3.40013908191317,
      "tokens_seen": 2470903808
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025402206619859583,
      "loss": 1.3967,
      "theoretical_loss": 3.400131932015613,
      "tokens_seen": 2470969344
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002540020060180542,
      "loss": 1.2748,
      "theoretical_loss": 3.4001247823607814,
      "tokens_seen": 2471034880
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025398194583751255,
      "loss": 1.3549,
      "theoretical_loss": 3.40011763294866,
      "tokens_seen": 2471100416
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002539618856569709,
      "loss": 1.2406,
      "theoretical_loss": 3.4001104837792346,
      "tokens_seen": 2471165952
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002539418254764293,
      "loss": 1.3719,
      "theoretical_loss": 3.4001033348524907,
      "tokens_seen": 2471231488
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002539217652958877,
      "loss": 1.4515,
      "theoretical_loss": 3.400096186168413,
      "tokens_seen": 2471297024
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025390170511534606,
      "loss": 1.3926,
      "theoretical_loss": 3.4000890377269877,
      "tokens_seen": 2471362560
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002538816449348044,
      "loss": 1.3353,
      "theoretical_loss": 3.4000818895281992,
      "tokens_seen": 2471428096
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002538615847542628,
      "loss": 1.3,
      "theoretical_loss": 3.4000747415720336,
      "tokens_seen": 2471493632
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002538415245737212,
      "loss": 1.2988,
      "theoretical_loss": 3.4000675938584757,
      "tokens_seen": 2471559168
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025382146439317956,
      "loss": 1.4186,
      "theoretical_loss": 3.400060446387511,
      "tokens_seen": 2471624704
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002538014042126379,
      "loss": 1.4615,
      "theoretical_loss": 3.400053299159125,
      "tokens_seen": 2471690240
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002537813440320963,
      "loss": 1.3819,
      "theoretical_loss": 3.4000461521733025,
      "tokens_seen": 2471755776
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025376128385155465,
      "loss": 1.293,
      "theoretical_loss": 3.4000390054300293,
      "tokens_seen": 2471821312
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025374122367101306,
      "loss": 1.3961,
      "theoretical_loss": 3.400031858929291,
      "tokens_seen": 2471886848
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002537211634904714,
      "loss": 1.3492,
      "theoretical_loss": 3.4000247126710725,
      "tokens_seen": 2471952384
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002537011033099298,
      "loss": 1.3707,
      "theoretical_loss": 3.400017566655359,
      "tokens_seen": 2472017920
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025368104312938815,
      "loss": 1.3786,
      "theoretical_loss": 3.4000104208821367,
      "tokens_seen": 2472083456
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025366098294884657,
      "loss": 1.2985,
      "theoretical_loss": 3.4000032753513896,
      "tokens_seen": 2472148992
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025364092276830493,
      "loss": 1.2483,
      "theoretical_loss": 3.399996130063104,
      "tokens_seen": 2472214528
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002536208625877633,
      "loss": 1.4151,
      "theoretical_loss": 3.3999889850172655,
      "tokens_seen": 2472280064
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025360080240722165,
      "loss": 1.3511,
      "theoretical_loss": 3.3999818402138584,
      "tokens_seen": 2472345600
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025358074222668,
      "loss": 1.338,
      "theoretical_loss": 3.399974695652869,
      "tokens_seen": 2472411136
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025356068204613843,
      "loss": 1.2979,
      "theoretical_loss": 3.399967551334282,
      "tokens_seen": 2472476672
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002535406218655968,
      "loss": 1.3212,
      "theoretical_loss": 3.399960407258083,
      "tokens_seen": 2472542208
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025352056168505516,
      "loss": 1.3896,
      "theoretical_loss": 3.3999532634242575,
      "tokens_seen": 2472607744
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002535005015045135,
      "loss": 1.4183,
      "theoretical_loss": 3.3999461198327907,
      "tokens_seen": 2472673280
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025348044132397193,
      "loss": 1.2561,
      "theoretical_loss": 3.399938976483668,
      "tokens_seen": 2472738816
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002534603811434303,
      "loss": 1.279,
      "theoretical_loss": 3.3999318333768747,
      "tokens_seen": 2472804352
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025344032096288866,
      "loss": 1.2252,
      "theoretical_loss": 3.399924690512396,
      "tokens_seen": 2472869888
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000253420260782347,
      "loss": 1.3246,
      "theoretical_loss": 3.3999175478902175,
      "tokens_seen": 2472935424
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002534002006018054,
      "loss": 1.2484,
      "theoretical_loss": 3.3999104055103246,
      "tokens_seen": 2473000960
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002533801404212638,
      "loss": 1.3323,
      "theoretical_loss": 3.3999032633727024,
      "tokens_seen": 2473066496
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025336008024072216,
      "loss": 1.3987,
      "theoretical_loss": 3.3998961214773367,
      "tokens_seen": 2473132032
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002533400200601805,
      "loss": 1.3845,
      "theoretical_loss": 3.3998889798242127,
      "tokens_seen": 2473197568
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002533199598796389,
      "loss": 1.3394,
      "theoretical_loss": 3.3998818384133154,
      "tokens_seen": 2473263104
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002532998996990973,
      "loss": 1.466,
      "theoretical_loss": 3.3998746972446305,
      "tokens_seen": 2473328640
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025327983951855566,
      "loss": 1.3671,
      "theoretical_loss": 3.399867556318143,
      "tokens_seen": 2473394176
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000253259779338014,
      "loss": 1.4764,
      "theoretical_loss": 3.3998604156338392,
      "tokens_seen": 2473459712
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002532397191574724,
      "loss": 1.2956,
      "theoretical_loss": 3.3998532751917034,
      "tokens_seen": 2473525248
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025321965897693075,
      "loss": 1.4286,
      "theoretical_loss": 3.3998461349917215,
      "tokens_seen": 2473590784
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025319959879638917,
      "loss": 1.3033,
      "theoretical_loss": 3.3998389950338788,
      "tokens_seen": 2473656320
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025317953861584753,
      "loss": 1.3654,
      "theoretical_loss": 3.3998318553181606,
      "tokens_seen": 2473721856
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002531594784353059,
      "loss": 1.2541,
      "theoretical_loss": 3.3998247158445523,
      "tokens_seen": 2473787392
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025313941825476425,
      "loss": 1.3976,
      "theoretical_loss": 3.3998175766130396,
      "tokens_seen": 2473852928
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025311935807422267,
      "loss": 1.3321,
      "theoretical_loss": 3.399810437623607,
      "tokens_seen": 2473918464
    },
    {
      "epoch": 0.75,
      "objective/train/advantage_avg": 0.49347853660583496,
      "objective/train/docs_used": 900350,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2348965406417847,
      "objective/train/original_loss": 1.2348964214324951,
      "objective/train/theoretical_loss": 3.399803298876241,
      "objective/train/tokens_used": 2494444000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24382860958576202,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505874156951904,
      "objective/train/weighted_lm_loss": 1.297237753868103,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.399803298876241,
      "tokens_seen": 2473984000
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025309929789368103,
      "loss": 1.3475,
      "theoretical_loss": 3.399803298876241,
      "tokens_seen": 2473984000
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002530792377131394,
      "loss": 1.5128,
      "theoretical_loss": 3.399796160370926,
      "tokens_seen": 2474049536
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025305917753259776,
      "loss": 1.4054,
      "theoretical_loss": 3.399789022107648,
      "tokens_seen": 2474115072
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002530391173520562,
      "loss": 1.2707,
      "theoretical_loss": 3.399781884086392,
      "tokens_seen": 2474180608
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025301905717151454,
      "loss": 1.5024,
      "theoretical_loss": 3.399774746307144,
      "tokens_seen": 2474246144
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025299899699097295,
      "loss": 1.3068,
      "theoretical_loss": 3.399767608769889,
      "tokens_seen": 2474311680
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002529789368104313,
      "loss": 1.3489,
      "theoretical_loss": 3.399760471474612,
      "tokens_seen": 2474377216
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002529588766298897,
      "loss": 1.3161,
      "theoretical_loss": 3.3997533344212987,
      "tokens_seen": 2474442752
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002529388164493481,
      "loss": 1.2925,
      "theoretical_loss": 3.3997461976099346,
      "tokens_seen": 2474508288
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025291875626880646,
      "loss": 1.3495,
      "theoretical_loss": 3.3997390610405054,
      "tokens_seen": 2474573824
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002528986960882648,
      "loss": 1.4577,
      "theoretical_loss": 3.399731924712996,
      "tokens_seen": 2474639360
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002528786359077232,
      "loss": 1.3949,
      "theoretical_loss": 3.3997247886273914,
      "tokens_seen": 2474704896
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002528585757271816,
      "loss": 1.3362,
      "theoretical_loss": 3.399717652783678,
      "tokens_seen": 2474770432
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025283851554663996,
      "loss": 1.3025,
      "theoretical_loss": 3.3997105171818403,
      "tokens_seen": 2474835968
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002528184553660983,
      "loss": 1.3243,
      "theoretical_loss": 3.3997033818218645,
      "tokens_seen": 2474901504
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002527983951855567,
      "loss": 1.4086,
      "theoretical_loss": 3.3996962467037353,
      "tokens_seen": 2474967040
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025277833500501505,
      "loss": 1.3656,
      "theoretical_loss": 3.3996891118274384,
      "tokens_seen": 2475032576
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025275827482447346,
      "loss": 1.4466,
      "theoretical_loss": 3.3996819771929596,
      "tokens_seen": 2475098112
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002527382146439318,
      "loss": 1.3642,
      "theoretical_loss": 3.3996748428002834,
      "tokens_seen": 2475163648
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002527181544633902,
      "loss": 1.4364,
      "theoretical_loss": 3.399667708649396,
      "tokens_seen": 2475229184
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025269809428284855,
      "loss": 1.1929,
      "theoretical_loss": 3.3996605747402824,
      "tokens_seen": 2475294720
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025267803410230697,
      "loss": 1.4504,
      "theoretical_loss": 3.399653441072928,
      "tokens_seen": 2475360256
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025265797392176533,
      "loss": 1.4428,
      "theoretical_loss": 3.3996463076473185,
      "tokens_seen": 2475425792
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002526379137412237,
      "loss": 1.3408,
      "theoretical_loss": 3.399639174463439,
      "tokens_seen": 2475491328
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025261785356068205,
      "loss": 1.2921,
      "theoretical_loss": 3.399632041521275,
      "tokens_seen": 2475556864
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002525977933801404,
      "loss": 1.4072,
      "theoretical_loss": 3.399624908820812,
      "tokens_seen": 2475622400
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025257773319959883,
      "loss": 1.4185,
      "theoretical_loss": 3.399617776362035,
      "tokens_seen": 2475687936
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002525576730190572,
      "loss": 1.253,
      "theoretical_loss": 3.3996106441449307,
      "tokens_seen": 2475753472
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025253761283851556,
      "loss": 1.2536,
      "theoretical_loss": 3.3996035121694828,
      "tokens_seen": 2475819008
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002525175526579739,
      "loss": 1.4228,
      "theoretical_loss": 3.399596380435678,
      "tokens_seen": 2475884544
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025249749247743233,
      "loss": 1.3019,
      "theoretical_loss": 3.3995892489435007,
      "tokens_seen": 2475950080
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002524774322968907,
      "loss": 1.3737,
      "theoretical_loss": 3.399582117692937,
      "tokens_seen": 2476015616
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025245737211634906,
      "loss": 1.2584,
      "theoretical_loss": 3.399574986683972,
      "tokens_seen": 2476081152
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002524373119358074,
      "loss": 1.4154,
      "theoretical_loss": 3.3995678559165916,
      "tokens_seen": 2476146688
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002524172517552658,
      "loss": 1.3305,
      "theoretical_loss": 3.3995607253907805,
      "tokens_seen": 2476212224
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002523971915747242,
      "loss": 1.4096,
      "theoretical_loss": 3.399553595106525,
      "tokens_seen": 2476277760
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025237713139418256,
      "loss": 1.3329,
      "theoretical_loss": 3.3995464650638096,
      "tokens_seen": 2476343296
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002523570712136409,
      "loss": 1.3701,
      "theoretical_loss": 3.39953933526262,
      "tokens_seen": 2476408832
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002523370110330993,
      "loss": 1.3391,
      "theoretical_loss": 3.3995322057029425,
      "tokens_seen": 2476474368
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002523169508525577,
      "loss": 1.3623,
      "theoretical_loss": 3.3995250763847613,
      "tokens_seen": 2476539904
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025229689067201606,
      "loss": 1.1808,
      "theoretical_loss": 3.3995179473080626,
      "tokens_seen": 2476605440
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025227683049147443,
      "loss": 1.3904,
      "theoretical_loss": 3.3995108184728315,
      "tokens_seen": 2476670976
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002522567703109328,
      "loss": 1.3609,
      "theoretical_loss": 3.3995036898790536,
      "tokens_seen": 2476736512
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025223671013039115,
      "loss": 1.3087,
      "theoretical_loss": 3.399496561526714,
      "tokens_seen": 2476802048
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025221664994984957,
      "loss": 1.2625,
      "theoretical_loss": 3.399489433415799,
      "tokens_seen": 2476867584
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025219658976930793,
      "loss": 1.3773,
      "theoretical_loss": 3.3994823055462926,
      "tokens_seen": 2476933120
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002521765295887663,
      "loss": 1.343,
      "theoretical_loss": 3.3994751779181813,
      "tokens_seen": 2476998656
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025215646940822465,
      "loss": 1.2568,
      "theoretical_loss": 3.3994680505314507,
      "tokens_seen": 2477064192
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025213640922768307,
      "loss": 1.3754,
      "theoretical_loss": 3.3994609233860853,
      "tokens_seen": 2477129728
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025211634904714143,
      "loss": 1.3659,
      "theoretical_loss": 3.3994537964820712,
      "tokens_seen": 2477195264
    },
    {
      "epoch": 0.75,
      "objective/train/advantage_avg": 0.4926232099533081,
      "objective/train/docs_used": 901501,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.221693515777588,
      "objective/train/original_loss": 1.2216936349868774,
      "objective/train/theoretical_loss": 3.399446669819394,
      "objective/train/tokens_used": 2497720800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24308888614177704,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504980087280273,
      "objective/train/weighted_lm_loss": 1.2833781242370605,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.399446669819394,
      "tokens_seen": 2477260800
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002520962888665998,
      "loss": 1.2928,
      "theoretical_loss": 3.399446669819394,
      "tokens_seen": 2477260800
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025207622868605816,
      "loss": 1.2627,
      "theoretical_loss": 3.3994395433980387,
      "tokens_seen": 2477326336
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002520561685055166,
      "loss": 1.4069,
      "theoretical_loss": 3.399432417217991,
      "tokens_seen": 2477391872
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025203610832497494,
      "loss": 1.4081,
      "theoretical_loss": 3.399425291279236,
      "tokens_seen": 2477457408
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002520160481444333,
      "loss": 1.374,
      "theoretical_loss": 3.3994181655817592,
      "tokens_seen": 2477522944
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025199598796389166,
      "loss": 1.3435,
      "theoretical_loss": 3.399411040125547,
      "tokens_seen": 2477588480
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025197592778335,
      "loss": 1.2313,
      "theoretical_loss": 3.3994039149105832,
      "tokens_seen": 2477654016
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025195586760280844,
      "loss": 1.3176,
      "theoretical_loss": 3.399396789936855,
      "tokens_seen": 2477719552
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002519358074222668,
      "loss": 1.3582,
      "theoretical_loss": 3.3993896652043465,
      "tokens_seen": 2477785088
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025191574724172516,
      "loss": 1.4526,
      "theoretical_loss": 3.399382540713044,
      "tokens_seen": 2477850624
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002518956870611835,
      "loss": 1.3123,
      "theoretical_loss": 3.3993754164629317,
      "tokens_seen": 2477916160
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025187562688064194,
      "loss": 1.4311,
      "theoretical_loss": 3.399368292453997,
      "tokens_seen": 2477981696
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002518555667001003,
      "loss": 1.2865,
      "theoretical_loss": 3.399361168686224,
      "tokens_seen": 2478047232
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025183550651955867,
      "loss": 1.3965,
      "theoretical_loss": 3.3993540451595985,
      "tokens_seen": 2478112768
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025181544633901703,
      "loss": 1.4441,
      "theoretical_loss": 3.3993469218741055,
      "tokens_seen": 2478178304
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002517953861584754,
      "loss": 1.2558,
      "theoretical_loss": 3.3993397988297316,
      "tokens_seen": 2478243840
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002517753259779338,
      "loss": 1.313,
      "theoretical_loss": 3.399332676026461,
      "tokens_seen": 2478309376
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025175526579739217,
      "loss": 1.4156,
      "theoretical_loss": 3.39932555346428,
      "tokens_seen": 2478374912
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025173520561685053,
      "loss": 1.4254,
      "theoretical_loss": 3.3993184311431737,
      "tokens_seen": 2478440448
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002517151454363089,
      "loss": 1.3584,
      "theoretical_loss": 3.3993113090631275,
      "tokens_seen": 2478505984
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002516950852557673,
      "loss": 1.4862,
      "theoretical_loss": 3.399304187224127,
      "tokens_seen": 2478571520
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002516750250752257,
      "loss": 1.3521,
      "theoretical_loss": 3.399297065626158,
      "tokens_seen": 2478637056
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025165496489468404,
      "loss": 1.4195,
      "theoretical_loss": 3.3992899442692055,
      "tokens_seen": 2478702592
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002516349047141424,
      "loss": 1.322,
      "theoretical_loss": 3.3992828231532553,
      "tokens_seen": 2478768128
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025161484453360076,
      "loss": 1.3431,
      "theoretical_loss": 3.3992757022782927,
      "tokens_seen": 2478833664
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002515947843530592,
      "loss": 1.3729,
      "theoretical_loss": 3.399268581644303,
      "tokens_seen": 2478899200
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025157472417251754,
      "loss": 1.4922,
      "theoretical_loss": 3.399261461251272,
      "tokens_seen": 2478964736
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002515546639919759,
      "loss": 1.4025,
      "theoretical_loss": 3.399254341099185,
      "tokens_seen": 2479030272
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025153460381143426,
      "loss": 1.241,
      "theoretical_loss": 3.3992472211880274,
      "tokens_seen": 2479095808
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002515145436308927,
      "loss": 1.3674,
      "theoretical_loss": 3.3992401015177847,
      "tokens_seen": 2479161344
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002514944834503511,
      "loss": 1.3496,
      "theoretical_loss": 3.3992329820884426,
      "tokens_seen": 2479226880
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025147442326980946,
      "loss": 1.3117,
      "theoretical_loss": 3.3992258628999865,
      "tokens_seen": 2479292416
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002514543630892678,
      "loss": 1.4904,
      "theoretical_loss": 3.3992187439524018,
      "tokens_seen": 2479357952
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002514343029087262,
      "loss": 1.2901,
      "theoretical_loss": 3.399211625245674,
      "tokens_seen": 2479423488
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002514142427281846,
      "loss": 1.3343,
      "theoretical_loss": 3.3992045067797885,
      "tokens_seen": 2479489024
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025139418254764296,
      "loss": 1.3384,
      "theoretical_loss": 3.399197388554731,
      "tokens_seen": 2479554560
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002513741223671013,
      "loss": 1.4953,
      "theoretical_loss": 3.399190270570487,
      "tokens_seen": 2479620096
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002513540621865597,
      "loss": 1.2851,
      "theoretical_loss": 3.3991831528270415,
      "tokens_seen": 2479685632
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002513340020060181,
      "loss": 1.4847,
      "theoretical_loss": 3.3991760353243805,
      "tokens_seen": 2479751168
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025131394182547647,
      "loss": 1.364,
      "theoretical_loss": 3.3991689180624896,
      "tokens_seen": 2479816704
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025129388164493483,
      "loss": 1.3697,
      "theoretical_loss": 3.3991618010413536,
      "tokens_seen": 2479882240
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002512738214643932,
      "loss": 1.3511,
      "theoretical_loss": 3.3991546842609592,
      "tokens_seen": 2479947776
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025125376128385155,
      "loss": 1.3491,
      "theoretical_loss": 3.39914756772129,
      "tokens_seen": 2480013312
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025123370110330997,
      "loss": 1.4489,
      "theoretical_loss": 3.3991404514223333,
      "tokens_seen": 2480078848
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025121364092276833,
      "loss": 1.4414,
      "theoretical_loss": 3.3991333353640742,
      "tokens_seen": 2480144384
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002511935807422267,
      "loss": 1.4565,
      "theoretical_loss": 3.3991262195464973,
      "tokens_seen": 2480209920
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025117352056168506,
      "loss": 1.4199,
      "theoretical_loss": 3.399119103969589,
      "tokens_seen": 2480275456
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025115346038114347,
      "loss": 1.3633,
      "theoretical_loss": 3.3991119886333347,
      "tokens_seen": 2480340992
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025113340020060183,
      "loss": 1.4671,
      "theoretical_loss": 3.3991048735377194,
      "tokens_seen": 2480406528
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002511133400200602,
      "loss": 1.3703,
      "theoretical_loss": 3.3990977586827293,
      "tokens_seen": 2480472064
    },
    {
      "epoch": 0.75,
      "objective/train/advantage_avg": 0.4946509301662445,
      "objective/train/docs_used": 902720,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.386062741279602,
      "objective/train/original_loss": 1.3860626220703125,
      "objective/train/theoretical_loss": 3.399090644068349,
      "objective/train/tokens_used": 2500997600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2451120764017105,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507112741470337,
      "objective/train/weighted_lm_loss": 1.456169843673706,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.399090644068349,
      "tokens_seen": 2480537600
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025109327983951856,
      "loss": 1.3914,
      "theoretical_loss": 3.399090644068349,
      "tokens_seen": 2480537600
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000251073219658977,
      "loss": 1.3911,
      "theoretical_loss": 3.3990835296945647,
      "tokens_seen": 2480603136
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025105315947843534,
      "loss": 1.3339,
      "theoretical_loss": 3.399076415561362,
      "tokens_seen": 2480668672
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002510330992978937,
      "loss": 1.3478,
      "theoretical_loss": 3.399069301668726,
      "tokens_seen": 2480734208
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025101303911735206,
      "loss": 1.3532,
      "theoretical_loss": 3.3990621880166425,
      "tokens_seen": 2480799744
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002509929789368104,
      "loss": 1.3771,
      "theoretical_loss": 3.3990550746050965,
      "tokens_seen": 2480865280
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025097291875626884,
      "loss": 1.4639,
      "theoretical_loss": 3.399047961434074,
      "tokens_seen": 2480930816
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002509528585757272,
      "loss": 1.2857,
      "theoretical_loss": 3.399040848503561,
      "tokens_seen": 2480996352
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025093279839518556,
      "loss": 1.2257,
      "theoretical_loss": 3.399033735813542,
      "tokens_seen": 2481061888
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002509127382146439,
      "loss": 1.4461,
      "theoretical_loss": 3.3990266233640023,
      "tokens_seen": 2481127424
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025089267803410234,
      "loss": 1.4529,
      "theoretical_loss": 3.399019511154929,
      "tokens_seen": 2481192960
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002508726178535607,
      "loss": 1.3259,
      "theoretical_loss": 3.399012399186306,
      "tokens_seen": 2481258496
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025085255767301907,
      "loss": 1.3738,
      "theoretical_loss": 3.3990052874581203,
      "tokens_seen": 2481324032
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025083249749247743,
      "loss": 1.3406,
      "theoretical_loss": 3.398998175970356,
      "tokens_seen": 2481389568
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002508124373119358,
      "loss": 1.3219,
      "theoretical_loss": 3.398991064722999,
      "tokens_seen": 2481455104
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002507923771313942,
      "loss": 1.3017,
      "theoretical_loss": 3.3989839537160353,
      "tokens_seen": 2481520640
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025077231695085257,
      "loss": 1.5166,
      "theoretical_loss": 3.3989768429494505,
      "tokens_seen": 2481586176
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025075225677031093,
      "loss": 1.3611,
      "theoretical_loss": 3.3989697324232298,
      "tokens_seen": 2481651712
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002507321965897693,
      "loss": 1.4058,
      "theoretical_loss": 3.3989626221373586,
      "tokens_seen": 2481717248
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002507121364092277,
      "loss": 1.4439,
      "theoretical_loss": 3.3989555120918222,
      "tokens_seen": 2481782784
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002506920762286861,
      "loss": 1.2653,
      "theoretical_loss": 3.3989484022866066,
      "tokens_seen": 2481848320
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025067201604814444,
      "loss": 1.3482,
      "theoretical_loss": 3.3989412927216978,
      "tokens_seen": 2481913856
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002506519558676028,
      "loss": 1.3744,
      "theoretical_loss": 3.3989341833970803,
      "tokens_seen": 2481979392
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025063189568706116,
      "loss": 1.2381,
      "theoretical_loss": 3.39892707431274,
      "tokens_seen": 2482044928
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002506118355065196,
      "loss": 1.3989,
      "theoretical_loss": 3.398919965468663,
      "tokens_seen": 2482110464
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025059177532597794,
      "loss": 1.4613,
      "theoretical_loss": 3.398912856864834,
      "tokens_seen": 2482176000
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002505717151454363,
      "loss": 1.2244,
      "theoretical_loss": 3.398905748501239,
      "tokens_seen": 2482241536
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025055165496489466,
      "loss": 1.3535,
      "theoretical_loss": 3.3988986403778636,
      "tokens_seen": 2482307072
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002505315947843531,
      "loss": 1.282,
      "theoretical_loss": 3.3988915324946927,
      "tokens_seen": 2482372608
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025051153460381144,
      "loss": 1.363,
      "theoretical_loss": 3.3988844248517127,
      "tokens_seen": 2482438144
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002504914744232698,
      "loss": 1.4264,
      "theoretical_loss": 3.3988773174489086,
      "tokens_seen": 2482503680
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025047141424272817,
      "loss": 1.4206,
      "theoretical_loss": 3.3988702102862662,
      "tokens_seen": 2482569216
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025045135406218653,
      "loss": 1.3739,
      "theoretical_loss": 3.3988631033637713,
      "tokens_seen": 2482634752
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025043129388164495,
      "loss": 1.3599,
      "theoretical_loss": 3.3988559966814087,
      "tokens_seen": 2482700288
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002504112337011033,
      "loss": 1.3692,
      "theoretical_loss": 3.3988488902391643,
      "tokens_seen": 2482765824
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025039117352056167,
      "loss": 1.4445,
      "theoretical_loss": 3.3988417840370237,
      "tokens_seen": 2482831360
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025037111334002003,
      "loss": 1.3794,
      "theoretical_loss": 3.3988346780749725,
      "tokens_seen": 2482896896
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025035105315947845,
      "loss": 1.3285,
      "theoretical_loss": 3.3988275723529964,
      "tokens_seen": 2482962432
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002503309929789368,
      "loss": 1.2404,
      "theoretical_loss": 3.3988204668710806,
      "tokens_seen": 2483027968
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002503109327983952,
      "loss": 1.3615,
      "theoretical_loss": 3.398813361629211,
      "tokens_seen": 2483093504
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025029087261785354,
      "loss": 1.3832,
      "theoretical_loss": 3.3988062566273727,
      "tokens_seen": 2483159040
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002502708124373119,
      "loss": 1.2511,
      "theoretical_loss": 3.3987991518655516,
      "tokens_seen": 2483224576
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002502507522567703,
      "loss": 1.3613,
      "theoretical_loss": 3.398792047343733,
      "tokens_seen": 2483290112
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002502306920762287,
      "loss": 1.4277,
      "theoretical_loss": 3.3987849430619033,
      "tokens_seen": 2483355648
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025021063189568704,
      "loss": 1.3592,
      "theoretical_loss": 3.3987778390200467,
      "tokens_seen": 2483421184
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002501905717151454,
      "loss": 1.4065,
      "theoretical_loss": 3.3987707352181498,
      "tokens_seen": 2483486720
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002501705115346038,
      "loss": 1.342,
      "theoretical_loss": 3.3987636316561978,
      "tokens_seen": 2483552256
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002501504513540622,
      "loss": 1.4207,
      "theoretical_loss": 3.398756528334176,
      "tokens_seen": 2483617792
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025013039117352054,
      "loss": 1.3373,
      "theoretical_loss": 3.398749425252071,
      "tokens_seen": 2483683328
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002501103309929789,
      "loss": 1.3755,
      "theoretical_loss": 3.3987423224098667,
      "tokens_seen": 2483748864
    },
    {
      "epoch": 0.75,
      "objective/train/advantage_avg": 0.4943104386329651,
      "objective/train/docs_used": 903880,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3607133626937866,
      "objective/train/original_loss": 1.360713243484497,
      "objective/train/theoretical_loss": 3.3987352198075502,
      "objective/train/tokens_used": 2504274400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24462854862213135,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506747961044312,
      "objective/train/weighted_lm_loss": 1.4297703504562378,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3987352198075502,
      "tokens_seen": 2483814400
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025009027081243727,
      "loss": 1.3526,
      "theoretical_loss": 3.3987352198075502,
      "tokens_seen": 2483814400
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002500702106318957,
      "loss": 1.3969,
      "theoretical_loss": 3.3987281174451063,
      "tokens_seen": 2483879936
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025005015045135405,
      "loss": 1.3596,
      "theoretical_loss": 3.3987210153225207,
      "tokens_seen": 2483945472
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002500300902708124,
      "loss": 1.3483,
      "theoretical_loss": 3.398713913439779,
      "tokens_seen": 2484011008
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00025001003009027077,
      "loss": 1.4031,
      "theoretical_loss": 3.3987068117968664,
      "tokens_seen": 2484076544
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002499899699097292,
      "loss": 1.4167,
      "theoretical_loss": 3.3986997103937693,
      "tokens_seen": 2484142080
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024996990972918755,
      "loss": 1.2916,
      "theoretical_loss": 3.398692609230473,
      "tokens_seen": 2484207616
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024994984954864597,
      "loss": 1.3431,
      "theoretical_loss": 3.3986855083069623,
      "tokens_seen": 2484273152
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024992978936810433,
      "loss": 1.2035,
      "theoretical_loss": 3.3986784076232235,
      "tokens_seen": 2484338688
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002499097291875627,
      "loss": 1.3129,
      "theoretical_loss": 3.3986713071792423,
      "tokens_seen": 2484404224
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024988966900702105,
      "loss": 1.3049,
      "theoretical_loss": 3.398664206975004,
      "tokens_seen": 2484469760
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002498696088264794,
      "loss": 1.227,
      "theoretical_loss": 3.3986571070104943,
      "tokens_seen": 2484535296
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024984954864593783,
      "loss": 1.3188,
      "theoretical_loss": 3.3986500072856987,
      "tokens_seen": 2484600832
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002498294884653962,
      "loss": 1.3854,
      "theoretical_loss": 3.3986429078006024,
      "tokens_seen": 2484666368
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024980942828485455,
      "loss": 1.292,
      "theoretical_loss": 3.3986358085551918,
      "tokens_seen": 2484731904
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002497893681043129,
      "loss": 1.312,
      "theoretical_loss": 3.3986287095494516,
      "tokens_seen": 2484797440
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024976930792377133,
      "loss": 1.3317,
      "theoretical_loss": 3.398621610783368,
      "tokens_seen": 2484862976
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002497492477432297,
      "loss": 1.3452,
      "theoretical_loss": 3.398614512256927,
      "tokens_seen": 2484928512
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024972918756268806,
      "loss": 1.3601,
      "theoretical_loss": 3.398607413970113,
      "tokens_seen": 2484994048
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002497091273821464,
      "loss": 1.3256,
      "theoretical_loss": 3.3986003159229123,
      "tokens_seen": 2485059584
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002496890672016048,
      "loss": 1.4817,
      "theoretical_loss": 3.3985932181153107,
      "tokens_seen": 2485125120
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002496690070210632,
      "loss": 1.4036,
      "theoretical_loss": 3.398586120547293,
      "tokens_seen": 2485190656
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024964894684052156,
      "loss": 1.3465,
      "theoretical_loss": 3.3985790232188458,
      "tokens_seen": 2485256192
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002496288866599799,
      "loss": 1.3061,
      "theoretical_loss": 3.3985719261299536,
      "tokens_seen": 2485321728
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024960882647943834,
      "loss": 1.3651,
      "theoretical_loss": 3.398564829280603,
      "tokens_seen": 2485387264
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002495887662988967,
      "loss": 1.4218,
      "theoretical_loss": 3.3985577326707794,
      "tokens_seen": 2485452800
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024956870611835506,
      "loss": 1.2979,
      "theoretical_loss": 3.3985506363004676,
      "tokens_seen": 2485518336
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002495486459378135,
      "loss": 1.2827,
      "theoretical_loss": 3.3985435401696544,
      "tokens_seen": 2485583872
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024952858575727184,
      "loss": 1.356,
      "theoretical_loss": 3.3985364442783244,
      "tokens_seen": 2485649408
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002495085255767302,
      "loss": 1.4041,
      "theoretical_loss": 3.3985293486264636,
      "tokens_seen": 2485714944
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024948846539618857,
      "loss": 1.3226,
      "theoretical_loss": 3.398522253214058,
      "tokens_seen": 2485780480
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024946840521564693,
      "loss": 1.3681,
      "theoretical_loss": 3.3985151580410924,
      "tokens_seen": 2485846016
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024944834503510535,
      "loss": 1.3112,
      "theoretical_loss": 3.398508063107553,
      "tokens_seen": 2485911552
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002494282848545637,
      "loss": 1.2449,
      "theoretical_loss": 3.3985009684134253,
      "tokens_seen": 2485977088
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024940822467402207,
      "loss": 1.407,
      "theoretical_loss": 3.3984938739586945,
      "tokens_seen": 2486042624
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024938816449348043,
      "loss": 1.3049,
      "theoretical_loss": 3.3984867797433465,
      "tokens_seen": 2486108160
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024936810431293885,
      "loss": 1.517,
      "theoretical_loss": 3.3984796857673674,
      "tokens_seen": 2486173696
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002493480441323972,
      "loss": 1.3088,
      "theoretical_loss": 3.398472592030742,
      "tokens_seen": 2486239232
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002493279839518556,
      "loss": 1.3799,
      "theoretical_loss": 3.3984654985334566,
      "tokens_seen": 2486304768
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024930792377131394,
      "loss": 1.3718,
      "theoretical_loss": 3.3984584052754965,
      "tokens_seen": 2486370304
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002492878635907723,
      "loss": 1.3573,
      "theoretical_loss": 3.3984513122568467,
      "tokens_seen": 2486435840
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002492678034102307,
      "loss": 1.4669,
      "theoretical_loss": 3.3984442194774944,
      "tokens_seen": 2486501376
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002492477432296891,
      "loss": 1.4213,
      "theoretical_loss": 3.3984371269374236,
      "tokens_seen": 2486566912
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024922768304914744,
      "loss": 1.3451,
      "theoretical_loss": 3.398430034636621,
      "tokens_seen": 2486632448
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002492076228686058,
      "loss": 1.3716,
      "theoretical_loss": 3.3984229425750714,
      "tokens_seen": 2486697984
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002491875626880642,
      "loss": 1.3729,
      "theoretical_loss": 3.3984158507527606,
      "tokens_seen": 2486763520
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002491675025075226,
      "loss": 1.3152,
      "theoretical_loss": 3.398408759169675,
      "tokens_seen": 2486829056
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024914744232698094,
      "loss": 1.3673,
      "theoretical_loss": 3.3984016678257998,
      "tokens_seen": 2486894592
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002491273821464393,
      "loss": 1.3452,
      "theoretical_loss": 3.3983945767211203,
      "tokens_seen": 2486960128
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024910732196589767,
      "loss": 1.3082,
      "theoretical_loss": 3.398387485855622,
      "tokens_seen": 2487025664
    },
    {
      "epoch": 0.75,
      "objective/train/advantage_avg": 0.49270322918891907,
      "objective/train/docs_used": 905256,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3241273164749146,
      "objective/train/original_loss": 1.324127435684204,
      "objective/train/theoretical_loss": 3.398380395229291,
      "objective/train/tokens_used": 2507551200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2430727779865265,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505061149597168,
      "objective/train/weighted_lm_loss": 1.3910661935806274,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9753099083900452,
      "theoretical_loss": 3.398380395229291,
      "tokens_seen": 2487091200
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002490872617853561,
      "loss": 1.3639,
      "theoretical_loss": 3.398380395229291,
      "tokens_seen": 2487091200
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024906720160481445,
      "loss": 1.3707,
      "theoretical_loss": 3.398373304842113,
      "tokens_seen": 2487156736
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002490471414242728,
      "loss": 1.4217,
      "theoretical_loss": 3.3983662146940734,
      "tokens_seen": 2487222272
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024902708124373117,
      "loss": 1.4039,
      "theoretical_loss": 3.398359124785158,
      "tokens_seen": 2487287808
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002490070210631896,
      "loss": 1.361,
      "theoretical_loss": 3.398352035115352,
      "tokens_seen": 2487353344
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024898696088264795,
      "loss": 1.2029,
      "theoretical_loss": 3.3983449456846415,
      "tokens_seen": 2487418880
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002489669007021063,
      "loss": 1.4749,
      "theoretical_loss": 3.398337856493012,
      "tokens_seen": 2487484416
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002489468405215647,
      "loss": 1.4302,
      "theoretical_loss": 3.3983307675404486,
      "tokens_seen": 2487549952
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024892678034102304,
      "loss": 1.2633,
      "theoretical_loss": 3.398323678826938,
      "tokens_seen": 2487615488
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024890672016048145,
      "loss": 1.435,
      "theoretical_loss": 3.398316590352465,
      "tokens_seen": 2487681024
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002488866599799398,
      "loss": 1.3621,
      "theoretical_loss": 3.3983095021170158,
      "tokens_seen": 2487746560
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024886659979939823,
      "loss": 1.3997,
      "theoretical_loss": 3.3983024141205758,
      "tokens_seen": 2487812096
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002488465396188566,
      "loss": 1.3365,
      "theoretical_loss": 3.3982953263631304,
      "tokens_seen": 2487877632
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024882647943831496,
      "loss": 1.328,
      "theoretical_loss": 3.398288238844666,
      "tokens_seen": 2487943168
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002488064192577733,
      "loss": 1.4266,
      "theoretical_loss": 3.398281151565167,
      "tokens_seen": 2488008704
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024878635907723173,
      "loss": 1.5241,
      "theoretical_loss": 3.39827406452462,
      "tokens_seen": 2488074240
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002487662988966901,
      "loss": 1.442,
      "theoretical_loss": 3.3982669777230106,
      "tokens_seen": 2488139776
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024874623871614846,
      "loss": 1.4326,
      "theoretical_loss": 3.398259891160324,
      "tokens_seen": 2488205312
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002487261785356068,
      "loss": 1.3634,
      "theoretical_loss": 3.398252804836546,
      "tokens_seen": 2488270848
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002487061183550652,
      "loss": 1.4293,
      "theoretical_loss": 3.398245718751663,
      "tokens_seen": 2488336384
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002486860581745236,
      "loss": 1.3228,
      "theoretical_loss": 3.3982386329056595,
      "tokens_seen": 2488401920
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024866599799398196,
      "loss": 1.3728,
      "theoretical_loss": 3.398231547298522,
      "tokens_seen": 2488467456
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002486459378134403,
      "loss": 1.3593,
      "theoretical_loss": 3.3982244619302358,
      "tokens_seen": 2488532992
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002486258776328987,
      "loss": 1.426,
      "theoretical_loss": 3.3982173768007864,
      "tokens_seen": 2488598528
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002486058174523571,
      "loss": 1.4213,
      "theoretical_loss": 3.3982102919101598,
      "tokens_seen": 2488664064
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024858575727181546,
      "loss": 1.4008,
      "theoretical_loss": 3.3982032072583412,
      "tokens_seen": 2488729600
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002485656970912738,
      "loss": 1.3715,
      "theoretical_loss": 3.398196122845317,
      "tokens_seen": 2488795136
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002485456369107322,
      "loss": 1.4205,
      "theoretical_loss": 3.398189038671072,
      "tokens_seen": 2488860672
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024852557673019055,
      "loss": 1.3605,
      "theoretical_loss": 3.398181954735593,
      "tokens_seen": 2488926208
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024850551654964897,
      "loss": 1.4315,
      "theoretical_loss": 3.3981748710388646,
      "tokens_seen": 2488991744
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024848545636910733,
      "loss": 1.3554,
      "theoretical_loss": 3.3981677875808725,
      "tokens_seen": 2489057280
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002484653961885657,
      "loss": 1.3832,
      "theoretical_loss": 3.398160704361603,
      "tokens_seen": 2489122816
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024844533600802405,
      "loss": 1.3098,
      "theoretical_loss": 3.398153621381042,
      "tokens_seen": 2489188352
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024842527582748247,
      "loss": 1.386,
      "theoretical_loss": 3.398146538639174,
      "tokens_seen": 2489253888
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024840521564694083,
      "loss": 1.2865,
      "theoretical_loss": 3.3981394561359854,
      "tokens_seen": 2489319424
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002483851554663992,
      "loss": 1.3635,
      "theoretical_loss": 3.398132373871462,
      "tokens_seen": 2489384960
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024836509528585756,
      "loss": 1.2786,
      "theoretical_loss": 3.398125291845589,
      "tokens_seen": 2489450496
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.000248345035105316,
      "loss": 1.4983,
      "theoretical_loss": 3.3981182100583527,
      "tokens_seen": 2489516032
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024832497492477434,
      "loss": 1.4931,
      "theoretical_loss": 3.398111128509738,
      "tokens_seen": 2489581568
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002483049147442327,
      "loss": 1.3707,
      "theoretical_loss": 3.3981040471997312,
      "tokens_seen": 2489647104
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024828485456369106,
      "loss": 1.4064,
      "theoretical_loss": 3.398096966128318,
      "tokens_seen": 2489712640
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002482647943831494,
      "loss": 1.2822,
      "theoretical_loss": 3.3980898852954837,
      "tokens_seen": 2489778176
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024824473420260784,
      "loss": 1.3189,
      "theoretical_loss": 3.3980828047012146,
      "tokens_seen": 2489843712
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002482246740220662,
      "loss": 1.2603,
      "theoretical_loss": 3.398075724345495,
      "tokens_seen": 2489909248
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024820461384152456,
      "loss": 1.3636,
      "theoretical_loss": 3.398068644228312,
      "tokens_seen": 2489974784
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002481845536609829,
      "loss": 1.3411,
      "theoretical_loss": 3.398061564349651,
      "tokens_seen": 2490040320
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024816449348044134,
      "loss": 1.2931,
      "theoretical_loss": 3.3980544847094976,
      "tokens_seen": 2490105856
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002481444332998997,
      "loss": 1.4103,
      "theoretical_loss": 3.3980474053078367,
      "tokens_seen": 2490171392
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024812437311935807,
      "loss": 1.2463,
      "theoretical_loss": 3.3980403261446552,
      "tokens_seen": 2490236928
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002481043129388165,
      "loss": 1.4068,
      "theoretical_loss": 3.398033247219938,
      "tokens_seen": 2490302464
    },
    {
      "epoch": 0.75,
      "objective/train/advantage_avg": 0.49699094891548157,
      "objective/train/docs_used": 906478,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2137683629989624,
      "objective/train/original_loss": 1.2137683629989624,
      "objective/train/theoretical_loss": 3.3980261685336712,
      "objective/train/tokens_used": 2510828000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24724505841732025,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050956130027771,
      "objective/train/weighted_lm_loss": 1.275539755821228,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3980261685336712,
      "tokens_seen": 2490368000
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024808425275827485,
      "loss": 1.3534,
      "theoretical_loss": 3.3980261685336712,
      "tokens_seen": 2490368000
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002480641925777332,
      "loss": 1.3324,
      "theoretical_loss": 3.3980190900858402,
      "tokens_seen": 2490433536
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024804413239719157,
      "loss": 1.3816,
      "theoretical_loss": 3.398012011876431,
      "tokens_seen": 2490499072
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024802407221665,
      "loss": 1.2427,
      "theoretical_loss": 3.398004933905429,
      "tokens_seen": 2490564608
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024800401203610835,
      "loss": 1.2912,
      "theoretical_loss": 3.39799785617282,
      "tokens_seen": 2490630144
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002479839518555667,
      "loss": 1.4194,
      "theoretical_loss": 3.39799077867859,
      "tokens_seen": 2490695680
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002479638916750251,
      "loss": 1.2943,
      "theoretical_loss": 3.3979837014227243,
      "tokens_seen": 2490761216
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024794383149448344,
      "loss": 1.4161,
      "theoretical_loss": 3.3979766244052083,
      "tokens_seen": 2490826752
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024792377131394185,
      "loss": 1.3614,
      "theoretical_loss": 3.3979695476260288,
      "tokens_seen": 2490892288
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002479037111334002,
      "loss": 1.4165,
      "theoretical_loss": 3.3979624710851706,
      "tokens_seen": 2490957824
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002478836509528586,
      "loss": 1.4117,
      "theoretical_loss": 3.3979553947826195,
      "tokens_seen": 2491023360
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024786359077231694,
      "loss": 1.3574,
      "theoretical_loss": 3.3979483187183614,
      "tokens_seen": 2491088896
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024784353059177536,
      "loss": 1.3297,
      "theoretical_loss": 3.397941242892382,
      "tokens_seen": 2491154432
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002478234704112337,
      "loss": 1.4557,
      "theoretical_loss": 3.397934167304667,
      "tokens_seen": 2491219968
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002478034102306921,
      "loss": 1.4354,
      "theoretical_loss": 3.397927091955202,
      "tokens_seen": 2491285504
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024778335005015044,
      "loss": 1.4712,
      "theoretical_loss": 3.3979200168439725,
      "tokens_seen": 2491351040
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.00024776328986960886,
      "loss": 1.3918,
      "theoretical_loss": 3.397912941970965,
      "tokens_seen": 2491416576
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002477432296890672,
      "loss": 1.3868,
      "theoretical_loss": 3.3979058673361644,
      "tokens_seen": 2491482112
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002477231695085256,
      "loss": 1.3157,
      "theoretical_loss": 3.3978987929395568,
      "tokens_seen": 2491547648
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024770310932798395,
      "loss": 1.3658,
      "theoretical_loss": 3.3978917187811275,
      "tokens_seen": 2491613184
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002476830491474423,
      "loss": 1.5501,
      "theoretical_loss": 3.3978846448608633,
      "tokens_seen": 2491678720
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002476629889669007,
      "loss": 1.3444,
      "theoretical_loss": 3.3978775711787486,
      "tokens_seen": 2491744256
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002476429287863591,
      "loss": 1.4137,
      "theoretical_loss": 3.39787049773477,
      "tokens_seen": 2491809792
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024762286860581745,
      "loss": 1.3903,
      "theoretical_loss": 3.3978634245289125,
      "tokens_seen": 2491875328
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002476028084252758,
      "loss": 1.3878,
      "theoretical_loss": 3.3978563515611624,
      "tokens_seen": 2491940864
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024758274824473423,
      "loss": 1.3608,
      "theoretical_loss": 3.3978492788315053,
      "tokens_seen": 2492006400
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002475626880641926,
      "loss": 1.4133,
      "theoretical_loss": 3.3978422063399267,
      "tokens_seen": 2492071936
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024754262788365095,
      "loss": 1.3729,
      "theoretical_loss": 3.397835134086413,
      "tokens_seen": 2492137472
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002475225677031093,
      "loss": 1.43,
      "theoretical_loss": 3.397828062070949,
      "tokens_seen": 2492203008
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002475025075225677,
      "loss": 1.3277,
      "theoretical_loss": 3.397820990293521,
      "tokens_seen": 2492268544
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002474824473420261,
      "loss": 1.4495,
      "theoretical_loss": 3.3978139187541143,
      "tokens_seen": 2492334080
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024746238716148445,
      "loss": 1.2547,
      "theoretical_loss": 3.397806847452715,
      "tokens_seen": 2492399616
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002474423269809428,
      "loss": 1.2428,
      "theoretical_loss": 3.3977997763893093,
      "tokens_seen": 2492465152
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002474222668004012,
      "loss": 1.4737,
      "theoretical_loss": 3.397792705563882,
      "tokens_seen": 2492530688
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002474022066198596,
      "loss": 1.3812,
      "theoretical_loss": 3.397785634976419,
      "tokens_seen": 2492596224
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024738214643931796,
      "loss": 1.4076,
      "theoretical_loss": 3.3977785646269063,
      "tokens_seen": 2492661760
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002473620862587764,
      "loss": 1.44,
      "theoretical_loss": 3.3977714945153297,
      "tokens_seen": 2492727296
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024734202607823474,
      "loss": 1.3478,
      "theoretical_loss": 3.397764424641675,
      "tokens_seen": 2492792832
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002473219658976931,
      "loss": 1.3399,
      "theoretical_loss": 3.3977573550059272,
      "tokens_seen": 2492858368
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024730190571715146,
      "loss": 1.4173,
      "theoretical_loss": 3.3977502856080735,
      "tokens_seen": 2492923904
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002472818455366098,
      "loss": 1.2876,
      "theoretical_loss": 3.397743216448098,
      "tokens_seen": 2492989440
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024726178535606824,
      "loss": 1.3347,
      "theoretical_loss": 3.3977361475259875,
      "tokens_seen": 2493054976
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002472417251755266,
      "loss": 1.4549,
      "theoretical_loss": 3.397729078841727,
      "tokens_seen": 2493120512
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024722166499498496,
      "loss": 1.3467,
      "theoretical_loss": 3.397722010395303,
      "tokens_seen": 2493186048
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002472016048144433,
      "loss": 1.4042,
      "theoretical_loss": 3.3977149421867012,
      "tokens_seen": 2493251584
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024718154463390174,
      "loss": 1.4241,
      "theoretical_loss": 3.397707874215907,
      "tokens_seen": 2493317120
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002471614844533601,
      "loss": 1.3516,
      "theoretical_loss": 3.3977008064829057,
      "tokens_seen": 2493382656
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024714142427281847,
      "loss": 1.4737,
      "theoretical_loss": 3.397693738987684,
      "tokens_seen": 2493448192
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024712136409227683,
      "loss": 1.4058,
      "theoretical_loss": 3.397686671730227,
      "tokens_seen": 2493513728
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002471013039117352,
      "loss": 1.2834,
      "theoretical_loss": 3.397679604710521,
      "tokens_seen": 2493579264
    },
    {
      "epoch": 0.76,
      "objective/train/advantage_avg": 0.4931562542915344,
      "objective/train/docs_used": 907814,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2678353786468506,
      "objective/train/original_loss": 1.267835259437561,
      "objective/train/theoretical_loss": 3.3976725379285506,
      "objective/train/tokens_used": 2514104800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24355097115039825,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050553798675537,
      "objective/train/weighted_lm_loss": 1.3317376375198364,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3976725379285506,
      "tokens_seen": 2493644800
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002470812437311936,
      "loss": 1.3706,
      "theoretical_loss": 3.3976725379285506,
      "tokens_seen": 2493644800
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024706118355065197,
      "loss": 1.4019,
      "theoretical_loss": 3.397665471384303,
      "tokens_seen": 2493710336
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024704112337011033,
      "loss": 1.3809,
      "theoretical_loss": 3.3976584050777636,
      "tokens_seen": 2493775872
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002470210631895687,
      "loss": 1.2416,
      "theoretical_loss": 3.3976513390089176,
      "tokens_seen": 2493841408
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002470010030090271,
      "loss": 1.3581,
      "theoretical_loss": 3.3976442731777508,
      "tokens_seen": 2493906944
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002469809428284855,
      "loss": 1.3485,
      "theoretical_loss": 3.3976372075842494,
      "tokens_seen": 2493972480
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024696088264794384,
      "loss": 1.3823,
      "theoretical_loss": 3.397630142228399,
      "tokens_seen": 2494038016
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002469408224674022,
      "loss": 1.3751,
      "theoretical_loss": 3.397623077110185,
      "tokens_seen": 2494103552
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024692076228686056,
      "loss": 1.3931,
      "theoretical_loss": 3.3976160122295944,
      "tokens_seen": 2494169088
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000246900702106319,
      "loss": 1.2801,
      "theoretical_loss": 3.3976089475866114,
      "tokens_seen": 2494234624
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024688064192577734,
      "loss": 1.383,
      "theoretical_loss": 3.3976018831812222,
      "tokens_seen": 2494300160
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002468605817452357,
      "loss": 1.3462,
      "theoretical_loss": 3.3975948190134133,
      "tokens_seen": 2494365696
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024684052156469406,
      "loss": 1.4577,
      "theoretical_loss": 3.39758775508317,
      "tokens_seen": 2494431232
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002468204613841525,
      "loss": 1.3491,
      "theoretical_loss": 3.3975806913904774,
      "tokens_seen": 2494496768
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024680040120361084,
      "loss": 1.3267,
      "theoretical_loss": 3.3975736279353224,
      "tokens_seen": 2494562304
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002467803410230692,
      "loss": 1.296,
      "theoretical_loss": 3.39756656471769,
      "tokens_seen": 2494627840
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024676028084252757,
      "loss": 1.2136,
      "theoretical_loss": 3.3975595017375664,
      "tokens_seen": 2494693376
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024674022066198593,
      "loss": 1.3989,
      "theoretical_loss": 3.3975524389949374,
      "tokens_seen": 2494758912
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024672016048144435,
      "loss": 1.3071,
      "theoretical_loss": 3.3975453764897883,
      "tokens_seen": 2494824448
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002467001003009027,
      "loss": 1.3621,
      "theoretical_loss": 3.397538314222105,
      "tokens_seen": 2494889984
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024668004012036107,
      "loss": 1.3493,
      "theoretical_loss": 3.397531252191874,
      "tokens_seen": 2494955520
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024665997993981943,
      "loss": 1.3521,
      "theoretical_loss": 3.3975241903990803,
      "tokens_seen": 2495021056
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024663991975927785,
      "loss": 1.3278,
      "theoretical_loss": 3.3975171288437096,
      "tokens_seen": 2495086592
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002466198595787362,
      "loss": 1.3636,
      "theoretical_loss": 3.397510067525748,
      "tokens_seen": 2495152128
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024659979939819463,
      "loss": 1.3695,
      "theoretical_loss": 3.397503006445182,
      "tokens_seen": 2495217664
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000246579739217653,
      "loss": 1.2823,
      "theoretical_loss": 3.397495945601996,
      "tokens_seen": 2495283200
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024655967903711135,
      "loss": 1.3273,
      "theoretical_loss": 3.397488884996177,
      "tokens_seen": 2495348736
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002465396188565697,
      "loss": 1.3452,
      "theoretical_loss": 3.3974818246277096,
      "tokens_seen": 2495414272
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002465195586760281,
      "loss": 1.3746,
      "theoretical_loss": 3.3974747644965806,
      "tokens_seen": 2495479808
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002464994984954865,
      "loss": 1.3968,
      "theoretical_loss": 3.397467704602775,
      "tokens_seen": 2495545344
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024647943831494486,
      "loss": 1.3228,
      "theoretical_loss": 3.3974606449462796,
      "tokens_seen": 2495610880
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002464593781344032,
      "loss": 1.4128,
      "theoretical_loss": 3.397453585527079,
      "tokens_seen": 2495676416
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002464393179538616,
      "loss": 1.3111,
      "theoretical_loss": 3.39744652634516,
      "tokens_seen": 2495741952
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024641925777332,
      "loss": 1.3539,
      "theoretical_loss": 3.397439467400508,
      "tokens_seen": 2495807488
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024639919759277836,
      "loss": 1.3404,
      "theoretical_loss": 3.3974324086931085,
      "tokens_seen": 2495873024
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002463791374122367,
      "loss": 1.4095,
      "theoretical_loss": 3.3974253502229477,
      "tokens_seen": 2495938560
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002463590772316951,
      "loss": 1.3151,
      "theoretical_loss": 3.397418291990011,
      "tokens_seen": 2496004096
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024633901705115345,
      "loss": 1.2933,
      "theoretical_loss": 3.397411233994285,
      "tokens_seen": 2496069632
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024631895687061186,
      "loss": 1.4704,
      "theoretical_loss": 3.3974041762357547,
      "tokens_seen": 2496135168
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002462988966900702,
      "loss": 1.4086,
      "theoretical_loss": 3.397397118714406,
      "tokens_seen": 2496200704
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002462788365095286,
      "loss": 1.3524,
      "theoretical_loss": 3.397390061430225,
      "tokens_seen": 2496266240
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024625877632898695,
      "loss": 1.3536,
      "theoretical_loss": 3.3973830043831974,
      "tokens_seen": 2496331776
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024623871614844536,
      "loss": 1.3243,
      "theoretical_loss": 3.397375947573309,
      "tokens_seen": 2496397312
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002462186559679037,
      "loss": 1.3076,
      "theoretical_loss": 3.3973688910005455,
      "tokens_seen": 2496462848
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002461985957873621,
      "loss": 1.412,
      "theoretical_loss": 3.397361834664893,
      "tokens_seen": 2496528384
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024617853560682045,
      "loss": 1.2632,
      "theoretical_loss": 3.3973547785663367,
      "tokens_seen": 2496593920
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002461584754262788,
      "loss": 1.3722,
      "theoretical_loss": 3.397347722704863,
      "tokens_seen": 2496659456
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024613841524573723,
      "loss": 1.4076,
      "theoretical_loss": 3.3973406670804573,
      "tokens_seen": 2496724992
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002461183550651956,
      "loss": 1.3167,
      "theoretical_loss": 3.397333611693106,
      "tokens_seen": 2496790528
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024609829488465395,
      "loss": 1.4401,
      "theoretical_loss": 3.3973265565427946,
      "tokens_seen": 2496856064
    },
    {
      "epoch": 0.76,
      "objective/train/advantage_avg": 0.49027153849601746,
      "objective/train/docs_used": 908925,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2945362329483032,
      "objective/train/original_loss": 1.2945363521575928,
      "objective/train/theoretical_loss": 3.397319501629509,
      "objective/train/tokens_used": 2517381600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.240862637758255,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502514839172363,
      "objective/train/weighted_lm_loss": 1.3596596717834473,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0125783681869507,
      "theoretical_loss": 3.397319501629509,
      "tokens_seen": 2496921600
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002460782347041123,
      "loss": 1.3313,
      "theoretical_loss": 3.397319501629509,
      "tokens_seen": 2496921600
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024605817452357073,
      "loss": 1.2922,
      "theoretical_loss": 3.3973124469532343,
      "tokens_seen": 2496987136
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002460381143430291,
      "loss": 1.3547,
      "theoretical_loss": 3.397305392513957,
      "tokens_seen": 2497052672
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024601805416248746,
      "loss": 1.358,
      "theoretical_loss": 3.3972983383116633,
      "tokens_seen": 2497118208
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002459979939819458,
      "loss": 1.3508,
      "theoretical_loss": 3.397291284346338,
      "tokens_seen": 2497183744
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002459779338014042,
      "loss": 1.4135,
      "theoretical_loss": 3.397284230617968,
      "tokens_seen": 2497249280
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002459578736208626,
      "loss": 1.2752,
      "theoretical_loss": 3.397277177126538,
      "tokens_seen": 2497314816
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024593781344032096,
      "loss": 1.4193,
      "theoretical_loss": 3.397270123872035,
      "tokens_seen": 2497380352
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002459177532597793,
      "loss": 1.2955,
      "theoretical_loss": 3.397263070854444,
      "tokens_seen": 2497445888
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002458976930792377,
      "loss": 1.4002,
      "theoretical_loss": 3.3972560180737505,
      "tokens_seen": 2497511424
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002458776328986961,
      "loss": 1.3191,
      "theoretical_loss": 3.3972489655299416,
      "tokens_seen": 2497576960
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024585757271815446,
      "loss": 1.3634,
      "theoretical_loss": 3.397241913223002,
      "tokens_seen": 2497642496
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002458375125376129,
      "loss": 1.3721,
      "theoretical_loss": 3.397234861152918,
      "tokens_seen": 2497708032
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024581745235707124,
      "loss": 1.3905,
      "theoretical_loss": 3.3972278093196753,
      "tokens_seen": 2497773568
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002457973921765296,
      "loss": 1.3233,
      "theoretical_loss": 3.39722075772326,
      "tokens_seen": 2497839104
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024577733199598797,
      "loss": 1.4271,
      "theoretical_loss": 3.3972137063636576,
      "tokens_seen": 2497904640
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024575727181544633,
      "loss": 1.3418,
      "theoretical_loss": 3.397206655240854,
      "tokens_seen": 2497970176
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024573721163490475,
      "loss": 1.3733,
      "theoretical_loss": 3.397199604354835,
      "tokens_seen": 2498035712
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002457171514543631,
      "loss": 1.4042,
      "theoretical_loss": 3.397192553705587,
      "tokens_seen": 2498101248
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024569709127382147,
      "loss": 1.2492,
      "theoretical_loss": 3.3971855032930955,
      "tokens_seen": 2498166784
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024567703109327983,
      "loss": 1.2598,
      "theoretical_loss": 3.3971784531173457,
      "tokens_seen": 2498232320
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024565697091273825,
      "loss": 1.3484,
      "theoretical_loss": 3.397171403178324,
      "tokens_seen": 2498297856
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002456369107321966,
      "loss": 1.3998,
      "theoretical_loss": 3.3971643534760165,
      "tokens_seen": 2498363392
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000245616850551655,
      "loss": 1.258,
      "theoretical_loss": 3.3971573040104084,
      "tokens_seen": 2498428928
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024559679037111334,
      "loss": 1.2647,
      "theoretical_loss": 3.397150254781486,
      "tokens_seen": 2498494464
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002455767301905717,
      "loss": 1.4487,
      "theoretical_loss": 3.3971432057892352,
      "tokens_seen": 2498560000
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002455566700100301,
      "loss": 1.4521,
      "theoretical_loss": 3.3971361570336414,
      "tokens_seen": 2498625536
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002455366098294885,
      "loss": 1.2531,
      "theoretical_loss": 3.397129108514691,
      "tokens_seen": 2498691072
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024551654964894684,
      "loss": 1.4001,
      "theoretical_loss": 3.3971220602323693,
      "tokens_seen": 2498756608
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002454964894684052,
      "loss": 1.4036,
      "theoretical_loss": 3.3971150121866627,
      "tokens_seen": 2498822144
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002454764292878636,
      "loss": 1.3591,
      "theoretical_loss": 3.3971079643775566,
      "tokens_seen": 2498887680
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000245456369107322,
      "loss": 1.4354,
      "theoretical_loss": 3.397100916805037,
      "tokens_seen": 2498953216
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024543630892678034,
      "loss": 1.3194,
      "theoretical_loss": 3.3970938694690895,
      "tokens_seen": 2499018752
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002454162487462387,
      "loss": 1.3417,
      "theoretical_loss": 3.3970868223697006,
      "tokens_seen": 2499084288
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024539618856569707,
      "loss": 1.2017,
      "theoretical_loss": 3.397079775506856,
      "tokens_seen": 2499149824
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002453761283851555,
      "loss": 1.4023,
      "theoretical_loss": 3.397072728880541,
      "tokens_seen": 2499215360
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024535606820461385,
      "loss": 1.3207,
      "theoretical_loss": 3.3970656824907417,
      "tokens_seen": 2499280896
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002453360080240722,
      "loss": 1.3112,
      "theoretical_loss": 3.3970586363374444,
      "tokens_seen": 2499346432
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024531594784353057,
      "loss": 1.3217,
      "theoretical_loss": 3.3970515904206344,
      "tokens_seen": 2499411968
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000245295887662989,
      "loss": 1.3568,
      "theoretical_loss": 3.397044544740298,
      "tokens_seen": 2499477504
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024527582748244735,
      "loss": 1.3786,
      "theoretical_loss": 3.3970374992964207,
      "tokens_seen": 2499543040
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002452557673019057,
      "loss": 1.2628,
      "theoretical_loss": 3.3970304540889886,
      "tokens_seen": 2499608576
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002452357071213641,
      "loss": 1.417,
      "theoretical_loss": 3.3970234091179874,
      "tokens_seen": 2499674112
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024521564694082244,
      "loss": 1.4031,
      "theoretical_loss": 3.3970163643834033,
      "tokens_seen": 2499739648
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024519558676028085,
      "loss": 1.4084,
      "theoretical_loss": 3.3970093198852216,
      "tokens_seen": 2499805184
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002451755265797392,
      "loss": 1.3975,
      "theoretical_loss": 3.3970022756234286,
      "tokens_seen": 2499870720
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002451554663991976,
      "loss": 1.4449,
      "theoretical_loss": 3.39699523159801,
      "tokens_seen": 2499936256
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024513540621865594,
      "loss": 1.3173,
      "theoretical_loss": 3.396988187808952,
      "tokens_seen": 2500001792
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024511534603811435,
      "loss": 1.3868,
      "theoretical_loss": 3.3969811442562396,
      "tokens_seen": 2500067328
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002450952858575727,
      "loss": 1.3342,
      "theoretical_loss": 3.3969741009398597,
      "tokens_seen": 2500132864
    },
    {
      "epoch": 0.76,
      "objective/train/advantage_avg": 0.4939713478088379,
      "objective/train/docs_used": 910166,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.221418023109436,
      "objective/train/original_loss": 1.2214179039001465,
      "objective/train/theoretical_loss": 3.396967057859798,
      "objective/train/tokens_used": 2520658400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2442205399274826,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506387948989868,
      "objective/train/weighted_lm_loss": 1.2832096815109253,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.396967057859798,
      "tokens_seen": 2500198400
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024507522567703113,
      "loss": 1.3148,
      "theoretical_loss": 3.396967057859798,
      "tokens_seen": 2500198400
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002450551654964895,
      "loss": 1.2087,
      "theoretical_loss": 3.39696001501604,
      "tokens_seen": 2500263936
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024503510531594786,
      "loss": 1.3278,
      "theoretical_loss": 3.396952972408571,
      "tokens_seen": 2500329472
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002450150451354062,
      "loss": 1.4017,
      "theoretical_loss": 3.3969459300373783,
      "tokens_seen": 2500395008
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002449949849548646,
      "loss": 1.3474,
      "theoretical_loss": 3.396938887902447,
      "tokens_seen": 2500460544
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000244974924774323,
      "loss": 1.216,
      "theoretical_loss": 3.3969318460037634,
      "tokens_seen": 2500526080
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024495486459378136,
      "loss": 1.4634,
      "theoretical_loss": 3.3969248043413125,
      "tokens_seen": 2500591616
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002449348044132397,
      "loss": 1.3059,
      "theoretical_loss": 3.3969177629150806,
      "tokens_seen": 2500657152
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002449147442326981,
      "loss": 1.2993,
      "theoretical_loss": 3.396910721725054,
      "tokens_seen": 2500722688
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002448946840521565,
      "loss": 1.2889,
      "theoretical_loss": 3.3969036807712185,
      "tokens_seen": 2500788224
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024487462387161486,
      "loss": 1.4096,
      "theoretical_loss": 3.3968966400535594,
      "tokens_seen": 2500853760
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002448545636910732,
      "loss": 1.3534,
      "theoretical_loss": 3.396889599572063,
      "tokens_seen": 2500919296
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002448345035105316,
      "loss": 1.317,
      "theoretical_loss": 3.3968825593267153,
      "tokens_seen": 2500984832
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024481444332998995,
      "loss": 1.3652,
      "theoretical_loss": 3.3968755193175024,
      "tokens_seen": 2501050368
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024479438314944837,
      "loss": 1.3235,
      "theoretical_loss": 3.396868479544409,
      "tokens_seen": 2501115904
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024477432296890673,
      "loss": 1.4154,
      "theoretical_loss": 3.396861440007423,
      "tokens_seen": 2501181440
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002447542627883651,
      "loss": 1.2825,
      "theoretical_loss": 3.3968544007065278,
      "tokens_seen": 2501246976
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024473420260782345,
      "loss": 1.2679,
      "theoretical_loss": 3.3968473616417114,
      "tokens_seen": 2501312512
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024471414242728187,
      "loss": 1.2576,
      "theoretical_loss": 3.3968403228129587,
      "tokens_seen": 2501378048
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024469408224674023,
      "loss": 1.2852,
      "theoretical_loss": 3.396833284220256,
      "tokens_seen": 2501443584
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002446740220661986,
      "loss": 1.2228,
      "theoretical_loss": 3.396826245863589,
      "tokens_seen": 2501509120
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024465396188565696,
      "loss": 1.2493,
      "theoretical_loss": 3.3968192077429435,
      "tokens_seen": 2501574656
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002446339017051154,
      "loss": 1.3242,
      "theoretical_loss": 3.396812169858306,
      "tokens_seen": 2501640192
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024461384152457374,
      "loss": 1.3762,
      "theoretical_loss": 3.396805132209661,
      "tokens_seen": 2501705728
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002445937813440321,
      "loss": 1.3743,
      "theoretical_loss": 3.396798094796996,
      "tokens_seen": 2501771264
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024457372116349046,
      "loss": 1.2388,
      "theoretical_loss": 3.396791057620296,
      "tokens_seen": 2501836800
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002445536609829488,
      "loss": 1.3683,
      "theoretical_loss": 3.396784020679547,
      "tokens_seen": 2501902336
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024453360080240724,
      "loss": 1.3329,
      "theoretical_loss": 3.3967769839747355,
      "tokens_seen": 2501967872
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002445135406218656,
      "loss": 1.4435,
      "theoretical_loss": 3.396769947505847,
      "tokens_seen": 2502033408
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024449348044132396,
      "loss": 1.316,
      "theoretical_loss": 3.3967629112728672,
      "tokens_seen": 2502098944
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002444734202607823,
      "loss": 1.3819,
      "theoretical_loss": 3.3967558752757823,
      "tokens_seen": 2502164480
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024445336008024074,
      "loss": 1.298,
      "theoretical_loss": 3.3967488395145775,
      "tokens_seen": 2502230016
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002444332998996991,
      "loss": 1.4388,
      "theoretical_loss": 3.39674180398924,
      "tokens_seen": 2502295552
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024441323971915747,
      "loss": 1.4579,
      "theoretical_loss": 3.396734768699755,
      "tokens_seen": 2502361088
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024439317953861583,
      "loss": 1.4526,
      "theoretical_loss": 3.396727733646108,
      "tokens_seen": 2502426624
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002443731193580742,
      "loss": 1.3932,
      "theoretical_loss": 3.396720698828286,
      "tokens_seen": 2502492160
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002443530591775326,
      "loss": 1.3051,
      "theoretical_loss": 3.3967136642462736,
      "tokens_seen": 2502557696
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024433299899699097,
      "loss": 1.3191,
      "theoretical_loss": 3.396706629900058,
      "tokens_seen": 2502623232
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002443129388164494,
      "loss": 1.3818,
      "theoretical_loss": 3.396699595789624,
      "tokens_seen": 2502688768
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024429287863590775,
      "loss": 1.3693,
      "theoretical_loss": 3.3966925619149584,
      "tokens_seen": 2502754304
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002442728184553661,
      "loss": 1.3714,
      "theoretical_loss": 3.3966855282760466,
      "tokens_seen": 2502819840
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002442527582748245,
      "loss": 1.3942,
      "theoretical_loss": 3.3966784948728748,
      "tokens_seen": 2502885376
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024423269809428284,
      "loss": 1.3826,
      "theoretical_loss": 3.396671461705429,
      "tokens_seen": 2502950912
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024421263791374125,
      "loss": 1.2961,
      "theoretical_loss": 3.396664428773695,
      "tokens_seen": 2503016448
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002441925777331996,
      "loss": 1.3394,
      "theoretical_loss": 3.3966573960776585,
      "tokens_seen": 2503081984
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000244172517552658,
      "loss": 1.3721,
      "theoretical_loss": 3.3966503636173053,
      "tokens_seen": 2503147520
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024415245737211634,
      "loss": 1.3773,
      "theoretical_loss": 3.396643331392622,
      "tokens_seen": 2503213056
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024413239719157473,
      "loss": 1.4379,
      "theoretical_loss": 3.3966362994035944,
      "tokens_seen": 2503278592
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024411233701103312,
      "loss": 1.3718,
      "theoretical_loss": 3.3966292676502077,
      "tokens_seen": 2503344128
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024409227683049148,
      "loss": 1.3047,
      "theoretical_loss": 3.3966222361324485,
      "tokens_seen": 2503409664
    },
    {
      "epoch": 0.76,
      "objective/train/advantage_avg": 0.4895636737346649,
      "objective/train/docs_used": 911421,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3965566158294678,
      "objective/train/original_loss": 1.3965566158294678,
      "objective/train/theoretical_loss": 3.3966152048503027,
      "objective/train/tokens_used": 2523935200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24006232619285583,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0501766204833984,
      "objective/train/weighted_lm_loss": 1.46641206741333,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3966152048503027,
      "tokens_seen": 2503475200
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024407221664994987,
      "loss": 1.4043,
      "theoretical_loss": 3.3966152048503027,
      "tokens_seen": 2503475200
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024405215646940823,
      "loss": 1.3466,
      "theoretical_loss": 3.3966081738037563,
      "tokens_seen": 2503540736
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002440320962888666,
      "loss": 1.2678,
      "theoretical_loss": 3.396601142992795,
      "tokens_seen": 2503606272
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024401203610832498,
      "loss": 1.2752,
      "theoretical_loss": 3.3965941124174046,
      "tokens_seen": 2503671808
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024399197592778335,
      "loss": 1.3005,
      "theoretical_loss": 3.396587082077571,
      "tokens_seen": 2503737344
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024397191574724173,
      "loss": 1.3842,
      "theoretical_loss": 3.396580051973281,
      "tokens_seen": 2503802880
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002439518555667001,
      "loss": 1.3392,
      "theoretical_loss": 3.3965730221045196,
      "tokens_seen": 2503868416
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024393179538615849,
      "loss": 1.3726,
      "theoretical_loss": 3.396565992471273,
      "tokens_seen": 2503933952
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024391173520561685,
      "loss": 1.4574,
      "theoretical_loss": 3.3965589630735273,
      "tokens_seen": 2503999488
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024389167502507524,
      "loss": 1.3057,
      "theoretical_loss": 3.3965519339112684,
      "tokens_seen": 2504065024
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002438716148445336,
      "loss": 1.2926,
      "theoretical_loss": 3.3965449049844825,
      "tokens_seen": 2504130560
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024385155466399196,
      "loss": 1.2415,
      "theoretical_loss": 3.396537876293155,
      "tokens_seen": 2504196096
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024383149448345035,
      "loss": 1.3154,
      "theoretical_loss": 3.3965308478372718,
      "tokens_seen": 2504261632
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024381143430290871,
      "loss": 1.3069,
      "theoretical_loss": 3.3965238196168195,
      "tokens_seen": 2504327168
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002437913741223671,
      "loss": 1.3421,
      "theoretical_loss": 3.3965167916317833,
      "tokens_seen": 2504392704
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024377131394182547,
      "loss": 1.4225,
      "theoretical_loss": 3.3965097638821504,
      "tokens_seen": 2504458240
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024375125376128385,
      "loss": 1.3408,
      "theoretical_loss": 3.3965027363679052,
      "tokens_seen": 2504523776
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024373119358074222,
      "loss": 1.3653,
      "theoretical_loss": 3.396495709089035,
      "tokens_seen": 2504589312
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002437111334002006,
      "loss": 1.2519,
      "theoretical_loss": 3.3964886820455247,
      "tokens_seen": 2504654848
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024369107321965897,
      "loss": 1.3715,
      "theoretical_loss": 3.396481655237361,
      "tokens_seen": 2504720384
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024367101303911736,
      "loss": 1.4093,
      "theoretical_loss": 3.3964746286645293,
      "tokens_seen": 2504785920
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024365095285857572,
      "loss": 1.4161,
      "theoretical_loss": 3.3964676023270157,
      "tokens_seen": 2504851456
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024363089267803408,
      "loss": 1.3363,
      "theoretical_loss": 3.396460576224807,
      "tokens_seen": 2504916992
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002436108324974925,
      "loss": 1.2699,
      "theoretical_loss": 3.3964535503578874,
      "tokens_seen": 2504982528
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024359077231695086,
      "loss": 1.1938,
      "theoretical_loss": 3.3964465247262448,
      "tokens_seen": 2505048064
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024357071213640925,
      "loss": 1.4629,
      "theoretical_loss": 3.3964394993298637,
      "tokens_seen": 2505113600
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002435506519558676,
      "loss": 1.3464,
      "theoretical_loss": 3.396432474168731,
      "tokens_seen": 2505179136
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000243530591775326,
      "loss": 1.3041,
      "theoretical_loss": 3.396425449242832,
      "tokens_seen": 2505244672
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024351053159478436,
      "loss": 1.2481,
      "theoretical_loss": 3.3964184245521536,
      "tokens_seen": 2505310208
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024349047141424275,
      "loss": 1.3911,
      "theoretical_loss": 3.3964114000966807,
      "tokens_seen": 2505375744
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024347041123370112,
      "loss": 1.3311,
      "theoretical_loss": 3.3964043758763998,
      "tokens_seen": 2505441280
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024345035105315948,
      "loss": 1.4589,
      "theoretical_loss": 3.396397351891297,
      "tokens_seen": 2505506816
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024343029087261787,
      "loss": 1.404,
      "theoretical_loss": 3.396390328141358,
      "tokens_seen": 2505572352
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024341023069207623,
      "loss": 1.4199,
      "theoretical_loss": 3.3963833046265686,
      "tokens_seen": 2505637888
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024339017051153462,
      "loss": 1.2668,
      "theoretical_loss": 3.396376281346915,
      "tokens_seen": 2505703424
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024337011033099298,
      "loss": 1.4121,
      "theoretical_loss": 3.3963692583023835,
      "tokens_seen": 2505768960
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024335005015045137,
      "loss": 1.325,
      "theoretical_loss": 3.3963622354929597,
      "tokens_seen": 2505834496
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024332998996990973,
      "loss": 1.3202,
      "theoretical_loss": 3.3963552129186296,
      "tokens_seen": 2505900032
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024330992978936812,
      "loss": 1.3785,
      "theoretical_loss": 3.3963481905793795,
      "tokens_seen": 2505965568
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024328986960882648,
      "loss": 1.3687,
      "theoretical_loss": 3.396341168475195,
      "tokens_seen": 2506031104
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024326980942828487,
      "loss": 1.3313,
      "theoretical_loss": 3.396334146606062,
      "tokens_seen": 2506096640
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024324974924774324,
      "loss": 1.4606,
      "theoretical_loss": 3.3963271249719673,
      "tokens_seen": 2506162176
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002432296890672016,
      "loss": 1.3983,
      "theoretical_loss": 3.396320103572896,
      "tokens_seen": 2506227712
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024320962888666,
      "loss": 1.1791,
      "theoretical_loss": 3.396313082408834,
      "tokens_seen": 2506293248
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024318956870611835,
      "loss": 1.2031,
      "theoretical_loss": 3.396306061479768,
      "tokens_seen": 2506358784
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024316950852557674,
      "loss": 1.4747,
      "theoretical_loss": 3.3962990407856837,
      "tokens_seen": 2506424320
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002431494483450351,
      "loss": 1.3525,
      "theoretical_loss": 3.396292020326567,
      "tokens_seen": 2506489856
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002431293881644935,
      "loss": 1.3927,
      "theoretical_loss": 3.396285000102404,
      "tokens_seen": 2506555392
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024310932798395185,
      "loss": 1.3329,
      "theoretical_loss": 3.3962779801131804,
      "tokens_seen": 2506620928
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024308926780341024,
      "loss": 1.3601,
      "theoretical_loss": 3.396270960358883,
      "tokens_seen": 2506686464
    },
    {
      "epoch": 0.76,
      "objective/train/advantage_avg": 0.4927324056625366,
      "objective/train/docs_used": 912553,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3673406839370728,
      "objective/train/original_loss": 1.3673408031463623,
      "objective/train/theoretical_loss": 3.396263940839497,
      "objective/train/tokens_used": 2527212000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2432987093925476,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050510048866272,
      "objective/train/weighted_lm_loss": 1.4364067316055298,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.396263940839497,
      "tokens_seen": 2506752000
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002430692076228686,
      "loss": 1.4214,
      "theoretical_loss": 3.396263940839497,
      "tokens_seen": 2506752000
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024304914744232697,
      "loss": 1.2727,
      "theoretical_loss": 3.3962569215550085,
      "tokens_seen": 2506817536
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024302908726178536,
      "loss": 1.4662,
      "theoretical_loss": 3.3962499025054034,
      "tokens_seen": 2506883072
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024300902708124372,
      "loss": 1.1925,
      "theoretical_loss": 3.396242883690668,
      "tokens_seen": 2506948608
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002429889669007021,
      "loss": 1.3576,
      "theoretical_loss": 3.3962358651107887,
      "tokens_seen": 2507014144
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024296890672016047,
      "loss": 1.3049,
      "theoretical_loss": 3.3962288467657507,
      "tokens_seen": 2507079680
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024294884653961886,
      "loss": 1.426,
      "theoretical_loss": 3.3962218286555403,
      "tokens_seen": 2507145216
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024292878635907722,
      "loss": 1.2552,
      "theoretical_loss": 3.3962148107801435,
      "tokens_seen": 2507210752
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002429087261785356,
      "loss": 1.2805,
      "theoretical_loss": 3.3962077931395465,
      "tokens_seen": 2507276288
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024288866599799397,
      "loss": 1.3306,
      "theoretical_loss": 3.3962007757337354,
      "tokens_seen": 2507341824
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024286860581745234,
      "loss": 1.2841,
      "theoretical_loss": 3.3961937585626956,
      "tokens_seen": 2507407360
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024284854563691075,
      "loss": 1.3266,
      "theoretical_loss": 3.3961867416264138,
      "tokens_seen": 2507472896
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024282848545636911,
      "loss": 1.3687,
      "theoretical_loss": 3.3961797249248753,
      "tokens_seen": 2507538432
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002428084252758275,
      "loss": 1.3158,
      "theoretical_loss": 3.3961727084580664,
      "tokens_seen": 2507603968
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024278836509528587,
      "loss": 1.4134,
      "theoretical_loss": 3.3961656922259738,
      "tokens_seen": 2507669504
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024276830491474426,
      "loss": 1.4106,
      "theoretical_loss": 3.3961586762285827,
      "tokens_seen": 2507735040
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024274824473420262,
      "loss": 1.3464,
      "theoretical_loss": 3.396151660465879,
      "tokens_seen": 2507800576
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000242728184553661,
      "loss": 1.3795,
      "theoretical_loss": 3.3961446449378494,
      "tokens_seen": 2507866112
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024270812437311937,
      "loss": 1.354,
      "theoretical_loss": 3.3961376296444796,
      "tokens_seen": 2507931648
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024268806419257776,
      "loss": 1.2854,
      "theoretical_loss": 3.3961306145857555,
      "tokens_seen": 2507997184
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024266800401203612,
      "loss": 1.4789,
      "theoretical_loss": 3.3961235997616632,
      "tokens_seen": 2508062720
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024264794383149448,
      "loss": 1.3419,
      "theoretical_loss": 3.3961165851721886,
      "tokens_seen": 2508128256
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024262788365095287,
      "loss": 1.3837,
      "theoretical_loss": 3.396109570817318,
      "tokens_seen": 2508193792
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024260782347041123,
      "loss": 1.4143,
      "theoretical_loss": 3.3961025566970373,
      "tokens_seen": 2508259328
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024258776328986962,
      "loss": 1.3671,
      "theoretical_loss": 3.3960955428113326,
      "tokens_seen": 2508324864
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024256770310932799,
      "loss": 1.2678,
      "theoretical_loss": 3.3960885291601897,
      "tokens_seen": 2508390400
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024254764292878638,
      "loss": 1.3619,
      "theoretical_loss": 3.3960815157435946,
      "tokens_seen": 2508455936
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024252758274824474,
      "loss": 1.3542,
      "theoretical_loss": 3.3960745025615338,
      "tokens_seen": 2508521472
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024250752256770313,
      "loss": 1.3201,
      "theoretical_loss": 3.396067489613993,
      "tokens_seen": 2508587008
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002424874623871615,
      "loss": 1.3802,
      "theoretical_loss": 3.396060476900958,
      "tokens_seen": 2508652544
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024246740220661985,
      "loss": 1.3346,
      "theoretical_loss": 3.3960534644224154,
      "tokens_seen": 2508718080
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024244734202607824,
      "loss": 1.4087,
      "theoretical_loss": 3.396046452178351,
      "tokens_seen": 2508783616
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002424272818455366,
      "loss": 1.3823,
      "theoretical_loss": 3.3960394401687504,
      "tokens_seen": 2508849152
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000242407221664995,
      "loss": 1.4289,
      "theoretical_loss": 3.3960324283936,
      "tokens_seen": 2508914688
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024238716148445335,
      "loss": 1.3205,
      "theoretical_loss": 3.3960254168528863,
      "tokens_seen": 2508980224
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024236710130391174,
      "loss": 1.2873,
      "theoretical_loss": 3.3960184055465943,
      "tokens_seen": 2509045760
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002423470411233701,
      "loss": 1.3248,
      "theoretical_loss": 3.396011394474711,
      "tokens_seen": 2509111296
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002423269809428285,
      "loss": 1.3329,
      "theoretical_loss": 3.3960043836372216,
      "tokens_seen": 2509176832
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024230692076228686,
      "loss": 1.2007,
      "theoretical_loss": 3.395997373034113,
      "tokens_seen": 2509242368
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024228686058174522,
      "loss": 1.3567,
      "theoretical_loss": 3.3959903626653705,
      "tokens_seen": 2509307904
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002422668004012036,
      "loss": 1.445,
      "theoretical_loss": 3.3959833525309806,
      "tokens_seen": 2509373440
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024224674022066197,
      "loss": 1.4103,
      "theoretical_loss": 3.3959763426309295,
      "tokens_seen": 2509438976
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024222668004012036,
      "loss": 1.454,
      "theoretical_loss": 3.395969332965203,
      "tokens_seen": 2509504512
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024220661985957872,
      "loss": 1.3337,
      "theoretical_loss": 3.395962323533787,
      "tokens_seen": 2509570048
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002421865596790371,
      "loss": 1.2817,
      "theoretical_loss": 3.395955314336667,
      "tokens_seen": 2509635584
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024216649949849547,
      "loss": 1.3345,
      "theoretical_loss": 3.3959483053738304,
      "tokens_seen": 2509701120
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024214643931795386,
      "loss": 1.4219,
      "theoretical_loss": 3.3959412966452627,
      "tokens_seen": 2509766656
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024212637913741223,
      "loss": 1.3719,
      "theoretical_loss": 3.3959342881509493,
      "tokens_seen": 2509832192
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024210631895687064,
      "loss": 1.2238,
      "theoretical_loss": 3.395927279890877,
      "tokens_seen": 2509897728
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000242086258776329,
      "loss": 1.4143,
      "theoretical_loss": 3.395920271865032,
      "tokens_seen": 2509963264
    },
    {
      "epoch": 0.76,
      "objective/train/advantage_avg": 0.4899941682815552,
      "objective/train/docs_used": 913305,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4074724912643433,
      "objective/train/original_loss": 1.4074723720550537,
      "objective/train/theoretical_loss": 3.3959132640733998,
      "objective/train/tokens_used": 2530488800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2409057915210724,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.05022394657135,
      "objective/train/weighted_lm_loss": 1.4781593084335327,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3959132640733998,
      "tokens_seen": 2510028800
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024206619859578737,
      "loss": 1.4285,
      "theoretical_loss": 3.3959132640733998,
      "tokens_seen": 2510028800
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024204613841524576,
      "loss": 1.4375,
      "theoretical_loss": 3.3959062565159663,
      "tokens_seen": 2510094336
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024202607823470412,
      "loss": 1.3464,
      "theoretical_loss": 3.395899249192718,
      "tokens_seen": 2510159872
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002420060180541625,
      "loss": 1.3622,
      "theoretical_loss": 3.395892242103641,
      "tokens_seen": 2510225408
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024198595787362087,
      "loss": 1.3691,
      "theoretical_loss": 3.395885235248721,
      "tokens_seen": 2510290944
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024196589769307926,
      "loss": 1.5016,
      "theoretical_loss": 3.3958782286279448,
      "tokens_seen": 2510356480
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024194583751253762,
      "loss": 1.2476,
      "theoretical_loss": 3.3958712222412974,
      "tokens_seen": 2510422016
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000241925777331996,
      "loss": 1.4411,
      "theoretical_loss": 3.3958642160887655,
      "tokens_seen": 2510487552
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024190571715145437,
      "loss": 1.3698,
      "theoretical_loss": 3.3958572101703353,
      "tokens_seen": 2510553088
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024188565697091274,
      "loss": 1.2983,
      "theoretical_loss": 3.395850204485993,
      "tokens_seen": 2510618624
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024186559679037113,
      "loss": 1.3032,
      "theoretical_loss": 3.3958431990357236,
      "tokens_seen": 2510684160
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002418455366098295,
      "loss": 1.2433,
      "theoretical_loss": 3.3958361938195143,
      "tokens_seen": 2510749696
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024182547642928788,
      "loss": 1.3493,
      "theoretical_loss": 3.3958291888373506,
      "tokens_seen": 2510815232
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024180541624874624,
      "loss": 1.4347,
      "theoretical_loss": 3.3958221840892184,
      "tokens_seen": 2510880768
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024178535606820463,
      "loss": 1.4156,
      "theoretical_loss": 3.395815179575105,
      "tokens_seen": 2510946304
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000241765295887663,
      "loss": 1.27,
      "theoretical_loss": 3.3958081752949947,
      "tokens_seen": 2511011840
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024174523570712138,
      "loss": 1.2885,
      "theoretical_loss": 3.3958011712488747,
      "tokens_seen": 2511077376
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024172517552657974,
      "loss": 1.4348,
      "theoretical_loss": 3.395794167436731,
      "tokens_seen": 2511142912
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024170511534603813,
      "loss": 1.2668,
      "theoretical_loss": 3.3957871638585493,
      "tokens_seen": 2511208448
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002416850551654965,
      "loss": 1.4431,
      "theoretical_loss": 3.3957801605143163,
      "tokens_seen": 2511273984
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024166499498495486,
      "loss": 1.3944,
      "theoretical_loss": 3.395773157404017,
      "tokens_seen": 2511339520
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024164493480441325,
      "loss": 1.3608,
      "theoretical_loss": 3.3957661545276387,
      "tokens_seen": 2511405056
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002416248746238716,
      "loss": 1.5143,
      "theoretical_loss": 3.3957591518851666,
      "tokens_seen": 2511470592
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024160481444333,
      "loss": 1.4155,
      "theoretical_loss": 3.3957521494765874,
      "tokens_seen": 2511536128
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024158475426278836,
      "loss": 1.4136,
      "theoretical_loss": 3.395745147301887,
      "tokens_seen": 2511601664
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024156469408224675,
      "loss": 1.3106,
      "theoretical_loss": 3.395738145361051,
      "tokens_seen": 2511667200
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002415446339017051,
      "loss": 1.5085,
      "theoretical_loss": 3.395731143654066,
      "tokens_seen": 2511732736
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002415245737211635,
      "loss": 1.3007,
      "theoretical_loss": 3.395724142180918,
      "tokens_seen": 2511798272
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024150451354062186,
      "loss": 1.3587,
      "theoretical_loss": 3.395717140941593,
      "tokens_seen": 2511863808
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024148445336008022,
      "loss": 1.3759,
      "theoretical_loss": 3.395710139936077,
      "tokens_seen": 2511929344
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024146439317953861,
      "loss": 1.3524,
      "theoretical_loss": 3.3957031391643566,
      "tokens_seen": 2511994880
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024144433299899698,
      "loss": 1.3094,
      "theoretical_loss": 3.395696138626417,
      "tokens_seen": 2512060416
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024142427281845537,
      "loss": 1.4954,
      "theoretical_loss": 3.395689138322245,
      "tokens_seen": 2512125952
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024140421263791373,
      "loss": 1.4268,
      "theoretical_loss": 3.3956821382518267,
      "tokens_seen": 2512191488
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024138415245737212,
      "loss": 1.3289,
      "theoretical_loss": 3.395675138415148,
      "tokens_seen": 2512257024
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002413640922768305,
      "loss": 1.301,
      "theoretical_loss": 3.395668138812195,
      "tokens_seen": 2512322560
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002413440320962889,
      "loss": 1.3633,
      "theoretical_loss": 3.3956611394429537,
      "tokens_seen": 2512388096
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024132397191574726,
      "loss": 1.2569,
      "theoretical_loss": 3.3956541403074105,
      "tokens_seen": 2512453632
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024130391173520562,
      "loss": 1.4158,
      "theoretical_loss": 3.395647141405551,
      "tokens_seen": 2512519168
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000241283851554664,
      "loss": 1.3323,
      "theoretical_loss": 3.395640142737362,
      "tokens_seen": 2512584704
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024126379137412237,
      "loss": 1.2839,
      "theoretical_loss": 3.395633144302829,
      "tokens_seen": 2512650240
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024124373119358076,
      "loss": 1.3931,
      "theoretical_loss": 3.3956261461019386,
      "tokens_seen": 2512715776
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024122367101303912,
      "loss": 1.2547,
      "theoretical_loss": 3.395619148134676,
      "tokens_seen": 2512781312
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002412036108324975,
      "loss": 1.4399,
      "theoretical_loss": 3.3956121504010284,
      "tokens_seen": 2512846848
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024118355065195587,
      "loss": 1.3141,
      "theoretical_loss": 3.3956051529009814,
      "tokens_seen": 2512912384
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024116349047141426,
      "loss": 1.3782,
      "theoretical_loss": 3.395598155634521,
      "tokens_seen": 2512977920
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024114343029087263,
      "loss": 1.37,
      "theoretical_loss": 3.3955911586016336,
      "tokens_seen": 2513043456
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024112337011033102,
      "loss": 1.375,
      "theoretical_loss": 3.3955841618023053,
      "tokens_seen": 2513108992
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024110330992978938,
      "loss": 1.3091,
      "theoretical_loss": 3.3955771652365216,
      "tokens_seen": 2513174528
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024108324974924774,
      "loss": 1.2455,
      "theoretical_loss": 3.39557016890427,
      "tokens_seen": 2513240064
    },
    {
      "epoch": 0.76,
      "objective/train/advantage_avg": 0.49562665820121765,
      "objective/train/docs_used": 914593,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4444080591201782,
      "objective/train/original_loss": 1.4444079399108887,
      "objective/train/theoretical_loss": 3.395563172805535,
      "objective/train/tokens_used": 2533765600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24579232931137085,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0508122444152832,
      "objective/train/weighted_lm_loss": 1.5177620649337769,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.395563172805535,
      "tokens_seen": 2513305600
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024106318956870613,
      "loss": 1.4213,
      "theoretical_loss": 3.395563172805535,
      "tokens_seen": 2513305600
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002410431293881645,
      "loss": 1.4963,
      "theoretical_loss": 3.3955561769403038,
      "tokens_seen": 2513371136
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024102306920762288,
      "loss": 1.3678,
      "theoretical_loss": 3.395549181308562,
      "tokens_seen": 2513436672
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024100300902708124,
      "loss": 1.4059,
      "theoretical_loss": 3.395542185910296,
      "tokens_seen": 2513502208
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024098294884653963,
      "loss": 1.5038,
      "theoretical_loss": 3.395535190745491,
      "tokens_seen": 2513567744
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000240962888665998,
      "loss": 1.3,
      "theoretical_loss": 3.395528195814135,
      "tokens_seen": 2513633280
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024094282848545638,
      "loss": 1.4557,
      "theoretical_loss": 3.3955212011162126,
      "tokens_seen": 2513698816
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024092276830491475,
      "loss": 1.3318,
      "theoretical_loss": 3.3955142066517103,
      "tokens_seen": 2513764352
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002409027081243731,
      "loss": 1.3623,
      "theoretical_loss": 3.3955072124206143,
      "tokens_seen": 2513829888
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002408826479438315,
      "loss": 1.3815,
      "theoretical_loss": 3.395500218422911,
      "tokens_seen": 2513895424
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024086258776328986,
      "loss": 1.3523,
      "theoretical_loss": 3.395493224658586,
      "tokens_seen": 2513960960
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024084252758274825,
      "loss": 1.3036,
      "theoretical_loss": 3.3954862311276255,
      "tokens_seen": 2514026496
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002408224674022066,
      "loss": 1.2484,
      "theoretical_loss": 3.395479237830016,
      "tokens_seen": 2514092032
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000240802407221665,
      "loss": 1.2302,
      "theoretical_loss": 3.3954722447657435,
      "tokens_seen": 2514157568
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024078234704112336,
      "loss": 1.4234,
      "theoretical_loss": 3.395465251934794,
      "tokens_seen": 2514223104
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024076228686058175,
      "loss": 1.3915,
      "theoretical_loss": 3.395458259337154,
      "tokens_seen": 2514288640
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024074222668004012,
      "loss": 1.2761,
      "theoretical_loss": 3.395451266972809,
      "tokens_seen": 2514354176
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024072216649949848,
      "loss": 1.2354,
      "theoretical_loss": 3.3954442748417457,
      "tokens_seen": 2514419712
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024070210631895687,
      "loss": 1.3613,
      "theoretical_loss": 3.39543728294395,
      "tokens_seen": 2514485248
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024068204613841523,
      "loss": 1.3231,
      "theoretical_loss": 3.395430291279408,
      "tokens_seen": 2514550784
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024066198595787362,
      "loss": 1.2737,
      "theoretical_loss": 3.3954232998481055,
      "tokens_seen": 2514616320
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024064192577733198,
      "loss": 1.2933,
      "theoretical_loss": 3.3954163086500295,
      "tokens_seen": 2514681856
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024062186559679037,
      "loss": 1.3265,
      "theoretical_loss": 3.395409317685165,
      "tokens_seen": 2514747392
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024060180541624876,
      "loss": 1.425,
      "theoretical_loss": 3.3954023269534996,
      "tokens_seen": 2514812928
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024058174523570715,
      "loss": 1.3644,
      "theoretical_loss": 3.3953953364550187,
      "tokens_seen": 2514878464
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002405616850551655,
      "loss": 1.3426,
      "theoretical_loss": 3.395388346189708,
      "tokens_seen": 2514944000
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002405416248746239,
      "loss": 1.3409,
      "theoretical_loss": 3.395381356157554,
      "tokens_seen": 2515009536
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024052156469408226,
      "loss": 1.4593,
      "theoretical_loss": 3.395374366358543,
      "tokens_seen": 2515075072
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024050150451354062,
      "loss": 1.4323,
      "theoretical_loss": 3.3953673767926613,
      "tokens_seen": 2515140608
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024048144433299901,
      "loss": 1.3787,
      "theoretical_loss": 3.3953603874598945,
      "tokens_seen": 2515206144
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024046138415245738,
      "loss": 1.3214,
      "theoretical_loss": 3.3953533983602293,
      "tokens_seen": 2515271680
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024044132397191577,
      "loss": 1.3259,
      "theoretical_loss": 3.3953464094936514,
      "tokens_seen": 2515337216
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024042126379137413,
      "loss": 1.3136,
      "theoretical_loss": 3.395339420860147,
      "tokens_seen": 2515402752
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024040120361083252,
      "loss": 1.4444,
      "theoretical_loss": 3.3953324324597025,
      "tokens_seen": 2515468288
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024038114343029088,
      "loss": 1.3826,
      "theoretical_loss": 3.3953254442923044,
      "tokens_seen": 2515533824
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024036108324974927,
      "loss": 1.3698,
      "theoretical_loss": 3.395318456357938,
      "tokens_seen": 2515599360
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024034102306920763,
      "loss": 1.4205,
      "theoretical_loss": 3.39531146865659,
      "tokens_seen": 2515664896
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000240320962888666,
      "loss": 1.4038,
      "theoretical_loss": 3.3953044811882465,
      "tokens_seen": 2515730432
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024030090270812438,
      "loss": 1.3669,
      "theoretical_loss": 3.3952974939528935,
      "tokens_seen": 2515795968
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024028084252758274,
      "loss": 1.3718,
      "theoretical_loss": 3.3952905069505173,
      "tokens_seen": 2515861504
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024026078234704113,
      "loss": 1.2665,
      "theoretical_loss": 3.395283520181104,
      "tokens_seen": 2515927040
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002402407221664995,
      "loss": 1.3758,
      "theoretical_loss": 3.39527653364464,
      "tokens_seen": 2515992576
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024022066198595789,
      "loss": 1.3009,
      "theoretical_loss": 3.395269547341111,
      "tokens_seen": 2516058112
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024020060180541625,
      "loss": 1.2625,
      "theoretical_loss": 3.3952625612705036,
      "tokens_seen": 2516123648
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024018054162487464,
      "loss": 1.269,
      "theoretical_loss": 3.3952555754328038,
      "tokens_seen": 2516189184
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000240160481444333,
      "loss": 1.3287,
      "theoretical_loss": 3.3952485898279976,
      "tokens_seen": 2516254720
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024014042126379136,
      "loss": 1.4764,
      "theoretical_loss": 3.3952416044560714,
      "tokens_seen": 2516320256
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024012036108324975,
      "loss": 1.4608,
      "theoretical_loss": 3.3952346193170113,
      "tokens_seen": 2516385792
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024010030090270811,
      "loss": 1.3897,
      "theoretical_loss": 3.3952276344108037,
      "tokens_seen": 2516451328
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002400802407221665,
      "loss": 1.4089,
      "theoretical_loss": 3.3952206497374346,
      "tokens_seen": 2516516864
    },
    {
      "epoch": 0.76,
      "objective/train/advantage_avg": 0.49223917722702026,
      "objective/train/docs_used": 916228,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4015480279922485,
      "objective/train/original_loss": 1.401547908782959,
      "objective/train/theoretical_loss": 3.3952136652968896,
      "objective/train/tokens_used": 2537042400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24264594912528992,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504573583602905,
      "objective/train/weighted_lm_loss": 1.4722973108291626,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3952136652968896,
      "tokens_seen": 2516582400
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024006018054162487,
      "loss": 1.4393,
      "theoretical_loss": 3.3952136652968896,
      "tokens_seen": 2516582400
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024004012036108325,
      "loss": 1.377,
      "theoretical_loss": 3.3952066810891557,
      "tokens_seen": 2516647936
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024002006018054162,
      "loss": 1.3828,
      "theoretical_loss": 3.3951996971142187,
      "tokens_seen": 2516713472
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00024,
      "loss": 1.3117,
      "theoretical_loss": 3.395192713372065,
      "tokens_seen": 2516779008
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023997993981945837,
      "loss": 1.3153,
      "theoretical_loss": 3.3951857298626806,
      "tokens_seen": 2516844544
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023995987963891676,
      "loss": 1.2682,
      "theoretical_loss": 3.395178746586052,
      "tokens_seen": 2516910080
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023993981945837512,
      "loss": 1.3443,
      "theoretical_loss": 3.3951717635421645,
      "tokens_seen": 2516975616
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023991975927783348,
      "loss": 1.3431,
      "theoretical_loss": 3.3951647807310055,
      "tokens_seen": 2517041152
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023989969909729187,
      "loss": 1.4136,
      "theoretical_loss": 3.3951577981525602,
      "tokens_seen": 2517106688
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023987963891675023,
      "loss": 1.3606,
      "theoretical_loss": 3.395150815806815,
      "tokens_seen": 2517172224
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023985957873620865,
      "loss": 1.2683,
      "theoretical_loss": 3.3951438336937567,
      "tokens_seen": 2517237760
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000239839518555667,
      "loss": 1.3899,
      "theoretical_loss": 3.395136851813371,
      "tokens_seen": 2517303296
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002398194583751254,
      "loss": 1.4163,
      "theoretical_loss": 3.395129870165644,
      "tokens_seen": 2517368832
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023979939819458376,
      "loss": 1.377,
      "theoretical_loss": 3.395122888750562,
      "tokens_seen": 2517434368
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023977933801404215,
      "loss": 1.3511,
      "theoretical_loss": 3.3951159075681105,
      "tokens_seen": 2517499904
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023975927783350052,
      "loss": 1.3472,
      "theoretical_loss": 3.3951089266182772,
      "tokens_seen": 2517565440
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023973921765295888,
      "loss": 1.4361,
      "theoretical_loss": 3.3951019459010476,
      "tokens_seen": 2517630976
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023971915747241727,
      "loss": 1.3735,
      "theoretical_loss": 3.395094965416407,
      "tokens_seen": 2517696512
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023969909729187563,
      "loss": 1.2798,
      "theoretical_loss": 3.395087985164343,
      "tokens_seen": 2517762048
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023967903711133402,
      "loss": 1.3357,
      "theoretical_loss": 3.3950810051448412,
      "tokens_seen": 2517827584
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023965897693079238,
      "loss": 1.2909,
      "theoretical_loss": 3.3950740253578875,
      "tokens_seen": 2517893120
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023963891675025077,
      "loss": 1.381,
      "theoretical_loss": 3.3950670458034686,
      "tokens_seen": 2517958656
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023961885656970913,
      "loss": 1.275,
      "theoretical_loss": 3.3950600664815704,
      "tokens_seen": 2518024192
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023959879638916752,
      "loss": 1.2742,
      "theoretical_loss": 3.3950530873921796,
      "tokens_seen": 2518089728
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023957873620862588,
      "loss": 1.4419,
      "theoretical_loss": 3.3950461085352814,
      "tokens_seen": 2518155264
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023955867602808427,
      "loss": 1.2895,
      "theoretical_loss": 3.3950391299108627,
      "tokens_seen": 2518220800
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023953861584754264,
      "loss": 1.3141,
      "theoretical_loss": 3.39503215151891,
      "tokens_seen": 2518286336
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000239518555667001,
      "loss": 1.4524,
      "theoretical_loss": 3.395025173359408,
      "tokens_seen": 2518351872
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002394984954864594,
      "loss": 1.4115,
      "theoretical_loss": 3.3950181954323453,
      "tokens_seen": 2518417408
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023947843530591775,
      "loss": 1.3714,
      "theoretical_loss": 3.3950112177377063,
      "tokens_seen": 2518482944
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023945837512537614,
      "loss": 1.288,
      "theoretical_loss": 3.3950042402754774,
      "tokens_seen": 2518548480
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002394383149448345,
      "loss": 1.5012,
      "theoretical_loss": 3.3949972630456458,
      "tokens_seen": 2518614016
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002394182547642929,
      "loss": 1.2662,
      "theoretical_loss": 3.3949902860481966,
      "tokens_seen": 2518679552
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023939819458375125,
      "loss": 1.3087,
      "theoretical_loss": 3.3949833092831163,
      "tokens_seen": 2518745088
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023937813440320964,
      "loss": 1.4363,
      "theoretical_loss": 3.394976332750392,
      "tokens_seen": 2518810624
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000239358074222668,
      "loss": 1.3927,
      "theoretical_loss": 3.3949693564500087,
      "tokens_seen": 2518876160
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023933801404212637,
      "loss": 1.2788,
      "theoretical_loss": 3.394962380381953,
      "tokens_seen": 2518941696
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023931795386158476,
      "loss": 1.4842,
      "theoretical_loss": 3.3949554045462116,
      "tokens_seen": 2519007232
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023929789368104312,
      "loss": 1.3158,
      "theoretical_loss": 3.39494842894277,
      "tokens_seen": 2519072768
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002392778335005015,
      "loss": 1.3117,
      "theoretical_loss": 3.394941453571615,
      "tokens_seen": 2519138304
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023925777331995987,
      "loss": 1.3552,
      "theoretical_loss": 3.3949344784327327,
      "tokens_seen": 2519203840
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023923771313941826,
      "loss": 1.381,
      "theoretical_loss": 3.394927503526109,
      "tokens_seen": 2519269376
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023921765295887662,
      "loss": 1.3126,
      "theoretical_loss": 3.3949205288517303,
      "tokens_seen": 2519334912
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000239197592778335,
      "loss": 1.3923,
      "theoretical_loss": 3.394913554409583,
      "tokens_seen": 2519400448
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023917753259779337,
      "loss": 1.4171,
      "theoretical_loss": 3.3949065801996534,
      "tokens_seen": 2519465984
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023915747241725174,
      "loss": 1.2547,
      "theoretical_loss": 3.394899606221927,
      "tokens_seen": 2519531520
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023913741223671012,
      "loss": 1.3499,
      "theoretical_loss": 3.394892632476391,
      "tokens_seen": 2519597056
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002391173520561685,
      "loss": 1.2692,
      "theoretical_loss": 3.394885658963031,
      "tokens_seen": 2519662592
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002390972918756269,
      "loss": 1.3438,
      "theoretical_loss": 3.3948786856818334,
      "tokens_seen": 2519728128
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023907723169508527,
      "loss": 1.3223,
      "theoretical_loss": 3.3948717126327845,
      "tokens_seen": 2519793664
    },
    {
      "epoch": 0.76,
      "objective/train/advantage_avg": 0.4879656136035919,
      "objective/train/docs_used": 917453,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2587946653366089,
      "objective/train/original_loss": 1.2587947845458984,
      "objective/train/theoretical_loss": 3.394864739815871,
      "objective/train/tokens_used": 2540319200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23904746770858765,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500117540359497,
      "objective/train/weighted_lm_loss": 1.3216274976730347,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.394864739815871,
      "tokens_seen": 2519859200
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023905717151454365,
      "loss": 1.2321,
      "theoretical_loss": 3.394864739815871,
      "tokens_seen": 2519859200
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023903711133400202,
      "loss": 1.3015,
      "theoretical_loss": 3.394857767231078,
      "tokens_seen": 2519924736
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002390170511534604,
      "loss": 1.495,
      "theoretical_loss": 3.3948507948783924,
      "tokens_seen": 2519990272
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023899699097291877,
      "loss": 1.2673,
      "theoretical_loss": 3.3948438227578004,
      "tokens_seen": 2520055808
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023897693079237716,
      "loss": 1.3068,
      "theoretical_loss": 3.3948368508692885,
      "tokens_seen": 2520121344
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023895687061183552,
      "loss": 1.2798,
      "theoretical_loss": 3.3948298792128426,
      "tokens_seen": 2520186880
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023893681043129388,
      "loss": 1.3945,
      "theoretical_loss": 3.3948229077884484,
      "tokens_seen": 2520252416
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023891675025075227,
      "loss": 1.386,
      "theoretical_loss": 3.3948159365960935,
      "tokens_seen": 2520317952
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023889669007021063,
      "loss": 1.3453,
      "theoretical_loss": 3.394808965635763,
      "tokens_seen": 2520383488
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023887662988966902,
      "loss": 1.3645,
      "theoretical_loss": 3.3948019949074437,
      "tokens_seen": 2520449024
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023885656970912739,
      "loss": 1.3593,
      "theoretical_loss": 3.394795024411122,
      "tokens_seen": 2520514560
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023883650952858577,
      "loss": 1.3579,
      "theoretical_loss": 3.3947880541467836,
      "tokens_seen": 2520580096
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023881644934804414,
      "loss": 1.4397,
      "theoretical_loss": 3.3947810841144146,
      "tokens_seen": 2520645632
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023879638916750253,
      "loss": 1.3697,
      "theoretical_loss": 3.3947741143140022,
      "tokens_seen": 2520711168
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002387763289869609,
      "loss": 1.3379,
      "theoretical_loss": 3.394767144745532,
      "tokens_seen": 2520776704
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023875626880641925,
      "loss": 1.3823,
      "theoretical_loss": 3.3947601754089898,
      "tokens_seen": 2520842240
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023873620862587764,
      "loss": 1.4195,
      "theoretical_loss": 3.394753206304363,
      "tokens_seen": 2520907776
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000238716148445336,
      "loss": 1.4885,
      "theoretical_loss": 3.3947462374316366,
      "tokens_seen": 2520973312
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002386960882647944,
      "loss": 1.4323,
      "theoretical_loss": 3.394739268790798,
      "tokens_seen": 2521038848
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023867602808425275,
      "loss": 1.3637,
      "theoretical_loss": 3.394732300381833,
      "tokens_seen": 2521104384
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023865596790371114,
      "loss": 1.2839,
      "theoretical_loss": 3.394725332204728,
      "tokens_seen": 2521169920
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002386359077231695,
      "loss": 1.3777,
      "theoretical_loss": 3.3947183642594685,
      "tokens_seen": 2521235456
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002386158475426279,
      "loss": 1.3125,
      "theoretical_loss": 3.3947113965460414,
      "tokens_seen": 2521300992
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023859578736208626,
      "loss": 1.3866,
      "theoretical_loss": 3.3947044290644333,
      "tokens_seen": 2521366528
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023857572718154462,
      "loss": 1.3886,
      "theoretical_loss": 3.39469746181463,
      "tokens_seen": 2521432064
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000238555667001003,
      "loss": 1.4346,
      "theoretical_loss": 3.394690494796617,
      "tokens_seen": 2521497600
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023853560682046137,
      "loss": 1.3502,
      "theoretical_loss": 3.3946835280103826,
      "tokens_seen": 2521563136
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023851554663991976,
      "loss": 1.3017,
      "theoretical_loss": 3.394676561455911,
      "tokens_seen": 2521628672
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023849548645937812,
      "loss": 1.2868,
      "theoretical_loss": 3.3946695951331893,
      "tokens_seen": 2521694208
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002384754262788365,
      "loss": 1.4419,
      "theoretical_loss": 3.3946626290422044,
      "tokens_seen": 2521759744
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023845536609829487,
      "loss": 1.4405,
      "theoretical_loss": 3.3946556631829417,
      "tokens_seen": 2521825280
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023843530591775326,
      "loss": 1.3288,
      "theoretical_loss": 3.3946486975553873,
      "tokens_seen": 2521890816
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023841524573721163,
      "loss": 1.3608,
      "theoretical_loss": 3.3946417321595286,
      "tokens_seen": 2521956352
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023839518555667002,
      "loss": 1.4336,
      "theoretical_loss": 3.3946347669953507,
      "tokens_seen": 2522021888
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023837512537612838,
      "loss": 1.4155,
      "theoretical_loss": 3.3946278020628404,
      "tokens_seen": 2522087424
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023835506519558677,
      "loss": 1.2745,
      "theoretical_loss": 3.394620837361984,
      "tokens_seen": 2522152960
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023833500501504516,
      "loss": 1.3099,
      "theoretical_loss": 3.3946138728927675,
      "tokens_seen": 2522218496
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023831494483450352,
      "loss": 1.3188,
      "theoretical_loss": 3.3946069086551773,
      "tokens_seen": 2522284032
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002382948846539619,
      "loss": 1.413,
      "theoretical_loss": 3.3945999446492,
      "tokens_seen": 2522349568
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023827482447342027,
      "loss": 1.3419,
      "theoretical_loss": 3.3945929808748216,
      "tokens_seen": 2522415104
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023825476429287866,
      "loss": 1.3178,
      "theoretical_loss": 3.394586017332028,
      "tokens_seen": 2522480640
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023823470411233702,
      "loss": 1.3696,
      "theoretical_loss": 3.3945790540208067,
      "tokens_seen": 2522546176
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002382146439317954,
      "loss": 1.3726,
      "theoretical_loss": 3.3945720909411428,
      "tokens_seen": 2522611712
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023819458375125377,
      "loss": 1.3569,
      "theoretical_loss": 3.3945651280930225,
      "tokens_seen": 2522677248
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023817452357071214,
      "loss": 1.2762,
      "theoretical_loss": 3.394558165476433,
      "tokens_seen": 2522742784
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023815446339017052,
      "loss": 1.3918,
      "theoretical_loss": 3.3945512030913605,
      "tokens_seen": 2522808320
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002381344032096289,
      "loss": 1.3732,
      "theoretical_loss": 3.3945442409377904,
      "tokens_seen": 2522873856
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023811434302908728,
      "loss": 1.3331,
      "theoretical_loss": 3.3945372790157093,
      "tokens_seen": 2522939392
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023809428284854564,
      "loss": 1.3826,
      "theoretical_loss": 3.394530317325104,
      "tokens_seen": 2523004928
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023807422266800403,
      "loss": 1.4064,
      "theoretical_loss": 3.39452335586596,
      "tokens_seen": 2523070464
    },
    {
      "epoch": 0.76,
      "objective/train/advantage_avg": 0.4928901493549347,
      "objective/train/docs_used": 918647,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.367674708366394,
      "objective/train/original_loss": 1.3676745891571045,
      "objective/train/theoretical_loss": 3.394516394638265,
      "objective/train/tokens_used": 2543596000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24333374202251434,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505261421203613,
      "objective/train/weighted_lm_loss": 1.436649203300476,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.394516394638265,
      "tokens_seen": 2523136000
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002380541624874624,
      "loss": 1.436,
      "theoretical_loss": 3.394516394638265,
      "tokens_seen": 2523136000
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023803410230692078,
      "loss": 1.3038,
      "theoretical_loss": 3.394509433642004,
      "tokens_seen": 2523201536
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023801404212637914,
      "loss": 1.2647,
      "theoretical_loss": 3.3945024728771633,
      "tokens_seen": 2523267072
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023799398194583753,
      "loss": 1.4249,
      "theoretical_loss": 3.39449551234373,
      "tokens_seen": 2523332608
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002379739217652959,
      "loss": 1.3325,
      "theoretical_loss": 3.3944885520416896,
      "tokens_seen": 2523398144
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023795386158475426,
      "loss": 1.3451,
      "theoretical_loss": 3.394481591971029,
      "tokens_seen": 2523463680
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023793380140421265,
      "loss": 1.5213,
      "theoretical_loss": 3.3944746321317343,
      "tokens_seen": 2523529216
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.000237913741223671,
      "loss": 1.3965,
      "theoretical_loss": 3.394467672523792,
      "tokens_seen": 2523594752
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002378936810431294,
      "loss": 1.3832,
      "theoretical_loss": 3.3944607131471876,
      "tokens_seen": 2523660288
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023787362086258776,
      "loss": 1.3241,
      "theoretical_loss": 3.394453754001908,
      "tokens_seen": 2523725824
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023785356068204615,
      "loss": 1.4536,
      "theoretical_loss": 3.39444679508794,
      "tokens_seen": 2523791360
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002378335005015045,
      "loss": 1.4937,
      "theoretical_loss": 3.394439836405269,
      "tokens_seen": 2523856896
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002378134403209629,
      "loss": 1.3293,
      "theoretical_loss": 3.3944328779538813,
      "tokens_seen": 2523922432
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023779338014042126,
      "loss": 1.3948,
      "theoretical_loss": 3.3944259197337643,
      "tokens_seen": 2523987968
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023777331995987962,
      "loss": 1.221,
      "theoretical_loss": 3.394418961744903,
      "tokens_seen": 2524053504
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023775325977933801,
      "loss": 1.3917,
      "theoretical_loss": 3.3944120039872847,
      "tokens_seen": 2524119040
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023773319959879638,
      "loss": 1.2758,
      "theoretical_loss": 3.3944050464608955,
      "tokens_seen": 2524184576
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023771313941825477,
      "loss": 1.3048,
      "theoretical_loss": 3.394398089165721,
      "tokens_seen": 2524250112
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023769307923771313,
      "loss": 1.3651,
      "theoretical_loss": 3.3943911321017484,
      "tokens_seen": 2524315648
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023767301905717152,
      "loss": 1.305,
      "theoretical_loss": 3.3943841752689634,
      "tokens_seen": 2524381184
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.00023765295887662988,
      "loss": 1.4184,
      "theoretical_loss": 3.394377218667353,
      "tokens_seen": 2524446720
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023763289869608827,
      "loss": 1.2434,
      "theoretical_loss": 3.3943702622969028,
      "tokens_seen": 2524512256
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023761283851554663,
      "loss": 1.2971,
      "theoretical_loss": 3.394363306157599,
      "tokens_seen": 2524577792
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023759277833500502,
      "loss": 1.3646,
      "theoretical_loss": 3.3943563502494287,
      "tokens_seen": 2524643328
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002375727181544634,
      "loss": 1.2872,
      "theoretical_loss": 3.3943493945723784,
      "tokens_seen": 2524708864
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023755265797392177,
      "loss": 1.4303,
      "theoretical_loss": 3.394342439126433,
      "tokens_seen": 2524774400
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023753259779338016,
      "loss": 1.3994,
      "theoretical_loss": 3.39433548391158,
      "tokens_seen": 2524839936
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023751253761283852,
      "loss": 1.356,
      "theoretical_loss": 3.394328528927806,
      "tokens_seen": 2524905472
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002374924774322969,
      "loss": 1.4102,
      "theoretical_loss": 3.394321574175096,
      "tokens_seen": 2524971008
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023747241725175527,
      "loss": 1.347,
      "theoretical_loss": 3.3943146196534375,
      "tokens_seen": 2525036544
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023745235707121366,
      "loss": 1.4622,
      "theoretical_loss": 3.394307665362816,
      "tokens_seen": 2525102080
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023743229689067203,
      "loss": 1.3634,
      "theoretical_loss": 3.3943007113032184,
      "tokens_seen": 2525167616
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023741223671013042,
      "loss": 1.2432,
      "theoretical_loss": 3.394293757474631,
      "tokens_seen": 2525233152
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023739217652958878,
      "loss": 1.4559,
      "theoretical_loss": 3.3942868038770397,
      "tokens_seen": 2525298688
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023737211634904714,
      "loss": 1.3653,
      "theoretical_loss": 3.3942798505104315,
      "tokens_seen": 2525364224
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023735205616850553,
      "loss": 1.3428,
      "theoretical_loss": 3.3942728973747918,
      "tokens_seen": 2525429760
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002373319959879639,
      "loss": 1.4723,
      "theoretical_loss": 3.394265944470108,
      "tokens_seen": 2525495296
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023731193580742228,
      "loss": 1.221,
      "theoretical_loss": 3.3942589917963657,
      "tokens_seen": 2525560832
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023729187562688064,
      "loss": 1.2913,
      "theoretical_loss": 3.3942520393535514,
      "tokens_seen": 2525626368
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023727181544633903,
      "loss": 1.352,
      "theoretical_loss": 3.394245087141652,
      "tokens_seen": 2525691904
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002372517552657974,
      "loss": 1.3586,
      "theoretical_loss": 3.3942381351606525,
      "tokens_seen": 2525757440
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023723169508525578,
      "loss": 1.3368,
      "theoretical_loss": 3.3942311834105405,
      "tokens_seen": 2525822976
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023721163490471415,
      "loss": 1.4461,
      "theoretical_loss": 3.394224231891302,
      "tokens_seen": 2525888512
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002371915747241725,
      "loss": 1.4224,
      "theoretical_loss": 3.3942172806029234,
      "tokens_seen": 2525954048
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002371715145436309,
      "loss": 1.3435,
      "theoretical_loss": 3.3942103295453903,
      "tokens_seen": 2526019584
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023715145436308926,
      "loss": 1.2905,
      "theoretical_loss": 3.3942033787186903,
      "tokens_seen": 2526085120
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023713139418254765,
      "loss": 1.2509,
      "theoretical_loss": 3.3941964281228088,
      "tokens_seen": 2526150656
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.000237111334002006,
      "loss": 1.3134,
      "theoretical_loss": 3.3941894777577324,
      "tokens_seen": 2526216192
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002370912738214644,
      "loss": 1.4383,
      "theoretical_loss": 3.394182527623448,
      "tokens_seen": 2526281728
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023707121364092276,
      "loss": 1.3642,
      "theoretical_loss": 3.3941755777199405,
      "tokens_seen": 2526347264
    },
    {
      "epoch": 0.77,
      "objective/train/advantage_avg": 0.4895476996898651,
      "objective/train/docs_used": 919806,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.163002610206604,
      "objective/train/original_loss": 1.1630024909973145,
      "objective/train/theoretical_loss": 3.394168628047198,
      "objective/train/tokens_used": 2546872800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24154788255691528,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050182580947876,
      "objective/train/weighted_lm_loss": 1.2214832305908203,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.394168628047198,
      "tokens_seen": 2526412800
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023705115346038115,
      "loss": 1.4056,
      "theoretical_loss": 3.394168628047198,
      "tokens_seen": 2526412800
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023703109327983952,
      "loss": 1.3597,
      "theoretical_loss": 3.3941616786052053,
      "tokens_seen": 2526478336
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023701103309929788,
      "loss": 1.4579,
      "theoretical_loss": 3.39415472939395,
      "tokens_seen": 2526543872
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023699097291875627,
      "loss": 1.2833,
      "theoretical_loss": 3.394147780413418,
      "tokens_seen": 2526609408
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023697091273821463,
      "loss": 1.3735,
      "theoretical_loss": 3.3941408316635955,
      "tokens_seen": 2526674944
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023695085255767302,
      "loss": 1.3719,
      "theoretical_loss": 3.3941338831444687,
      "tokens_seen": 2526740480
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023693079237713138,
      "loss": 1.3179,
      "theoretical_loss": 3.3941269348560246,
      "tokens_seen": 2526806016
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023691073219658977,
      "loss": 1.4957,
      "theoretical_loss": 3.3941199867982492,
      "tokens_seen": 2526871552
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023689067201604813,
      "loss": 1.2763,
      "theoretical_loss": 3.394113038971129,
      "tokens_seen": 2526937088
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023687061183550652,
      "loss": 1.3625,
      "theoretical_loss": 3.3941060913746495,
      "tokens_seen": 2527002624
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002368505516549649,
      "loss": 1.3627,
      "theoretical_loss": 3.3940991440087984,
      "tokens_seen": 2527068160
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002368304914744233,
      "loss": 1.4245,
      "theoretical_loss": 3.394092196873561,
      "tokens_seen": 2527133696
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023681043129388166,
      "loss": 1.4029,
      "theoretical_loss": 3.3940852499689242,
      "tokens_seen": 2527199232
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023679037111334002,
      "loss": 1.3987,
      "theoretical_loss": 3.3940783032948745,
      "tokens_seen": 2527264768
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023677031093279841,
      "loss": 1.4165,
      "theoretical_loss": 3.3940713568513976,
      "tokens_seen": 2527330304
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023675025075225678,
      "loss": 1.3383,
      "theoretical_loss": 3.3940644106384807,
      "tokens_seen": 2527395840
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023673019057171517,
      "loss": 1.3775,
      "theoretical_loss": 3.3940574646561097,
      "tokens_seen": 2527461376
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023671013039117353,
      "loss": 1.3502,
      "theoretical_loss": 3.3940505189042707,
      "tokens_seen": 2527526912
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023669007021063192,
      "loss": 1.303,
      "theoretical_loss": 3.394043573382951,
      "tokens_seen": 2527592448
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023667001003009028,
      "loss": 1.3378,
      "theoretical_loss": 3.394036628092136,
      "tokens_seen": 2527657984
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023664994984954867,
      "loss": 1.4002,
      "theoretical_loss": 3.394029683031812,
      "tokens_seen": 2527723520
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023662988966900703,
      "loss": 1.3188,
      "theoretical_loss": 3.3940227382019668,
      "tokens_seen": 2527789056
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002366098294884654,
      "loss": 1.3631,
      "theoretical_loss": 3.394015793602585,
      "tokens_seen": 2527854592
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023658976930792378,
      "loss": 1.3192,
      "theoretical_loss": 3.3940088492336544,
      "tokens_seen": 2527920128
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023656970912738214,
      "loss": 1.4729,
      "theoretical_loss": 3.39400190509516,
      "tokens_seen": 2527985664
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023654964894684053,
      "loss": 1.3378,
      "theoretical_loss": 3.3939949611870897,
      "tokens_seen": 2528051200
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002365295887662989,
      "loss": 1.2916,
      "theoretical_loss": 3.393988017509429,
      "tokens_seen": 2528116736
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023650952858575729,
      "loss": 1.4233,
      "theoretical_loss": 3.3939810740621636,
      "tokens_seen": 2528182272
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023648946840521565,
      "loss": 1.2592,
      "theoretical_loss": 3.3939741308452813,
      "tokens_seen": 2528247808
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023646940822467404,
      "loss": 1.4796,
      "theoretical_loss": 3.393967187858768,
      "tokens_seen": 2528313344
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002364493480441324,
      "loss": 1.3463,
      "theoretical_loss": 3.3939602451026096,
      "tokens_seen": 2528378880
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023642928786359076,
      "loss": 1.3864,
      "theoretical_loss": 3.393953302576793,
      "tokens_seen": 2528444416
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023640922768304915,
      "loss": 1.3005,
      "theoretical_loss": 3.393946360281305,
      "tokens_seen": 2528509952
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002363891675025075,
      "loss": 1.3949,
      "theoretical_loss": 3.3939394182161307,
      "tokens_seen": 2528575488
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002363691073219659,
      "loss": 1.2651,
      "theoretical_loss": 3.393932476381257,
      "tokens_seen": 2528641024
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023634904714142426,
      "loss": 1.4426,
      "theoretical_loss": 3.3939255347766712,
      "tokens_seen": 2528706560
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023632898696088265,
      "loss": 1.3128,
      "theoretical_loss": 3.3939185934023586,
      "tokens_seen": 2528772096
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023630892678034102,
      "loss": 1.4068,
      "theoretical_loss": 3.393911652258306,
      "tokens_seen": 2528837632
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002362888665997994,
      "loss": 1.2877,
      "theoretical_loss": 3.3939047113445,
      "tokens_seen": 2528903168
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023626880641925777,
      "loss": 1.2906,
      "theoretical_loss": 3.393897770660926,
      "tokens_seen": 2528968704
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023624874623871616,
      "loss": 1.3877,
      "theoretical_loss": 3.3938908302075723,
      "tokens_seen": 2529034240
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023622868605817452,
      "loss": 1.4063,
      "theoretical_loss": 3.3938838899844237,
      "tokens_seen": 2529099776
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023620862587763288,
      "loss": 1.3728,
      "theoretical_loss": 3.393876949991467,
      "tokens_seen": 2529165312
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023618856569709127,
      "loss": 1.4176,
      "theoretical_loss": 3.3938700102286887,
      "tokens_seen": 2529230848
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023616850551654963,
      "loss": 1.3866,
      "theoretical_loss": 3.393863070696075,
      "tokens_seen": 2529296384
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023614844533600802,
      "loss": 1.3813,
      "theoretical_loss": 3.393856131393613,
      "tokens_seen": 2529361920
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023612838515546639,
      "loss": 1.2731,
      "theoretical_loss": 3.393849192321288,
      "tokens_seen": 2529427456
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002361083249749248,
      "loss": 1.3867,
      "theoretical_loss": 3.393842253479087,
      "tokens_seen": 2529492992
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023608826479438316,
      "loss": 1.3858,
      "theoretical_loss": 3.3938353148669966,
      "tokens_seen": 2529558528
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023606820461384155,
      "loss": 1.4437,
      "theoretical_loss": 3.3938283764850032,
      "tokens_seen": 2529624064
    },
    {
      "epoch": 0.77,
      "objective/train/advantage_avg": 0.4953210651874542,
      "objective/train/docs_used": 920541,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.251931071281433,
      "objective/train/original_loss": 1.2519309520721436,
      "objective/train/theoretical_loss": 3.3938214383330925,
      "objective/train/tokens_used": 2550149600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24554577469825745,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507805347442627,
      "objective/train/weighted_lm_loss": 1.3153451681137085,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3938214383330925,
      "tokens_seen": 2529689600
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023604814443329992,
      "loss": 1.3758,
      "theoretical_loss": 3.3938214383330925,
      "tokens_seen": 2529689600
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023602808425275828,
      "loss": 1.3818,
      "theoretical_loss": 3.393814500411252,
      "tokens_seen": 2529755136
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023600802407221667,
      "loss": 1.4806,
      "theoretical_loss": 3.3938075627194673,
      "tokens_seen": 2529820672
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023598796389167503,
      "loss": 1.2475,
      "theoretical_loss": 3.3938006252577244,
      "tokens_seen": 2529886208
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023596790371113342,
      "loss": 1.3792,
      "theoretical_loss": 3.3937936880260113,
      "tokens_seen": 2529951744
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023594784353059178,
      "loss": 1.4567,
      "theoretical_loss": 3.393786751024313,
      "tokens_seen": 2530017280
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023592778335005017,
      "loss": 1.3778,
      "theoretical_loss": 3.3937798142526168,
      "tokens_seen": 2530082816
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023590772316950853,
      "loss": 1.347,
      "theoretical_loss": 3.393772877710908,
      "tokens_seen": 2530148352
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023588766298896692,
      "loss": 1.4035,
      "theoretical_loss": 3.3937659413991743,
      "tokens_seen": 2530213888
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023586760280842528,
      "loss": 1.3792,
      "theoretical_loss": 3.3937590053174014,
      "tokens_seen": 2530279424
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023584754262788367,
      "loss": 1.3198,
      "theoretical_loss": 3.3937520694655756,
      "tokens_seen": 2530344960
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023582748244734204,
      "loss": 1.4624,
      "theoretical_loss": 3.3937451338436837,
      "tokens_seen": 2530410496
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002358074222668004,
      "loss": 1.2999,
      "theoretical_loss": 3.3937381984517123,
      "tokens_seen": 2530476032
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002357873620862588,
      "loss": 1.2851,
      "theoretical_loss": 3.3937312632896472,
      "tokens_seen": 2530541568
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023576730190571715,
      "loss": 1.3884,
      "theoretical_loss": 3.393724328357475,
      "tokens_seen": 2530607104
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023574724172517554,
      "loss": 1.3642,
      "theoretical_loss": 3.3937173936551828,
      "tokens_seen": 2530672640
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002357271815446339,
      "loss": 1.2142,
      "theoretical_loss": 3.3937104591827563,
      "tokens_seen": 2530738176
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002357071213640923,
      "loss": 1.3705,
      "theoretical_loss": 3.393703524940182,
      "tokens_seen": 2530803712
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023568706118355065,
      "loss": 1.3898,
      "theoretical_loss": 3.3936965909274464,
      "tokens_seen": 2530869248
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023566700100300904,
      "loss": 1.3151,
      "theoretical_loss": 3.3936896571445363,
      "tokens_seen": 2530934784
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002356469408224674,
      "loss": 1.3696,
      "theoretical_loss": 3.3936827235914375,
      "tokens_seen": 2531000320
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023562688064192577,
      "loss": 1.3633,
      "theoretical_loss": 3.3936757902681367,
      "tokens_seen": 2531065856
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023560682046138416,
      "loss": 1.3774,
      "theoretical_loss": 3.3936688571746205,
      "tokens_seen": 2531131392
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023558676028084252,
      "loss": 1.4213,
      "theoretical_loss": 3.3936619243108757,
      "tokens_seen": 2531196928
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002355667001003009,
      "loss": 1.3431,
      "theoretical_loss": 3.3936549916768874,
      "tokens_seen": 2531262464
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023554663991975927,
      "loss": 1.4483,
      "theoretical_loss": 3.3936480592726435,
      "tokens_seen": 2531328000
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023552657973921766,
      "loss": 1.3979,
      "theoretical_loss": 3.39364112709813,
      "tokens_seen": 2531393536
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023550651955867602,
      "loss": 1.4634,
      "theoretical_loss": 3.3936341951533326,
      "tokens_seen": 2531459072
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002354864593781344,
      "loss": 1.3484,
      "theoretical_loss": 3.3936272634382387,
      "tokens_seen": 2531524608
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023546639919759277,
      "loss": 1.3454,
      "theoretical_loss": 3.3936203319528344,
      "tokens_seen": 2531590144
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023544633901705113,
      "loss": 1.3996,
      "theoretical_loss": 3.393613400697106,
      "tokens_seen": 2531655680
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023542627883650952,
      "loss": 1.419,
      "theoretical_loss": 3.39360646967104,
      "tokens_seen": 2531721216
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002354062186559679,
      "loss": 1.3159,
      "theoretical_loss": 3.3935995388746227,
      "tokens_seen": 2531786752
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023538615847542628,
      "loss": 1.3216,
      "theoretical_loss": 3.393592608307841,
      "tokens_seen": 2531852288
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023536609829488464,
      "loss": 1.2959,
      "theoretical_loss": 3.393585677970681,
      "tokens_seen": 2531917824
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023534603811434305,
      "loss": 1.2772,
      "theoretical_loss": 3.393578747863129,
      "tokens_seen": 2531983360
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023532597793380142,
      "loss": 1.3417,
      "theoretical_loss": 3.393571817985172,
      "tokens_seen": 2532048896
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002353059177532598,
      "loss": 1.4,
      "theoretical_loss": 3.393564888336796,
      "tokens_seen": 2532114432
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023528585757271817,
      "loss": 1.3943,
      "theoretical_loss": 3.393557958917988,
      "tokens_seen": 2532179968
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023526579739217656,
      "loss": 1.3742,
      "theoretical_loss": 3.3935510297287337,
      "tokens_seen": 2532245504
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023524573721163492,
      "loss": 1.3944,
      "theoretical_loss": 3.3935441007690197,
      "tokens_seen": 2532311040
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023522567703109328,
      "loss": 1.2909,
      "theoretical_loss": 3.393537172038833,
      "tokens_seen": 2532376576
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023520561685055167,
      "loss": 1.3245,
      "theoretical_loss": 3.3935302435381596,
      "tokens_seen": 2532442112
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023518555667001003,
      "loss": 1.4435,
      "theoretical_loss": 3.393523315266986,
      "tokens_seen": 2532507648
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023516549648946842,
      "loss": 1.3534,
      "theoretical_loss": 3.393516387225299,
      "tokens_seen": 2532573184
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023514543630892679,
      "loss": 1.3777,
      "theoretical_loss": 3.3935094594130844,
      "tokens_seen": 2532638720
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023512537612838517,
      "loss": 1.3131,
      "theoretical_loss": 3.393502531830329,
      "tokens_seen": 2532704256
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023510531594784354,
      "loss": 1.3992,
      "theoretical_loss": 3.3934956044770197,
      "tokens_seen": 2532769792
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023508525576730193,
      "loss": 1.3251,
      "theoretical_loss": 3.393488677353142,
      "tokens_seen": 2532835328
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002350651955867603,
      "loss": 1.3201,
      "theoretical_loss": 3.3934817504586836,
      "tokens_seen": 2532900864
    },
    {
      "epoch": 0.77,
      "objective/train/advantage_avg": 0.4878202974796295,
      "objective/train/docs_used": 921695,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.311432123184204,
      "objective/train/original_loss": 1.3114320039749146,
      "objective/train/theoretical_loss": 3.39347482379363,
      "objective/train/tokens_used": 2553426400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23883184790611267,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0499961376190186,
      "objective/train/weighted_lm_loss": 1.376983404159546,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.39347482379363,
      "tokens_seen": 2532966400
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023504513540621865,
      "loss": 1.3626,
      "theoretical_loss": 3.39347482379363,
      "tokens_seen": 2532966400
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023502507522567704,
      "loss": 1.4016,
      "theoretical_loss": 3.393467897357968,
      "tokens_seen": 2533031936
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002350050150451354,
      "loss": 1.3453,
      "theoretical_loss": 3.3934609711516845,
      "tokens_seen": 2533097472
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002349849548645938,
      "loss": 1.3122,
      "theoretical_loss": 3.393454045174765,
      "tokens_seen": 2533163008
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023496489468405215,
      "loss": 1.3465,
      "theoretical_loss": 3.3934471194271967,
      "tokens_seen": 2533228544
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023494483450351054,
      "loss": 1.3309,
      "theoretical_loss": 3.393440193908966,
      "tokens_seen": 2533294080
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002349247743229689,
      "loss": 1.3265,
      "theoretical_loss": 3.3934332686200586,
      "tokens_seen": 2533359616
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002349047141424273,
      "loss": 1.4206,
      "theoretical_loss": 3.3934263435604617,
      "tokens_seen": 2533425152
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023488465396188566,
      "loss": 1.3381,
      "theoretical_loss": 3.3934194187301623,
      "tokens_seen": 2533490688
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023486459378134402,
      "loss": 1.3655,
      "theoretical_loss": 3.3934124941291457,
      "tokens_seen": 2533556224
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002348445336008024,
      "loss": 1.3851,
      "theoretical_loss": 3.3934055697573995,
      "tokens_seen": 2533621760
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023482447342026077,
      "loss": 1.2804,
      "theoretical_loss": 3.393398645614909,
      "tokens_seen": 2533687296
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023480441323971916,
      "loss": 1.4142,
      "theoretical_loss": 3.393391721701662,
      "tokens_seen": 2533752832
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023478435305917752,
      "loss": 1.3,
      "theoretical_loss": 3.3933847980176433,
      "tokens_seen": 2533818368
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002347642928786359,
      "loss": 1.5024,
      "theoretical_loss": 3.393377874562841,
      "tokens_seen": 2533883904
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023474423269809427,
      "loss": 1.2847,
      "theoretical_loss": 3.393370951337241,
      "tokens_seen": 2533949440
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023472417251755266,
      "loss": 1.41,
      "theoretical_loss": 3.3933640283408293,
      "tokens_seen": 2534014976
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023470411233701103,
      "loss": 1.3367,
      "theoretical_loss": 3.3933571055735934,
      "tokens_seen": 2534080512
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023468405215646942,
      "loss": 1.4329,
      "theoretical_loss": 3.393350183035519,
      "tokens_seen": 2534146048
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023466399197592778,
      "loss": 1.3866,
      "theoretical_loss": 3.3933432607265925,
      "tokens_seen": 2534211584
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023464393179538614,
      "loss": 1.2821,
      "theoretical_loss": 3.393336338646801,
      "tokens_seen": 2534277120
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023462387161484453,
      "loss": 1.3802,
      "theoretical_loss": 3.3933294167961305,
      "tokens_seen": 2534342656
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023460381143430292,
      "loss": 1.2822,
      "theoretical_loss": 3.3933224951745675,
      "tokens_seen": 2534408192
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002345837512537613,
      "loss": 1.4266,
      "theoretical_loss": 3.3933155737820986,
      "tokens_seen": 2534473728
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023456369107321967,
      "loss": 1.2886,
      "theoretical_loss": 3.393308652618711,
      "tokens_seen": 2534539264
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023454363089267806,
      "loss": 1.3492,
      "theoretical_loss": 3.3933017316843896,
      "tokens_seen": 2534604800
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023452357071213642,
      "loss": 1.3482,
      "theoretical_loss": 3.3932948109791226,
      "tokens_seen": 2534670336
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002345035105315948,
      "loss": 1.3743,
      "theoretical_loss": 3.3932878905028954,
      "tokens_seen": 2534735872
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023448345035105317,
      "loss": 1.3947,
      "theoretical_loss": 3.393280970255695,
      "tokens_seen": 2534801408
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023446339017051154,
      "loss": 1.2994,
      "theoretical_loss": 3.3932740502375074,
      "tokens_seen": 2534866944
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023444332998996992,
      "loss": 1.4401,
      "theoretical_loss": 3.39326713044832,
      "tokens_seen": 2534932480
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002344232698094283,
      "loss": 1.2451,
      "theoretical_loss": 3.3932602108881182,
      "tokens_seen": 2534998016
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023440320962888668,
      "loss": 1.4127,
      "theoretical_loss": 3.3932532915568894,
      "tokens_seen": 2535063552
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023438314944834504,
      "loss": 1.4395,
      "theoretical_loss": 3.3932463724546196,
      "tokens_seen": 2535129088
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023436308926780343,
      "loss": 1.4363,
      "theoretical_loss": 3.393239453581295,
      "tokens_seen": 2535194624
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002343430290872618,
      "loss": 1.3894,
      "theoretical_loss": 3.393232534936903,
      "tokens_seen": 2535260160
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023432296890672018,
      "loss": 1.3882,
      "theoretical_loss": 3.3932256165214296,
      "tokens_seen": 2535325696
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023430290872617854,
      "loss": 1.4404,
      "theoretical_loss": 3.3932186983348616,
      "tokens_seen": 2535391232
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002342828485456369,
      "loss": 1.3377,
      "theoretical_loss": 3.393211780377185,
      "tokens_seen": 2535456768
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002342627883650953,
      "loss": 1.3469,
      "theoretical_loss": 3.393204862648387,
      "tokens_seen": 2535522304
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023424272818455366,
      "loss": 1.3591,
      "theoretical_loss": 3.393197945148453,
      "tokens_seen": 2535587840
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023422266800401204,
      "loss": 1.3436,
      "theoretical_loss": 3.3931910278773705,
      "tokens_seen": 2535653376
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002342026078234704,
      "loss": 1.2725,
      "theoretical_loss": 3.3931841108351257,
      "tokens_seen": 2535718912
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002341825476429288,
      "loss": 1.2775,
      "theoretical_loss": 3.3931771940217055,
      "tokens_seen": 2535784448
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023416248746238716,
      "loss": 1.3319,
      "theoretical_loss": 3.393170277437096,
      "tokens_seen": 2535849984
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023414242728184555,
      "loss": 1.3835,
      "theoretical_loss": 3.3931633610812835,
      "tokens_seen": 2535915520
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002341223671013039,
      "loss": 1.3624,
      "theoretical_loss": 3.393156444954255,
      "tokens_seen": 2535981056
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002341023069207623,
      "loss": 1.3015,
      "theoretical_loss": 3.3931495290559965,
      "tokens_seen": 2536046592
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023408224674022066,
      "loss": 1.4085,
      "theoretical_loss": 3.393142613386495,
      "tokens_seen": 2536112128
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023406218655967902,
      "loss": 1.3854,
      "theoretical_loss": 3.3931356979457368,
      "tokens_seen": 2536177664
    },
    {
      "epoch": 0.77,
      "objective/train/advantage_avg": 0.4929909110069275,
      "objective/train/docs_used": 922895,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3860373497009277,
      "objective/train/original_loss": 1.3860375881195068,
      "objective/train/theoretical_loss": 3.393128782733709,
      "objective/train/tokens_used": 2556703200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24366992712020874,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505378246307373,
      "objective/train/weighted_lm_loss": 1.456054925918579,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.393128782733709,
      "tokens_seen": 2536243200
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002340421263791374,
      "loss": 1.4572,
      "theoretical_loss": 3.393128782733709,
      "tokens_seen": 2536243200
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023402206619859578,
      "loss": 1.3692,
      "theoretical_loss": 3.3931218677503967,
      "tokens_seen": 2536308736
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023400200601805416,
      "loss": 1.4476,
      "theoretical_loss": 3.393114952995788,
      "tokens_seen": 2536374272
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023398194583751253,
      "loss": 1.3353,
      "theoretical_loss": 3.3931080384698684,
      "tokens_seen": 2536439808
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023396188565697092,
      "loss": 1.4522,
      "theoretical_loss": 3.3931011241726248,
      "tokens_seen": 2536505344
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023394182547642928,
      "loss": 1.3416,
      "theoretical_loss": 3.393094210104044,
      "tokens_seen": 2536570880
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023392176529588767,
      "loss": 1.3218,
      "theoretical_loss": 3.3930872962641123,
      "tokens_seen": 2536636416
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023390170511534603,
      "loss": 1.3036,
      "theoretical_loss": 3.393080382652816,
      "tokens_seen": 2536701952
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002338816449348044,
      "loss": 1.3448,
      "theoretical_loss": 3.393073469270142,
      "tokens_seen": 2536767488
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023386158475426278,
      "loss": 1.3944,
      "theoretical_loss": 3.3930665561160764,
      "tokens_seen": 2536833024
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023384152457372117,
      "loss": 1.4135,
      "theoretical_loss": 3.3930596431906057,
      "tokens_seen": 2536898560
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023382146439317956,
      "loss": 1.3398,
      "theoretical_loss": 3.3930527304937175,
      "tokens_seen": 2536964096
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023380140421263792,
      "loss": 1.246,
      "theoretical_loss": 3.393045818025397,
      "tokens_seen": 2537029632
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002337813440320963,
      "loss": 1.3409,
      "theoretical_loss": 3.393038905785631,
      "tokens_seen": 2537095168
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023376128385155467,
      "loss": 1.3287,
      "theoretical_loss": 3.3930319937744073,
      "tokens_seen": 2537160704
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023374122367101306,
      "loss": 1.3728,
      "theoretical_loss": 3.393025081991711,
      "tokens_seen": 2537226240
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023372116349047143,
      "loss": 1.4576,
      "theoretical_loss": 3.393018170437529,
      "tokens_seen": 2537291776
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023370110330992982,
      "loss": 1.4091,
      "theoretical_loss": 3.393011259111848,
      "tokens_seen": 2537357312
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023368104312938818,
      "loss": 1.3722,
      "theoretical_loss": 3.3930043480146543,
      "tokens_seen": 2537422848
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023366098294884654,
      "loss": 1.3158,
      "theoretical_loss": 3.392997437145935,
      "tokens_seen": 2537488384
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023364092276830493,
      "loss": 1.4488,
      "theoretical_loss": 3.392990526505676,
      "tokens_seen": 2537553920
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002336208625877633,
      "loss": 1.3985,
      "theoretical_loss": 3.3929836160938645,
      "tokens_seen": 2537619456
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023360080240722168,
      "loss": 1.4396,
      "theoretical_loss": 3.3929767059104865,
      "tokens_seen": 2537684992
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023358074222668004,
      "loss": 1.2279,
      "theoretical_loss": 3.392969795955529,
      "tokens_seen": 2537750528
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023356068204613843,
      "loss": 1.3959,
      "theoretical_loss": 3.3929628862289776,
      "tokens_seen": 2537816064
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002335406218655968,
      "loss": 1.3802,
      "theoretical_loss": 3.39295597673082,
      "tokens_seen": 2537881600
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023352056168505518,
      "loss": 1.2298,
      "theoretical_loss": 3.3929490674610423,
      "tokens_seen": 2537947136
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023350050150451355,
      "loss": 1.4904,
      "theoretical_loss": 3.392942158419631,
      "tokens_seen": 2538012672
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002334804413239719,
      "loss": 1.367,
      "theoretical_loss": 3.392935249606573,
      "tokens_seen": 2538078208
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002334603811434303,
      "loss": 1.4625,
      "theoretical_loss": 3.392928341021854,
      "tokens_seen": 2538143744
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023344032096288866,
      "loss": 1.3612,
      "theoretical_loss": 3.392921432665461,
      "tokens_seen": 2538209280
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023342026078234705,
      "loss": 1.4011,
      "theoretical_loss": 3.3929145245373813,
      "tokens_seen": 2538274816
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002334002006018054,
      "loss": 1.3448,
      "theoretical_loss": 3.3929076166376007,
      "tokens_seen": 2538340352
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002333801404212638,
      "loss": 1.3017,
      "theoretical_loss": 3.392900708966106,
      "tokens_seen": 2538405888
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023336008024072216,
      "loss": 1.3073,
      "theoretical_loss": 3.3928938015228836,
      "tokens_seen": 2538471424
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023334002006018055,
      "loss": 1.292,
      "theoretical_loss": 3.39288689430792,
      "tokens_seen": 2538536960
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023331995987963891,
      "loss": 1.3346,
      "theoretical_loss": 3.3928799873212014,
      "tokens_seen": 2538602496
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023329989969909728,
      "loss": 1.3451,
      "theoretical_loss": 3.3928730805627154,
      "tokens_seen": 2538668032
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023327983951855567,
      "loss": 1.2998,
      "theoretical_loss": 3.392866174032448,
      "tokens_seen": 2538733568
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023325977933801403,
      "loss": 1.4135,
      "theoretical_loss": 3.392859267730386,
      "tokens_seen": 2538799104
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023323971915747242,
      "loss": 1.3255,
      "theoretical_loss": 3.3928523616565154,
      "tokens_seen": 2538864640
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023321965897693078,
      "loss": 1.2906,
      "theoretical_loss": 3.392845455810823,
      "tokens_seen": 2538930176
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023319959879638917,
      "loss": 1.348,
      "theoretical_loss": 3.392838550193296,
      "tokens_seen": 2538995712
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023317953861584753,
      "loss": 1.2719,
      "theoretical_loss": 3.3928316448039206,
      "tokens_seen": 2539061248
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023315947843530592,
      "loss": 1.3679,
      "theoretical_loss": 3.3928247396426827,
      "tokens_seen": 2539126784
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023313941825476428,
      "loss": 1.3927,
      "theoretical_loss": 3.3928178347095694,
      "tokens_seen": 2539192320
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023311935807422267,
      "loss": 1.2568,
      "theoretical_loss": 3.3928109300045675,
      "tokens_seen": 2539257856
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023309929789368106,
      "loss": 1.2848,
      "theoretical_loss": 3.3928040255276635,
      "tokens_seen": 2539323392
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023307923771313942,
      "loss": 1.2563,
      "theoretical_loss": 3.3927971212788437,
      "tokens_seen": 2539388928
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023305917753259781,
      "loss": 1.3469,
      "theoretical_loss": 3.392790217258095,
      "tokens_seen": 2539454464
    },
    {
      "epoch": 0.77,
      "objective/train/advantage_avg": 0.49056997895240784,
      "objective/train/docs_used": 924078,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2718769311904907,
      "objective/train/original_loss": 1.2718766927719116,
      "objective/train/theoretical_loss": 3.3927833134654035,
      "objective/train/tokens_used": 2559980000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2412417083978653,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502833127975464,
      "objective/train/weighted_lm_loss": 1.3357346057891846,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3927833134654035,
      "tokens_seen": 2539520000
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023303911735205618,
      "loss": 1.4163,
      "theoretical_loss": 3.3927833134654035,
      "tokens_seen": 2539520000
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023301905717151457,
      "loss": 1.3285,
      "theoretical_loss": 3.3927764099007565,
      "tokens_seen": 2539585536
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023299899699097293,
      "loss": 1.346,
      "theoretical_loss": 3.3927695065641394,
      "tokens_seen": 2539651072
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023297893681043132,
      "loss": 1.4368,
      "theoretical_loss": 3.3927626034555405,
      "tokens_seen": 2539716608
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023295887662988968,
      "loss": 1.314,
      "theoretical_loss": 3.392755700574945,
      "tokens_seen": 2539782144
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023293881644934807,
      "loss": 1.2209,
      "theoretical_loss": 3.39274879792234,
      "tokens_seen": 2539847680
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023291875626880643,
      "loss": 1.3649,
      "theoretical_loss": 3.392741895497712,
      "tokens_seen": 2539913216
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002328986960882648,
      "loss": 1.3072,
      "theoretical_loss": 3.3927349933010476,
      "tokens_seen": 2539978752
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023287863590772318,
      "loss": 1.389,
      "theoretical_loss": 3.3927280913323337,
      "tokens_seen": 2540044288
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023285857572718154,
      "loss": 1.1659,
      "theoretical_loss": 3.392721189591556,
      "tokens_seen": 2540109824
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023283851554663993,
      "loss": 1.3335,
      "theoretical_loss": 3.392714288078702,
      "tokens_seen": 2540175360
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002328184553660983,
      "loss": 1.3425,
      "theoretical_loss": 3.392707386793758,
      "tokens_seen": 2540240896
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023279839518555669,
      "loss": 1.2781,
      "theoretical_loss": 3.3927004857367105,
      "tokens_seen": 2540306432
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023277833500501505,
      "loss": 1.3596,
      "theoretical_loss": 3.3926935849075464,
      "tokens_seen": 2540371968
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023275827482447344,
      "loss": 1.3722,
      "theoretical_loss": 3.3926866843062515,
      "tokens_seen": 2540437504
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002327382146439318,
      "loss": 1.334,
      "theoretical_loss": 3.3926797839328136,
      "tokens_seen": 2540503040
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023271815446339016,
      "loss": 1.2858,
      "theoretical_loss": 3.392672883787218,
      "tokens_seen": 2540568576
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023269809428284855,
      "loss": 1.3268,
      "theoretical_loss": 3.3926659838694526,
      "tokens_seen": 2540634112
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002326780341023069,
      "loss": 1.3692,
      "theoretical_loss": 3.392659084179503,
      "tokens_seen": 2540699648
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002326579739217653,
      "loss": 1.3032,
      "theoretical_loss": 3.3926521847173565,
      "tokens_seen": 2540765184
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023263791374122366,
      "loss": 1.2923,
      "theoretical_loss": 3.392645285482999,
      "tokens_seen": 2540830720
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023261785356068205,
      "loss": 1.2734,
      "theoretical_loss": 3.3926383864764174,
      "tokens_seen": 2540896256
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023259779338014042,
      "loss": 1.3461,
      "theoretical_loss": 3.3926314876975985,
      "tokens_seen": 2540961792
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002325777331995988,
      "loss": 1.2292,
      "theoretical_loss": 3.392624589146529,
      "tokens_seen": 2541027328
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023255767301905717,
      "loss": 1.3982,
      "theoretical_loss": 3.392617690823195,
      "tokens_seen": 2541092864
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023253761283851556,
      "loss": 1.4014,
      "theoretical_loss": 3.3926107927275835,
      "tokens_seen": 2541158400
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023251755265797392,
      "loss": 1.2848,
      "theoretical_loss": 3.3926038948596813,
      "tokens_seen": 2541223936
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023249749247743228,
      "loss": 1.3773,
      "theoretical_loss": 3.3925969972194743,
      "tokens_seen": 2541289472
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023247743229689067,
      "loss": 1.3627,
      "theoretical_loss": 3.3925900998069496,
      "tokens_seen": 2541355008
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023245737211634903,
      "loss": 1.2603,
      "theoretical_loss": 3.392583202622094,
      "tokens_seen": 2541420544
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023243731193580742,
      "loss": 1.361,
      "theoretical_loss": 3.3925763056648934,
      "tokens_seen": 2541486080
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023241725175526578,
      "loss": 1.3392,
      "theoretical_loss": 3.3925694089353353,
      "tokens_seen": 2541551616
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023239719157472417,
      "loss": 1.3959,
      "theoretical_loss": 3.3925625124334053,
      "tokens_seen": 2541617152
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023237713139418254,
      "loss": 1.2652,
      "theoretical_loss": 3.392555616159091,
      "tokens_seen": 2541682688
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023235707121364095,
      "loss": 1.3456,
      "theoretical_loss": 3.3925487201123787,
      "tokens_seen": 2541748224
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023233701103309932,
      "loss": 1.4024,
      "theoretical_loss": 3.392541824293255,
      "tokens_seen": 2541813760
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023231695085255768,
      "loss": 1.3819,
      "theoretical_loss": 3.392534928701706,
      "tokens_seen": 2541879296
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023229689067201607,
      "loss": 1.3976,
      "theoretical_loss": 3.3925280333377197,
      "tokens_seen": 2541944832
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023227683049147443,
      "loss": 1.3725,
      "theoretical_loss": 3.392521138201281,
      "tokens_seen": 2542010368
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023225677031093282,
      "loss": 1.3531,
      "theoretical_loss": 3.3925142432923776,
      "tokens_seen": 2542075904
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023223671013039118,
      "loss": 1.3061,
      "theoretical_loss": 3.392507348610996,
      "tokens_seen": 2542141440
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023221664994984957,
      "loss": 1.3453,
      "theoretical_loss": 3.392500454157122,
      "tokens_seen": 2542206976
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023219658976930793,
      "loss": 1.3762,
      "theoretical_loss": 3.392493559930744,
      "tokens_seen": 2542272512
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023217652958876632,
      "loss": 1.4996,
      "theoretical_loss": 3.392486665931847,
      "tokens_seen": 2542338048
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023215646940822468,
      "loss": 1.2486,
      "theoretical_loss": 3.392479772160418,
      "tokens_seen": 2542403584
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023213640922768307,
      "loss": 1.3835,
      "theoretical_loss": 3.392472878616444,
      "tokens_seen": 2542469120
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023211634904714144,
      "loss": 1.3858,
      "theoretical_loss": 3.3924659852999115,
      "tokens_seen": 2542534656
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002320962888665998,
      "loss": 1.3521,
      "theoretical_loss": 3.392459092210807,
      "tokens_seen": 2542600192
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002320762286860582,
      "loss": 1.4662,
      "theoretical_loss": 3.3924521993491172,
      "tokens_seen": 2542665728
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023205616850551655,
      "loss": 1.2915,
      "theoretical_loss": 3.3924453067148286,
      "tokens_seen": 2542731264
    },
    {
      "epoch": 0.77,
      "objective/train/advantage_avg": 0.49153921008110046,
      "objective/train/docs_used": 925229,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.0878748893737793,
      "objective/train/original_loss": 1.0878746509552002,
      "objective/train/theoretical_loss": 3.3924384143079283,
      "objective/train/tokens_used": 2563256800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2419668585062027,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503840446472168,
      "objective/train/weighted_lm_loss": 1.1427239179611206,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3924384143079283,
      "tokens_seen": 2542796800
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023203610832497494,
      "loss": 1.2531,
      "theoretical_loss": 3.3924384143079283,
      "tokens_seen": 2542796800
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002320160481444333,
      "loss": 1.4138,
      "theoretical_loss": 3.3924315221284025,
      "tokens_seen": 2542862336
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002319959879638917,
      "loss": 1.3994,
      "theoretical_loss": 3.392424630176238,
      "tokens_seen": 2542927872
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023197592778335005,
      "loss": 1.3664,
      "theoretical_loss": 3.392417738451421,
      "tokens_seen": 2542993408
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023195586760280844,
      "loss": 1.4915,
      "theoretical_loss": 3.392410846953939,
      "tokens_seen": 2543058944
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002319358074222668,
      "loss": 1.3743,
      "theoretical_loss": 3.392403955683778,
      "tokens_seen": 2543124480
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023191574724172517,
      "loss": 1.375,
      "theoretical_loss": 3.392397064640925,
      "tokens_seen": 2543190016
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023189568706118356,
      "loss": 1.3652,
      "theoretical_loss": 3.3923901738253663,
      "tokens_seen": 2543255552
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023187562688064192,
      "loss": 1.3564,
      "theoretical_loss": 3.3923832832370886,
      "tokens_seen": 2543321088
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002318555667001003,
      "loss": 1.326,
      "theoretical_loss": 3.3923763928760793,
      "tokens_seen": 2543386624
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023183550651955867,
      "loss": 1.3933,
      "theoretical_loss": 3.3923695027423237,
      "tokens_seen": 2543452160
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023181544633901706,
      "loss": 1.4591,
      "theoretical_loss": 3.3923626128358095,
      "tokens_seen": 2543517696
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023179538615847542,
      "loss": 1.3871,
      "theoretical_loss": 3.392355723156523,
      "tokens_seen": 2543583232
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002317753259779338,
      "loss": 1.3631,
      "theoretical_loss": 3.3923488337044505,
      "tokens_seen": 2543648768
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023175526579739217,
      "loss": 1.2903,
      "theoretical_loss": 3.3923419444795795,
      "tokens_seen": 2543714304
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023173520561685053,
      "loss": 1.3843,
      "theoretical_loss": 3.392335055481896,
      "tokens_seen": 2543779840
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023171514543630892,
      "loss": 1.4421,
      "theoretical_loss": 3.392328166711387,
      "tokens_seen": 2543845376
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023169508525576729,
      "loss": 1.3911,
      "theoretical_loss": 3.3923212781680387,
      "tokens_seen": 2543910912
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023167502507522568,
      "loss": 1.3774,
      "theoretical_loss": 3.392314389851838,
      "tokens_seen": 2543976448
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023165496489468404,
      "loss": 1.328,
      "theoretical_loss": 3.392307501762772,
      "tokens_seen": 2544041984
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023163490471414243,
      "loss": 1.3023,
      "theoretical_loss": 3.3923006139008267,
      "tokens_seen": 2544107520
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002316148445336008,
      "loss": 1.3338,
      "theoretical_loss": 3.392293726265989,
      "tokens_seen": 2544173056
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002315947843530592,
      "loss": 1.4133,
      "theoretical_loss": 3.392286838858246,
      "tokens_seen": 2544238592
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023157472417251757,
      "loss": 1.3053,
      "theoretical_loss": 3.3922799516775832,
      "tokens_seen": 2544304128
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023155466399197596,
      "loss": 1.2647,
      "theoretical_loss": 3.3922730647239887,
      "tokens_seen": 2544369664
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023153460381143432,
      "loss": 1.3383,
      "theoretical_loss": 3.392266177997448,
      "tokens_seen": 2544435200
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023151454363089268,
      "loss": 1.2097,
      "theoretical_loss": 3.3922592914979486,
      "tokens_seen": 2544500736
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023149448345035107,
      "loss": 1.3598,
      "theoretical_loss": 3.3922524052254763,
      "tokens_seen": 2544566272
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023147442326980943,
      "loss": 1.3216,
      "theoretical_loss": 3.392245519180019,
      "tokens_seen": 2544631808
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023145436308926782,
      "loss": 1.3839,
      "theoretical_loss": 3.392238633361562,
      "tokens_seen": 2544697344
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023143430290872619,
      "loss": 1.2673,
      "theoretical_loss": 3.392231747770093,
      "tokens_seen": 2544762880
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023141424272818457,
      "loss": 1.3379,
      "theoretical_loss": 3.392224862405598,
      "tokens_seen": 2544828416
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023139418254764294,
      "loss": 1.3386,
      "theoretical_loss": 3.3922179772680643,
      "tokens_seen": 2544893952
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023137412236710133,
      "loss": 1.2764,
      "theoretical_loss": 3.392211092357478,
      "tokens_seen": 2544959488
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002313540621865597,
      "loss": 1.5141,
      "theoretical_loss": 3.392204207673826,
      "tokens_seen": 2545025024
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023133400200601805,
      "loss": 1.3639,
      "theoretical_loss": 3.392197323217095,
      "tokens_seen": 2545090560
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023131394182547644,
      "loss": 1.3007,
      "theoretical_loss": 3.392190438987272,
      "tokens_seen": 2545156096
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002312938816449348,
      "loss": 1.393,
      "theoretical_loss": 3.392183554984343,
      "tokens_seen": 2545221632
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002312738214643932,
      "loss": 1.3765,
      "theoretical_loss": 3.3921766712082952,
      "tokens_seen": 2545287168
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023125376128385155,
      "loss": 1.4754,
      "theoretical_loss": 3.3921697876591153,
      "tokens_seen": 2545352704
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023123370110330994,
      "loss": 1.2673,
      "theoretical_loss": 3.39216290433679,
      "tokens_seen": 2545418240
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002312136409227683,
      "loss": 1.3728,
      "theoretical_loss": 3.392156021241305,
      "tokens_seen": 2545483776
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002311935807422267,
      "loss": 1.3481,
      "theoretical_loss": 3.3921491383726483,
      "tokens_seen": 2545549312
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023117352056168506,
      "loss": 1.2592,
      "theoretical_loss": 3.3921422557308056,
      "tokens_seen": 2545614848
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023115346038114342,
      "loss": 1.3331,
      "theoretical_loss": 3.392135373315764,
      "tokens_seen": 2545680384
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002311334002006018,
      "loss": 1.3218,
      "theoretical_loss": 3.3921284911275107,
      "tokens_seen": 2545745920
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023111334002006017,
      "loss": 1.4039,
      "theoretical_loss": 3.392121609166032,
      "tokens_seen": 2545811456
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023109327983951856,
      "loss": 1.2678,
      "theoretical_loss": 3.392114727431314,
      "tokens_seen": 2545876992
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023107321965897692,
      "loss": 1.2607,
      "theoretical_loss": 3.392107845923344,
      "tokens_seen": 2545942528
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002310531594784353,
      "loss": 1.2556,
      "theoretical_loss": 3.3921009646421085,
      "tokens_seen": 2546008064
    },
    {
      "epoch": 0.77,
      "objective/train/advantage_avg": 0.49309083819389343,
      "objective/train/docs_used": 926243,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.202783465385437,
      "objective/train/original_loss": 1.2027833461761475,
      "objective/train/theoretical_loss": 3.392094083587595,
      "objective/train/tokens_used": 2566533600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24342311918735504,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050546646118164,
      "objective/train/weighted_lm_loss": 1.2635185718536377,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.392094083587595,
      "tokens_seen": 2546073600
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023103309929789367,
      "loss": 1.2677,
      "theoretical_loss": 3.392094083587595,
      "tokens_seen": 2546073600
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023101303911735206,
      "loss": 1.3263,
      "theoretical_loss": 3.3920872027597886,
      "tokens_seen": 2546139136
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023099297893681043,
      "loss": 1.4076,
      "theoretical_loss": 3.3920803221586775,
      "tokens_seen": 2546204672
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023097291875626881,
      "loss": 1.2816,
      "theoretical_loss": 3.3920734417842473,
      "tokens_seen": 2546270208
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023095285857572718,
      "loss": 1.3822,
      "theoretical_loss": 3.392066561636485,
      "tokens_seen": 2546335744
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023093279839518554,
      "loss": 1.2785,
      "theoretical_loss": 3.392059681715378,
      "tokens_seen": 2546401280
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023091273821464393,
      "loss": 1.3184,
      "theoretical_loss": 3.3920528020209124,
      "tokens_seen": 2546466816
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002308926780341023,
      "loss": 1.2253,
      "theoretical_loss": 3.392045922553075,
      "tokens_seen": 2546532352
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023087261785356068,
      "loss": 1.3016,
      "theoretical_loss": 3.392039043311852,
      "tokens_seen": 2546597888
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023085255767301907,
      "loss": 1.2756,
      "theoretical_loss": 3.392032164297231,
      "tokens_seen": 2546663424
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023083249749247746,
      "loss": 1.2988,
      "theoretical_loss": 3.392025285509198,
      "tokens_seen": 2546728960
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023081243731193582,
      "loss": 1.3004,
      "theoretical_loss": 3.39201840694774,
      "tokens_seen": 2546794496
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002307923771313942,
      "loss": 1.3498,
      "theoretical_loss": 3.392011528612844,
      "tokens_seen": 2546860032
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023077231695085257,
      "loss": 1.2625,
      "theoretical_loss": 3.392004650504496,
      "tokens_seen": 2546925568
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023075225677031094,
      "loss": 1.2983,
      "theoretical_loss": 3.391997772622683,
      "tokens_seen": 2546991104
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023073219658976932,
      "loss": 1.2459,
      "theoretical_loss": 3.3919908949673925,
      "tokens_seen": 2547056640
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002307121364092277,
      "loss": 1.317,
      "theoretical_loss": 3.39198401753861,
      "tokens_seen": 2547122176
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023069207622868608,
      "loss": 1.313,
      "theoretical_loss": 3.391977140336323,
      "tokens_seen": 2547187712
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023067201604814444,
      "loss": 1.2569,
      "theoretical_loss": 3.3919702633605175,
      "tokens_seen": 2547253248
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023065195586760283,
      "loss": 1.1677,
      "theoretical_loss": 3.391963386611181,
      "tokens_seen": 2547318784
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002306318956870612,
      "loss": 1.4167,
      "theoretical_loss": 3.3919565100883,
      "tokens_seen": 2547384320
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023061183550651958,
      "loss": 1.2999,
      "theoretical_loss": 3.391949633791861,
      "tokens_seen": 2547449856
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023059177532597794,
      "loss": 1.2363,
      "theoretical_loss": 3.391942757721851,
      "tokens_seen": 2547515392
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002305717151454363,
      "loss": 1.3469,
      "theoretical_loss": 3.391935881878256,
      "tokens_seen": 2547580928
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002305516549648947,
      "loss": 1.3335,
      "theoretical_loss": 3.3919290062610634,
      "tokens_seen": 2547646464
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023053159478435306,
      "loss": 1.2525,
      "theoretical_loss": 3.39192213087026,
      "tokens_seen": 2547712000
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023051153460381144,
      "loss": 1.4335,
      "theoretical_loss": 3.3919152557058325,
      "tokens_seen": 2547777536
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002304914744232698,
      "loss": 1.404,
      "theoretical_loss": 3.391908380767767,
      "tokens_seen": 2547843072
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002304714142427282,
      "loss": 1.3297,
      "theoretical_loss": 3.3919015060560507,
      "tokens_seen": 2547908608
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023045135406218656,
      "loss": 1.2419,
      "theoretical_loss": 3.39189463157067,
      "tokens_seen": 2547974144
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023043129388164495,
      "loss": 1.3038,
      "theoretical_loss": 3.3918877573116126,
      "tokens_seen": 2548039680
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002304112337011033,
      "loss": 1.3876,
      "theoretical_loss": 3.3918808832788643,
      "tokens_seen": 2548105216
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002303911735205617,
      "loss": 1.3682,
      "theoretical_loss": 3.391874009472412,
      "tokens_seen": 2548170752
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023037111334002006,
      "loss": 1.2984,
      "theoretical_loss": 3.3918671358922423,
      "tokens_seen": 2548236288
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023035105315947842,
      "loss": 1.3958,
      "theoretical_loss": 3.3918602625383425,
      "tokens_seen": 2548301824
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002303309929789368,
      "loss": 1.362,
      "theoretical_loss": 3.3918533894106986,
      "tokens_seen": 2548367360
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023031093279839518,
      "loss": 1.3726,
      "theoretical_loss": 3.3918465165092977,
      "tokens_seen": 2548432896
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023029087261785356,
      "loss": 1.2612,
      "theoretical_loss": 3.391839643834127,
      "tokens_seen": 2548498432
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023027081243731193,
      "loss": 1.3612,
      "theoretical_loss": 3.391832771385172,
      "tokens_seen": 2548563968
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023025075225677032,
      "loss": 1.3532,
      "theoretical_loss": 3.391825899162421,
      "tokens_seen": 2548629504
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023023069207622868,
      "loss": 1.2982,
      "theoretical_loss": 3.3918190271658593,
      "tokens_seen": 2548695040
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023021063189568707,
      "loss": 1.2719,
      "theoretical_loss": 3.3918121553954745,
      "tokens_seen": 2548760576
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023019057171514543,
      "loss": 1.3243,
      "theoretical_loss": 3.3918052838512533,
      "tokens_seen": 2548826112
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002301705115346038,
      "loss": 1.3922,
      "theoretical_loss": 3.391798412533182,
      "tokens_seen": 2548891648
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023015045135406218,
      "loss": 1.2516,
      "theoretical_loss": 3.391791541441248,
      "tokens_seen": 2548957184
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023013039117352054,
      "loss": 1.3467,
      "theoretical_loss": 3.391784670575437,
      "tokens_seen": 2549022720
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023011033099297893,
      "loss": 1.3179,
      "theoretical_loss": 3.3917777999357366,
      "tokens_seen": 2549088256
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023009027081243732,
      "loss": 1.3457,
      "theoretical_loss": 3.3917709295221337,
      "tokens_seen": 2549153792
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002300702106318957,
      "loss": 1.4223,
      "theoretical_loss": 3.3917640593346143,
      "tokens_seen": 2549219328
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023005015045135407,
      "loss": 1.3274,
      "theoretical_loss": 3.3917571893731657,
      "tokens_seen": 2549284864
    },
    {
      "epoch": 0.77,
      "objective/train/advantage_avg": 0.49393928050994873,
      "objective/train/docs_used": 927515,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.270937204360962,
      "objective/train/original_loss": 1.2709370851516724,
      "objective/train/theoretical_loss": 3.3917503196377745,
      "objective/train/tokens_used": 2569810400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2442527711391449,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506356954574585,
      "objective/train/weighted_lm_loss": 1.3353575468063354,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3917503196377745,
      "tokens_seen": 2549350400
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023003009027081246,
      "loss": 1.3757,
      "theoretical_loss": 3.3917503196377745,
      "tokens_seen": 2549350400
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00023001003009027083,
      "loss": 1.3399,
      "theoretical_loss": 3.3917434501284274,
      "tokens_seen": 2549415936
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022998996990972922,
      "loss": 1.1963,
      "theoretical_loss": 3.391736580845111,
      "tokens_seen": 2549481472
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022996990972918758,
      "loss": 1.4945,
      "theoretical_loss": 3.3917297117878125,
      "tokens_seen": 2549547008
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022994984954864594,
      "loss": 1.3165,
      "theoretical_loss": 3.391722842956518,
      "tokens_seen": 2549612544
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022992978936810433,
      "loss": 1.317,
      "theoretical_loss": 3.391715974351215,
      "tokens_seen": 2549678080
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002299097291875627,
      "loss": 1.2227,
      "theoretical_loss": 3.3917091059718896,
      "tokens_seen": 2549743616
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022988966900702108,
      "loss": 1.3493,
      "theoretical_loss": 3.3917022378185293,
      "tokens_seen": 2549809152
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022986960882647944,
      "loss": 1.3008,
      "theoretical_loss": 3.3916953698911203,
      "tokens_seen": 2549874688
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022984954864593783,
      "loss": 1.3135,
      "theoretical_loss": 3.3916885021896492,
      "tokens_seen": 2549940224
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002298294884653962,
      "loss": 1.3667,
      "theoretical_loss": 3.391681634714103,
      "tokens_seen": 2550005760
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022980942828485458,
      "loss": 1.3472,
      "theoretical_loss": 3.391674767464469,
      "tokens_seen": 2550071296
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022978936810431295,
      "loss": 1.2562,
      "theoretical_loss": 3.391667900440733,
      "tokens_seen": 2550136832
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002297693079237713,
      "loss": 1.286,
      "theoretical_loss": 3.3916610336428823,
      "tokens_seen": 2550202368
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002297492477432297,
      "loss": 1.4338,
      "theoretical_loss": 3.3916541670709037,
      "tokens_seen": 2550267904
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022972918756268806,
      "loss": 1.3475,
      "theoretical_loss": 3.391647300724784,
      "tokens_seen": 2550333440
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022970912738214645,
      "loss": 1.3017,
      "theoretical_loss": 3.3916404346045095,
      "tokens_seen": 2550398976
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002296890672016048,
      "loss": 1.3821,
      "theoretical_loss": 3.3916335687100676,
      "tokens_seen": 2550464512
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002296690070210632,
      "loss": 1.2639,
      "theoretical_loss": 3.391626703041444,
      "tokens_seen": 2550530048
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022964894684052156,
      "loss": 1.2889,
      "theoretical_loss": 3.3916198375986273,
      "tokens_seen": 2550595584
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022962888665997995,
      "loss": 1.3688,
      "theoretical_loss": 3.3916129723816026,
      "tokens_seen": 2550661120
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022960882647943831,
      "loss": 1.336,
      "theoretical_loss": 3.3916061073903574,
      "tokens_seen": 2550726656
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022958876629889668,
      "loss": 1.4028,
      "theoretical_loss": 3.3915992426248787,
      "tokens_seen": 2550792192
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022956870611835507,
      "loss": 1.471,
      "theoretical_loss": 3.3915923780851522,
      "tokens_seen": 2550857728
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022954864593781343,
      "loss": 1.2054,
      "theoretical_loss": 3.3915855137711657,
      "tokens_seen": 2550923264
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022952858575727182,
      "loss": 1.2508,
      "theoretical_loss": 3.3915786496829057,
      "tokens_seen": 2550988800
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022950852557673018,
      "loss": 1.1792,
      "theoretical_loss": 3.391571785820359,
      "tokens_seen": 2551054336
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022948846539618857,
      "loss": 1.3005,
      "theoretical_loss": 3.3915649221835125,
      "tokens_seen": 2551119872
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022946840521564693,
      "loss": 1.3874,
      "theoretical_loss": 3.3915580587723526,
      "tokens_seen": 2551185408
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022944834503510532,
      "loss": 1.2947,
      "theoretical_loss": 3.3915511955868665,
      "tokens_seen": 2551250944
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022942828485456368,
      "loss": 1.3029,
      "theoretical_loss": 3.3915443326270407,
      "tokens_seen": 2551316480
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022940822467402207,
      "loss": 1.2321,
      "theoretical_loss": 3.391537469892862,
      "tokens_seen": 2551382016
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022938816449348043,
      "loss": 1.318,
      "theoretical_loss": 3.391530607384317,
      "tokens_seen": 2551447552
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002293681043129388,
      "loss": 1.1894,
      "theoretical_loss": 3.3915237451013933,
      "tokens_seen": 2551513088
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002293480441323972,
      "loss": 1.4127,
      "theoretical_loss": 3.3915168830440767,
      "tokens_seen": 2551578624
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022932798395185558,
      "loss": 1.3659,
      "theoretical_loss": 3.3915100212123543,
      "tokens_seen": 2551644160
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022930792377131397,
      "loss": 1.2576,
      "theoretical_loss": 3.3915031596062133,
      "tokens_seen": 2551709696
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022928786359077233,
      "loss": 1.2983,
      "theoretical_loss": 3.3914962982256402,
      "tokens_seen": 2551775232
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022926780341023072,
      "loss": 1.2187,
      "theoretical_loss": 3.3914894370706214,
      "tokens_seen": 2551840768
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022924774322968908,
      "loss": 1.3743,
      "theoretical_loss": 3.3914825761411445,
      "tokens_seen": 2551906304
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022922768304914747,
      "loss": 1.3388,
      "theoretical_loss": 3.391475715437196,
      "tokens_seen": 2551971840
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022920762286860583,
      "loss": 1.3192,
      "theoretical_loss": 3.391468854958762,
      "tokens_seen": 2552037376
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002291875626880642,
      "loss": 1.3603,
      "theoretical_loss": 3.3914619947058307,
      "tokens_seen": 2552102912
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022916750250752258,
      "loss": 1.2407,
      "theoretical_loss": 3.391455134678387,
      "tokens_seen": 2552168448
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022914744232698094,
      "loss": 1.2301,
      "theoretical_loss": 3.39144827487642,
      "tokens_seen": 2552233984
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022912738214643933,
      "loss": 1.3565,
      "theoretical_loss": 3.3914414152999144,
      "tokens_seen": 2552299520
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002291073219658977,
      "loss": 1.3952,
      "theoretical_loss": 3.391434555948858,
      "tokens_seen": 2552365056
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022908726178535609,
      "loss": 1.4507,
      "theoretical_loss": 3.3914276968232375,
      "tokens_seen": 2552430592
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022906720160481445,
      "loss": 1.238,
      "theoretical_loss": 3.3914208379230395,
      "tokens_seen": 2552496128
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022904714142427284,
      "loss": 1.3803,
      "theoretical_loss": 3.3914139792482514,
      "tokens_seen": 2552561664
    },
    {
      "epoch": 0.77,
      "objective/train/advantage_avg": 0.4878837764263153,
      "objective/train/docs_used": 928605,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2220240831375122,
      "objective/train/original_loss": 1.2220239639282227,
      "objective/train/theoretical_loss": 3.391407120798859,
      "objective/train/tokens_used": 2573087200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.23892426490783691,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500028133392334,
      "objective/train/weighted_lm_loss": 1.2831807136535645,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.391407120798859,
      "tokens_seen": 2552627200
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002290270812437312,
      "loss": 1.2491,
      "theoretical_loss": 3.391407120798859,
      "tokens_seen": 2552627200
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022900702106318956,
      "loss": 1.4281,
      "theoretical_loss": 3.39140026257485,
      "tokens_seen": 2552692736
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022898696088264795,
      "loss": 1.3518,
      "theoretical_loss": 3.3913934045762106,
      "tokens_seen": 2552758272
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002289669007021063,
      "loss": 1.3101,
      "theoretical_loss": 3.3913865468029285,
      "tokens_seen": 2552823808
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002289468405215647,
      "loss": 1.3418,
      "theoretical_loss": 3.3913796892549897,
      "tokens_seen": 2552889344
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022892678034102306,
      "loss": 1.2848,
      "theoretical_loss": 3.3913728319323813,
      "tokens_seen": 2552954880
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022890672016048145,
      "loss": 1.2313,
      "theoretical_loss": 3.3913659748350895,
      "tokens_seen": 2553020416
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022888665997993982,
      "loss": 1.3703,
      "theoretical_loss": 3.391359117963102,
      "tokens_seen": 2553085952
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002288665997993982,
      "loss": 1.405,
      "theoretical_loss": 3.3913522613164053,
      "tokens_seen": 2553151488
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022884653961885657,
      "loss": 1.3229,
      "theoretical_loss": 3.391345404894986,
      "tokens_seen": 2553217024
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022882647943831496,
      "loss": 1.3356,
      "theoretical_loss": 3.391338548698831,
      "tokens_seen": 2553282560
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022880641925777332,
      "loss": 1.3087,
      "theoretical_loss": 3.3913316927279276,
      "tokens_seen": 2553348096
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022878635907723168,
      "loss": 1.3281,
      "theoretical_loss": 3.3913248369822617,
      "tokens_seen": 2553413632
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022876629889669007,
      "loss": 1.2432,
      "theoretical_loss": 3.391317981461821,
      "tokens_seen": 2553479168
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022874623871614843,
      "loss": 1.446,
      "theoretical_loss": 3.391311126166592,
      "tokens_seen": 2553544704
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022872617853560682,
      "loss": 1.2623,
      "theoretical_loss": 3.3913042710965615,
      "tokens_seen": 2553610240
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022870611835506518,
      "loss": 1.2475,
      "theoretical_loss": 3.391297416251716,
      "tokens_seen": 2553675776
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022868605817452357,
      "loss": 1.3002,
      "theoretical_loss": 3.391290561632043,
      "tokens_seen": 2553741312
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022866599799398194,
      "loss": 1.3395,
      "theoretical_loss": 3.3912837072375286,
      "tokens_seen": 2553806848
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022864593781344033,
      "loss": 1.3801,
      "theoretical_loss": 3.39127685306816,
      "tokens_seen": 2553872384
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002286258776328987,
      "loss": 1.3415,
      "theoretical_loss": 3.3912699991239244,
      "tokens_seen": 2553937920
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022860581745235705,
      "loss": 1.2577,
      "theoretical_loss": 3.391263145404808,
      "tokens_seen": 2554003456
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022858575727181547,
      "loss": 1.373,
      "theoretical_loss": 3.391256291910798,
      "tokens_seen": 2554068992
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022856569709127383,
      "loss": 1.3385,
      "theoretical_loss": 3.391249438641881,
      "tokens_seen": 2554134528
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022854563691073222,
      "loss": 1.2631,
      "theoretical_loss": 3.391242585598044,
      "tokens_seen": 2554200064
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022852557673019058,
      "loss": 1.3301,
      "theoretical_loss": 3.3912357327792737,
      "tokens_seen": 2554265600
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022850551654964897,
      "loss": 1.3572,
      "theoretical_loss": 3.391228880185557,
      "tokens_seen": 2554331136
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022848545636910733,
      "loss": 1.3029,
      "theoretical_loss": 3.3912220278168803,
      "tokens_seen": 2554396672
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022846539618856572,
      "loss": 1.2952,
      "theoretical_loss": 3.3912151756732314,
      "tokens_seen": 2554462208
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022844533600802408,
      "loss": 1.2283,
      "theoretical_loss": 3.3912083237545967,
      "tokens_seen": 2554527744
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022842527582748247,
      "loss": 1.3247,
      "theoretical_loss": 3.3912014720609625,
      "tokens_seen": 2554593280
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022840521564694084,
      "loss": 1.3697,
      "theoretical_loss": 3.3911946205923167,
      "tokens_seen": 2554658816
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002283851554663992,
      "loss": 1.2779,
      "theoretical_loss": 3.3911877693486447,
      "tokens_seen": 2554724352
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002283650952858576,
      "loss": 1.4198,
      "theoretical_loss": 3.391180918329935,
      "tokens_seen": 2554789888
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022834503510531595,
      "loss": 1.3809,
      "theoretical_loss": 3.391174067536173,
      "tokens_seen": 2554855424
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022832497492477434,
      "loss": 1.3659,
      "theoretical_loss": 3.3911672169673466,
      "tokens_seen": 2554920960
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002283049147442327,
      "loss": 1.4065,
      "theoretical_loss": 3.391160366623442,
      "tokens_seen": 2554986496
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002282848545636911,
      "loss": 1.2797,
      "theoretical_loss": 3.391153516504446,
      "tokens_seen": 2555052032
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022826479438314945,
      "loss": 1.4519,
      "theoretical_loss": 3.391146666610346,
      "tokens_seen": 2555117568
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022824473420260784,
      "loss": 1.3263,
      "theoretical_loss": 3.3911398169411284,
      "tokens_seen": 2555183104
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002282246740220662,
      "loss": 1.3535,
      "theoretical_loss": 3.3911329674967803,
      "tokens_seen": 2555248640
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022820461384152457,
      "loss": 1.3906,
      "theoretical_loss": 3.3911261182772883,
      "tokens_seen": 2555314176
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022818455366098296,
      "loss": 1.4268,
      "theoretical_loss": 3.3911192692826395,
      "tokens_seen": 2555379712
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022816449348044132,
      "loss": 1.3263,
      "theoretical_loss": 3.3911124205128207,
      "tokens_seen": 2555445248
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002281444332998997,
      "loss": 1.4129,
      "theoretical_loss": 3.3911055719678185,
      "tokens_seen": 2555510784
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022812437311935807,
      "loss": 1.2495,
      "theoretical_loss": 3.39109872364762,
      "tokens_seen": 2555576320
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022810431293881646,
      "loss": 1.3503,
      "theoretical_loss": 3.3910918755522124,
      "tokens_seen": 2555641856
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022808425275827482,
      "loss": 1.383,
      "theoretical_loss": 3.391085027681582,
      "tokens_seen": 2555707392
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002280641925777332,
      "loss": 1.2931,
      "theoretical_loss": 3.3910781800357155,
      "tokens_seen": 2555772928
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022804413239719157,
      "loss": 1.3067,
      "theoretical_loss": 3.3910713326146,
      "tokens_seen": 2555838464
    },
    {
      "epoch": 0.77,
      "objective/train/advantage_avg": 0.4952106177806854,
      "objective/train/docs_used": 929757,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3001960515975952,
      "objective/train/original_loss": 1.3001959323883057,
      "objective/train/theoretical_loss": 3.391064485418223,
      "objective/train/tokens_used": 2576364000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2455010563135147,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0507692098617554,
      "objective/train/weighted_lm_loss": 1.3661308288574219,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0125783681869507,
      "theoretical_loss": 3.391064485418223,
      "tokens_seen": 2555904000
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022802407221664993,
      "loss": 1.3404,
      "theoretical_loss": 3.391064485418223,
      "tokens_seen": 2555904000
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022800401203610832,
      "loss": 1.4333,
      "theoretical_loss": 3.3910576384465703,
      "tokens_seen": 2555969536
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022798395185556669,
      "loss": 1.3257,
      "theoretical_loss": 3.3910507916996298,
      "tokens_seen": 2556035072
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022796389167502508,
      "loss": 1.4168,
      "theoretical_loss": 3.3910439451773873,
      "tokens_seen": 2556100608
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022794383149448344,
      "loss": 1.3382,
      "theoretical_loss": 3.3910370988798304,
      "tokens_seen": 2556166144
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022792377131394183,
      "loss": 1.3918,
      "theoretical_loss": 3.391030252806946,
      "tokens_seen": 2556231680
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002279037111334002,
      "loss": 1.4613,
      "theoretical_loss": 3.3910234069587206,
      "tokens_seen": 2556297216
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022788365095285858,
      "loss": 1.2924,
      "theoretical_loss": 3.391016561335141,
      "tokens_seen": 2556362752
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022786359077231694,
      "loss": 1.365,
      "theoretical_loss": 3.3910097159361943,
      "tokens_seen": 2556428288
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022784353059177536,
      "loss": 1.3939,
      "theoretical_loss": 3.391002870761868,
      "tokens_seen": 2556493824
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022782347041123372,
      "loss": 1.1848,
      "theoretical_loss": 3.3909960258121474,
      "tokens_seen": 2556559360
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022780341023069208,
      "loss": 1.3867,
      "theoretical_loss": 3.3909891810870207,
      "tokens_seen": 2556624896
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022778335005015047,
      "loss": 1.2771,
      "theoretical_loss": 3.3909823365864744,
      "tokens_seen": 2556690432
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022776328986960883,
      "loss": 1.3594,
      "theoretical_loss": 3.390975492310495,
      "tokens_seen": 2556755968
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022774322968906722,
      "loss": 1.4116,
      "theoretical_loss": 3.39096864825907,
      "tokens_seen": 2556821504
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022772316950852558,
      "loss": 1.3568,
      "theoretical_loss": 3.390961804432186,
      "tokens_seen": 2556887040
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022770310932798397,
      "loss": 1.2044,
      "theoretical_loss": 3.3909549608298297,
      "tokens_seen": 2556952576
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022768304914744234,
      "loss": 1.2624,
      "theoretical_loss": 3.3909481174519884,
      "tokens_seen": 2557018112
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022766298896690073,
      "loss": 1.3372,
      "theoretical_loss": 3.3909412742986484,
      "tokens_seen": 2557083648
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002276429287863591,
      "loss": 1.3642,
      "theoretical_loss": 3.3909344313697973,
      "tokens_seen": 2557149184
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022762286860581745,
      "loss": 1.3689,
      "theoretical_loss": 3.3909275886654213,
      "tokens_seen": 2557214720
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022760280842527584,
      "loss": 1.3675,
      "theoretical_loss": 3.3909207461855075,
      "tokens_seen": 2557280256
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002275827482447342,
      "loss": 1.3945,
      "theoretical_loss": 3.390913903930043,
      "tokens_seen": 2557345792
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002275626880641926,
      "loss": 1.2977,
      "theoretical_loss": 3.3909070618990147,
      "tokens_seen": 2557411328
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.00022754262788365095,
      "loss": 1.3436,
      "theoretical_loss": 3.3909002200924094,
      "tokens_seen": 2557476864
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022752256770310934,
      "loss": 1.3926,
      "theoretical_loss": 3.3908933785102136,
      "tokens_seen": 2557542400
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002275025075225677,
      "loss": 1.4244,
      "theoretical_loss": 3.3908865371524146,
      "tokens_seen": 2557607936
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002274824473420261,
      "loss": 1.2704,
      "theoretical_loss": 3.3908796960189993,
      "tokens_seen": 2557673472
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022746238716148446,
      "loss": 1.3393,
      "theoretical_loss": 3.3908728551099543,
      "tokens_seen": 2557739008
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022744232698094282,
      "loss": 1.3382,
      "theoretical_loss": 3.390866014425267,
      "tokens_seen": 2557804544
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002274222668004012,
      "loss": 1.3821,
      "theoretical_loss": 3.3908591739649236,
      "tokens_seen": 2557870080
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022740220661985957,
      "loss": 1.4173,
      "theoretical_loss": 3.390852333728912,
      "tokens_seen": 2557935616
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022738214643931796,
      "loss": 1.2728,
      "theoretical_loss": 3.390845493717218,
      "tokens_seen": 2558001152
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022736208625877632,
      "loss": 1.3764,
      "theoretical_loss": 3.3908386539298285,
      "tokens_seen": 2558066688
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002273420260782347,
      "loss": 1.421,
      "theoretical_loss": 3.3908318143667318,
      "tokens_seen": 2558132224
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022732196589769307,
      "loss": 1.3248,
      "theoretical_loss": 3.3908249750279134,
      "tokens_seen": 2558197760
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022730190571715146,
      "loss": 1.4286,
      "theoretical_loss": 3.3908181359133605,
      "tokens_seen": 2558263296
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022728184553660983,
      "loss": 1.2547,
      "theoretical_loss": 3.39081129702306,
      "tokens_seen": 2558328832
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022726178535606821,
      "loss": 1.3334,
      "theoretical_loss": 3.3908044583569996,
      "tokens_seen": 2558394368
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022724172517552658,
      "loss": 1.3679,
      "theoretical_loss": 3.3907976199151655,
      "tokens_seen": 2558459904
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022722166499498494,
      "loss": 1.3413,
      "theoretical_loss": 3.390790781697544,
      "tokens_seen": 2558525440
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022720160481444333,
      "loss": 1.4692,
      "theoretical_loss": 3.3907839437041236,
      "tokens_seen": 2558590976
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002271815446339017,
      "loss": 1.2927,
      "theoretical_loss": 3.3907771059348897,
      "tokens_seen": 2558656512
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022716148445336008,
      "loss": 1.3285,
      "theoretical_loss": 3.39077026838983,
      "tokens_seen": 2558722048
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022714142427281844,
      "loss": 1.4078,
      "theoretical_loss": 3.390763431068931,
      "tokens_seen": 2558787584
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022712136409227683,
      "loss": 1.3139,
      "theoretical_loss": 3.39075659397218,
      "tokens_seen": 2558853120
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022710130391173522,
      "loss": 1.3709,
      "theoretical_loss": 3.3907497570995635,
      "tokens_seen": 2558918656
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002270812437311936,
      "loss": 1.3312,
      "theoretical_loss": 3.390742920451069,
      "tokens_seen": 2558984192
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022706118355065197,
      "loss": 1.2741,
      "theoretical_loss": 3.390736084026683,
      "tokens_seen": 2559049728
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022704112337011033,
      "loss": 1.2813,
      "theoretical_loss": 3.390729247826392,
      "tokens_seen": 2559115264
    },
    {
      "epoch": 0.78,
      "objective/train/advantage_avg": 0.49315059185028076,
      "objective/train/docs_used": 930868,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1144838333129883,
      "objective/train/original_loss": 1.1144835948944092,
      "objective/train/theoretical_loss": 3.390722411850184,
      "objective/train/tokens_used": 2579640800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2437744438648224,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505543947219849,
      "objective/train/weighted_lm_loss": 1.17076575756073,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.390722411850184,
      "tokens_seen": 2559180800
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022702106318956872,
      "loss": 1.2169,
      "theoretical_loss": 3.390722411850184,
      "tokens_seen": 2559180800
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022700100300902709,
      "loss": 1.2583,
      "theoretical_loss": 3.390715576098045,
      "tokens_seen": 2559246336
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022698094282848548,
      "loss": 1.2621,
      "theoretical_loss": 3.390708740569962,
      "tokens_seen": 2559311872
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022696088264794384,
      "loss": 1.3767,
      "theoretical_loss": 3.3907019052659226,
      "tokens_seen": 2559377408
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022694082246740223,
      "loss": 1.3971,
      "theoretical_loss": 3.390695070185913,
      "tokens_seen": 2559442944
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002269207622868606,
      "loss": 1.3707,
      "theoretical_loss": 3.3906882353299204,
      "tokens_seen": 2559508480
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022690070210631898,
      "loss": 1.3317,
      "theoretical_loss": 3.3906814006979316,
      "tokens_seen": 2559574016
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022688064192577734,
      "loss": 1.3261,
      "theoretical_loss": 3.390674566289934,
      "tokens_seen": 2559639552
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002268605817452357,
      "loss": 1.2802,
      "theoretical_loss": 3.3906677321059133,
      "tokens_seen": 2559705088
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002268405215646941,
      "loss": 1.4078,
      "theoretical_loss": 3.390660898145858,
      "tokens_seen": 2559770624
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022682046138415245,
      "loss": 1.2691,
      "theoretical_loss": 3.390654064409754,
      "tokens_seen": 2559836160
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022680040120361084,
      "loss": 1.2439,
      "theoretical_loss": 3.3906472308975886,
      "tokens_seen": 2559901696
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002267803410230692,
      "loss": 1.2464,
      "theoretical_loss": 3.390640397609349,
      "tokens_seen": 2559967232
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002267602808425276,
      "loss": 1.2346,
      "theoretical_loss": 3.3906335645450216,
      "tokens_seen": 2560032768
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022674022066198596,
      "loss": 1.2442,
      "theoretical_loss": 3.3906267317045935,
      "tokens_seen": 2560098304
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022672016048144435,
      "loss": 1.2128,
      "theoretical_loss": 3.390619899088051,
      "tokens_seen": 2560163840
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002267001003009027,
      "loss": 1.4595,
      "theoretical_loss": 3.3906130666953826,
      "tokens_seen": 2560229376
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002266800401203611,
      "loss": 1.2942,
      "theoretical_loss": 3.390606234526574,
      "tokens_seen": 2560294912
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022665997993981946,
      "loss": 1.3291,
      "theoretical_loss": 3.3905994025816124,
      "tokens_seen": 2560360448
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022663991975927782,
      "loss": 1.3323,
      "theoretical_loss": 3.390592570860485,
      "tokens_seen": 2560425984
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002266198595787362,
      "loss": 1.2451,
      "theoretical_loss": 3.3905857393631784,
      "tokens_seen": 2560491520
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022659979939819458,
      "loss": 1.3648,
      "theoretical_loss": 3.3905789080896795,
      "tokens_seen": 2560557056
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022657973921765296,
      "loss": 1.3743,
      "theoretical_loss": 3.3905720770399754,
      "tokens_seen": 2560622592
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022655967903711133,
      "loss": 1.246,
      "theoretical_loss": 3.390565246214053,
      "tokens_seen": 2560688128
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022653961885656972,
      "loss": 1.2591,
      "theoretical_loss": 3.3905584156118995,
      "tokens_seen": 2560753664
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022651955867602808,
      "loss": 1.2593,
      "theoretical_loss": 3.390551585233502,
      "tokens_seen": 2560819200
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022649949849548647,
      "loss": 1.3128,
      "theoretical_loss": 3.3905447550788463,
      "tokens_seen": 2560884736
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022647943831494483,
      "loss": 1.2681,
      "theoretical_loss": 3.3905379251479206,
      "tokens_seen": 2560950272
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002264593781344032,
      "loss": 1.3796,
      "theoretical_loss": 3.3905310954407115,
      "tokens_seen": 2561015808
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022643931795386158,
      "loss": 1.3366,
      "theoretical_loss": 3.3905242659572052,
      "tokens_seen": 2561081344
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022641925777331994,
      "loss": 1.2649,
      "theoretical_loss": 3.39051743669739,
      "tokens_seen": 2561146880
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022639919759277833,
      "loss": 1.2846,
      "theoretical_loss": 3.3905106076612515,
      "tokens_seen": 2561212416
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002263791374122367,
      "loss": 1.3234,
      "theoretical_loss": 3.3905037788487773,
      "tokens_seen": 2561277952
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022635907723169508,
      "loss": 1.3578,
      "theoretical_loss": 3.390496950259955,
      "tokens_seen": 2561343488
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022633901705115347,
      "loss": 1.3163,
      "theoretical_loss": 3.39049012189477,
      "tokens_seen": 2561409024
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022631895687061186,
      "loss": 1.2265,
      "theoretical_loss": 3.390483293753211,
      "tokens_seen": 2561474560
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022629889669007023,
      "loss": 1.2529,
      "theoretical_loss": 3.3904764658352633,
      "tokens_seen": 2561540096
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022627883650952861,
      "loss": 1.3035,
      "theoretical_loss": 3.3904696381409147,
      "tokens_seen": 2561605632
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022625877632898698,
      "loss": 1.3929,
      "theoretical_loss": 3.3904628106701526,
      "tokens_seen": 2561671168
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022623871614844534,
      "loss": 1.3828,
      "theoretical_loss": 3.390455983422963,
      "tokens_seen": 2561736704
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022621865596790373,
      "loss": 1.3895,
      "theoretical_loss": 3.3904491563993338,
      "tokens_seen": 2561802240
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002261985957873621,
      "loss": 1.3055,
      "theoretical_loss": 3.390442329599251,
      "tokens_seen": 2561867776
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022617853560682048,
      "loss": 1.3691,
      "theoretical_loss": 3.3904355030227022,
      "tokens_seen": 2561933312
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022615847542627884,
      "loss": 1.2759,
      "theoretical_loss": 3.390428676669674,
      "tokens_seen": 2561998848
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022613841524573723,
      "loss": 1.3195,
      "theoretical_loss": 3.3904218505401538,
      "tokens_seen": 2562064384
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002261183550651956,
      "loss": 1.403,
      "theoretical_loss": 3.3904150246341285,
      "tokens_seen": 2562129920
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022609829488465398,
      "loss": 1.3179,
      "theoretical_loss": 3.3904081989515844,
      "tokens_seen": 2562195456
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022607823470411235,
      "loss": 1.2441,
      "theoretical_loss": 3.3904013734925096,
      "tokens_seen": 2562260992
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002260581745235707,
      "loss": 1.3445,
      "theoretical_loss": 3.39039454825689,
      "tokens_seen": 2562326528
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002260381143430291,
      "loss": 1.2844,
      "theoretical_loss": 3.3903877232447126,
      "tokens_seen": 2562392064
    },
    {
      "epoch": 0.78,
      "objective/train/advantage_avg": 0.49449002742767334,
      "objective/train/docs_used": 931999,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.209947109222412,
      "objective/train/original_loss": 1.209947109222412,
      "objective/train/theoretical_loss": 3.390380898455965,
      "objective/train/tokens_used": 2582917600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24471138417720795,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050693154335022,
      "objective/train/weighted_lm_loss": 1.271156907081604,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0304546356201172,
      "theoretical_loss": 3.390380898455965,
      "tokens_seen": 2562457600
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022601805416248746,
      "loss": 1.3488,
      "theoretical_loss": 3.390380898455965,
      "tokens_seen": 2562457600
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022599799398194585,
      "loss": 1.3586,
      "theoretical_loss": 3.3903740738906345,
      "tokens_seen": 2562523136
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002259779338014042,
      "loss": 1.3377,
      "theoretical_loss": 3.3903672495487074,
      "tokens_seen": 2562588672
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002259578736208626,
      "loss": 1.4641,
      "theoretical_loss": 3.3903604254301705,
      "tokens_seen": 2562654208
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022593781344032096,
      "loss": 1.3334,
      "theoretical_loss": 3.390353601535011,
      "tokens_seen": 2562719744
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022591775325977935,
      "loss": 1.3732,
      "theoretical_loss": 3.3903467778632157,
      "tokens_seen": 2562785280
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022589769307923771,
      "loss": 1.3148,
      "theoretical_loss": 3.390339954414772,
      "tokens_seen": 2562850816
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022587763289869608,
      "loss": 1.357,
      "theoretical_loss": 3.390333131189667,
      "tokens_seen": 2562916352
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022585757271815447,
      "loss": 1.2637,
      "theoretical_loss": 3.3903263081878876,
      "tokens_seen": 2562981888
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022583751253761283,
      "loss": 1.3026,
      "theoretical_loss": 3.39031948540942,
      "tokens_seen": 2563047424
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022581745235707122,
      "loss": 1.3339,
      "theoretical_loss": 3.3903126628542517,
      "tokens_seen": 2563112960
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022579739217652958,
      "loss": 1.3098,
      "theoretical_loss": 3.39030584052237,
      "tokens_seen": 2563178496
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022577733199598797,
      "loss": 1.2933,
      "theoretical_loss": 3.390299018413762,
      "tokens_seen": 2563244032
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022575727181544633,
      "loss": 1.3387,
      "theoretical_loss": 3.390292196528413,
      "tokens_seen": 2563309568
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022573721163490472,
      "loss": 1.3703,
      "theoretical_loss": 3.390285374866312,
      "tokens_seen": 2563375104
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022571715145436308,
      "loss": 1.2804,
      "theoretical_loss": 3.3902785534274456,
      "tokens_seen": 2563440640
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022569709127382147,
      "loss": 1.2756,
      "theoretical_loss": 3.3902717322118,
      "tokens_seen": 2563506176
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022567703109327983,
      "loss": 1.2947,
      "theoretical_loss": 3.390264911219363,
      "tokens_seen": 2563571712
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002256569709127382,
      "loss": 1.4511,
      "theoretical_loss": 3.390258090450121,
      "tokens_seen": 2563637248
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022563691073219659,
      "loss": 1.3286,
      "theoretical_loss": 3.390251269904061,
      "tokens_seen": 2563702784
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022561685055165495,
      "loss": 1.3751,
      "theoretical_loss": 3.3902444495811705,
      "tokens_seen": 2563768320
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022559679037111336,
      "loss": 1.3343,
      "theoretical_loss": 3.390237629481436,
      "tokens_seen": 2563833856
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022557673019057173,
      "loss": 1.3021,
      "theoretical_loss": 3.390230809604845,
      "tokens_seen": 2563899392
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022555667001003012,
      "loss": 1.3763,
      "theoretical_loss": 3.3902239899513837,
      "tokens_seen": 2563964928
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022553660982948848,
      "loss": 1.3532,
      "theoretical_loss": 3.3902171705210398,
      "tokens_seen": 2564030464
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022551654964894687,
      "loss": 1.3462,
      "theoretical_loss": 3.3902103513138004,
      "tokens_seen": 2564096000
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022549648946840523,
      "loss": 1.1662,
      "theoretical_loss": 3.390203532329652,
      "tokens_seen": 2564161536
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002254764292878636,
      "loss": 1.2807,
      "theoretical_loss": 3.3901967135685815,
      "tokens_seen": 2564227072
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022545636910732198,
      "loss": 1.3828,
      "theoretical_loss": 3.390189895030576,
      "tokens_seen": 2564292608
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022543630892678034,
      "loss": 1.2047,
      "theoretical_loss": 3.3901830767156236,
      "tokens_seen": 2564358144
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022541624874623873,
      "loss": 1.3374,
      "theoretical_loss": 3.39017625862371,
      "tokens_seen": 2564423680
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002253961885656971,
      "loss": 1.3173,
      "theoretical_loss": 3.390169440754822,
      "tokens_seen": 2564489216
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022537612838515548,
      "loss": 1.393,
      "theoretical_loss": 3.390162623108948,
      "tokens_seen": 2564554752
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022535606820461385,
      "loss": 1.3877,
      "theoretical_loss": 3.390155805686074,
      "tokens_seen": 2564620288
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022533600802407224,
      "loss": 1.3335,
      "theoretical_loss": 3.390148988486187,
      "tokens_seen": 2564685824
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002253159478435306,
      "loss": 1.3415,
      "theoretical_loss": 3.3901421715092743,
      "tokens_seen": 2564751360
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022529588766298896,
      "loss": 1.3351,
      "theoretical_loss": 3.390135354755323,
      "tokens_seen": 2564816896
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022527582748244735,
      "loss": 1.3514,
      "theoretical_loss": 3.3901285382243196,
      "tokens_seen": 2564882432
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002252557673019057,
      "loss": 1.4267,
      "theoretical_loss": 3.3901217219162514,
      "tokens_seen": 2564947968
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002252357071213641,
      "loss": 1.304,
      "theoretical_loss": 3.3901149058311058,
      "tokens_seen": 2565013504
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022521564694082246,
      "loss": 1.2343,
      "theoretical_loss": 3.3901080899688694,
      "tokens_seen": 2565079040
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022519558676028085,
      "loss": 1.2705,
      "theoretical_loss": 3.3901012743295293,
      "tokens_seen": 2565144576
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022517552657973922,
      "loss": 1.4273,
      "theoretical_loss": 3.3900944589130724,
      "tokens_seen": 2565210112
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002251554663991976,
      "loss": 1.2491,
      "theoretical_loss": 3.390087643719486,
      "tokens_seen": 2565275648
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022513540621865597,
      "loss": 1.3329,
      "theoretical_loss": 3.390080828748757,
      "tokens_seen": 2565341184
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022511534603811436,
      "loss": 1.432,
      "theoretical_loss": 3.390074014000872,
      "tokens_seen": 2565406720
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022509528585757272,
      "loss": 1.281,
      "theoretical_loss": 3.3900671994758187,
      "tokens_seen": 2565472256
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022507522567703108,
      "loss": 1.2746,
      "theoretical_loss": 3.3900603851735838,
      "tokens_seen": 2565537792
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022505516549648947,
      "loss": 1.3409,
      "theoretical_loss": 3.390053571094154,
      "tokens_seen": 2565603328
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022503510531594783,
      "loss": 1.3431,
      "theoretical_loss": 3.390046757237517,
      "tokens_seen": 2565668864
    },
    {
      "epoch": 0.78,
      "objective/train/advantage_avg": 0.4931109547615051,
      "objective/train/docs_used": 933221,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4775112867355347,
      "objective/train/original_loss": 1.4775111675262451,
      "objective/train/theoretical_loss": 3.3900399436036595,
      "objective/train/tokens_used": 2586194400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24357615411281586,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505495071411133,
      "objective/train/weighted_lm_loss": 1.5520991086959839,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3900399436036595,
      "tokens_seen": 2565734400
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022501504513540622,
      "loss": 1.4214,
      "theoretical_loss": 3.3900399436036595,
      "tokens_seen": 2565734400
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022499498495486458,
      "loss": 1.3115,
      "theoretical_loss": 3.390033130192568,
      "tokens_seen": 2565799936
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022497492477432297,
      "loss": 1.3313,
      "theoretical_loss": 3.3900263170042306,
      "tokens_seen": 2565865472
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022495486459378134,
      "loss": 1.3188,
      "theoretical_loss": 3.3900195040386336,
      "tokens_seen": 2565931008
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022493480441323973,
      "loss": 1.3185,
      "theoretical_loss": 3.390012691295764,
      "tokens_seen": 2565996544
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002249147442326981,
      "loss": 1.2951,
      "theoretical_loss": 3.3900058787756095,
      "tokens_seen": 2566062080
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022489468405215645,
      "loss": 1.3237,
      "theoretical_loss": 3.3899990664781563,
      "tokens_seen": 2566127616
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022487462387161484,
      "loss": 1.3686,
      "theoretical_loss": 3.389992254403392,
      "tokens_seen": 2566193152
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002248545636910732,
      "loss": 1.3341,
      "theoretical_loss": 3.3899854425513034,
      "tokens_seen": 2566258688
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022483450351053162,
      "loss": 1.3682,
      "theoretical_loss": 3.389978630921877,
      "tokens_seen": 2566324224
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022481444332998998,
      "loss": 1.3834,
      "theoretical_loss": 3.389971819515101,
      "tokens_seen": 2566389760
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022479438314944837,
      "loss": 1.3364,
      "theoretical_loss": 3.389965008330962,
      "tokens_seen": 2566455296
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022477432296890673,
      "loss": 1.2818,
      "theoretical_loss": 3.389958197369446,
      "tokens_seen": 2566520832
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022475426278836512,
      "loss": 1.4423,
      "theoretical_loss": 3.3899513866305417,
      "tokens_seen": 2566586368
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022473420260782348,
      "loss": 1.4093,
      "theoretical_loss": 3.389944576114235,
      "tokens_seen": 2566651904
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022471414242728187,
      "loss": 1.3204,
      "theoretical_loss": 3.3899377658205134,
      "tokens_seen": 2566717440
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022469408224674023,
      "loss": 1.255,
      "theoretical_loss": 3.3899309557493638,
      "tokens_seen": 2566782976
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002246740220661986,
      "loss": 1.3005,
      "theoretical_loss": 3.389924145900773,
      "tokens_seen": 2566848512
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022465396188565699,
      "loss": 1.2886,
      "theoretical_loss": 3.3899173362747286,
      "tokens_seen": 2566914048
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022463390170511535,
      "loss": 1.2918,
      "theoretical_loss": 3.3899105268712173,
      "tokens_seen": 2566979584
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022461384152457374,
      "loss": 1.4959,
      "theoretical_loss": 3.3899037176902262,
      "tokens_seen": 2567045120
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002245937813440321,
      "loss": 1.2354,
      "theoretical_loss": 3.3898969087317425,
      "tokens_seen": 2567110656
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002245737211634905,
      "loss": 1.3486,
      "theoretical_loss": 3.389890099995753,
      "tokens_seen": 2567176192
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022455366098294885,
      "loss": 1.4283,
      "theoretical_loss": 3.3898832914822448,
      "tokens_seen": 2567241728
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022453360080240724,
      "loss": 1.2875,
      "theoretical_loss": 3.3898764831912054,
      "tokens_seen": 2567307264
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002245135406218656,
      "loss": 1.3591,
      "theoretical_loss": 3.389869675122621,
      "tokens_seen": 2567372800
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022449348044132397,
      "loss": 1.4103,
      "theoretical_loss": 3.389862867276479,
      "tokens_seen": 2567438336
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022447342026078236,
      "loss": 1.3247,
      "theoretical_loss": 3.389856059652767,
      "tokens_seen": 2567503872
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022445336008024072,
      "loss": 1.2924,
      "theoretical_loss": 3.389849252251471,
      "tokens_seen": 2567569408
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002244332998996991,
      "loss": 1.3166,
      "theoretical_loss": 3.3898424450725795,
      "tokens_seen": 2567634944
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022441323971915747,
      "loss": 1.2576,
      "theoretical_loss": 3.3898356381160784,
      "tokens_seen": 2567700480
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022439317953861586,
      "loss": 1.3558,
      "theoretical_loss": 3.389828831381955,
      "tokens_seen": 2567766016
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022437311935807422,
      "loss": 1.4236,
      "theoretical_loss": 3.3898220248701962,
      "tokens_seen": 2567831552
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002243530591775326,
      "loss": 1.3352,
      "theoretical_loss": 3.38981521858079,
      "tokens_seen": 2567897088
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022433299899699097,
      "loss": 1.3522,
      "theoretical_loss": 3.3898084125137222,
      "tokens_seen": 2567962624
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022431293881644933,
      "loss": 1.296,
      "theoretical_loss": 3.3898016066689807,
      "tokens_seen": 2568028160
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022429287863590772,
      "loss": 1.3216,
      "theoretical_loss": 3.389794801046552,
      "tokens_seen": 2568093696
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022427281845536609,
      "loss": 1.3679,
      "theoretical_loss": 3.389787995646424,
      "tokens_seen": 2568159232
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022425275827482448,
      "loss": 1.3503,
      "theoretical_loss": 3.389781190468583,
      "tokens_seen": 2568224768
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022423269809428284,
      "loss": 1.4008,
      "theoretical_loss": 3.3897743855130162,
      "tokens_seen": 2568290304
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022421263791374123,
      "loss": 1.4535,
      "theoretical_loss": 3.3897675807797105,
      "tokens_seen": 2568355840
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002241925777331996,
      "loss": 1.2863,
      "theoretical_loss": 3.3897607762686537,
      "tokens_seen": 2568421376
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022417251755265798,
      "loss": 1.2308,
      "theoretical_loss": 3.3897539719798324,
      "tokens_seen": 2568486912
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022415245737211634,
      "loss": 1.3682,
      "theoretical_loss": 3.3897471679132334,
      "tokens_seen": 2568552448
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002241323971915747,
      "loss": 1.3422,
      "theoretical_loss": 3.389740364068844,
      "tokens_seen": 2568617984
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002241123370110331,
      "loss": 1.3076,
      "theoretical_loss": 3.389733560446652,
      "tokens_seen": 2568683520
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022409227683049148,
      "loss": 1.4284,
      "theoretical_loss": 3.389726757046643,
      "tokens_seen": 2568749056
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022407221664994987,
      "loss": 1.3638,
      "theoretical_loss": 3.389719953868805,
      "tokens_seen": 2568814592
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022405215646940823,
      "loss": 1.3545,
      "theoretical_loss": 3.3897131509131255,
      "tokens_seen": 2568880128
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022403209628886662,
      "loss": 1.2493,
      "theoretical_loss": 3.3897063481795904,
      "tokens_seen": 2568945664
    },
    {
      "epoch": 0.78,
      "objective/train/advantage_avg": 0.4915311336517334,
      "objective/train/docs_used": 934338,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.1632598638534546,
      "objective/train/original_loss": 1.1632598638534546,
      "objective/train/theoretical_loss": 3.3896995456681878,
      "objective/train/tokens_used": 2589471200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24213433265686035,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503840446472168,
      "objective/train/weighted_lm_loss": 1.2219172716140747,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3896995456681878,
      "tokens_seen": 2569011200
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022401203610832498,
      "loss": 1.2635,
      "theoretical_loss": 3.3896995456681878,
      "tokens_seen": 2569011200
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022399197592778337,
      "loss": 1.4096,
      "theoretical_loss": 3.3896927433789044,
      "tokens_seen": 2569076736
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022397191574724174,
      "loss": 1.3342,
      "theoretical_loss": 3.3896859413117273,
      "tokens_seen": 2569142272
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022395185556670013,
      "loss": 1.5266,
      "theoretical_loss": 3.3896791394666432,
      "tokens_seen": 2569207808
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002239317953861585,
      "loss": 1.4456,
      "theoretical_loss": 3.38967233784364,
      "tokens_seen": 2569273344
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022391173520561685,
      "loss": 1.2921,
      "theoretical_loss": 3.389665536442704,
      "tokens_seen": 2569338880
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022389167502507524,
      "loss": 1.2401,
      "theoretical_loss": 3.389658735263823,
      "tokens_seen": 2569404416
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002238716148445336,
      "loss": 1.3217,
      "theoretical_loss": 3.389651934306983,
      "tokens_seen": 2569469952
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000223851554663992,
      "loss": 1.3921,
      "theoretical_loss": 3.3896451335721727,
      "tokens_seen": 2569535488
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022383149448345035,
      "loss": 1.3145,
      "theoretical_loss": 3.3896383330593776,
      "tokens_seen": 2569601024
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022381143430290874,
      "loss": 1.4239,
      "theoretical_loss": 3.3896315327685858,
      "tokens_seen": 2569666560
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002237913741223671,
      "loss": 1.248,
      "theoretical_loss": 3.389624732699784,
      "tokens_seen": 2569732096
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002237713139418255,
      "loss": 1.3475,
      "theoretical_loss": 3.3896179328529588,
      "tokens_seen": 2569797632
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022375125376128386,
      "loss": 1.5097,
      "theoretical_loss": 3.3896111332280983,
      "tokens_seen": 2569863168
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022373119358074222,
      "loss": 1.3605,
      "theoretical_loss": 3.389604333825189,
      "tokens_seen": 2569928704
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002237111334002006,
      "loss": 1.2753,
      "theoretical_loss": 3.389597534644219,
      "tokens_seen": 2569994240
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022369107321965897,
      "loss": 1.3427,
      "theoretical_loss": 3.3895907356851733,
      "tokens_seen": 2570059776
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022367101303911736,
      "loss": 1.2991,
      "theoretical_loss": 3.389583936948041,
      "tokens_seen": 2570125312
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022365095285857572,
      "loss": 1.3189,
      "theoretical_loss": 3.389577138432808,
      "tokens_seen": 2570190848
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002236308926780341,
      "loss": 1.3566,
      "theoretical_loss": 3.389570340139462,
      "tokens_seen": 2570256384
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022361083249749247,
      "loss": 1.408,
      "theoretical_loss": 3.38956354206799,
      "tokens_seen": 2570321920
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022359077231695086,
      "loss": 1.3475,
      "theoretical_loss": 3.3895567442183787,
      "tokens_seen": 2570387456
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022357071213640923,
      "loss": 1.3647,
      "theoretical_loss": 3.389549946590616,
      "tokens_seen": 2570452992
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022355065195586761,
      "loss": 1.3299,
      "theoretical_loss": 3.3895431491846884,
      "tokens_seen": 2570518528
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022353059177532598,
      "loss": 1.4623,
      "theoretical_loss": 3.389536352000583,
      "tokens_seen": 2570584064
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022351053159478434,
      "loss": 1.3445,
      "theoretical_loss": 3.389529555038287,
      "tokens_seen": 2570649600
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022349047141424273,
      "loss": 1.414,
      "theoretical_loss": 3.3895227582977876,
      "tokens_seen": 2570715136
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002234704112337011,
      "loss": 1.433,
      "theoretical_loss": 3.3895159617790718,
      "tokens_seen": 2570780672
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022345035105315948,
      "loss": 1.4669,
      "theoretical_loss": 3.389509165482127,
      "tokens_seen": 2570846208
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022343029087261784,
      "loss": 1.3121,
      "theoretical_loss": 3.38950236940694,
      "tokens_seen": 2570911744
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022341023069207623,
      "loss": 1.3206,
      "theoretical_loss": 3.3894955735534977,
      "tokens_seen": 2570977280
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002233901705115346,
      "loss": 1.439,
      "theoretical_loss": 3.389488777921788,
      "tokens_seen": 2571042816
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022337011033099298,
      "loss": 1.4057,
      "theoretical_loss": 3.389481982511797,
      "tokens_seen": 2571108352
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022335005015045135,
      "loss": 1.346,
      "theoretical_loss": 3.3894751873235127,
      "tokens_seen": 2571173888
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022332998996990973,
      "loss": 1.5669,
      "theoretical_loss": 3.3894683923569215,
      "tokens_seen": 2571239424
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022330992978936812,
      "loss": 1.3113,
      "theoretical_loss": 3.3894615976120113,
      "tokens_seen": 2571304960
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022328986960882649,
      "loss": 1.2951,
      "theoretical_loss": 3.3894548030887686,
      "tokens_seen": 2571370496
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022326980942828488,
      "loss": 1.3859,
      "theoretical_loss": 3.3894480087871806,
      "tokens_seen": 2571436032
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022324974924774324,
      "loss": 1.3097,
      "theoretical_loss": 3.389441214707235,
      "tokens_seen": 2571501568
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022322968906720163,
      "loss": 1.2407,
      "theoretical_loss": 3.389434420848918,
      "tokens_seen": 2571567104
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022320962888666,
      "loss": 1.3786,
      "theoretical_loss": 3.3894276272122172,
      "tokens_seen": 2571632640
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022318956870611838,
      "loss": 1.2859,
      "theoretical_loss": 3.3894208337971197,
      "tokens_seen": 2571698176
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022316950852557674,
      "loss": 1.2486,
      "theoretical_loss": 3.3894140406036124,
      "tokens_seen": 2571763712
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002231494483450351,
      "loss": 1.3974,
      "theoretical_loss": 3.389407247631683,
      "tokens_seen": 2571829248
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002231293881644935,
      "loss": 1.3574,
      "theoretical_loss": 3.389400454881318,
      "tokens_seen": 2571894784
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022310932798395185,
      "loss": 1.3851,
      "theoretical_loss": 3.3893936623525054,
      "tokens_seen": 2571960320
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022308926780341024,
      "loss": 1.3882,
      "theoretical_loss": 3.3893868700452314,
      "tokens_seen": 2572025856
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002230692076228686,
      "loss": 1.4201,
      "theoretical_loss": 3.3893800779594834,
      "tokens_seen": 2572091392
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000223049147442327,
      "loss": 1.3355,
      "theoretical_loss": 3.389373286095248,
      "tokens_seen": 2572156928
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022302908726178536,
      "loss": 1.5767,
      "theoretical_loss": 3.3893664944525135,
      "tokens_seen": 2572222464
    },
    {
      "epoch": 0.78,
      "objective/train/advantage_avg": 0.49347570538520813,
      "objective/train/docs_used": 935480,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2209218740463257,
      "objective/train/original_loss": 1.2209217548370361,
      "objective/train/theoretical_loss": 3.3893597030312663,
      "objective/train/tokens_used": 2592748000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2438102662563324,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505869388580322,
      "objective/train/weighted_lm_loss": 1.2825274467468262,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3893597030312663,
      "tokens_seen": 2572288000
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022300902708124375,
      "loss": 1.3902,
      "theoretical_loss": 3.3893597030312663,
      "tokens_seen": 2572288000
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002229889669007021,
      "loss": 1.4193,
      "theoretical_loss": 3.389352911831494,
      "tokens_seen": 2572353536
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002229689067201605,
      "loss": 1.4555,
      "theoretical_loss": 3.389346120853183,
      "tokens_seen": 2572419072
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022294884653961886,
      "loss": 1.407,
      "theoretical_loss": 3.3893393300963215,
      "tokens_seen": 2572484608
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022292878635907722,
      "loss": 1.4078,
      "theoretical_loss": 3.389332539560895,
      "tokens_seen": 2572550144
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002229087261785356,
      "loss": 1.4243,
      "theoretical_loss": 3.3893257492468924,
      "tokens_seen": 2572615680
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022288866599799397,
      "loss": 1.3772,
      "theoretical_loss": 3.3893189591543,
      "tokens_seen": 2572681216
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022286860581745236,
      "loss": 1.492,
      "theoretical_loss": 3.3893121692831047,
      "tokens_seen": 2572746752
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022284854563691073,
      "loss": 1.2554,
      "theoretical_loss": 3.389305379633294,
      "tokens_seen": 2572812288
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022282848545636912,
      "loss": 1.2969,
      "theoretical_loss": 3.3892985902048554,
      "tokens_seen": 2572877824
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022280842527582748,
      "loss": 1.6253,
      "theoretical_loss": 3.3892918009977753,
      "tokens_seen": 2572943360
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022278836509528587,
      "loss": 1.3611,
      "theoretical_loss": 3.3892850120120412,
      "tokens_seen": 2573008896
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022276830491474423,
      "loss": 1.3394,
      "theoretical_loss": 3.38927822324764,
      "tokens_seen": 2573074432
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002227482447342026,
      "loss": 1.2645,
      "theoretical_loss": 3.3892714347045594,
      "tokens_seen": 2573139968
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022272818455366098,
      "loss": 1.3413,
      "theoretical_loss": 3.3892646463827862,
      "tokens_seen": 2573205504
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022270812437311934,
      "loss": 1.2552,
      "theoretical_loss": 3.3892578582823076,
      "tokens_seen": 2573271040
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022268806419257773,
      "loss": 1.3776,
      "theoretical_loss": 3.3892510704031107,
      "tokens_seen": 2573336576
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002226680040120361,
      "loss": 1.3414,
      "theoretical_loss": 3.3892442827451825,
      "tokens_seen": 2573402112
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022264794383149448,
      "loss": 1.268,
      "theoretical_loss": 3.3892374953085107,
      "tokens_seen": 2573467648
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022262788365095285,
      "loss": 1.4651,
      "theoretical_loss": 3.389230708093082,
      "tokens_seen": 2573533184
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022260782347041124,
      "loss": 1.4501,
      "theoretical_loss": 3.3892239210988837,
      "tokens_seen": 2573598720
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022258776328986963,
      "loss": 1.3877,
      "theoretical_loss": 3.3892171343259023,
      "tokens_seen": 2573664256
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022256770310932801,
      "loss": 1.4919,
      "theoretical_loss": 3.389210347774126,
      "tokens_seen": 2573729792
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022254764292878638,
      "loss": 1.2667,
      "theoretical_loss": 3.3892035614435416,
      "tokens_seen": 2573795328
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022252758274824474,
      "loss": 1.4134,
      "theoretical_loss": 3.389196775334136,
      "tokens_seen": 2573860864
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022250752256770313,
      "loss": 1.2616,
      "theoretical_loss": 3.3891899894458968,
      "tokens_seen": 2573926400
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002224874623871615,
      "loss": 1.4031,
      "theoretical_loss": 3.3891832037788103,
      "tokens_seen": 2573991936
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022246740220661988,
      "loss": 1.1536,
      "theoretical_loss": 3.3891764183328648,
      "tokens_seen": 2574057472
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022244734202607824,
      "loss": 1.2734,
      "theoretical_loss": 3.3891696331080468,
      "tokens_seen": 2574123008
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022242728184553663,
      "loss": 1.3883,
      "theoretical_loss": 3.3891628481043434,
      "tokens_seen": 2574188544
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000222407221664995,
      "loss": 1.4004,
      "theoretical_loss": 3.389156063321742,
      "tokens_seen": 2574254080
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022238716148445338,
      "loss": 1.3888,
      "theoretical_loss": 3.3891492787602298,
      "tokens_seen": 2574319616
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022236710130391175,
      "loss": 1.3798,
      "theoretical_loss": 3.3891424944197936,
      "tokens_seen": 2574385152
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002223470411233701,
      "loss": 1.3298,
      "theoretical_loss": 3.389135710300421,
      "tokens_seen": 2574450688
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002223269809428285,
      "loss": 1.2769,
      "theoretical_loss": 3.3891289264020994,
      "tokens_seen": 2574516224
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022230692076228686,
      "loss": 1.3848,
      "theoretical_loss": 3.389122142724815,
      "tokens_seen": 2574581760
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022228686058174525,
      "loss": 1.3699,
      "theoretical_loss": 3.3891153592685557,
      "tokens_seen": 2574647296
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002222668004012036,
      "loss": 1.3722,
      "theoretical_loss": 3.389108576033309,
      "tokens_seen": 2574712832
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000222246740220662,
      "loss": 1.3257,
      "theoretical_loss": 3.3891017930190612,
      "tokens_seen": 2574778368
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022222668004012036,
      "loss": 1.3599,
      "theoretical_loss": 3.3890950102258,
      "tokens_seen": 2574843904
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022220661985957875,
      "loss": 1.3605,
      "theoretical_loss": 3.389088227653512,
      "tokens_seen": 2574909440
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022218655967903711,
      "loss": 1.3665,
      "theoretical_loss": 3.3890814453021854,
      "tokens_seen": 2574974976
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022216649949849548,
      "loss": 1.4709,
      "theoretical_loss": 3.3890746631718067,
      "tokens_seen": 2575040512
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022214643931795387,
      "loss": 1.1926,
      "theoretical_loss": 3.3890678812623634,
      "tokens_seen": 2575106048
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022212637913741223,
      "loss": 1.1589,
      "theoretical_loss": 3.389061099573842,
      "tokens_seen": 2575171584
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022210631895687062,
      "loss": 1.4106,
      "theoretical_loss": 3.3890543181062305,
      "tokens_seen": 2575237120
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022208625877632898,
      "loss": 1.2967,
      "theoretical_loss": 3.3890475368595157,
      "tokens_seen": 2575302656
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022206619859578737,
      "loss": 1.4865,
      "theoretical_loss": 3.3890407558336846,
      "tokens_seen": 2575368192
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022204613841524573,
      "loss": 1.2805,
      "theoretical_loss": 3.3890339750287244,
      "tokens_seen": 2575433728
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022202607823470412,
      "loss": 1.4039,
      "theoretical_loss": 3.3890271944446226,
      "tokens_seen": 2575499264
    },
    {
      "epoch": 0.78,
      "objective/train/advantage_avg": 0.4925488531589508,
      "objective/train/docs_used": 936685,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3424147367477417,
      "objective/train/original_loss": 1.3424148559570312,
      "objective/train/theoretical_loss": 3.389020414081367,
      "objective/train/tokens_used": 2596024800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24291947484016418,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504897832870483,
      "objective/train/weighted_lm_loss": 1.4100499153137207,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.389020414081367,
      "tokens_seen": 2575564800
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022200601805416248,
      "loss": 1.3208,
      "theoretical_loss": 3.389020414081367,
      "tokens_seen": 2575564800
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022198595787362084,
      "loss": 1.2281,
      "theoretical_loss": 3.3890136339389434,
      "tokens_seen": 2575630336
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022196589769307923,
      "loss": 1.407,
      "theoretical_loss": 3.3890068540173397,
      "tokens_seen": 2575695872
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002219458375125376,
      "loss": 1.3997,
      "theoretical_loss": 3.389000074316543,
      "tokens_seen": 2575761408
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022192577733199599,
      "loss": 1.4182,
      "theoretical_loss": 3.3889932948365407,
      "tokens_seen": 2575826944
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022190571715145435,
      "loss": 1.3037,
      "theoretical_loss": 3.3889865155773196,
      "tokens_seen": 2575892480
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022188565697091274,
      "loss": 1.3306,
      "theoretical_loss": 3.3889797365388676,
      "tokens_seen": 2575958016
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002218655967903711,
      "loss": 1.2042,
      "theoretical_loss": 3.388972957721171,
      "tokens_seen": 2576023552
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022184553660982952,
      "loss": 1.3688,
      "theoretical_loss": 3.3889661791242176,
      "tokens_seen": 2576089088
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022182547642928788,
      "loss": 1.3854,
      "theoretical_loss": 3.3889594007479946,
      "tokens_seen": 2576154624
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022180541624874627,
      "loss": 1.3702,
      "theoretical_loss": 3.3889526225924884,
      "tokens_seen": 2576220160
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022178535606820463,
      "loss": 1.3359,
      "theoretical_loss": 3.388945844657687,
      "tokens_seen": 2576285696
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000221765295887663,
      "loss": 1.5032,
      "theoretical_loss": 3.3889390669435775,
      "tokens_seen": 2576351232
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022174523570712138,
      "loss": 1.4172,
      "theoretical_loss": 3.388932289450147,
      "tokens_seen": 2576416768
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022172517552657974,
      "loss": 1.3537,
      "theoretical_loss": 3.3889255121773827,
      "tokens_seen": 2576482304
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022170511534603813,
      "loss": 1.4347,
      "theoretical_loss": 3.388918735125272,
      "tokens_seen": 2576547840
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002216850551654965,
      "loss": 1.4137,
      "theoretical_loss": 3.388911958293802,
      "tokens_seen": 2576613376
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022166499498495488,
      "loss": 1.2026,
      "theoretical_loss": 3.3889051816829596,
      "tokens_seen": 2576678912
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022164493480441325,
      "loss": 1.3345,
      "theoretical_loss": 3.388898405292732,
      "tokens_seen": 2576744448
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022162487462387164,
      "loss": 1.3101,
      "theoretical_loss": 3.3888916291231066,
      "tokens_seen": 2576809984
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022160481444333,
      "loss": 1.2428,
      "theoretical_loss": 3.388884853174071,
      "tokens_seen": 2576875520
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022158475426278836,
      "loss": 1.4349,
      "theoretical_loss": 3.3888780774456118,
      "tokens_seen": 2576941056
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022156469408224675,
      "loss": 1.4642,
      "theoretical_loss": 3.3888713019377166,
      "tokens_seen": 2577006592
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002215446339017051,
      "loss": 1.4501,
      "theoretical_loss": 3.388864526650372,
      "tokens_seen": 2577072128
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002215245737211635,
      "loss": 1.3501,
      "theoretical_loss": 3.3888577515835663,
      "tokens_seen": 2577137664
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022150451354062186,
      "loss": 1.3144,
      "theoretical_loss": 3.388850976737286,
      "tokens_seen": 2577203200
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022148445336008025,
      "loss": 1.2302,
      "theoretical_loss": 3.388844202111518,
      "tokens_seen": 2577268736
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022146439317953862,
      "loss": 1.3156,
      "theoretical_loss": 3.3888374277062505,
      "tokens_seen": 2577334272
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000221444332998997,
      "loss": 1.2245,
      "theoretical_loss": 3.3888306535214694,
      "tokens_seen": 2577399808
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022142427281845537,
      "loss": 1.1696,
      "theoretical_loss": 3.3888238795571635,
      "tokens_seen": 2577465344
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022140421263791376,
      "loss": 1.5058,
      "theoretical_loss": 3.3888171058133185,
      "tokens_seen": 2577530880
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022138415245737212,
      "loss": 1.3874,
      "theoretical_loss": 3.3888103322899226,
      "tokens_seen": 2577596416
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022136409227683048,
      "loss": 1.3618,
      "theoretical_loss": 3.3888035589869627,
      "tokens_seen": 2577661952
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022134403209628887,
      "loss": 1.4544,
      "theoretical_loss": 3.388796785904426,
      "tokens_seen": 2577727488
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022132397191574723,
      "loss": 1.3693,
      "theoretical_loss": 3.3887900130422994,
      "tokens_seen": 2577793024
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022130391173520562,
      "loss": 1.2819,
      "theoretical_loss": 3.388783240400571,
      "tokens_seen": 2577858560
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022128385155466398,
      "loss": 1.2825,
      "theoretical_loss": 3.3887764679792274,
      "tokens_seen": 2577924096
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022126379137412237,
      "loss": 1.4289,
      "theoretical_loss": 3.388769695778256,
      "tokens_seen": 2577989632
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022124373119358074,
      "loss": 1.371,
      "theoretical_loss": 3.3887629237976435,
      "tokens_seen": 2578055168
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022122367101303913,
      "loss": 1.2207,
      "theoretical_loss": 3.388756152037378,
      "tokens_seen": 2578120704
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002212036108324975,
      "loss": 1.3981,
      "theoretical_loss": 3.3887493804974462,
      "tokens_seen": 2578186240
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022118355065195585,
      "loss": 1.5046,
      "theoretical_loss": 3.388742609177836,
      "tokens_seen": 2578251776
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022116349047141424,
      "loss": 1.3614,
      "theoretical_loss": 3.3887358380785333,
      "tokens_seen": 2578317312
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002211434302908726,
      "loss": 1.4122,
      "theoretical_loss": 3.388729067199526,
      "tokens_seen": 2578382848
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000221123370110331,
      "loss": 1.4414,
      "theoretical_loss": 3.388722296540802,
      "tokens_seen": 2578448384
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022110330992978935,
      "loss": 1.5236,
      "theoretical_loss": 3.388715526102348,
      "tokens_seen": 2578513920
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022108324974924777,
      "loss": 1.2517,
      "theoretical_loss": 3.388708755884151,
      "tokens_seen": 2578579456
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022106318956870613,
      "loss": 1.2881,
      "theoretical_loss": 3.3887019858861986,
      "tokens_seen": 2578644992
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022104312938816452,
      "loss": 1.4924,
      "theoretical_loss": 3.3886952161084776,
      "tokens_seen": 2578710528
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022102306920762288,
      "loss": 1.3528,
      "theoretical_loss": 3.388688446550976,
      "tokens_seen": 2578776064
    },
    {
      "epoch": 0.78,
      "objective/train/advantage_avg": 0.49278950691223145,
      "objective/train/docs_used": 937774,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.184408187866211,
      "objective/train/original_loss": 1.184408187866211,
      "objective/train/theoretical_loss": 3.38868167721368,
      "objective/train/tokens_used": 2599301600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24345184862613678,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505166053771973,
      "objective/train/weighted_lm_loss": 1.2443217039108276,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.38868167721368,
      "tokens_seen": 2578841600
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022100300902708125,
      "loss": 1.3899,
      "theoretical_loss": 3.38868167721368,
      "tokens_seen": 2578841600
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022098294884653963,
      "loss": 1.4451,
      "theoretical_loss": 3.3886749080965775,
      "tokens_seen": 2578907136
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000220962888665998,
      "loss": 1.1915,
      "theoretical_loss": 3.3886681391996563,
      "tokens_seen": 2578972672
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022094282848545639,
      "loss": 1.3243,
      "theoretical_loss": 3.388661370522902,
      "tokens_seen": 2579038208
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022092276830491475,
      "loss": 1.338,
      "theoretical_loss": 3.3886546020663038,
      "tokens_seen": 2579103744
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022090270812437314,
      "loss": 1.4975,
      "theoretical_loss": 3.3886478338298476,
      "tokens_seen": 2579169280
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002208826479438315,
      "loss": 1.4287,
      "theoretical_loss": 3.3886410658135206,
      "tokens_seen": 2579234816
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002208625877632899,
      "loss": 1.3099,
      "theoretical_loss": 3.3886342980173114,
      "tokens_seen": 2579300352
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022084252758274825,
      "loss": 1.2938,
      "theoretical_loss": 3.3886275304412057,
      "tokens_seen": 2579365888
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022082246740220664,
      "loss": 1.526,
      "theoretical_loss": 3.3886207630851914,
      "tokens_seen": 2579431424
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000220802407221665,
      "loss": 1.4907,
      "theoretical_loss": 3.388613995949256,
      "tokens_seen": 2579496960
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022078234704112337,
      "loss": 1.4644,
      "theoretical_loss": 3.3886072290333864,
      "tokens_seen": 2579562496
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022076228686058175,
      "loss": 1.4542,
      "theoretical_loss": 3.3886004623375694,
      "tokens_seen": 2579628032
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022074222668004012,
      "loss": 1.3303,
      "theoretical_loss": 3.3885936958617933,
      "tokens_seen": 2579693568
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002207221664994985,
      "loss": 1.3269,
      "theoretical_loss": 3.388586929606045,
      "tokens_seen": 2579759104
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022070210631895687,
      "loss": 1.3812,
      "theoretical_loss": 3.388580163570311,
      "tokens_seen": 2579824640
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022068204613841526,
      "loss": 1.3152,
      "theoretical_loss": 3.38857339775458,
      "tokens_seen": 2579890176
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022066198595787362,
      "loss": 1.3409,
      "theoretical_loss": 3.388566632158838,
      "tokens_seen": 2579955712
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000220641925777332,
      "loss": 1.2826,
      "theoretical_loss": 3.3885598667830727,
      "tokens_seen": 2580021248
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022062186559679037,
      "loss": 1.4448,
      "theoretical_loss": 3.388553101627271,
      "tokens_seen": 2580086784
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022060180541624873,
      "loss": 1.2862,
      "theoretical_loss": 3.3885463366914212,
      "tokens_seen": 2580152320
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022058174523570712,
      "loss": 1.0768,
      "theoretical_loss": 3.3885395719755094,
      "tokens_seen": 2580217856
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022056168505516549,
      "loss": 1.3178,
      "theoretical_loss": 3.3885328074795233,
      "tokens_seen": 2580283392
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022054162487462387,
      "loss": 1.3745,
      "theoretical_loss": 3.3885260432034503,
      "tokens_seen": 2580348928
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022052156469408224,
      "loss": 1.3065,
      "theoretical_loss": 3.3885192791472774,
      "tokens_seen": 2580414464
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022050150451354063,
      "loss": 1.3837,
      "theoretical_loss": 3.388512515310992,
      "tokens_seen": 2580480000
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000220481444332999,
      "loss": 1.2983,
      "theoretical_loss": 3.3885057516945816,
      "tokens_seen": 2580545536
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022046138415245738,
      "loss": 1.3246,
      "theoretical_loss": 3.388498988298033,
      "tokens_seen": 2580611072
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022044132397191574,
      "loss": 1.4207,
      "theoretical_loss": 3.388492225121334,
      "tokens_seen": 2580676608
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002204212637913741,
      "loss": 1.4933,
      "theoretical_loss": 3.3884854621644713,
      "tokens_seen": 2580742144
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002204012036108325,
      "loss": 1.3291,
      "theoretical_loss": 3.388478699427433,
      "tokens_seen": 2580807680
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022038114343029085,
      "loss": 1.3124,
      "theoretical_loss": 3.3884719369102054,
      "tokens_seen": 2580873216
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022036108324974924,
      "loss": 1.3324,
      "theoretical_loss": 3.3884651746127763,
      "tokens_seen": 2580938752
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022034102306920763,
      "loss": 1.2969,
      "theoretical_loss": 3.3884584125351327,
      "tokens_seen": 2581004288
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022032096288866602,
      "loss": 1.3389,
      "theoretical_loss": 3.388451650677262,
      "tokens_seen": 2581069824
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022030090270812438,
      "loss": 1.4314,
      "theoretical_loss": 3.3884448890391523,
      "tokens_seen": 2581135360
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022028084252758277,
      "loss": 1.3369,
      "theoretical_loss": 3.3884381276207893,
      "tokens_seen": 2581200896
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022026078234704114,
      "loss": 1.3828,
      "theoretical_loss": 3.3884313664221617,
      "tokens_seen": 2581266432
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022024072216649953,
      "loss": 1.365,
      "theoretical_loss": 3.388424605443256,
      "tokens_seen": 2581331968
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002202206619859579,
      "loss": 1.2963,
      "theoretical_loss": 3.388417844684059,
      "tokens_seen": 2581397504
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022020060180541625,
      "loss": 1.4249,
      "theoretical_loss": 3.38841108414456,
      "tokens_seen": 2581463040
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022018054162487464,
      "loss": 1.3517,
      "theoretical_loss": 3.388404323824744,
      "tokens_seen": 2581528576
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000220160481444333,
      "loss": 1.3351,
      "theoretical_loss": 3.3883975637245993,
      "tokens_seen": 2581594112
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002201404212637914,
      "loss": 1.2454,
      "theoretical_loss": 3.3883908038441133,
      "tokens_seen": 2581659648
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022012036108324975,
      "loss": 1.4247,
      "theoretical_loss": 3.3883840441832724,
      "tokens_seen": 2581725184
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022010030090270814,
      "loss": 1.3053,
      "theoretical_loss": 3.3883772847420657,
      "tokens_seen": 2581790720
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002200802407221665,
      "loss": 1.3088,
      "theoretical_loss": 3.388370525520479,
      "tokens_seen": 2581856256
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002200601805416249,
      "loss": 1.3466,
      "theoretical_loss": 3.3883637665184994,
      "tokens_seen": 2581921792
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022004012036108326,
      "loss": 1.3545,
      "theoretical_loss": 3.388357007736115,
      "tokens_seen": 2581987328
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022002006018054162,
      "loss": 1.2751,
      "theoretical_loss": 3.388350249173313,
      "tokens_seen": 2582052864
    },
    {
      "epoch": 0.78,
      "objective/train/advantage_avg": 0.488750696182251,
      "objective/train/docs_used": 938961,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3585911989212036,
      "objective/train/original_loss": 1.3585909605026245,
      "objective/train/theoretical_loss": 3.3883434908300805,
      "objective/train/tokens_used": 2602578400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24045901000499725,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0500972270965576,
      "objective/train/weighted_lm_loss": 1.4263198375701904,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.7788007855415344,
      "theoretical_loss": 3.3883434908300805,
      "tokens_seen": 2582118400
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00022,
      "loss": 1.424,
      "theoretical_loss": 3.3883434908300805,
      "tokens_seen": 2582118400
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021997993981945837,
      "loss": 1.3848,
      "theoretical_loss": 3.3883367327064047,
      "tokens_seen": 2582183936
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021995987963891676,
      "loss": 1.4393,
      "theoretical_loss": 3.388329974802273,
      "tokens_seen": 2582249472
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021993981945837512,
      "loss": 1.3696,
      "theoretical_loss": 3.388323217117673,
      "tokens_seen": 2582315008
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002199197592778335,
      "loss": 1.3929,
      "theoretical_loss": 3.388316459652591,
      "tokens_seen": 2582380544
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021989969909729187,
      "loss": 1.2826,
      "theoretical_loss": 3.3883097024070157,
      "tokens_seen": 2582446080
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021987963891675026,
      "loss": 1.3088,
      "theoretical_loss": 3.3883029453809335,
      "tokens_seen": 2582511616
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021985957873620862,
      "loss": 1.4104,
      "theoretical_loss": 3.388296188574332,
      "tokens_seen": 2582577152
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021983951855566701,
      "loss": 1.2753,
      "theoretical_loss": 3.3882894319871983,
      "tokens_seen": 2582642688
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021981945837512538,
      "loss": 1.2961,
      "theoretical_loss": 3.38828267561952,
      "tokens_seen": 2582708224
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021979939819458374,
      "loss": 1.328,
      "theoretical_loss": 3.388275919471284,
      "tokens_seen": 2582773760
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021977933801404213,
      "loss": 1.453,
      "theoretical_loss": 3.388269163542478,
      "tokens_seen": 2582839296
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002197592778335005,
      "loss": 1.3283,
      "theoretical_loss": 3.388262407833089,
      "tokens_seen": 2582904832
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021973921765295888,
      "loss": 1.2912,
      "theoretical_loss": 3.3882556523431044,
      "tokens_seen": 2582970368
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021971915747241724,
      "loss": 1.4404,
      "theoretical_loss": 3.3882488970725118,
      "tokens_seen": 2583035904
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021969909729187563,
      "loss": 1.4612,
      "theoretical_loss": 3.388242142021298,
      "tokens_seen": 2583101440
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000219679037111334,
      "loss": 1.366,
      "theoretical_loss": 3.3882353871894506,
      "tokens_seen": 2583166976
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021965897693079238,
      "loss": 1.4078,
      "theoretical_loss": 3.388228632576957,
      "tokens_seen": 2583232512
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021963891675025075,
      "loss": 1.3765,
      "theoretical_loss": 3.388221878183804,
      "tokens_seen": 2583298048
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002196188565697091,
      "loss": 1.3525,
      "theoretical_loss": 3.38821512400998,
      "tokens_seen": 2583363584
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002195987963891675,
      "loss": 1.277,
      "theoretical_loss": 3.388208370055471,
      "tokens_seen": 2583429120
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021957873620862589,
      "loss": 1.3387,
      "theoretical_loss": 3.388201616320265,
      "tokens_seen": 2583494656
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021955867602808428,
      "loss": 1.3894,
      "theoretical_loss": 3.3881948628043492,
      "tokens_seen": 2583560192
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021953861584754264,
      "loss": 1.3165,
      "theoretical_loss": 3.3881881095077113,
      "tokens_seen": 2583625728
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021951855566700103,
      "loss": 1.3328,
      "theoretical_loss": 3.388181356430338,
      "tokens_seen": 2583691264
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002194984954864594,
      "loss": 1.3017,
      "theoretical_loss": 3.3881746035722173,
      "tokens_seen": 2583756800
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021947843530591778,
      "loss": 1.2678,
      "theoretical_loss": 3.3881678509333355,
      "tokens_seen": 2583822336
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021945837512537614,
      "loss": 1.3237,
      "theoretical_loss": 3.388161098513681,
      "tokens_seen": 2583887872
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002194383149448345,
      "loss": 1.4,
      "theoretical_loss": 3.3881543463132404,
      "tokens_seen": 2583953408
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002194182547642929,
      "loss": 1.4276,
      "theoretical_loss": 3.3881475943320014,
      "tokens_seen": 2584018944
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021939819458375125,
      "loss": 1.2092,
      "theoretical_loss": 3.3881408425699515,
      "tokens_seen": 2584084480
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021937813440320964,
      "loss": 1.3565,
      "theoretical_loss": 3.3881340910270774,
      "tokens_seen": 2584150016
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000219358074222668,
      "loss": 1.3306,
      "theoretical_loss": 3.3881273397033667,
      "tokens_seen": 2584215552
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002193380140421264,
      "loss": 1.404,
      "theoretical_loss": 3.3881205885988073,
      "tokens_seen": 2584281088
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021931795386158476,
      "loss": 1.3913,
      "theoretical_loss": 3.3881138377133855,
      "tokens_seen": 2584346624
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021929789368104315,
      "loss": 1.3273,
      "theoretical_loss": 3.3881070870470893,
      "tokens_seen": 2584412160
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002192778335005015,
      "loss": 1.3134,
      "theoretical_loss": 3.388100336599906,
      "tokens_seen": 2584477696
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002192577733199599,
      "loss": 1.2803,
      "theoretical_loss": 3.3880935863718227,
      "tokens_seen": 2584543232
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021923771313941826,
      "loss": 1.3505,
      "theoretical_loss": 3.3880868363628265,
      "tokens_seen": 2584608768
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021921765295887662,
      "loss": 1.4275,
      "theoretical_loss": 3.3880800865729057,
      "tokens_seen": 2584674304
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000219197592778335,
      "loss": 1.3171,
      "theoretical_loss": 3.388073337002047,
      "tokens_seen": 2584739840
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021917753259779337,
      "loss": 1.2504,
      "theoretical_loss": 3.388066587650237,
      "tokens_seen": 2584805376
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021915747241725176,
      "loss": 1.4307,
      "theoretical_loss": 3.3880598385174645,
      "tokens_seen": 2584870912
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021913741223671013,
      "loss": 1.2885,
      "theoretical_loss": 3.3880530896037158,
      "tokens_seen": 2584936448
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021911735205616852,
      "loss": 1.3998,
      "theoretical_loss": 3.388046340908979,
      "tokens_seen": 2585001984
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021909729187562688,
      "loss": 1.3477,
      "theoretical_loss": 3.3880395924332407,
      "tokens_seen": 2585067520
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021907723169508527,
      "loss": 1.2134,
      "theoretical_loss": 3.3880328441764886,
      "tokens_seen": 2585133056
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021905717151454363,
      "loss": 1.4055,
      "theoretical_loss": 3.38802609613871,
      "tokens_seen": 2585198592
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000219037111334002,
      "loss": 1.4373,
      "theoretical_loss": 3.388019348319892,
      "tokens_seen": 2585264128
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021901705115346038,
      "loss": 1.2697,
      "theoretical_loss": 3.3880126007200224,
      "tokens_seen": 2585329664
    },
    {
      "epoch": 0.78,
      "objective/train/advantage_avg": 0.4940796494483948,
      "objective/train/docs_used": 939988,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2108778953552246,
      "objective/train/original_loss": 1.2108778953552246,
      "objective/train/theoretical_loss": 3.3880058533390884,
      "objective/train/tokens_used": 2605855200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24459028244018555,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506514310836792,
      "objective/train/weighted_lm_loss": 1.2722145318984985,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3880058533390884,
      "tokens_seen": 2585395200
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021899699097291874,
      "loss": 1.2704,
      "theoretical_loss": 3.3880058533390884,
      "tokens_seen": 2585395200
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021897693079237713,
      "loss": 1.2768,
      "theoretical_loss": 3.3879991061770776,
      "tokens_seen": 2585460736
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002189568706118355,
      "loss": 1.3385,
      "theoretical_loss": 3.3879923592339765,
      "tokens_seen": 2585526272
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021893681043129388,
      "loss": 1.4287,
      "theoretical_loss": 3.3879856125097727,
      "tokens_seen": 2585591808
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021891675025075225,
      "loss": 1.3467,
      "theoretical_loss": 3.3879788660044543,
      "tokens_seen": 2585657344
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021889669007021064,
      "loss": 1.3812,
      "theoretical_loss": 3.3879721197180084,
      "tokens_seen": 2585722880
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000218876629889669,
      "loss": 1.1413,
      "theoretical_loss": 3.3879653736504216,
      "tokens_seen": 2585788416
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021885656970912736,
      "loss": 1.2604,
      "theoretical_loss": 3.387958627801682,
      "tokens_seen": 2585853952
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021883650952858578,
      "loss": 1.3127,
      "theoretical_loss": 3.3879518821717767,
      "tokens_seen": 2585919488
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021881644934804414,
      "loss": 1.306,
      "theoretical_loss": 3.387945136760693,
      "tokens_seen": 2585985024
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021879638916750253,
      "loss": 1.2845,
      "theoretical_loss": 3.3879383915684187,
      "tokens_seen": 2586050560
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002187763289869609,
      "loss": 1.4562,
      "theoretical_loss": 3.3879316465949407,
      "tokens_seen": 2586116096
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021875626880641928,
      "loss": 1.4842,
      "theoretical_loss": 3.387924901840246,
      "tokens_seen": 2586181632
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021873620862587764,
      "loss": 1.3161,
      "theoretical_loss": 3.387918157304323,
      "tokens_seen": 2586247168
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021871614844533603,
      "loss": 1.4826,
      "theoretical_loss": 3.387911412987158,
      "tokens_seen": 2586312704
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002186960882647944,
      "loss": 1.3999,
      "theoretical_loss": 3.3879046688887393,
      "tokens_seen": 2586378240
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021867602808425278,
      "loss": 1.2589,
      "theoretical_loss": 3.3878979250090535,
      "tokens_seen": 2586443776
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021865596790371115,
      "loss": 1.3459,
      "theoretical_loss": 3.3878911813480883,
      "tokens_seen": 2586509312
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002186359077231695,
      "loss": 1.3803,
      "theoretical_loss": 3.3878844379058313,
      "tokens_seen": 2586574848
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002186158475426279,
      "loss": 1.4392,
      "theoretical_loss": 3.3878776946822695,
      "tokens_seen": 2586640384
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021859578736208626,
      "loss": 1.362,
      "theoretical_loss": 3.38787095167739,
      "tokens_seen": 2586705920
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021857572718154465,
      "loss": 1.3589,
      "theoretical_loss": 3.387864208891181,
      "tokens_seen": 2586771456
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000218555667001003,
      "loss": 1.428,
      "theoretical_loss": 3.387857466323629,
      "tokens_seen": 2586836992
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002185356068204614,
      "loss": 1.5137,
      "theoretical_loss": 3.387850723974722,
      "tokens_seen": 2586902528
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021851554663991976,
      "loss": 1.4968,
      "theoretical_loss": 3.3878439818444472,
      "tokens_seen": 2586968064
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021849548645937815,
      "loss": 1.2643,
      "theoretical_loss": 3.387837239932792,
      "tokens_seen": 2587033600
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021847542627883651,
      "loss": 1.3182,
      "theoretical_loss": 3.3878304982397434,
      "tokens_seen": 2587099136
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021845536609829488,
      "loss": 1.4416,
      "theoretical_loss": 3.387823756765289,
      "tokens_seen": 2587164672
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021843530591775327,
      "loss": 1.3872,
      "theoretical_loss": 3.3878170155094165,
      "tokens_seen": 2587230208
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021841524573721163,
      "loss": 1.3035,
      "theoretical_loss": 3.387810274472113,
      "tokens_seen": 2587295744
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021839518555667002,
      "loss": 1.2197,
      "theoretical_loss": 3.387803533653366,
      "tokens_seen": 2587361280
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021837512537612838,
      "loss": 1.4573,
      "theoretical_loss": 3.3877967930531625,
      "tokens_seen": 2587426816
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021835506519558677,
      "loss": 1.3035,
      "theoretical_loss": 3.38779005267149,
      "tokens_seen": 2587492352
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021833500501504513,
      "loss": 1.4432,
      "theoretical_loss": 3.3877833125083368,
      "tokens_seen": 2587557888
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021831494483450352,
      "loss": 1.45,
      "theoretical_loss": 3.3877765725636886,
      "tokens_seen": 2587623424
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021829488465396188,
      "loss": 1.1906,
      "theoretical_loss": 3.3877698328375345,
      "tokens_seen": 2587688960
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021827482447342024,
      "loss": 1.42,
      "theoretical_loss": 3.387763093329861,
      "tokens_seen": 2587754496
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021825476429287863,
      "loss": 1.4689,
      "theoretical_loss": 3.387756354040655,
      "tokens_seen": 2587820032
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.000218234704112337,
      "loss": 1.3651,
      "theoretical_loss": 3.3877496149699047,
      "tokens_seen": 2587885568
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021821464393179539,
      "loss": 1.3851,
      "theoretical_loss": 3.3877428761175974,
      "tokens_seen": 2587951104
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021819458375125375,
      "loss": 1.306,
      "theoretical_loss": 3.38773613748372,
      "tokens_seen": 2588016640
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021817452357071214,
      "loss": 1.3107,
      "theoretical_loss": 3.3877293990682604,
      "tokens_seen": 2588082176
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002181544633901705,
      "loss": 1.2848,
      "theoretical_loss": 3.387722660871206,
      "tokens_seen": 2588147712
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002181344032096289,
      "loss": 1.3545,
      "theoretical_loss": 3.3877159228925433,
      "tokens_seen": 2588213248
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021811434302908725,
      "loss": 1.4676,
      "theoretical_loss": 3.3877091851322607,
      "tokens_seen": 2588278784
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021809428284854564,
      "loss": 1.3306,
      "theoretical_loss": 3.3877024475903457,
      "tokens_seen": 2588344320
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021807422266800403,
      "loss": 1.301,
      "theoretical_loss": 3.387695710266785,
      "tokens_seen": 2588409856
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002180541624874624,
      "loss": 1.4607,
      "theoretical_loss": 3.3876889731615663,
      "tokens_seen": 2588475392
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021803410230692078,
      "loss": 1.3741,
      "theoretical_loss": 3.387682236274677,
      "tokens_seen": 2588540928
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021801404212637914,
      "loss": 1.4778,
      "theoretical_loss": 3.3876754996061047,
      "tokens_seen": 2588606464
    },
    {
      "epoch": 0.78,
      "objective/train/advantage_avg": 0.4922197461128235,
      "objective/train/docs_used": 941211,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2892658710479736,
      "objective/train/original_loss": 1.289265751838684,
      "objective/train/theoretical_loss": 3.387668763155836,
      "objective/train/tokens_used": 2609132000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24262723326683044,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0504553318023682,
      "objective/train/weighted_lm_loss": 1.3542470932006836,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.387668763155836,
      "tokens_seen": 2588672000
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021799398194583753,
      "loss": 1.3262,
      "theoretical_loss": 3.387668763155836,
      "tokens_seen": 2588672000
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002179739217652959,
      "loss": 1.4298,
      "theoretical_loss": 3.3876620269238593,
      "tokens_seen": 2588737536
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021795386158475428,
      "loss": 1.3744,
      "theoretical_loss": 3.3876552909101614,
      "tokens_seen": 2588803072
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021793380140421265,
      "loss": 1.4272,
      "theoretical_loss": 3.3876485551147297,
      "tokens_seen": 2588868608
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021791374122367104,
      "loss": 1.2353,
      "theoretical_loss": 3.387641819537552,
      "tokens_seen": 2588934144
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002178936810431294,
      "loss": 1.3706,
      "theoretical_loss": 3.3876350841786156,
      "tokens_seen": 2588999680
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021787362086258776,
      "loss": 1.3619,
      "theoretical_loss": 3.3876283490379073,
      "tokens_seen": 2589065216
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021785356068204615,
      "loss": 1.3808,
      "theoretical_loss": 3.3876216141154156,
      "tokens_seen": 2589130752
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002178335005015045,
      "loss": 1.5203,
      "theoretical_loss": 3.3876148794111267,
      "tokens_seen": 2589196288
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002178134403209629,
      "loss": 1.3548,
      "theoretical_loss": 3.3876081449250286,
      "tokens_seen": 2589261824
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021779338014042126,
      "loss": 1.3385,
      "theoretical_loss": 3.3876014106571093,
      "tokens_seen": 2589327360
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021777331995987965,
      "loss": 1.3199,
      "theoretical_loss": 3.387594676607355,
      "tokens_seen": 2589392896
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021775325977933802,
      "loss": 1.276,
      "theoretical_loss": 3.3875879427757543,
      "tokens_seen": 2589458432
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002177331995987964,
      "loss": 1.3011,
      "theoretical_loss": 3.3875812091622937,
      "tokens_seen": 2589523968
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021771313941825477,
      "loss": 1.4287,
      "theoretical_loss": 3.387574475766961,
      "tokens_seen": 2589589504
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021769307923771316,
      "loss": 1.2998,
      "theoretical_loss": 3.3875677425897432,
      "tokens_seen": 2589655040
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021767301905717152,
      "loss": 1.3726,
      "theoretical_loss": 3.3875610096306286,
      "tokens_seen": 2589720576
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021765295887662988,
      "loss": 1.2781,
      "theoretical_loss": 3.387554276889604,
      "tokens_seen": 2589786112
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021763289869608827,
      "loss": 1.34,
      "theoretical_loss": 3.3875475443666563,
      "tokens_seen": 2589851648
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021761283851554663,
      "loss": 1.4061,
      "theoretical_loss": 3.3875408120617743,
      "tokens_seen": 2589917184
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021759277833500502,
      "loss": 1.4241,
      "theoretical_loss": 3.3875340799749445,
      "tokens_seen": 2589982720
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021757271815446338,
      "loss": 1.3467,
      "theoretical_loss": 3.387527348106154,
      "tokens_seen": 2590048256
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021755265797392177,
      "loss": 1.2541,
      "theoretical_loss": 3.387520616455391,
      "tokens_seen": 2590113792
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021753259779338014,
      "loss": 1.2687,
      "theoretical_loss": 3.387513885022643,
      "tokens_seen": 2590179328
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021751253761283852,
      "loss": 1.1872,
      "theoretical_loss": 3.3875071538078965,
      "tokens_seen": 2590244864
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002174924774322969,
      "loss": 1.2427,
      "theoretical_loss": 3.3875004228111396,
      "tokens_seen": 2590310400
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021747241725175525,
      "loss": 1.2912,
      "theoretical_loss": 3.3874936920323595,
      "tokens_seen": 2590375936
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.00021745235707121364,
      "loss": 1.3978,
      "theoretical_loss": 3.387486961471544,
      "tokens_seen": 2590441472
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.000217432296890672,
      "loss": 1.196,
      "theoretical_loss": 3.38748023112868,
      "tokens_seen": 2590507008
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002174122367101304,
      "loss": 1.3795,
      "theoretical_loss": 3.387473501003755,
      "tokens_seen": 2590572544
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021739217652958875,
      "loss": 1.3139,
      "theoretical_loss": 3.387466771096757,
      "tokens_seen": 2590638080
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021737211634904714,
      "loss": 1.2589,
      "theoretical_loss": 3.3874600414076728,
      "tokens_seen": 2590703616
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002173520561685055,
      "loss": 1.3384,
      "theoretical_loss": 3.3874533119364902,
      "tokens_seen": 2590769152
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021733199598796392,
      "loss": 1.4183,
      "theoretical_loss": 3.3874465826831965,
      "tokens_seen": 2590834688
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021731193580742228,
      "loss": 1.4252,
      "theoretical_loss": 3.3874398536477788,
      "tokens_seen": 2590900224
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021729187562688065,
      "loss": 1.3267,
      "theoretical_loss": 3.387433124830225,
      "tokens_seen": 2590965760
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021727181544633903,
      "loss": 1.3617,
      "theoretical_loss": 3.3874263962305227,
      "tokens_seen": 2591031296
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002172517552657974,
      "loss": 1.2539,
      "theoretical_loss": 3.3874196678486586,
      "tokens_seen": 2591096832
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021723169508525579,
      "loss": 1.2011,
      "theoretical_loss": 3.387412939684621,
      "tokens_seen": 2591162368
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021721163490471415,
      "loss": 1.4163,
      "theoretical_loss": 3.3874062117383965,
      "tokens_seen": 2591227904
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021719157472417254,
      "loss": 1.3745,
      "theoretical_loss": 3.3873994840099733,
      "tokens_seen": 2591293440
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002171715145436309,
      "loss": 1.237,
      "theoretical_loss": 3.387392756499338,
      "tokens_seen": 2591358976
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002171514543630893,
      "loss": 1.4821,
      "theoretical_loss": 3.3873860292064792,
      "tokens_seen": 2591424512
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021713139418254765,
      "loss": 1.3135,
      "theoretical_loss": 3.387379302131383,
      "tokens_seen": 2591490048
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021711133400200604,
      "loss": 1.4461,
      "theoretical_loss": 3.3873725752740382,
      "tokens_seen": 2591555584
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002170912738214644,
      "loss": 1.2783,
      "theoretical_loss": 3.387365848634431,
      "tokens_seen": 2591621120
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021707121364092277,
      "loss": 1.2558,
      "theoretical_loss": 3.38735912221255,
      "tokens_seen": 2591686656
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021705115346038115,
      "loss": 1.3489,
      "theoretical_loss": 3.3873523960083816,
      "tokens_seen": 2591752192
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021703109327983952,
      "loss": 1.3855,
      "theoretical_loss": 3.387345670021914,
      "tokens_seen": 2591817728
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002170110330992979,
      "loss": 1.372,
      "theoretical_loss": 3.387338944253134,
      "tokens_seen": 2591883264
    },
    {
      "epoch": 0.79,
      "objective/train/advantage_avg": 0.4944056570529938,
      "objective/train/docs_used": 942495,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2341057062149048,
      "objective/train/original_loss": 1.2341053485870361,
      "objective/train/theoretical_loss": 3.3873322187020296,
      "objective/train/tokens_used": 2612408800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24471396207809448,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506846904754639,
      "objective/train/weighted_lm_loss": 1.296675682067871,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3873322187020296,
      "tokens_seen": 2591948800
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021699097291875627,
      "loss": 1.2928,
      "theoretical_loss": 3.3873322187020296,
      "tokens_seen": 2591948800
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021697091273821466,
      "loss": 1.3298,
      "theoretical_loss": 3.387325493368588,
      "tokens_seen": 2592014336
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021695085255767302,
      "loss": 1.437,
      "theoretical_loss": 3.387318768252797,
      "tokens_seen": 2592079872
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002169307923771314,
      "loss": 1.3417,
      "theoretical_loss": 3.387312043354643,
      "tokens_seen": 2592145408
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021691073219658977,
      "loss": 1.4195,
      "theoretical_loss": 3.387305318674115,
      "tokens_seen": 2592210944
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021689067201604813,
      "loss": 1.2775,
      "theoretical_loss": 3.3872985942111993,
      "tokens_seen": 2592276480
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021687061183550652,
      "loss": 1.3549,
      "theoretical_loss": 3.387291869965884,
      "tokens_seen": 2592342016
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021685055165496489,
      "loss": 1.4596,
      "theoretical_loss": 3.387285145938156,
      "tokens_seen": 2592407552
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021683049147442327,
      "loss": 1.4046,
      "theoretical_loss": 3.3872784221280035,
      "tokens_seen": 2592473088
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021681043129388164,
      "loss": 1.3632,
      "theoretical_loss": 3.3872716985354128,
      "tokens_seen": 2592538624
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021679037111334003,
      "loss": 1.3646,
      "theoretical_loss": 3.3872649751603725,
      "tokens_seen": 2592604160
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002167703109327984,
      "loss": 1.4041,
      "theoretical_loss": 3.38725825200287,
      "tokens_seen": 2592669696
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021675025075225678,
      "loss": 1.336,
      "theoretical_loss": 3.3872515290628917,
      "tokens_seen": 2592735232
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021673019057171514,
      "loss": 1.3875,
      "theoretical_loss": 3.387244806340426,
      "tokens_seen": 2592800768
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002167101303911735,
      "loss": 1.3561,
      "theoretical_loss": 3.3872380838354603,
      "tokens_seen": 2592866304
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002166900702106319,
      "loss": 1.2574,
      "theoretical_loss": 3.3872313615479817,
      "tokens_seen": 2592931840
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021667001003009025,
      "loss": 1.3063,
      "theoretical_loss": 3.387224639477978,
      "tokens_seen": 2592997376
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021664994984954864,
      "loss": 1.2853,
      "theoretical_loss": 3.3872179176254362,
      "tokens_seen": 2593062912
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.000216629889669007,
      "loss": 1.2539,
      "theoretical_loss": 3.3872111959903446,
      "tokens_seen": 2593128448
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002166098294884654,
      "loss": 1.2727,
      "theoretical_loss": 3.38720447457269,
      "tokens_seen": 2593193984
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021658976930792378,
      "loss": 1.3688,
      "theoretical_loss": 3.38719775337246,
      "tokens_seen": 2593259520
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021656970912738217,
      "loss": 1.2169,
      "theoretical_loss": 3.3871910323896417,
      "tokens_seen": 2593325056
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021654964894684054,
      "loss": 1.372,
      "theoretical_loss": 3.3871843116242237,
      "tokens_seen": 2593390592
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021652958876629893,
      "loss": 1.3508,
      "theoretical_loss": 3.3871775910761923,
      "tokens_seen": 2593456128
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002165095285857573,
      "loss": 1.3096,
      "theoretical_loss": 3.3871708707455355,
      "tokens_seen": 2593521664
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021648946840521565,
      "loss": 1.4485,
      "theoretical_loss": 3.3871641506322407,
      "tokens_seen": 2593587200
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021646940822467404,
      "loss": 1.3107,
      "theoretical_loss": 3.3871574307362957,
      "tokens_seen": 2593652736
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002164493480441324,
      "loss": 1.3191,
      "theoretical_loss": 3.387150711057687,
      "tokens_seen": 2593718272
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002164292878635908,
      "loss": 1.3313,
      "theoretical_loss": 3.3871439915964032,
      "tokens_seen": 2593783808
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021640922768304915,
      "loss": 1.3261,
      "theoretical_loss": 3.3871372723524313,
      "tokens_seen": 2593849344
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021638916750250754,
      "loss": 1.4154,
      "theoretical_loss": 3.387130553325759,
      "tokens_seen": 2593914880
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002163691073219659,
      "loss": 1.213,
      "theoretical_loss": 3.387123834516373,
      "tokens_seen": 2593980416
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002163490471414243,
      "loss": 1.3603,
      "theoretical_loss": 3.387117115924262,
      "tokens_seen": 2594045952
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021632898696088266,
      "loss": 1.4757,
      "theoretical_loss": 3.3871103975494123,
      "tokens_seen": 2594111488
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021630892678034102,
      "loss": 1.3957,
      "theoretical_loss": 3.3871036793918123,
      "tokens_seen": 2594177024
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002162888665997994,
      "loss": 1.3838,
      "theoretical_loss": 3.3870969614514492,
      "tokens_seen": 2594242560
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021626880641925777,
      "loss": 1.3914,
      "theoretical_loss": 3.38709024372831,
      "tokens_seen": 2594308096
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021624874623871616,
      "loss": 1.2812,
      "theoretical_loss": 3.387083526222383,
      "tokens_seen": 2594373632
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021622868605817452,
      "loss": 1.3986,
      "theoretical_loss": 3.3870768089336556,
      "tokens_seen": 2594439168
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002162086258776329,
      "loss": 1.2774,
      "theoretical_loss": 3.3870700918621144,
      "tokens_seen": 2594504704
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021618856569709127,
      "loss": 1.392,
      "theoretical_loss": 3.387063375007748,
      "tokens_seen": 2594570240
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021616850551654966,
      "loss": 1.3955,
      "theoretical_loss": 3.3870566583705424,
      "tokens_seen": 2594635776
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021614844533600802,
      "loss": 1.4165,
      "theoretical_loss": 3.3870499419504867,
      "tokens_seen": 2594701312
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021612838515546641,
      "loss": 1.3593,
      "theoretical_loss": 3.387043225747568,
      "tokens_seen": 2594766848
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021610832497492478,
      "loss": 1.3799,
      "theoretical_loss": 3.3870365097617734,
      "tokens_seen": 2594832384
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021608826479438314,
      "loss": 1.3252,
      "theoretical_loss": 3.3870297939930905,
      "tokens_seen": 2594897920
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021606820461384153,
      "loss": 1.3783,
      "theoretical_loss": 3.3870230784415067,
      "tokens_seen": 2594963456
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002160481444332999,
      "loss": 1.407,
      "theoretical_loss": 3.38701636310701,
      "tokens_seen": 2595028992
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021602808425275828,
      "loss": 1.3875,
      "theoretical_loss": 3.3870096479895873,
      "tokens_seen": 2595094528
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021600802407221664,
      "loss": 1.3734,
      "theoretical_loss": 3.387002933089226,
      "tokens_seen": 2595160064
    },
    {
      "epoch": 0.79,
      "objective/train/advantage_avg": 0.4897902309894562,
      "objective/train/docs_used": 943650,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3126845359802246,
      "objective/train/original_loss": 1.312684416770935,
      "objective/train/theoretical_loss": 3.3869962184059146,
      "objective/train/tokens_used": 2615685600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24117735028266907,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.050204873085022,
      "objective/train/weighted_lm_loss": 1.3786314725875854,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.9048373699188232,
      "theoretical_loss": 3.3869962184059146,
      "tokens_seen": 2595225600
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021598796389167503,
      "loss": 1.4937,
      "theoretical_loss": 3.3869962184059146,
      "tokens_seen": 2595225600
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002159679037111334,
      "loss": 1.4249,
      "theoretical_loss": 3.3869895039396396,
      "tokens_seen": 2595291136
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021594784353059178,
      "loss": 1.3561,
      "theoretical_loss": 3.3869827896903892,
      "tokens_seen": 2595356672
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021592778335005014,
      "loss": 1.5021,
      "theoretical_loss": 3.38697607565815,
      "tokens_seen": 2595422208
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002159077231695085,
      "loss": 1.4742,
      "theoretical_loss": 3.386969361842911,
      "tokens_seen": 2595487744
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002158876629889669,
      "loss": 1.4509,
      "theoretical_loss": 3.386962648244658,
      "tokens_seen": 2595553280
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021586760280842526,
      "loss": 1.361,
      "theoretical_loss": 3.386955934863379,
      "tokens_seen": 2595618816
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021584754262788365,
      "loss": 1.3274,
      "theoretical_loss": 3.3869492216990627,
      "tokens_seen": 2595684352
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021582748244734204,
      "loss": 1.4105,
      "theoretical_loss": 3.3869425087516953,
      "tokens_seen": 2595749888
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021580742226680043,
      "loss": 1.4191,
      "theoretical_loss": 3.3869357960212647,
      "tokens_seen": 2595815424
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002157873620862588,
      "loss": 1.3826,
      "theoretical_loss": 3.3869290835077583,
      "tokens_seen": 2595880960
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021576730190571718,
      "loss": 1.2759,
      "theoretical_loss": 3.3869223712111642,
      "tokens_seen": 2595946496
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021574724172517554,
      "loss": 1.4924,
      "theoretical_loss": 3.386915659131469,
      "tokens_seen": 2596012032
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002157271815446339,
      "loss": 1.3318,
      "theoretical_loss": 3.386908947268661,
      "tokens_seen": 2596077568
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002157071213640923,
      "loss": 1.3348,
      "theoretical_loss": 3.386902235622727,
      "tokens_seen": 2596143104
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021568706118355065,
      "loss": 1.3836,
      "theoretical_loss": 3.3868955241936556,
      "tokens_seen": 2596208640
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021566700100300904,
      "loss": 1.5604,
      "theoretical_loss": 3.386888812981433,
      "tokens_seen": 2596274176
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002156469408224674,
      "loss": 1.395,
      "theoretical_loss": 3.3868821019860476,
      "tokens_seen": 2596339712
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002156268806419258,
      "loss": 1.3514,
      "theoretical_loss": 3.3868753912074867,
      "tokens_seen": 2596405248
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021560682046138416,
      "loss": 1.3234,
      "theoretical_loss": 3.3868686806457378,
      "tokens_seen": 2596470784
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021558676028084255,
      "loss": 1.2358,
      "theoretical_loss": 3.3868619703007883,
      "tokens_seen": 2596536320
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002155667001003009,
      "loss": 1.344,
      "theoretical_loss": 3.3868552601726263,
      "tokens_seen": 2596601856
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002155466399197593,
      "loss": 1.3369,
      "theoretical_loss": 3.3868485502612384,
      "tokens_seen": 2596667392
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021552657973921766,
      "loss": 1.3294,
      "theoretical_loss": 3.3868418405666127,
      "tokens_seen": 2596732928
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021550651955867602,
      "loss": 1.3084,
      "theoretical_loss": 3.386835131088737,
      "tokens_seen": 2596798464
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002154864593781344,
      "loss": 1.2664,
      "theoretical_loss": 3.3868284218275977,
      "tokens_seen": 2596864000
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021546639919759277,
      "loss": 1.3061,
      "theoretical_loss": 3.3868217127831834,
      "tokens_seen": 2596929536
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021544633901705116,
      "loss": 1.3855,
      "theoretical_loss": 3.3868150039554816,
      "tokens_seen": 2596995072
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021542627883650953,
      "loss": 1.4086,
      "theoretical_loss": 3.386808295344479,
      "tokens_seen": 2597060608
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021540621865596792,
      "loss": 1.217,
      "theoretical_loss": 3.386801586950164,
      "tokens_seen": 2597126144
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021538615847542628,
      "loss": 1.4545,
      "theoretical_loss": 3.386794878772524,
      "tokens_seen": 2597191680
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021536609829488467,
      "loss": 1.4049,
      "theoretical_loss": 3.3867881708115464,
      "tokens_seen": 2597257216
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021534603811434303,
      "loss": 1.3339,
      "theoretical_loss": 3.386781463067218,
      "tokens_seen": 2597322752
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002153259779338014,
      "loss": 1.3757,
      "theoretical_loss": 3.386774755539528,
      "tokens_seen": 2597388288
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021530591775325978,
      "loss": 1.3564,
      "theoretical_loss": 3.386768048228462,
      "tokens_seen": 2597453824
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021528585757271814,
      "loss": 1.3733,
      "theoretical_loss": 3.386761341134009,
      "tokens_seen": 2597519360
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021526579739217653,
      "loss": 1.2837,
      "theoretical_loss": 3.386754634256156,
      "tokens_seen": 2597584896
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002152457372116349,
      "loss": 1.3449,
      "theoretical_loss": 3.3867479275948904,
      "tokens_seen": 2597650432
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021522567703109328,
      "loss": 1.2409,
      "theoretical_loss": 3.3867412211502,
      "tokens_seen": 2597715968
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021520561685055165,
      "loss": 1.3434,
      "theoretical_loss": 3.3867345149220722,
      "tokens_seen": 2597781504
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021518555667001004,
      "loss": 1.3869,
      "theoretical_loss": 3.3867278089104946,
      "tokens_seen": 2597847040
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002151654964894684,
      "loss": 1.3619,
      "theoretical_loss": 3.386721103115455,
      "tokens_seen": 2597912576
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021514543630892676,
      "loss": 1.2428,
      "theoretical_loss": 3.3867143975369407,
      "tokens_seen": 2597978112
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021512537612838515,
      "loss": 1.4088,
      "theoretical_loss": 3.386707692174939,
      "tokens_seen": 2598043648
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002151053159478435,
      "loss": 1.4333,
      "theoretical_loss": 3.3867009870294376,
      "tokens_seen": 2598109184
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021508525576730193,
      "loss": 1.4904,
      "theoretical_loss": 3.3866942821004242,
      "tokens_seen": 2598174720
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002150651955867603,
      "loss": 1.3797,
      "theoretical_loss": 3.3866875773878866,
      "tokens_seen": 2598240256
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021504513540621868,
      "loss": 1.3739,
      "theoretical_loss": 3.3866808728918114,
      "tokens_seen": 2598305792
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021502507522567704,
      "loss": 1.4146,
      "theoretical_loss": 3.3866741686121875,
      "tokens_seen": 2598371328
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021500501504513543,
      "loss": 1.437,
      "theoretical_loss": 3.386667464549001,
      "tokens_seen": 2598436864
    },
    {
      "epoch": 0.79,
      "objective/train/advantage_avg": 0.4915306568145752,
      "objective/train/docs_used": 944930,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2090405225753784,
      "objective/train/original_loss": 1.209040641784668,
      "objective/train/theoretical_loss": 3.3866607607022408,
      "objective/train/tokens_used": 2618962400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24274511635303497,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0503870248794556,
      "objective/train/weighted_lm_loss": 1.2697603702545166,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.7788007855415344,
      "theoretical_loss": 3.3866607607022408,
      "tokens_seen": 2598502400
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002149849548645938,
      "loss": 1.2655,
      "theoretical_loss": 3.3866607607022408,
      "tokens_seen": 2598502400
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021496489468405218,
      "loss": 1.345,
      "theoretical_loss": 3.386654057071894,
      "tokens_seen": 2598567936
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021494483450351055,
      "loss": 1.2918,
      "theoretical_loss": 3.3866473536579473,
      "tokens_seen": 2598633472
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002149247743229689,
      "loss": 1.3934,
      "theoretical_loss": 3.3866406504603894,
      "tokens_seen": 2598699008
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002149047141424273,
      "loss": 1.4899,
      "theoretical_loss": 3.3866339474792078,
      "tokens_seen": 2598764544
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021488465396188566,
      "loss": 1.3646,
      "theoretical_loss": 3.386627244714389,
      "tokens_seen": 2598830080
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021486459378134405,
      "loss": 1.4034,
      "theoretical_loss": 3.3866205421659217,
      "tokens_seen": 2598895616
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002148445336008024,
      "loss": 1.3772,
      "theoretical_loss": 3.3866138398337924,
      "tokens_seen": 2598961152
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002148244734202608,
      "loss": 1.5088,
      "theoretical_loss": 3.3866071377179896,
      "tokens_seen": 2599026688
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021480441323971916,
      "loss": 1.3227,
      "theoretical_loss": 3.386600435818501,
      "tokens_seen": 2599092224
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021478435305917755,
      "loss": 1.3765,
      "theoretical_loss": 3.386593734135313,
      "tokens_seen": 2599157760
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021476429287863591,
      "loss": 1.3351,
      "theoretical_loss": 3.386587032668414,
      "tokens_seen": 2599223296
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021474423269809428,
      "loss": 1.3946,
      "theoretical_loss": 3.386580331417792,
      "tokens_seen": 2599288832
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021472417251755267,
      "loss": 1.5176,
      "theoretical_loss": 3.3865736303834333,
      "tokens_seen": 2599354368
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021470411233701103,
      "loss": 1.3294,
      "theoretical_loss": 3.3865669295653262,
      "tokens_seen": 2599419904
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021468405215646942,
      "loss": 1.2909,
      "theoretical_loss": 3.3865602289634587,
      "tokens_seen": 2599485440
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021466399197592778,
      "loss": 1.2635,
      "theoretical_loss": 3.386553528577817,
      "tokens_seen": 2599550976
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021464393179538617,
      "loss": 1.3454,
      "theoretical_loss": 3.386546828408391,
      "tokens_seen": 2599616512
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021462387161484453,
      "loss": 1.2541,
      "theoretical_loss": 3.3865401284551657,
      "tokens_seen": 2599682048
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021460381143430292,
      "loss": 1.331,
      "theoretical_loss": 3.3865334287181303,
      "tokens_seen": 2599747584
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021458375125376128,
      "loss": 1.3638,
      "theoretical_loss": 3.3865267291972714,
      "tokens_seen": 2599813120
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021456369107321964,
      "loss": 1.4142,
      "theoretical_loss": 3.3865200298925777,
      "tokens_seen": 2599878656
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021454363089267803,
      "loss": 1.3575,
      "theoretical_loss": 3.3865133308040356,
      "tokens_seen": 2599944192
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002145235707121364,
      "loss": 1.3329,
      "theoretical_loss": 3.386506631931634,
      "tokens_seen": 2600009728
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021450351053159479,
      "loss": 1.4595,
      "theoretical_loss": 3.386499933275359,
      "tokens_seen": 2600075264
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021448345035105315,
      "loss": 1.4322,
      "theoretical_loss": 3.386493234835199,
      "tokens_seen": 2600140800
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021446339017051154,
      "loss": 1.3488,
      "theoretical_loss": 3.3864865366111414,
      "tokens_seen": 2600206336
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002144433299899699,
      "loss": 1.3388,
      "theoretical_loss": 3.386479838603174,
      "tokens_seen": 2600271872
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002144232698094283,
      "loss": 1.2386,
      "theoretical_loss": 3.386473140811284,
      "tokens_seen": 2600337408
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021440320962888665,
      "loss": 1.3306,
      "theoretical_loss": 3.3864664432354594,
      "tokens_seen": 2600402944
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021438314944834504,
      "loss": 1.2646,
      "theoretical_loss": 3.3864597458756878,
      "tokens_seen": 2600468480
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002143630892678034,
      "loss": 1.4137,
      "theoretical_loss": 3.386453048731956,
      "tokens_seen": 2600534016
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021434302908726176,
      "loss": 1.3308,
      "theoretical_loss": 3.3864463518042527,
      "tokens_seen": 2600599552
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021432296890672018,
      "loss": 1.493,
      "theoretical_loss": 3.386439655092565,
      "tokens_seen": 2600665088
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021430290872617854,
      "loss": 1.3347,
      "theoretical_loss": 3.38643295859688,
      "tokens_seen": 2600730624
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021428284854563693,
      "loss": 1.4055,
      "theoretical_loss": 3.386426262317186,
      "tokens_seen": 2600796160
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002142627883650953,
      "loss": 1.3989,
      "theoretical_loss": 3.38641956625347,
      "tokens_seen": 2600861696
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021424272818455368,
      "loss": 1.4891,
      "theoretical_loss": 3.3864128704057204,
      "tokens_seen": 2600927232
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021422266800401205,
      "loss": 1.4508,
      "theoretical_loss": 3.3864061747739242,
      "tokens_seen": 2600992768
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021420260782347044,
      "loss": 1.4407,
      "theoretical_loss": 3.386399479358069,
      "tokens_seen": 2601058304
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002141825476429288,
      "loss": 1.5119,
      "theoretical_loss": 3.3863927841581427,
      "tokens_seen": 2601123840
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021416248746238716,
      "loss": 1.4054,
      "theoretical_loss": 3.3863860891741324,
      "tokens_seen": 2601189376
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021414242728184555,
      "loss": 1.3806,
      "theoretical_loss": 3.386379394406026,
      "tokens_seen": 2601254912
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002141223671013039,
      "loss": 1.3135,
      "theoretical_loss": 3.386372699853811,
      "tokens_seen": 2601320448
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002141023069207623,
      "loss": 1.3496,
      "theoretical_loss": 3.3863660055174756,
      "tokens_seen": 2601385984
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021408224674022066,
      "loss": 1.3201,
      "theoretical_loss": 3.3863593113970065,
      "tokens_seen": 2601451520
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021406218655967905,
      "loss": 1.379,
      "theoretical_loss": 3.3863526174923915,
      "tokens_seen": 2601517056
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021404212637913742,
      "loss": 1.4859,
      "theoretical_loss": 3.386345923803619,
      "tokens_seen": 2601582592
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002140220661985958,
      "loss": 1.3192,
      "theoretical_loss": 3.3863392303306754,
      "tokens_seen": 2601648128
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021400200601805417,
      "loss": 1.2361,
      "theoretical_loss": 3.386332537073549,
      "tokens_seen": 2601713664
    },
    {
      "epoch": 0.79,
      "objective/train/advantage_avg": 0.4939787685871124,
      "objective/train/docs_used": 946148,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.4858473539352417,
      "objective/train/original_loss": 1.4858472347259521,
      "objective/train/theoretical_loss": 3.3863258440322275,
      "objective/train/tokens_used": 2622239200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24438557028770447,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506402254104614,
      "objective/train/weighted_lm_loss": 1.5611227750778198,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0,
      "theoretical_loss": 3.3863258440322275,
      "tokens_seen": 2601779200
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021398194583751256,
      "loss": 1.4761,
      "theoretical_loss": 3.3863258440322275,
      "tokens_seen": 2601779200
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021396188565697092,
      "loss": 1.2838,
      "theoretical_loss": 3.386319151206698,
      "tokens_seen": 2601844736
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021394182547642928,
      "loss": 1.3263,
      "theoretical_loss": 3.386312458596949,
      "tokens_seen": 2601910272
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021392176529588767,
      "loss": 1.5412,
      "theoretical_loss": 3.386305766202967,
      "tokens_seen": 2601975808
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021390170511534603,
      "loss": 1.443,
      "theoretical_loss": 3.3862990740247403,
      "tokens_seen": 2602041344
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021388164493480442,
      "loss": 1.2961,
      "theoretical_loss": 3.3862923820622566,
      "tokens_seen": 2602106880
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021386158475426278,
      "loss": 1.2709,
      "theoretical_loss": 3.386285690315503,
      "tokens_seen": 2602172416
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021384152457372117,
      "loss": 1.455,
      "theoretical_loss": 3.3862789987844675,
      "tokens_seen": 2602237952
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021382146439317954,
      "loss": 1.3786,
      "theoretical_loss": 3.3862723074691377,
      "tokens_seen": 2602303488
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021380140421263792,
      "loss": 1.476,
      "theoretical_loss": 3.3862656163695006,
      "tokens_seen": 2602369024
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002137813440320963,
      "loss": 1.4006,
      "theoretical_loss": 3.386258925485545,
      "tokens_seen": 2602434560
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021376128385155465,
      "loss": 1.5007,
      "theoretical_loss": 3.386252234817257,
      "tokens_seen": 2602500096
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021374122367101304,
      "loss": 1.4419,
      "theoretical_loss": 3.3862455443646255,
      "tokens_seen": 2602565632
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002137211634904714,
      "loss": 1.2823,
      "theoretical_loss": 3.3862388541276376,
      "tokens_seen": 2602631168
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002137011033099298,
      "loss": 1.3107,
      "theoretical_loss": 3.386232164106281,
      "tokens_seen": 2602696704
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021368104312938815,
      "loss": 1.4309,
      "theoretical_loss": 3.3862254743005438,
      "tokens_seen": 2602762240
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021366098294884654,
      "loss": 1.6105,
      "theoretical_loss": 3.3862187847104126,
      "tokens_seen": 2602827776
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002136409227683049,
      "loss": 1.2367,
      "theoretical_loss": 3.3862120953358756,
      "tokens_seen": 2602893312
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002136208625877633,
      "loss": 1.3654,
      "theoretical_loss": 3.3862054061769205,
      "tokens_seen": 2602958848
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021360080240722166,
      "loss": 1.3825,
      "theoretical_loss": 3.3861987172335346,
      "tokens_seen": 2603024384
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021358074222668004,
      "loss": 1.4173,
      "theoretical_loss": 3.3861920285057057,
      "tokens_seen": 2603089920
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021356068204613843,
      "loss": 1.3238,
      "theoretical_loss": 3.386185339993422,
      "tokens_seen": 2603155456
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002135406218655968,
      "loss": 1.3294,
      "theoretical_loss": 3.38617865169667,
      "tokens_seen": 2603220992
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021352056168505519,
      "loss": 1.4553,
      "theoretical_loss": 3.386171963615438,
      "tokens_seen": 2603286528
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021350050150451355,
      "loss": 1.3592,
      "theoretical_loss": 3.386165275749714,
      "tokens_seen": 2603352064
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021348044132397194,
      "loss": 1.3935,
      "theoretical_loss": 3.3861585880994847,
      "tokens_seen": 2603417600
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002134603811434303,
      "loss": 1.3222,
      "theoretical_loss": 3.3861519006647383,
      "tokens_seen": 2603483136
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002134403209628887,
      "loss": 1.2875,
      "theoretical_loss": 3.386145213445462,
      "tokens_seen": 2603548672
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021342026078234705,
      "loss": 1.3887,
      "theoretical_loss": 3.3861385264416444,
      "tokens_seen": 2603614208
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021340020060180544,
      "loss": 1.2349,
      "theoretical_loss": 3.386131839653272,
      "tokens_seen": 2603679744
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002133801404212638,
      "loss": 1.4538,
      "theoretical_loss": 3.386125153080333,
      "tokens_seen": 2603745280
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021336008024072216,
      "loss": 1.3931,
      "theoretical_loss": 3.3861184667228157,
      "tokens_seen": 2603810816
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021334002006018055,
      "loss": 1.3622,
      "theoretical_loss": 3.3861117805807064,
      "tokens_seen": 2603876352
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021331995987963892,
      "loss": 1.4065,
      "theoretical_loss": 3.3861050946539932,
      "tokens_seen": 2603941888
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002132998996990973,
      "loss": 1.4332,
      "theoretical_loss": 3.3860984089426642,
      "tokens_seen": 2604007424
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021327983951855567,
      "loss": 1.295,
      "theoretical_loss": 3.3860917234467065,
      "tokens_seen": 2604072960
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021325977933801406,
      "loss": 1.3483,
      "theoretical_loss": 3.3860850381661085,
      "tokens_seen": 2604138496
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021323971915747242,
      "loss": 1.4142,
      "theoretical_loss": 3.386078353100857,
      "tokens_seen": 2604204032
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002132196589769308,
      "loss": 1.4746,
      "theoretical_loss": 3.3860716682509397,
      "tokens_seen": 2604269568
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021319959879638917,
      "loss": 1.3486,
      "theoretical_loss": 3.386064983616345,
      "tokens_seen": 2604335104
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021317953861584753,
      "loss": 1.3348,
      "theoretical_loss": 3.3860582991970594,
      "tokens_seen": 2604400640
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021315947843530592,
      "loss": 1.2278,
      "theoretical_loss": 3.3860516149930717,
      "tokens_seen": 2604466176
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021313941825476429,
      "loss": 1.3685,
      "theoretical_loss": 3.3860449310043688,
      "tokens_seen": 2604531712
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021311935807422267,
      "loss": 1.4223,
      "theoretical_loss": 3.386038247230939,
      "tokens_seen": 2604597248
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021309929789368104,
      "loss": 1.4932,
      "theoretical_loss": 3.3860315636727694,
      "tokens_seen": 2604662784
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021307923771313943,
      "loss": 1.3407,
      "theoretical_loss": 3.3860248803298476,
      "tokens_seen": 2604728320
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002130591775325978,
      "loss": 1.4423,
      "theoretical_loss": 3.3860181972021617,
      "tokens_seen": 2604793856
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021303911735205618,
      "loss": 1.4464,
      "theoretical_loss": 3.3860115142896987,
      "tokens_seen": 2604859392
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021301905717151454,
      "loss": 1.2905,
      "theoretical_loss": 3.386004831592447,
      "tokens_seen": 2604924928
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002129989969909729,
      "loss": 1.2745,
      "theoretical_loss": 3.385998149110394,
      "tokens_seen": 2604990464
    },
    {
      "epoch": 0.79,
      "objective/train/advantage_avg": 0.4931279420852661,
      "objective/train/docs_used": 947237,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.157224416732788,
      "objective/train/original_loss": 1.1572242975234985,
      "objective/train/theoretical_loss": 3.385991466843527,
      "objective/train/tokens_used": 2625516000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24348780512809753,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505506992340088,
      "objective/train/weighted_lm_loss": 1.2156673669815063,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.385991466843527,
      "tokens_seen": 2605056000
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002129789368104313,
      "loss": 1.3415,
      "theoretical_loss": 3.385991466843527,
      "tokens_seen": 2605056000
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021295887662988965,
      "loss": 1.3551,
      "theoretical_loss": 3.3859847847918343,
      "tokens_seen": 2605121536
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021293881644934804,
      "loss": 1.3051,
      "theoretical_loss": 3.385978102955303,
      "tokens_seen": 2605187072
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002129187562688064,
      "loss": 1.3124,
      "theoretical_loss": 3.385971421333921,
      "tokens_seen": 2605252608
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002128986960882648,
      "loss": 1.4232,
      "theoretical_loss": 3.3859647399276755,
      "tokens_seen": 2605318144
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021287863590772316,
      "loss": 1.3507,
      "theoretical_loss": 3.385958058736555,
      "tokens_seen": 2605383680
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021285857572718155,
      "loss": 1.2563,
      "theoretical_loss": 3.3859513777605468,
      "tokens_seen": 2605449216
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002128385155466399,
      "loss": 1.4364,
      "theoretical_loss": 3.3859446969996383,
      "tokens_seen": 2605514752
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021281845536609832,
      "loss": 1.3544,
      "theoretical_loss": 3.3859380164538173,
      "tokens_seen": 2605580288
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002127983951855567,
      "loss": 1.2804,
      "theoretical_loss": 3.385931336123072,
      "tokens_seen": 2605645824
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021277833500501505,
      "loss": 1.3041,
      "theoretical_loss": 3.385924656007389,
      "tokens_seen": 2605711360
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021275827482447344,
      "loss": 1.366,
      "theoretical_loss": 3.385917976106757,
      "tokens_seen": 2605776896
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002127382146439318,
      "loss": 1.4211,
      "theoretical_loss": 3.385911296421163,
      "tokens_seen": 2605842432
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002127181544633902,
      "loss": 1.3381,
      "theoretical_loss": 3.3859046169505946,
      "tokens_seen": 2605907968
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021269809428284855,
      "loss": 1.3177,
      "theoretical_loss": 3.38589793769504,
      "tokens_seen": 2605973504
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021267803410230694,
      "loss": 1.3876,
      "theoretical_loss": 3.385891258654487,
      "tokens_seen": 2606039040
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002126579739217653,
      "loss": 1.4389,
      "theoretical_loss": 3.3858845798289225,
      "tokens_seen": 2606104576
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002126379137412237,
      "loss": 1.3607,
      "theoretical_loss": 3.385877901218335,
      "tokens_seen": 2606170112
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021261785356068206,
      "loss": 1.3327,
      "theoretical_loss": 3.3858712228227117,
      "tokens_seen": 2606235648
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021259779338014042,
      "loss": 1.3716,
      "theoretical_loss": 3.38586454464204,
      "tokens_seen": 2606301184
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002125777331995988,
      "loss": 1.3187,
      "theoretical_loss": 3.385857866676308,
      "tokens_seen": 2606366720
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021255767301905717,
      "loss": 1.2883,
      "theoretical_loss": 3.3858511889255034,
      "tokens_seen": 2606432256
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021253761283851556,
      "loss": 1.2603,
      "theoretical_loss": 3.385844511389614,
      "tokens_seen": 2606497792
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021251755265797392,
      "loss": 1.3554,
      "theoretical_loss": 3.385837834068627,
      "tokens_seen": 2606563328
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002124974924774323,
      "loss": 1.6039,
      "theoretical_loss": 3.38583115696253,
      "tokens_seen": 2606628864
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021247743229689067,
      "loss": 1.4265,
      "theoretical_loss": 3.3858244800713115,
      "tokens_seen": 2606694400
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021245737211634906,
      "loss": 1.3454,
      "theoretical_loss": 3.385817803394959,
      "tokens_seen": 2606759936
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021243731193580742,
      "loss": 1.3276,
      "theoretical_loss": 3.385811126933459,
      "tokens_seen": 2606825472
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002124172517552658,
      "loss": 1.381,
      "theoretical_loss": 3.3858044506868006,
      "tokens_seen": 2606891008
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021239719157472418,
      "loss": 1.2741,
      "theoretical_loss": 3.385797774654971,
      "tokens_seen": 2606956544
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021237713139418254,
      "loss": 1.3895,
      "theoretical_loss": 3.3857910988379576,
      "tokens_seen": 2607022080
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021235707121364093,
      "loss": 1.3405,
      "theoretical_loss": 3.385784423235749,
      "tokens_seen": 2607087616
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002123370110330993,
      "loss": 1.4166,
      "theoretical_loss": 3.3857777478483313,
      "tokens_seen": 2607153152
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021231695085255768,
      "loss": 1.3328,
      "theoretical_loss": 3.3857710726756935,
      "tokens_seen": 2607218688
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021229689067201604,
      "loss": 1.329,
      "theoretical_loss": 3.385764397717823,
      "tokens_seen": 2607284224
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021227683049147443,
      "loss": 1.3051,
      "theoretical_loss": 3.385757722974707,
      "tokens_seen": 2607349760
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002122567703109328,
      "loss": 1.4376,
      "theoretical_loss": 3.385751048446334,
      "tokens_seen": 2607415296
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021223671013039118,
      "loss": 1.332,
      "theoretical_loss": 3.385744374132691,
      "tokens_seen": 2607480832
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021221664994984954,
      "loss": 1.3873,
      "theoretical_loss": 3.3857377000337663,
      "tokens_seen": 2607546368
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002121965897693079,
      "loss": 1.3658,
      "theoretical_loss": 3.3857310261495472,
      "tokens_seen": 2607611904
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002121765295887663,
      "loss": 1.306,
      "theoretical_loss": 3.3857243524800213,
      "tokens_seen": 2607677440
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021215646940822466,
      "loss": 1.2425,
      "theoretical_loss": 3.3857176790251766,
      "tokens_seen": 2607742976
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021213640922768305,
      "loss": 1.3289,
      "theoretical_loss": 3.3857110057850006,
      "tokens_seen": 2607808512
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002121163490471414,
      "loss": 1.3074,
      "theoretical_loss": 3.385704332759481,
      "tokens_seen": 2607874048
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002120962888665998,
      "loss": 1.3336,
      "theoretical_loss": 3.3856976599486055,
      "tokens_seen": 2607939584
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002120762286860582,
      "loss": 1.402,
      "theoretical_loss": 3.385690987352362,
      "tokens_seen": 2608005120
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021205616850551658,
      "loss": 1.4243,
      "theoretical_loss": 3.3856843149707383,
      "tokens_seen": 2608070656
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021203610832497494,
      "loss": 1.4324,
      "theoretical_loss": 3.3856776428037216,
      "tokens_seen": 2608136192
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002120160481444333,
      "loss": 1.2067,
      "theoretical_loss": 3.3856709708513,
      "tokens_seen": 2608201728
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002119959879638917,
      "loss": 1.2502,
      "theoretical_loss": 3.385664299113461,
      "tokens_seen": 2608267264
    },
    {
      "epoch": 0.79,
      "objective/train/advantage_avg": 0.49286285042762756,
      "objective/train/docs_used": 948381,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2887426614761353,
      "objective/train/original_loss": 1.2887426614761353,
      "objective/train/theoretical_loss": 3.3856576275901924,
      "objective/train/tokens_used": 2628792800,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24341760575771332,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505237579345703,
      "objective/train/weighted_lm_loss": 1.3537731170654297,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.3856576275901924,
      "tokens_seen": 2608332800
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021197592778335005,
      "loss": 1.3603,
      "theoretical_loss": 3.3856576275901924,
      "tokens_seen": 2608332800
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021195586760280844,
      "loss": 1.4405,
      "theoretical_loss": 3.3856509562814816,
      "tokens_seen": 2608398336
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002119358074222668,
      "loss": 1.3592,
      "theoretical_loss": 3.385644285187317,
      "tokens_seen": 2608463872
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002119157472417252,
      "loss": 1.3106,
      "theoretical_loss": 3.3856376143076856,
      "tokens_seen": 2608529408
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021189568706118356,
      "loss": 1.3773,
      "theoretical_loss": 3.385630943642576,
      "tokens_seen": 2608594944
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021187562688064195,
      "loss": 1.3792,
      "theoretical_loss": 3.385624273191975,
      "tokens_seen": 2608660480
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002118555667001003,
      "loss": 1.3717,
      "theoretical_loss": 3.3856176029558704,
      "tokens_seen": 2608726016
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002118355065195587,
      "loss": 1.2661,
      "theoretical_loss": 3.3856109329342505,
      "tokens_seen": 2608791552
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021181544633901706,
      "loss": 1.3186,
      "theoretical_loss": 3.3856042631271026,
      "tokens_seen": 2608857088
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021179538615847542,
      "loss": 1.4803,
      "theoretical_loss": 3.3855975935344147,
      "tokens_seen": 2608922624
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002117753259779338,
      "loss": 1.2668,
      "theoretical_loss": 3.385590924156174,
      "tokens_seen": 2608988160
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021175526579739217,
      "loss": 1.2727,
      "theoretical_loss": 3.3855842549923687,
      "tokens_seen": 2609053696
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021173520561685056,
      "loss": 1.4249,
      "theoretical_loss": 3.385577586042986,
      "tokens_seen": 2609119232
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021171514543630893,
      "loss": 1.4116,
      "theoretical_loss": 3.385570917308015,
      "tokens_seen": 2609184768
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021169508525576732,
      "loss": 1.3453,
      "theoretical_loss": 3.3855642487874418,
      "tokens_seen": 2609250304
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021167502507522568,
      "loss": 1.4517,
      "theoretical_loss": 3.3855575804812545,
      "tokens_seen": 2609315840
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021165496489468407,
      "loss": 1.256,
      "theoretical_loss": 3.385550912389441,
      "tokens_seen": 2609381376
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021163490471414243,
      "loss": 1.4333,
      "theoretical_loss": 3.385544244511989,
      "tokens_seen": 2609446912
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002116148445336008,
      "loss": 1.4089,
      "theoretical_loss": 3.385537576848887,
      "tokens_seen": 2609512448
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021159478435305918,
      "loss": 1.331,
      "theoretical_loss": 3.3855309094001216,
      "tokens_seen": 2609577984
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021157472417251754,
      "loss": 1.376,
      "theoretical_loss": 3.385524242165681,
      "tokens_seen": 2609643520
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021155466399197593,
      "loss": 1.2014,
      "theoretical_loss": 3.3855175751455526,
      "tokens_seen": 2609709056
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002115346038114343,
      "loss": 1.3008,
      "theoretical_loss": 3.385510908339725,
      "tokens_seen": 2609774592
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021151454363089268,
      "loss": 1.3615,
      "theoretical_loss": 3.3855042417481847,
      "tokens_seen": 2609840128
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021149448345035105,
      "loss": 1.3885,
      "theoretical_loss": 3.3854975753709207,
      "tokens_seen": 2609905664
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021147442326980944,
      "loss": 1.442,
      "theoretical_loss": 3.3854909092079195,
      "tokens_seen": 2609971200
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002114543630892678,
      "loss": 1.5015,
      "theoretical_loss": 3.3854842432591696,
      "tokens_seen": 2610036736
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021143430290872616,
      "loss": 1.2835,
      "theoretical_loss": 3.3854775775246586,
      "tokens_seen": 2610102272
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021141424272818455,
      "loss": 1.3221,
      "theoretical_loss": 3.3854709120043744,
      "tokens_seen": 2610167808
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002113941825476429,
      "loss": 1.3884,
      "theoretical_loss": 3.3854642466983043,
      "tokens_seen": 2610233344
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002113741223671013,
      "loss": 1.3391,
      "theoretical_loss": 3.3854575816064365,
      "tokens_seen": 2610298880
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021135406218655966,
      "loss": 1.3908,
      "theoretical_loss": 3.3854509167287583,
      "tokens_seen": 2610364416
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021133400200601808,
      "loss": 1.2656,
      "theoretical_loss": 3.3854442520652577,
      "tokens_seen": 2610429952
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021131394182547644,
      "loss": 1.404,
      "theoretical_loss": 3.3854375876159226,
      "tokens_seen": 2610495488
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021129388164493483,
      "loss": 1.4536,
      "theoretical_loss": 3.3854309233807403,
      "tokens_seen": 2610561024
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002112738214643932,
      "loss": 1.2682,
      "theoretical_loss": 3.385424259359699,
      "tokens_seen": 2610626560
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021125376128385158,
      "loss": 1.2741,
      "theoretical_loss": 3.385417595552786,
      "tokens_seen": 2610692096
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021123370110330994,
      "loss": 1.3643,
      "theoretical_loss": 3.3854109319599894,
      "tokens_seen": 2610757632
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002112136409227683,
      "loss": 1.4075,
      "theoretical_loss": 3.385404268581297,
      "tokens_seen": 2610823168
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002111935807422267,
      "loss": 1.3851,
      "theoretical_loss": 3.385397605416696,
      "tokens_seen": 2610888704
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021117352056168506,
      "loss": 1.376,
      "theoretical_loss": 3.3853909424661746,
      "tokens_seen": 2610954240
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021115346038114345,
      "loss": 1.3989,
      "theoretical_loss": 3.3853842797297204,
      "tokens_seen": 2611019776
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002111334002006018,
      "loss": 1.4555,
      "theoretical_loss": 3.3853776172073218,
      "tokens_seen": 2611085312
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002111133400200602,
      "loss": 1.3922,
      "theoretical_loss": 3.3853709548989652,
      "tokens_seen": 2611150848
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021109327983951856,
      "loss": 1.3469,
      "theoretical_loss": 3.3853642928046397,
      "tokens_seen": 2611216384
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021107321965897695,
      "loss": 1.3794,
      "theoretical_loss": 3.3853576309243323,
      "tokens_seen": 2611281920
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002110531594784353,
      "loss": 1.416,
      "theoretical_loss": 3.3853509692580306,
      "tokens_seen": 2611347456
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021103309929789368,
      "loss": 1.3441,
      "theoretical_loss": 3.385344307805723,
      "tokens_seen": 2611412992
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021101303911735207,
      "loss": 1.357,
      "theoretical_loss": 3.3853376465673968,
      "tokens_seen": 2611478528
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021099297893681043,
      "loss": 1.432,
      "theoretical_loss": 3.38533098554304,
      "tokens_seen": 2611544064
    },
    {
      "epoch": 0.79,
      "objective/train/advantage_avg": 0.4943682849407196,
      "objective/train/docs_used": 949550,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.2004292011260986,
      "objective/train/original_loss": 1.2004292011260986,
      "objective/train/theoretical_loss": 3.38532432473264,
      "objective/train/tokens_used": 2632069600,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2447984665632248,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0506813526153564,
      "objective/train/weighted_lm_loss": 1.2611896991729736,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.38532432473264,
      "tokens_seen": 2611609600
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021097291875626882,
      "loss": 1.4115,
      "theoretical_loss": 3.38532432473264,
      "tokens_seen": 2611609600
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021095285857572718,
      "loss": 1.5438,
      "theoretical_loss": 3.385317664136185,
      "tokens_seen": 2611675136
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021093279839518557,
      "loss": 1.394,
      "theoretical_loss": 3.3853110037536625,
      "tokens_seen": 2611740672
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021091273821464393,
      "loss": 1.3493,
      "theoretical_loss": 3.38530434358506,
      "tokens_seen": 2611806208
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021089267803410232,
      "loss": 1.5273,
      "theoretical_loss": 3.385297683630366,
      "tokens_seen": 2611871744
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021087261785356068,
      "loss": 1.3708,
      "theoretical_loss": 3.385291023889568,
      "tokens_seen": 2611937280
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021085255767301904,
      "loss": 1.3565,
      "theoretical_loss": 3.3852843643626533,
      "tokens_seen": 2612002816
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021083249749247743,
      "loss": 1.4446,
      "theoretical_loss": 3.38527770504961,
      "tokens_seen": 2612068352
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002108124373119358,
      "loss": 1.346,
      "theoretical_loss": 3.385271045950426,
      "tokens_seen": 2612133888
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021079237713139419,
      "loss": 1.3862,
      "theoretical_loss": 3.3852643870650887,
      "tokens_seen": 2612199424
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021077231695085255,
      "loss": 1.5359,
      "theoretical_loss": 3.3852577283935865,
      "tokens_seen": 2612264960
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021075225677031094,
      "loss": 1.3849,
      "theoretical_loss": 3.385251069935906,
      "tokens_seen": 2612330496
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002107321965897693,
      "loss": 1.4285,
      "theoretical_loss": 3.385244411692036,
      "tokens_seen": 2612396032
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002107121364092277,
      "loss": 1.3891,
      "theoretical_loss": 3.3852377536619644,
      "tokens_seen": 2612461568
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021069207622868605,
      "loss": 1.4198,
      "theoretical_loss": 3.385231095845678,
      "tokens_seen": 2612527104
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021067201604814444,
      "loss": 1.1622,
      "theoretical_loss": 3.3852244382431658,
      "tokens_seen": 2612592640
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002106519558676028,
      "loss": 1.3436,
      "theoretical_loss": 3.385217780854415,
      "tokens_seen": 2612658176
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021063189568706116,
      "loss": 1.2179,
      "theoretical_loss": 3.3852111236794125,
      "tokens_seen": 2612723712
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021061183550651955,
      "loss": 1.4202,
      "theoretical_loss": 3.385204466718147,
      "tokens_seen": 2612789248
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021059177532597792,
      "loss": 1.4675,
      "theoretical_loss": 3.3851978099706064,
      "tokens_seen": 2612854784
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021057171514543633,
      "loss": 1.3571,
      "theoretical_loss": 3.385191153436778,
      "tokens_seen": 2612920320
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002105516549648947,
      "loss": 1.2857,
      "theoretical_loss": 3.38518449711665,
      "tokens_seen": 2612985856
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021053159478435308,
      "loss": 1.3638,
      "theoretical_loss": 3.3851778410102096,
      "tokens_seen": 2613051392
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021051153460381145,
      "loss": 1.3423,
      "theoretical_loss": 3.3851711851174455,
      "tokens_seen": 2613116928
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021049147442326984,
      "loss": 1.3773,
      "theoretical_loss": 3.3851645294383443,
      "tokens_seen": 2613182464
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002104714142427282,
      "loss": 1.4113,
      "theoretical_loss": 3.385157873972895,
      "tokens_seen": 2613248000
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021045135406218656,
      "loss": 1.4319,
      "theoretical_loss": 3.3851512187210844,
      "tokens_seen": 2613313536
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021043129388164495,
      "loss": 1.335,
      "theoretical_loss": 3.3851445636829007,
      "tokens_seen": 2613379072
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002104112337011033,
      "loss": 1.2458,
      "theoretical_loss": 3.385137908858332,
      "tokens_seen": 2613444608
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002103911735205617,
      "loss": 1.5062,
      "theoretical_loss": 3.3851312542473657,
      "tokens_seen": 2613510144
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021037111334002006,
      "loss": 1.3777,
      "theoretical_loss": 3.3851245998499895,
      "tokens_seen": 2613575680
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021035105315947845,
      "loss": 1.3342,
      "theoretical_loss": 3.3851179456661913,
      "tokens_seen": 2613641216
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021033099297893681,
      "loss": 1.3587,
      "theoretical_loss": 3.385111291695959,
      "tokens_seen": 2613706752
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002103109327983952,
      "loss": 1.4029,
      "theoretical_loss": 3.38510463793928,
      "tokens_seen": 2613772288
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021029087261785357,
      "loss": 1.452,
      "theoretical_loss": 3.385097984396143,
      "tokens_seen": 2613837824
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021027081243731196,
      "loss": 1.3158,
      "theoretical_loss": 3.3850913310665347,
      "tokens_seen": 2613903360
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021025075225677032,
      "loss": 1.4054,
      "theoretical_loss": 3.3850846779504438,
      "tokens_seen": 2613968896
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021023069207622868,
      "loss": 1.3713,
      "theoretical_loss": 3.385078025047857,
      "tokens_seen": 2614034432
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021021063189568707,
      "loss": 1.336,
      "theoretical_loss": 3.3850713723587633,
      "tokens_seen": 2614099968
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021019057171514543,
      "loss": 1.4397,
      "theoretical_loss": 3.3850647198831503,
      "tokens_seen": 2614165504
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021017051153460382,
      "loss": 1.2338,
      "theoretical_loss": 3.3850580676210047,
      "tokens_seen": 2614231040
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021015045135406218,
      "loss": 1.342,
      "theoretical_loss": 3.385051415572316,
      "tokens_seen": 2614296576
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021013039117352057,
      "loss": 1.3445,
      "theoretical_loss": 3.38504476373707,
      "tokens_seen": 2614362112
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021011033099297894,
      "loss": 1.2803,
      "theoretical_loss": 3.3850381121152564,
      "tokens_seen": 2614427648
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021009027081243732,
      "loss": 1.4082,
      "theoretical_loss": 3.385031460706862,
      "tokens_seen": 2614493184
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002100702106318957,
      "loss": 1.3318,
      "theoretical_loss": 3.385024809511875,
      "tokens_seen": 2614558720
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021005015045135405,
      "loss": 1.3518,
      "theoretical_loss": 3.3850181585302823,
      "tokens_seen": 2614624256
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00021003009027081244,
      "loss": 1.3091,
      "theoretical_loss": 3.385011507762073,
      "tokens_seen": 2614689792
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002100100300902708,
      "loss": 1.2436,
      "theoretical_loss": 3.385004857207234,
      "tokens_seen": 2614755328
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002099899699097292,
      "loss": 1.2644,
      "theoretical_loss": 3.3849982068657534,
      "tokens_seen": 2614820864
    },
    {
      "epoch": 0.79,
      "objective/train/advantage_avg": 0.49283334612846375,
      "objective/train/docs_used": 950739,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.317156195640564,
      "objective/train/original_loss": 1.3171560764312744,
      "objective/train/theoretical_loss": 3.384991556737619,
      "objective/train/tokens_used": 2635346400,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24363349378108978,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505218505859375,
      "objective/train/weighted_lm_loss": 1.3836530447006226,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.384991556737619,
      "tokens_seen": 2614886400
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020996990972918755,
      "loss": 1.413,
      "theoretical_loss": 3.384991556737619,
      "tokens_seen": 2614886400
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020994984954864594,
      "loss": 1.254,
      "theoretical_loss": 3.384984906822819,
      "tokens_seen": 2614951936
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002099297893681043,
      "loss": 1.4133,
      "theoretical_loss": 3.3849782571213405,
      "tokens_seen": 2615017472
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002099097291875627,
      "loss": 1.4268,
      "theoretical_loss": 3.3849716076331715,
      "tokens_seen": 2615083008
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020988966900702106,
      "loss": 1.4069,
      "theoretical_loss": 3.3849649583583004,
      "tokens_seen": 2615148544
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020986960882647942,
      "loss": 1.3982,
      "theoretical_loss": 3.384958309296714,
      "tokens_seen": 2615214080
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002098495486459378,
      "loss": 1.3799,
      "theoretical_loss": 3.384951660448401,
      "tokens_seen": 2615279616
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002098294884653962,
      "loss": 1.3462,
      "theoretical_loss": 3.384945011813349,
      "tokens_seen": 2615345152
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020980942828485459,
      "loss": 1.29,
      "theoretical_loss": 3.3849383633915457,
      "tokens_seen": 2615410688
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020978936810431295,
      "loss": 1.3237,
      "theoretical_loss": 3.3849317151829785,
      "tokens_seen": 2615476224
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020976930792377134,
      "loss": 1.3986,
      "theoretical_loss": 3.3849250671876363,
      "tokens_seen": 2615541760
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002097492477432297,
      "loss": 1.3327,
      "theoretical_loss": 3.3849184194055058,
      "tokens_seen": 2615607296
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002097291875626881,
      "loss": 1.2941,
      "theoretical_loss": 3.3849117718365758,
      "tokens_seen": 2615672832
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020970912738214645,
      "loss": 1.3766,
      "theoretical_loss": 3.384905124480833,
      "tokens_seen": 2615738368
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020968906720160484,
      "loss": 1.4202,
      "theoretical_loss": 3.384898477338266,
      "tokens_seen": 2615803904
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002096690070210632,
      "loss": 1.3165,
      "theoretical_loss": 3.3848918304088627,
      "tokens_seen": 2615869440
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020964894684052156,
      "loss": 1.4139,
      "theoretical_loss": 3.38488518369261,
      "tokens_seen": 2615934976
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020962888665997995,
      "loss": 1.4075,
      "theoretical_loss": 3.3848785371894974,
      "tokens_seen": 2616000512
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020960882647943832,
      "loss": 1.2684,
      "theoretical_loss": 3.384871890899511,
      "tokens_seen": 2616066048
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002095887662988967,
      "loss": 1.3556,
      "theoretical_loss": 3.38486524482264,
      "tokens_seen": 2616131584
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020956870611835507,
      "loss": 1.3624,
      "theoretical_loss": 3.384858598958871,
      "tokens_seen": 2616197120
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020954864593781346,
      "loss": 1.3456,
      "theoretical_loss": 3.3848519533081927,
      "tokens_seen": 2616262656
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020952858575727182,
      "loss": 1.4025,
      "theoretical_loss": 3.3848453078705925,
      "tokens_seen": 2616328192
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002095085255767302,
      "loss": 1.3585,
      "theoretical_loss": 3.3848386626460583,
      "tokens_seen": 2616393728
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020948846539618857,
      "loss": 1.3101,
      "theoretical_loss": 3.384832017634578,
      "tokens_seen": 2616459264
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020946840521564693,
      "loss": 1.4279,
      "theoretical_loss": 3.3848253728361395,
      "tokens_seen": 2616524800
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020944834503510532,
      "loss": 1.3396,
      "theoretical_loss": 3.3848187282507305,
      "tokens_seen": 2616590336
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020942828485456368,
      "loss": 1.2074,
      "theoretical_loss": 3.384812083878339,
      "tokens_seen": 2616655872
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020940822467402207,
      "loss": 1.395,
      "theoretical_loss": 3.384805439718953,
      "tokens_seen": 2616721408
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020938816449348044,
      "loss": 1.4758,
      "theoretical_loss": 3.38479879577256,
      "tokens_seen": 2616786944
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020936810431293883,
      "loss": 1.4071,
      "theoretical_loss": 3.3847921520391475,
      "tokens_seen": 2616852480
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002093480441323972,
      "loss": 1.3424,
      "theoretical_loss": 3.384785508518704,
      "tokens_seen": 2616918016
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020932798395185558,
      "loss": 1.2049,
      "theoretical_loss": 3.384778865211217,
      "tokens_seen": 2616983552
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020930792377131394,
      "loss": 1.408,
      "theoretical_loss": 3.3847722221166747,
      "tokens_seen": 2617049088
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002092878635907723,
      "loss": 1.3973,
      "theoretical_loss": 3.3847655792350646,
      "tokens_seen": 2617114624
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002092678034102307,
      "loss": 1.2941,
      "theoretical_loss": 3.3847589365663744,
      "tokens_seen": 2617180160
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020924774322968905,
      "loss": 1.3703,
      "theoretical_loss": 3.3847522941105925,
      "tokens_seen": 2617245696
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020922768304914744,
      "loss": 1.34,
      "theoretical_loss": 3.384745651867706,
      "tokens_seen": 2617311232
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002092076228686058,
      "loss": 1.2997,
      "theoretical_loss": 3.3847390098377037,
      "tokens_seen": 2617376768
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002091875626880642,
      "loss": 1.3632,
      "theoretical_loss": 3.3847323680205728,
      "tokens_seen": 2617442304
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020916750250752256,
      "loss": 1.3167,
      "theoretical_loss": 3.384725726416301,
      "tokens_seen": 2617507840
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020914744232698095,
      "loss": 1.3618,
      "theoretical_loss": 3.3847190850248765,
      "tokens_seen": 2617573376
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002091273821464393,
      "loss": 1.3845,
      "theoretical_loss": 3.3847124438462868,
      "tokens_seen": 2617638912
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002091073219658977,
      "loss": 1.2566,
      "theoretical_loss": 3.38470580288052,
      "tokens_seen": 2617704448
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020908726178535606,
      "loss": 1.3482,
      "theoretical_loss": 3.384699162127564,
      "tokens_seen": 2617769984
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020906720160481445,
      "loss": 1.393,
      "theoretical_loss": 3.3846925215874073,
      "tokens_seen": 2617835520
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020904714142427284,
      "loss": 1.3774,
      "theoretical_loss": 3.3846858812600362,
      "tokens_seen": 2617901056
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002090270812437312,
      "loss": 1.4824,
      "theoretical_loss": 3.38467924114544,
      "tokens_seen": 2617966592
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002090070210631896,
      "loss": 1.2855,
      "theoretical_loss": 3.3846726012436057,
      "tokens_seen": 2618032128
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020898696088264795,
      "loss": 1.2816,
      "theoretical_loss": 3.3846659615545214,
      "tokens_seen": 2618097664
    },
    {
      "epoch": 0.79,
      "objective/train/advantage_avg": 0.490215927362442,
      "objective/train/docs_used": 951960,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3754942417144775,
      "objective/train/original_loss": 1.3754942417144775,
      "objective/train/theoretical_loss": 3.384659322078175,
      "objective/train/tokens_used": 2638623200,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.2407800555229187,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0502456426620483,
      "objective/train/weighted_lm_loss": 1.4445750713348389,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 0.951229453086853,
      "theoretical_loss": 3.384659322078175,
      "tokens_seen": 2618163200
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020896690070210634,
      "loss": 1.4405,
      "theoretical_loss": 3.384659322078175,
      "tokens_seen": 2618163200
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002089468405215647,
      "loss": 1.2487,
      "theoretical_loss": 3.3846526828145542,
      "tokens_seen": 2618228736
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002089267803410231,
      "loss": 1.2918,
      "theoretical_loss": 3.3846460437636474,
      "tokens_seen": 2618294272
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020890672016048146,
      "loss": 1.3629,
      "theoretical_loss": 3.384639404925442,
      "tokens_seen": 2618359808
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020888665997993982,
      "loss": 1.2875,
      "theoretical_loss": 3.3846327662999256,
      "tokens_seen": 2618425344
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002088665997993982,
      "loss": 1.2916,
      "theoretical_loss": 3.3846261278870866,
      "tokens_seen": 2618490880
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020884653961885657,
      "loss": 1.2997,
      "theoretical_loss": 3.3846194896869126,
      "tokens_seen": 2618556416
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020882647943831496,
      "loss": 1.3469,
      "theoretical_loss": 3.3846128516993916,
      "tokens_seen": 2618621952
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020880641925777332,
      "loss": 1.4813,
      "theoretical_loss": 3.3846062139245117,
      "tokens_seen": 2618687488
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002087863590772317,
      "loss": 1.3482,
      "theoretical_loss": 3.38459957636226,
      "tokens_seen": 2618753024
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020876629889669007,
      "loss": 1.4804,
      "theoretical_loss": 3.3845929390126246,
      "tokens_seen": 2618818560
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020874623871614846,
      "loss": 1.3366,
      "theoretical_loss": 3.384586301875594,
      "tokens_seen": 2618884096
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020872617853560682,
      "loss": 1.4277,
      "theoretical_loss": 3.3845796649511555,
      "tokens_seen": 2618949632
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020870611835506519,
      "loss": 1.4572,
      "theoretical_loss": 3.384573028239297,
      "tokens_seen": 2619015168
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020868605817452358,
      "loss": 1.3337,
      "theoretical_loss": 3.384566391740007,
      "tokens_seen": 2619080704
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020866599799398194,
      "loss": 1.3912,
      "theoretical_loss": 3.3845597554532727,
      "tokens_seen": 2619146240
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020864593781344033,
      "loss": 1.4983,
      "theoretical_loss": 3.384553119379082,
      "tokens_seen": 2619211776
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002086258776328987,
      "loss": 1.3819,
      "theoretical_loss": 3.384546483517423,
      "tokens_seen": 2619277312
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020860581745235708,
      "loss": 1.3047,
      "theoretical_loss": 3.3845398478682833,
      "tokens_seen": 2619342848
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020858575727181544,
      "loss": 1.3357,
      "theoretical_loss": 3.384533212431651,
      "tokens_seen": 2619408384
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020856569709127383,
      "loss": 1.4032,
      "theoretical_loss": 3.3845265772075144,
      "tokens_seen": 2619473920
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002085456369107322,
      "loss": 1.2474,
      "theoretical_loss": 3.3845199421958605,
      "tokens_seen": 2619539456
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020852557673019058,
      "loss": 1.2727,
      "theoretical_loss": 3.384513307396678,
      "tokens_seen": 2619604992
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020850551654964894,
      "loss": 1.3659,
      "theoretical_loss": 3.384506672809954,
      "tokens_seen": 2619670528
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002084854563691073,
      "loss": 1.3113,
      "theoretical_loss": 3.384500038435677,
      "tokens_seen": 2619736064
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002084653961885657,
      "loss": 1.3983,
      "theoretical_loss": 3.3844934042738344,
      "tokens_seen": 2619801600
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020844533600802406,
      "loss": 1.2768,
      "theoretical_loss": 3.384486770324415,
      "tokens_seen": 2619867136
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020842527582748245,
      "loss": 1.3357,
      "theoretical_loss": 3.3844801365874053,
      "tokens_seen": 2619932672
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002084052156469408,
      "loss": 1.3407,
      "theoretical_loss": 3.384473503062794,
      "tokens_seen": 2619998208
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002083851554663992,
      "loss": 1.4209,
      "theoretical_loss": 3.384466869750569,
      "tokens_seen": 2620063744
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020836509528585756,
      "loss": 1.339,
      "theoretical_loss": 3.3844602366507184,
      "tokens_seen": 2620129280
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020834503510531595,
      "loss": 1.3767,
      "theoretical_loss": 3.3844536037632293,
      "tokens_seen": 2620194816
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020832497492477434,
      "loss": 1.4248,
      "theoretical_loss": 3.3844469710880905,
      "tokens_seen": 2620260352
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002083049147442327,
      "loss": 1.3939,
      "theoretical_loss": 3.3844403386252893,
      "tokens_seen": 2620325888
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002082848545636911,
      "loss": 1.3271,
      "theoretical_loss": 3.3844337063748138,
      "tokens_seen": 2620391424
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020826479438314945,
      "loss": 1.4951,
      "theoretical_loss": 3.3844270743366516,
      "tokens_seen": 2620456960
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020824473420260784,
      "loss": 1.4954,
      "theoretical_loss": 3.384420442510791,
      "tokens_seen": 2620522496
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002082246740220662,
      "loss": 1.2763,
      "theoretical_loss": 3.38441381089722,
      "tokens_seen": 2620588032
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002082046138415246,
      "loss": 1.3857,
      "theoretical_loss": 3.384407179495926,
      "tokens_seen": 2620653568
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020818455366098296,
      "loss": 1.3509,
      "theoretical_loss": 3.384400548306897,
      "tokens_seen": 2620719104
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020816449348044135,
      "loss": 1.317,
      "theoretical_loss": 3.384393917330121,
      "tokens_seen": 2620784640
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002081444332998997,
      "loss": 1.2773,
      "theoretical_loss": 3.384387286565586,
      "tokens_seen": 2620850176
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002081243731193581,
      "loss": 1.2945,
      "theoretical_loss": 3.38438065601328,
      "tokens_seen": 2620915712
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020810431293881646,
      "loss": 1.3934,
      "theoretical_loss": 3.3843740256731905,
      "tokens_seen": 2620981248
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020808425275827482,
      "loss": 1.2578,
      "theoretical_loss": 3.3843673955453055,
      "tokens_seen": 2621046784
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002080641925777332,
      "loss": 1.2727,
      "theoretical_loss": 3.384360765629613,
      "tokens_seen": 2621112320
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020804413239719157,
      "loss": 1.2711,
      "theoretical_loss": 3.3843541359261016,
      "tokens_seen": 2621177856
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020802407221664996,
      "loss": 1.362,
      "theoretical_loss": 3.384347506434758,
      "tokens_seen": 2621243392
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020800401203610833,
      "loss": 1.3301,
      "theoretical_loss": 3.3843408771555707,
      "tokens_seen": 2621308928
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020798395185556671,
      "loss": 1.3946,
      "theoretical_loss": 3.3843342480885275,
      "tokens_seen": 2621374464
    },
    {
      "debugging/Compilability": 1.0,
      "debugging/distinct-1-grams": 0.7183788677974512,
      "debugging/entropy-1-grams": 5.311092611595205,
      "debugging/length": 475.35714285714283,
      "debugging/num_segments": 14,
      "debugging/raw_token_scores_avg": 0.007209094241261482,
      "debugging/raw_token_scores_std": 0.01626458764076233,
      "debugging/score": 0.005309415865714915,
      "debugging/score_std": 0.0067471201632161425,
      "epoch": 0.79,
      "objective/train/advantage_avg": 0.4927899241447449,
      "objective/train/docs_used": 953088,
      "objective/train/instantaneous_batch_size": 32,
      "objective/train/instantaneous_microbatch_size": 32768,
      "objective/train/lm_loss": 1.3089274168014526,
      "objective/train/original_loss": 1.3089275360107422,
      "objective/train/theoretical_loss": 3.3843276192336167,
      "objective/train/tokens_used": 2641900000,
      "objective/train/value_avg": -0.5,
      "objective/train/value_loss": 0.24310648441314697,
      "objective/train/value_max": -0.5,
      "objective/train/value_min": -0.5,
      "objective/train/value_reward_corr": NaN,
      "objective/train/value_std": 0.0,
      "objective/train/weight_avg": 1.0505149364471436,
      "objective/train/weighted_lm_loss": 1.3751276731491089,
      "objective/train/weights_max": 1.0512710809707642,
      "objective/train/weights_min": 1.0253151655197144,
      "theoretical_loss": 3.3843276192336167,
      "tokens_seen": 2621440000
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.00020796389167502508,
      "loss": 1.4026,
      "theoretical_loss": 3.3843276192336167,
      "tokens_seen": 2621440000
    }
  ],
  "max_steps": 50354,
  "num_train_epochs": 9223372036854775807,
  "total_flos": 1.33782728343552e+18,
  "trial_name": null,
  "trial_params": null
}